Quant

Quantum Theory at the Crossroads
arXiv:quant-ph/0609184v2 24 Oct 2009
Re onsidering the 1927 Solvay Conferen e
Guido Ba iagaluppi
Antony Valentini
Cambridge University Press

ISBN: 9780521814218
To the memory of James T. Cushing
Contents
Prefa e
Abbreviations
Typographi onventions
Note on the bibliography
Permissions and opyright noti es
page vii
xiii
xiv
xiv
xv
Part I
Perspe tives on the 1927 Solvay onferen e
1
Histori al introdu tion
1.1 Ernest Solvay and the Institute of Physi s

1.2 War and international relations
1.3 S ienti planning and ba kground
1.4 Further details of planning
1.5 The Solvay meeting
1.6 The editing of the pro eedings
1.7 Con lusion
Ar hival notes
De Broglie's pilot-wave theory
2.1
2.2
2.3
2.4
Ba kground
A new approa h to parti le dynami s: 192324
2.2.1 First papers on pilot-wave theory (1923)
2.2.2 Thesis (1924)
2.2.3 Opti al interferen e fringes: November 1924
Towards a omplete pilot-wave dynami s: 192527
2.3.1 `Stru ture': Journal de Physique, May 1927
2.3.2 Signi an e of de Broglie's `Stru ture' paper
1927 Solvay report: the new dynami s of quanta
ii
1
3
3
6
10
16
20
22
24
26
30
30
37
38
43
54
57
61
72
74
Contents
iii
2.5 Signi an e of de Broglie's work from 1923 to 1927

Ar hival notes
84
88
From matrix me hani s to quantum me hani s
3.1
3.2
3.3
Summary of Born and Heisenberg's report

Writing of the report
Formalism
3.3.1 Before matrix me hani s
3.3.2 Matrix me hani s
3.3.3 Formal extensions of matrix me hani s
3.4 Interpretation
3.4.1 Matrix me hani s, Born and Wiener
3.4.2 Born and Jordan on guiding elds, Bohr on
ollisions
3.4.3 Born's ollision papers
3.4.4 Heisenberg on energy u tuations
3.4.5 Transformation theory
3.4.6 Development of the `statisti al view' in the
report
3.4.7 Justi ation and overall on lusions
Ar hival notes
S hrdinger's wave me hani s
4.1
4.2
4.3
4.4
4.5
4.6
Planning of S hrdinger's report

Summary of the report
Parti les as wave pa kets
The problem of radiation
S hrdinger and de Broglie
The oni t with matrix me hani s
4.6.1 Early days
4.6.2 From Muni h to Copenhagen
4.6.3 Continuity and dis ontinuity
Ar hival notes
Part II
Quantum foundations and the 1927 Solvay onferen e
5
Quantum theory and the measurement problem

5.1
5.2
What is quantum theory?

The measurement problem today
5.2.1 A fundamental ambiguity
89
90
93
94
94
96
100
102
103
105
107
109
111
115
119
122
123
124
126
128
133
137
139
140
142
146
150
151
153
153
155
155
iv
Contents
5.2.2
5.2.3
5.2.4
6.1
6.3
6.4
10
157
162
164
Interferen e, superposition, and wave pa ket ollapse 168
6.2
Measurement as a physi al pro ess: quantum

theory `without observers'
Quantum osmology
The measurement problem in `statisti al'
interpretations of
Probability and interferen e

6.1.1 Interferen e in de Broglie's pilot-wave theory
6.1.2 Interferen e in the `quantum me hani s' of
Born and Heisenberg
Ma ros opi superposition: Born's dis ussion of the
loud hamber
6.2.1 Quantum me hani s without wave pa ket
ollapse?
Dira and Heisenberg: interferen e, state redu tion,
and delayed hoi e
Further remarks on Born and Heisenberg's quantum
me hani s
168
169
172
177
178
182
190
Lo ality and in ompleteness
194
194
198
202
Time, determinism, and the spa etime framework
204
204
209
213
Guiding elds in 3-spa e
218
9.2
218
221
7.1
7.2
7.3
8.1
8.2
8.3
9.1
Einstein's 1927 argument for in ompleteness

A pre ursor: Einstein at Salzburg in 1909
More on nonlo ality and relativity
Time in quantum theory
Determinism and probability
Visualisability and the spa etime framework
Einstein's early attempts to formulate a dynami al

theory of light quanta
The failure of energy-momentum onservation
S attering and measurement in de Broglie's pilot-wave

theory
227
10.1 S attering in pilot-wave theory
10.2 Elasti and inelasti s attering: Born and Brillouin,
Pauli and de Broglie
10.3 Quantum measurement in pilot-wave theory
10.4 Re oil of a single photon: Kramers and de Broglie
228
232
244
246
Contents
11
12
Pilot-wave theory in retrospe t
248
250
258
259
265
Beyond the Bohr-Einstein debate
268
269
272
11.1
11.2
11.3
11.4
Histori al mis on eptions

Why was de Broglie's theory reje ted?
Einstein's alternative pilot-wave theory (May 1927)
Obje tions: in 1927 and today
12.1 The standard histori al a ount

12.2 Towards a histori al revision
Part III
The pro eedings of the 1927 Solvay onferen e
H. A. Lorentz
Fifth physi s onferen e
The intensity of X-ray ree tion (W. L. Bragg )
The lassi al treatment of X-ray dira tion phenomena

History of the use of quantitative methods
Results of quantitative analysis
Interpretation of measurements of F
Examples of analysis
The me hanism of X-ray s attering
The analysis of atomi stru ture by X-ray intensity
measurements
The refra tion of X-rays
Referen es
Dis ussion of Mr Bragg's report
Notes to the translation
277
279
281
283
283
285
289
292
295
303
308
312
316
318
327
Disagreements between experiment and the ele tromagneti

theory of radiation (A. H. Compton )
329
Introdu tion
The problem of the ether
The emission of radiation
The photoele tri ee t
Phenomena asso iated with the s attering of X-rays
Intera tions between radiation and single ele trons
Reliability of experimental eviden e
Summary
Dis ussion of Mr Compton's report
329
332
333
335
342
347
352
353
355
371
vi
Contents
The new dynami s of quanta (L. de Broglie )
I. Prin ipal points of view

II. Probable meaning of the ontinuous waves
III. Experiments showing preliminary dire t eviden e for
the new Dynami s of the ele tron
Bibliography
Dis ussion of Mr de Broglie's report
Quantum me hani s (M. Born and W. Heisenberg )
373
373
381
389
396
398
406
407
407
409
419
Introdu tion
I. The mathemati al methods of quantum me hani s
II. Physi al interpretation
III. Formulation of the prin iples and delimitation of
their s ope
IV. Appli ations of quantum me hani s
Con lusion
Bibliography
Dis ussion of Messrs Born and Heisenberg's report
Wave me hani s (E. S hrdinger )
428
432
435
437
441
444
Introdu tion
I. Multi-dimensional theory
II. Four-dimensional theory
III. The many-ele tron problem
Dis ussion of Mr S hrdinger's report
447
447
448
456
461
468
474
General dis ussion of the new ideas presented
476
476
497
500
520
523
Causality, determinism, probability

Photons
Photons and ele trons
Bibliography
Prefa e
And they said one to another: Go to, let us build us a

tower, whose top may rea h unto heaven; and let us make
us a name. And the Lord said: Go to, let us go down, and
there onfound their language, that they may not
understand one another's spee h.
Genesis 11: 37
Anyone who has taken part in a debate on the interpretation of quantum

theory will re ognise how tting is the above quotation from the book of
Genesis, a ording to whi h the builders of the Tower of Babel found that
they ould no longer understand one another's spee h. For when it omes
to the interpretation of quantum theory, even the most lear-thinking
and apable physi ists are often unable to understand ea h other.
This state of aairs dates ba k to the genesis of quantum theory
itself. In O tober 1927, during the `general dis ussion' that took pla e
in Brussels at the end of the fth Solvay onferen e, Paul Ehrenfest
wrote the above lines on the bla kboard. As Langevin later remarked,
the Solvay meeting in 1927 was the onferen e where `the onfusion of
ideas rea hed its peak'.
Ehrenfest's per eptive gesture aptured the essen e of a situation that
has persisted for three-quarters of a entury. A ording to widespread
histori al folklore, the deep dieren es of opinion among the leading
physi ists of the day led to intense debates, whi h were satisfa torily
resolved by Bohr and Heisenberg around the time of the 1927 Solvay
meeting. But in fa t, at the end of 1927, a signi ant number of the
main parti ipants (in parti ular de Broglie, Einstein, and S hrdinger)
remained un onvin ed, and the deep dieren es of opinion were never
really resolved.
The interpretation of quantum theory seems as highly ontroversial
vii
viii
Prefa e
today as it was in 1927. There has also been riti ism on the part
of historians as well as physi ists of the ta ti s used by Bohr and
others to propagate their views in the late 1920s, and a realisation that
alternative ideas may have been dismissed or unfairly disparaged. For
many physi ists, a sense of unease lingers over the whole subje t. Might
it be that things are not as lear- ut as Bohr and Heisenberg would have
us believe? Might it be that their opponents had something important
to say after all? Be ause today there is no longer an established interpretation of quantum me hani s, we feel it is important to go ba k to
the sour es and re-evaluate them.
In this spirit, we oer the reader a return to a time just before the
Copenhagen interpretation was widely a epted, when the best physi ists of the day gathered to dis uss a range of views, on erning many topi s of interest today (measurement, determinism, nonlo ality, subje tivity, interferen e, and so on), and when three distin t theories de
Broglie's pilot-wave theory, Born and Heisenberg's quantum me hani s,
and S hrdinger's wave me hani s were presented and dis ussed on
an equal footing.
*
Sin e the 1930s, and espe ially sin e the Se ond World War, it has
been ommon to dismiss questions about the interpretation of quantum
theory as `metaphysi al' or `just philosophi al'. It will be lear from
the lively and wide-ranging dis ussions of 1927 that at that time, for
the most distinguished physi ists of the day, the issues were de idedly
physi al : Is the ele tron a point parti le with a ontinuous traje tory
(de Broglie), or a wave pa ket (S hrdinger), or neither (Born and
Heisenberg)? Do quantum out omes o ur when nature makes a hoi e
(Dira ), or when an observer de ides to re ord them (Heisenberg)? Is
the nonlo ality of quantum theory ompatible with relativity (Einstein)?
Can a theory with traje tories a ount for the re oil of a single photon
on a mirror (Kramers, de Broglie)? Is indeterminism a fundamental
limitation, or merely the out ome of oarse-graining over something
deeper and deterministi (Lorentz)?
After 1927, the Copenhagen interpretation be ame rmly established.
Rival views were marginalised, in parti ular those represented by de
Broglie, S hrdinger and Einstein, even though these s ientists were responsible for many of the major developments in quantum physi s itself.
(This marginalisation is apparent in most histori al a ounts written
throughout the twentieth entury.) From the very beginning, however,
Prefa e
ix
there were some notes of aution: for example, when Bohr's landmark
paper of 1928 (the English version of his famous Como le ture) was
published in Nature, an editorial prefa e expressed dissatisfa tion with
the `somewhat vague statisti al des ription' and ended with the hope
that this would not be the `last word on the subje t'. And there were a
few outstanding alarm bells, in parti ular the famous paper by Einstein,
Podolsky and Rosen in 1935, and the important papers by S hrdinger
(in the same year) on the at paradox and on entanglement. But on
the whole, the questioning eased in all but a few orners. A general
opinion arose that the questions had been essentially settled, and that a
satisfa tory point of view had been arrived at, prin ipally through the
work of Bohr and Heisenberg. For subsequent generations of physi ists,
`shut up and al ulate' emerged as the working rule among the vast
majority.
Despite this atmosphere, the questioning never ompletely died out,
and some very signi ant work was published, for example by Bohm in
1952, Everett in 1957, and Bell in 1964 and 1966. But attitudes hanged
very slowly. Younger physi ists were strongly dis ouraged from pursuing
su h questions. Those who persisted generally had di ult areers, and
mu h of the areful thinking about quantum foundations was relegated
to departments of philosophy.
Nevertheless, the losing de ade of the twentieth entury saw a resurgen e of interest in the foundations of quantum theory. At the time of
writing, a range of alternatives (su h as hidden variables, many worlds,
ollapse models, among others) are being a tively pursued, and the
Copenhagen interpretation an no longer laim to be the dominant or
òrthodox' interpretation.
The modern reader familiar with urrent debates and positions in
quantum foundations will re ognise many of the standard points of view
in the dis ussions reprodu ed here, though expressed with a remarkable
on ision and larity. This provides a wel ome ontrast with the generally poor level of debate today: as the distinguished osmologist Dennis
S iama was fond of pointing out, when it omes to the interpretation
of quantum theory `the standard of argument suddenly drops to zero'.
We hope that the publi ation of this book will ontribute to a revival of
sharp and informed debate about the meaning of quantum theory.
*
Remarkably, the pro eedings of the fth Solvay onferen e have not
re eived the attention they deserve, neither from physi ists nor from
Prefa e
historians, and the literature ontains numerous major misunderstandings about what took pla e there.
The fth Solvay onferen e is usually remembered for the lash that
took pla e between Einstein and Bohr over the un ertainty relations. It is
remarkable, then, to nd that not a word of these dis ussions appears in
the published pro eedings. It is known that Einstein and Bohr engaged
in vigorous informal dis ussions, but in the formal debates re orded
in the pro eedings they were relatively silent. Bohr did ontribute to
the general dis ussion, but this material was not published. Instead, at
Bohr's request, it was repla ed by a translation of the German version
of his Como le ture, whi h appeared in Naturwissens haften in 1928.
(We do not reprodu e this well-known paper here.) The appending of
this translation to the published pro eedings may be the ause of the
ommon misunderstanding that Bohr gave a report at the onferen e:
in fa t, he did not.
Born and Heisenberg present a number of unfamiliar viewpoints on erning, among other things, the nature of the wave fun tion and the role
of time and of probability in quantum theory. Parti ularly surprising is
the seeming absen e of a ollapse postulate in their formulation, and the
apparently phenomenologi al status of the time-dependent S hrdinger
equation. Born and Heisenberg's `quantum me hani s' seems remarkably
dierent from quantum me hani s (in the Dira -von Neumann formulation) as we know it today.
De Broglie's pilot-wave theory was the subje t of extensive and varied
dis ussions. This is rather startling in view of the laim in Max
Jammer's lassi histori al study The Philosophy of Quantum Me hani s
that de Broglie's theory `was hardly dis ussed at all' and that `the only
serious rea tion ame from Pauli' (Jammer 1974, pp. 11011). Jammer's
view is typi al even today. But in the published pro eedings, at the end of
de Broglie's report there are 9 pages of dis ussion devoted to de Broglie's
theory, and of the 42 pages of general dis ussion, 15 ontain dis ussion of
de Broglie's theory, with serious rea tions and omments oming not only
from Pauli but also from Born, Brillouin, Einstein, Kramers, Lorentz,
S hrdinger and others. Even the well-known ex hange between Pauli
and de Broglie has been widely misunderstood.
Finally, another surprise is that in his report de Broglie proposed the
many-body pilot-wave dynami s for a system of parti les, with the total
onguration guided by a wave in onguration spa e, and not just (as is
generally believed) the one-body theory in 3-spa e. De Broglie's theory
is essentially the same as that developed by Bohm in 1952, the only
Prefa e
xi
dieren e being that de Broglie's dynami s (like the form of pilot-wave

theory popularised by Bell) is formulated in terms of velo ity rather than
a eleration.
*
This work is a translation of and ommentary on the pro eedings of
the fth Solvay onferen e of 1927, whi h were published in Fren h in
1928 under the title le trons et Photons.
We have not attempted to give an exhaustive histori al analysis of
the fth Solvay onferen e. Rather, our main aims have been to present
the material in a manner a essible to the general physi ist, and to
situate the pro eedings in the ontext of urrent resear h in quantum
foundations. We hope that the book will ontribute to stimulating and
reviving serious debate about quantum foundations in the wider physi s
ommunity, and that making the pro eedings available in English will
en ourage historians and philosophers to re onsider their signi an e.
Part I begins with a histori al introdu tion and provides essays on
the three main theories presented at the onferen e (pilot-wave theory,
quantum me hani s, wave me hani s). The le tures and dis ussions that
took pla e at the fth Solvay onferen e ontain an extensive range
of material that is relevant to urrent resear h in the foundations of
quantum theory. In Part II, after a brief review of the status of quantum
foundations today, we summarise what seem to us to be the highlights
of the onferen e, from the point of view of urrent debates about the
meaning of quantum theory. Part III of the book onsists of translations
of the reports, of the dis ussions following them, and of the general
dis ussion. Wherever possible, the original (in parti ular English or German) texts have been used. We have ta itly orre ted minor mistakes in
pun tuation and spelling, and we have uniformised the style of equations,
referen es and footnotes. (Unless otherwise spe ied, all translations of
quotations are ours.)
Part I (ex ept for hapter 2) and the reports by Compton, by Born
and Heisenberg and by S hrdinger, are prin ipally the work of Guido
Ba iagaluppi. Chapter 2, Part II, and the reports by Bragg and by de
Broglie and the general dis ussion in Part III, are prin ipally the work
of Antony Valentini.
Chapters 2, 10 and 11 are based on a seminar, `The early history of
Louis de Broglie's pilot-wave dynami s', given by Antony Valentini at
the University of Notre Dame in September 1997, at a onferen e in
honour of the sixtieth birthday of the late James T. Cushing.
xii
Prefa e
*
To James T. Cushing, physi ist, philosopher, historian and gentleman,

we both owe a spe ial and heartfelt thanks. It was he who brought us together on this proje t, and to him we are indebted for his en ouragement
and, above all, his example. This book is dedi ated to his memory.
Guido Ba iagaluppi wishes to express his thanks to the Humboldt
Foundation, whi h supported the bulk of his work in the form of an
Alexander von Humboldt Fors hungsstipendium, and to his hosts in
Germany, Carsten Held and the Philosophis hes Seminar I, University
of Freiburg, and Harald Atmanspa her and the Institut fr Grenzgebiete
der Psy hologie und Psy hohygiene, Freiburg, as well as to Ja ques
Dubu s and the Institut d'Histoire et de Philosophie des S ien es et des
Te hniques (CNRS, Paris 1, ENS) for support during the nal phase. He
also wishes to thank Didier Devriese of the Universit Libre de Bruxelles,
who is in harge of the ar hives of the Instituts Internationaux de Physique et de Chimie Solvay, Universit Libre de Bruxelles, for his kindness
and availability, and Brigitte Parakenings (formerly Uhlemann) and her
sta at the Philosophis hes Ar hiv of the University of Konstanz, for
the ontinuous assistan e with the Ar hive for the History of Quantum
Physi s. Finally, he should wish to thank Je Barrett for suggesting
this proje t to him in Utre ht one day ba k in 1996, as well as Mark van
Atten, Jennifer Bailey, Olivier Darrigol, Feli ity Pors, Gregor S hiemann
and many others for dis ussions, suggestions, orresponden e, referen es
and other help.
Antony Valentini began studying these fas inating pro eedings while
holding a postdo toral position at the University of Rome `La Sapienza'
(199496), and is grateful to Mar ello Cini, Bruno Bertotti and Dennis
S iama for their support and en ouragement during that period. For
support in re ent years, he is grateful to Perimeter Institute, and wishes
to express a spe ial thanks to Howard Burton, Lu ien Hardy and Lee
Smolin.
We are both grateful to Tamsin van Essen at Cambridge University
Press for her support and en ouragement during most of the gestation
of this book, and to Augustus College for support during the nal stages
of this work.
Guido Ba iagaluppi
Antony Valentini
Lake Maggiore, August 2006
Abbreviations
xiii
Abbreviations
AEA: Albert Einstein Ar hives, Jewish National and University Library,
Hebrew University of Jerusalem.
AHQP: Ar hive for the History of Quantum Physi s.
AHQP-BSC: Bohr S ienti Corresponden e, mi rolmed from the Niels
Bohr Arkiv, Copenhagen.
AHQP-BMSS: Bohr S ienti Manus ripts, mi rolmed from the Niels
Bohr Arkiv, Copenhagen.
AHQP-EHR: Ehrenfest olle tion, mi rolmed from the Rijksmuseum
voor de Ges hiedenis van de Natuurwetens happen en van de Geneeskunde `Museum Boerhaave', Leiden.
AHQP-LTZ: Lorentz olle tion, mi rolmed from the Algemeen Rijksar hief, Den Haag.
AHQP-RDN: Ri hardson Colle tion, mi rolmed from the Harry Ransom Humanities Resear h Center, University of Texas at Austin.
AHQP-OHI: Oral history interview trans ripts.
IIPCS: Ar hives of the Instituts Internationaux de Physique et de Chimie
Solvay, Universit Libre de Bruxelles.
Ann. d. Phys. or Ann. der Phys.: Annalen der Physik.
Bayr. Akad. d. Wiss. Math. phys. Kl.: Sitzungsberi hte der Mathematis h-Physikalis hen Klasse der Knigli h-Bayeris hen Akademie der Wissens haften (Mn hen).
Berl. Ber.: Sitzungsberi hte der Preussis hen Akademie der Wissens haften (Berlin).
A ad. Roy. Belg. or Bull. A . R. Belg. or Bull. A . roy. de Belgique or
Bull. A . roy. Belgique or Bull. A . roy. Belg. or Bull. A . R. Belg., Cl.
des S ien es: Bulletin de l'A admie Royale des S ien es, des Lettres et
des Beaux-arts de Belgique. Classe des S ien es.
Bull. Natl. Res. Coun.: Bulletin of the National Resear h Coun il (U.S.).
Comm. Fenn.: Commentationes Physi o-mathemati ae, So ietas S ientiarum Fenni a.
C. R. or C. R. A ad. S . or Comptes Rendus A ad. S i. Paris: Comptes
Rendus Hebdomadaires des San es de l'A admie des S ien es (Paris).
Gtt. Na hr.: Na hri hten der Akademie der Wissens haften in Gttingen. II, Mathematis h-Physikalis he Klasse.
J. de Phys. or Jour. de Phys. or Journ. Physique or Journ. d. Phys.:
Journal de Physique (until 1919), then Journal de Physique et le Radium.
Jour. Frank. Inst.: Journal of the Franklin Institute.
xiv
Lin ei Rend.: Rendi onti Lin ei.

Man hester Memoirs: Man hester Literary and Philosophi al So iety,
Memoirs and Pro eedings.
Math. Ann. or Mathem. Ann.: Mathematis he Annalen.
Naturw. or Naturwiss. or Naturwissens h. or Naturwissens haften: Die
Naturwissens haften.
Nat. A ad. S i. Pro . or Pro . Nat. A ad. S i. or Pro . Nat. A ad.:
Pro eedings of the National A ademy of S ien es (U.S.).
Phil. Mag.: Philosophi al Magazine.
Phil. Trans. or Phil. Trans. Roy. So .: Philosophi al Transa tions of the
Royal So iety of London.
Phys. Rev.: Physi al Review.
Phys. Zeits. or Phys. Zeits h. or Physik. Zts.: Physikalis he Zeits hrift.
Pro . Camb. Phil. So . or Pro . Cambr. Phil. So . or Pro . Cambridge
Phil. So .: Pro eedings of the Cambridge Philosophi al So iety.
Pro . Phys. So .: Pro eedings of the Physi al So iety of London.
Pro . Roy. So . or Roy. So . Pro .: Pro eedings of the Royal So iety of
London.
Upsala Univ. rsskr.: Uppsala Universitets rsskrift.
Z. f. Phys. or Zts. f. Phys. or Zeit. f. Phys. or Zeits. f. Phys. or Zeits h.
f. Phys. or Zeits hr. f. Phys.: Zeits hrift fr Physik.
Typographi onventions
The following onventions have been used.
Square bra kets [ denote editorial amendments or (in the translations)
original wordings.
Curly bra kets { } denote additions (in original types ripts or manus ripts).
Angle bra kets < > denote an ellations (in original types ripts or manus ripts).

The referen es ited in Parts I and II, and in the endnotes and editorial
footnotes to Part III, are listed in our bibliography. The referen es
ited in the original Solvay volume are found in the translation of the
pro eedings in Part III.
xv
Part I
Perspe tives on the 1927 Solvay onferen e
1
Quantum re on iliation very [added, deleted unpleasant

[deleted tenden y [deleted retrograde [deleted
questionable [added, deleted idea [deleted ippant
[deleted title leads to misunderstanding.
Ehrenfest, on the onferen e plans
The onferen e was surely the most interesting s ienti

onferen e I have taken part in so far.
2
Heisenberg, upon re eipt of the onferen e photograph
The early Solvay onferen es were remarkable o asions, made possible

by the generosity of Belgian industrialist Ernest Solvay and, with the
ex eption of the rst onferen e in 1912, planned and organised by the
indefatigable Hendrik Antoon Lorentz. In this hapter, we shall rst
sket h the beginnings of the Solvay onferen es, Lorentz's involvement
and the situation in the years leading up to 1927 (se tions 1.1 and 1.2).
Then we shall des ribe spe i ally the planning of the fth Solvay onferen e, both in its s ienti aspe ts (se tion 1.3) and in its more pra ti al
aspe ts (se tion 1.4). Se tion 1.5 presents the day-by-day progress of
the onferen e as far as it an be re onstru ted from the sour es, while
se tion 1.6 follows the making of the volume of pro eedings, whi h is
the main sour e of original material from the fth Solvay onferen e
and forms Part III of this book.

Ernest Solvay had an extensive re ord of supporting s ienti , edu ational and so ial initiatives, as Lorentz emphasises in a two-page do u3
ment written in September 1914, during the rst months of the rst
world war:3
I feel bound to say some words in these days about one of Belgium's noblest
itizens, one of the men whom I admire and honour most highly. Mr Ernest
Solvay .... is the founder of one of the most ourishing industries of the world,
the soda manufa ture based on the pro ess invented by him and now spread
over Belgium, Fran e, England, Germany, Russia and the United States. ....
The fortune won by an a tivity of half a entury has been largely used by Mr
Solvay for the publi benet. In the rm onvi tion that a better understanding
of the laws of nature and of human so iety will prove one of the most powerful
means for promoting the happiness of mankind, he has in many ways and on
a large s ale en ouraged and supported s ienti resear h and tea hing.
Part of this a tivity was entred around the proje t of the Cit S ientique, a series of institutes in Brussels founded and endowed by Ernest
Solvay and by his brother Alfred Solvay, whi h ulminated in the founding of the Institutes of Physi s and of Chemistry in 1912 and 1913.a
This proje t had originally developed through the han e en ounter between Ernest Solvay and Paul Hger, physi ian and professor of
physiology at the Universit Libre de Bruxelles (ULB), and involved a
ollaboration between Solvay, the ULB and the ity of Brussels. In June
1892, it was agreed that Solvay would onstru t and equip two Institutes
of Physiology on land owned by the ity in the Par Lopold in Brussels.b
There soon followed in 189394 an Institute for Hygiene, Ba teriology
and Therapy, funded mainly by Alfred Solvay, and a S hool of Politi al
and So ial S ien es, founded by Ernest Solvay in 1894, whi h moved to
the Cit S ientique in 1901, and to whi h a S hool of Commer e was
added in 1904.
The idea for what be ame known as the rst Solvay onferen e in
physi s goes ba k to Wilhelm Nernst and Max Plan k,c who around 1910
onsidered that the urrent problems in the theory of radiation and in
the theory of spe i heats had be ome so serious that an international
meeting (indeed a ` oun il') should be onvened in order to attempt to
resolve the situation. The further en ounter between Nernst and Solvay
provided the material opportunity for the meeting, and by July 1910,
Nernst was sending Solvay the detailed proposals. He had also se ured
a The following material on the Cit S ientique is drawn mainly from Despy-Meyer
and Devriese (1997).
b One was to be ome property of the ity and given in use to the ULB, while the
other was to be leased for thirty years to and run by Solvay himself.
In the rest of this and in part of the following se tions, we draw on an unpublished
ompilation of the ontents of the Solvay ar hives by J. Pelseneer.4
the ollaboration of Lorentz (who was eventually asked to preside), of

Knudsen and naturally of Plan k, who wrote:
.... anything that may happen in this dire tion will ex ite my greatest interest
and .... I promise already my parti ipation in any su h endeavour. For I an
say without exaggeration that in fa t for the past 10 years nothing in physi s
has so ontinuously stimulated, ex ited and irritated me as mu h as these
a
quanta of a tion.
Lorentz set up a ommittee to onsider questions relating to the

new experimental resear h that had been deemed ne essary during the
onferen e (whi h took pla e between 30 O tober and 3 November 1911).
This ommittee in luded Marie Curie, Brillouin, Warburg, Kamerlingh
Onnes, Nernst, Rutherford and Knudsen. Lorentz in turn was asked to
be the president. Further, at the end of the onferen e, Solvay proposed
to Lorentz the idea of a s ienti foundation. Lorentz's reply to Solvay's
proposals, of 4 January 1912, in ludes extremely detailed suggestions on
the fun tions and stru ture of the foundation, all of whi h were put into
pra ti e and whi h an be summarised as follows.6
The foundation would be devoted prin ipally to physi s and physi al
hemistry, as well as to questions relating to physi s from other s ien es.
It would provide international support to resear hers (à Rutherford, a
Lenard, a Weiss') in the form of money or loan of s ienti instruments,
and it would provide s holarships for young Belgian s ientists (both
men and women) to work in the best laboratories or universities, mostly
abroad. The question of a link between the foundation and the `Conseil
de physique' was left open, but Lorentz suggested to provide meeting
fa ilities if Solvay wished to link the two. Lorentz suggested instituting
an administrative board ( onsisting of a Solvay family member or appointee, an appointee of the King, and a member of the Belgian s ienti
establishment) and a s ienti ommittee (whi h ould initially be the
one he had formed during the rst Solvay onferen e). Finally, Lorentz
suggested housing the foundation in an annex of one of the existing
institutes in the Cit Universitaire.
During January, Solvay sent Paul Hger to Leiden to work with Lorentz on the statutes of the foundation, whi h Lorentz sent to Solvay
on 2 February. Solvay approved them with hardly any modi ations
(only su h as were required by the Belgian legislation of the time). The
foundation, or rather the `Solvay International Institute of Physi s', was
a Exquisite ending in the original: `.... dass mi h seit 10 Jahren im Grunde ni hts in
der Physik so ununterbro hen an-, er-, und aufregt wie diese Wirkungsquanten'.5
o ially established on 1 May 1912, whi h predates by several years the

establishment of the omparable Belgian state institutions (Fondation
Universitaire: 1920; Fonds National de la Re her he S ientique: 1928).
In this onne tion, Lorentz hoped `that governments would understand
more and more the importan e of s ienti resear h and that in the
long run one will arrive at a satisfa tory organisation, independent of
the individual eorts of private persons',7 a sentiment e hoed by Solvay
himself.a
The institute, whi h Solvay had endowed for thirty years, ould soon
boast of remarkable a tivity in supporting s ienti resear h. The numerous re ipients of subsidies granted during the rst two years until the
rst world war in luded Lebedew's laboratory, von Laue, Sommerfeld,
Fran k and Hertz, W. L. Bragg (who was later to be ome president
of the s ienti ommittee), Stark, and Wien. In 1913, an Institute of
Chemistry followed suit, organised along similar lines to the Institute of
Physi s.

The rst meeting of the s ienti ommittee, for the planning of the
se ond Solvay onferen e, took pla e on 30 September and 1 O tober
1912. The onferen e was held the following year, but the a tivities of
the institute were soon disrupted by the start of the rst world war, in
parti ular the German invasion of Belgium.
Immediate pra ti al disruption in luded the fear of requisitions, the
di ulty of ommuni ation between the international membership of
the s ienti ommittee and, with regard to the publi ation of the pro eedings of the se ond Solvay onferen e, the impossibility of sending
Lorentz the proofs for orre tion and the eventual prospe t of German
ensorship.a
The war, however, had longer-term negative impli ations for international intelle tual ooperation. In O tober 1914, a group of 93 representatives of German s ien e and ulture signed the manifesto Àn die
a `Mr Solvay also thinks that it is the role of the state to subsidise and organise
s ienti institutions, and he hopes that in thirty years the state will fulll this
duty better than it does today.'8
a The pro eedings of the rst Solvay onferen e had had both a Fren h and a German
edition. Those of the se ond Solvay onferen e were printed in three languages in
1915, but never published in this form and later mostly destroyed. Only under
the hanged onditions after the war, in 1921, were the pro eedings published in
a Fren h translation ( arried out, as on later o asions, by J.-. Vers haelt).
Kulturwelt!', denying German responsibilities in the war.a Among the

signatories were both Nernst and Plan k. This manifesto was partly
responsible for the very strong hostility of Fren h and Belgian s ientists
and institutions towards renewal of s ienti relations with Germany
after the war.
No Germans or Austrians were invited to the third Solvay onferen e
of 1921. The only ex eption (whi h remained problemati until the last
minute) was Ehrenfest, who was Austrian, but who had remained in
Leiden throughout the war as Lorentz's su essor. Similarly, no Germans
parti ipated in the fourth Solvay onferen e of 1924. Fren h and Belgian
armies had o upied the Ruhr in January 1923, and the international
situation was parti ularly tense. Einstein had (temporarily) resigned
from the League of Nations' Committee on Intelle tual Cooperation, and
wrote to Lorentz that he would not parti ipate in the Solvay onferen e
be ause of the ex lusion of the German s ientists, and that he should
please make sure that no invitation was sent.9 Bohr also de lined to parti ipate in the onferen e apparently be ause of the ontinued ex lusion
of German s ientists (Moore 1989, p. 157). S hrdinger, however, who
was Austrian and working in Switzerland, was invited.a
Einstein had distinguished himself by assuming a pa ist position
during the war.b Lorentz was pointing out Einstein's ex eptional ase to
Solvay already in January 1919:
However, in talking about the Germans, we must not lose sight of the fa t that
they ome in all kinds of nuan es. A man like Einstein, the great and profound
physi ist, is not `German' in the sense one often atta hes to the word today;
his judgement on the events of the past years will not dier at all from yours
11
or mine.
a The main laims of the manifesto were: `.... It is not true that Germany is the
ause of this war. .... It is not true that we have wantonly [freventli h infringed
the neutrality of Belgium. .... It is not true that the life and property of a single
Belgian itizen has been tou hed by our soldiers, ex ept when utter self-defen e
required it. .... It is not true that our troops have raged brutally against Leuven. ....
It is not true that our ondu t of war disregards the laws of international right. ....
It is not true that the struggle against our so- alled militarism is not a struggle
against our ulture ....' (translated from Bhme 1975, pp. 479).
a Van Aubel (a member of the s ienti ommittee) obje ted strongly in 1923 to the
possibility of Einstein being invited to the fourth Solvay onferen e, and resigned
when it was de ided to invite him. It appears he was onvin ed to remain on the
ommittee.10
b For instan e, Einstein was one of only four signatories of the ounter-manifesto
Àufruf an die Europer' (Ni olai 1917). Note also that Einstein had renoun ed
his German itizenship and had be ome a Swiss itizen in 1901, although there
was some un ertainty about his itizenship when he was awarded the Nobel prize
(Pais 1982, pp. 45 and 5034).
In the meantime, after the treaty of Lo arno of 1925, Germany was

going to join the League of Nations, but the details of the negotiations
were problemati .a As early as February 1926, one nds mention of
the prospe t of renewed in lusion of German s ientists at the Solvay
onferen es.13 In the same month, Kamerlingh Onnes died, and at the
next meeting of the s ienti ommittee, in early April (at whi h the
fth Solvay onferen e was planned), it was de ided to propose both
to invite Einstein to repla e Onnes and to in lude again the German
s ientists.
On 1 April, Charles Lefbure, then se retary of the administrative
ommission, wrote to ommission members Armand Solvay and Jules
Bordet,a enquiring about the admissibility of `moderate gures like Einstein, Plan kb and others'16 (Bordet telegraphed ba k: `Germany will
soon be League of Nations therefore no obje tion'17 ). On 2 April, Lorentz
himself had a long interview with the King, who gave his approval.
Thus, nally, Lorentz wrote to Einstein on 6 April, informing him
of the unanimous de ision by the members of the ommittee present
at the meeting,c as well as of the whole administrative ommission, to
invite him to su eed Kamerlingh Onnes. The Solvay onferen es were
to readmit Germans, and if Einstein were a member of the ommittee,
Lorentz hoped this would en ourage the German s ientists to a ept the
invitation.18 Einstein was favourably impressed by the positive Belgian
attitude and glad to a ept under the altered onditions.19 Lorentz
pro eeded to invite the German s ientists, `not be ause there should
be su h a great haste in the thing, rather to show the Germans as soon
a Lorentz to Einstein on 14 Mar h 1926: `Things are bad with the League of Nations;
if only one ould yet nd a way out until the day after tomorrow'.12 Negotiations
provisionally broke down on 17 Mar h, but Germany eventually joined the League
in September 1926.
a Lefbure was the appointee of the Solvay family to the administrative ommission,
and as su h su eeded Eugne Tassel, who had died in O tober 1922 and had been
a long-standing ollaborator of Ernest Solvay sin e 1886. Armand Solvay was the
son of Ernest Solvay, who had died on 26 May 1922. Bordet was the royal appointee
to the ommission, and had just been appointed in February 1926, following the
death of Paul Hger.14
b A ording to Lorentz, Plan k had always been helpful to him when he had tried
to intervene with the German authorities during the war. Further, Plan k had
somewhat qualied his position with regard to the Kulturwelt manifesto in an
open letter, whi h he asked Lorentz to publish in the Dut h newspapers in 1916.
On the other hand, he expli itly ruled out a publi disavowal of the manifesto in
De ember 1923.15
Listed as Marie Curie, Langevin, Ri hardson, Guye and Knudsen (with two
members absent, W. H. Bragg and Van Aubel).
as possible our good will',20 and sent the informal invitations to Born,
Heisenberg and Plan k (as well as to Bohr) in or around June 1926.21
As late as O tober 1927, however, the issue was still a sensitive one.
Van Aubel (who had not been present at the April 1926 meeting of the
s ienti ommittee) replied in the negative to the o ial invitation
to the onferen e.a Furthermore, it was proposed to release the list
of parti ipants to the press only after the onferen e to avoid publi
demonstrations. Lorentz travelled in person to Brussels on 17 O tober
to dis uss the matter.23
Lorentz's own position during and immediately after the war, as a
physi ist from one of the neutral ountries, had possibly been rather
deli ate. In the text on Ernest Solvay from whi h we have quoted at the
beginning of this hapter, for instan e, he appears to be defending the
impartiality of the poli ies of the Institute of Physi s in the years leading
up to the war. Lorentz started working for some form of re on iliation as
soon as the war was over, writing as follows to Solvay in January 1919:
All things onsidered, I think I must propose to you not to ex lude formally
the Germans, that is, not to lose the door on them forever. I hope that it may
be open for a new generation, and even that maybe, in the ourse of the years,
one may admit those of today's s holars who one an believe regret sin erely
and honestly the events that have taken pla e. Thus German s ien e will be
24
able to regain the pla e that, despite everything, it deserves for its past.
It should be noted that Lorentz was not only the s ienti organiser of
the Solvay institute and the Solvay onferen es, but also a prime mover
behind eorts towards international intelle tual ooperation, through his
heavy involvement with the Conseil International de Re her hes, as well
as with the League of Nations' Committee on Intelle tual Cooperation,
of whi h he was a member from 1923 and president from 1925.b
Lorentz's gure and ontributions to the Solvay onferen es are movingly re alled by Marie Curie in her obituary of Lorentz in the pro eedings of the fth Solvay onferen e (whi h opens Part III of this volume).
a Lefbure's omment was: `be ause there are Germans! Then why does he stay in
the Institute of Physi s?'22
b The Conseil International de Re her hes (founded in 1919) has today be ome
the International Coun il for S ien e (ICSU). The Committee on Intelle tual
Cooperation (founded in 1922) and the related International Institute of
Intelle tual Cooperation (inaugurated in Paris in 1926) were the forerunners of
UNESCO.25
10

What was at issue in the remark that heads this hapter,a s ribbled by
Ehrenfest in the margin of a letter from Lorentz, was the proposed topi
for the fth Solvay onferen e, namely `the oni t and the possible
re on iliation between the lassi al theories and the theory of quanta'.26
Ehrenfest found the phrasing obje tionable in that it en ouraged one
to `swindle away the fruitful and suggestive harshness of the oni t by
most slimy un lear thinking, quite in analogy with what happened also
even after 1900 with the me hani al ether theories of the Maxwell equations', pointing out that `Bohr feels even more strongly than me against
this slogan [S hlagwort, pre isely be ause he takes it so parti ularly to
heart to nd the foundations of the future theory'.27
Lorentz took Ehrenfest's suggestion into a ount, and dropped the
referen e to re on iliation both from the title and from later des riptions
of the fo us of the meeting.a
The meeting of the s ienti ommittee for the planning of the fth
Solvay onferen e took pla e in Brussels on 1 and 2 April 1926. Lorentz
reported a few days later to Einstein:
As the topi for 1927 we have hosen `The quantum theory and the lassi al
theories of radiation', and we hope to have the following reports or le tures:
1
2
3
W. L. Bragg. New tests of the lassi al theory.

A. H. Compton. Compton ee t and its onsequen es.
C. T. R. Wilson. Observations on photoele trons and ollision ele trons by
the ondensation method.
4
5
L. de Broglie. Interferen e and light quanta.

(short note): Kramers. Theory of Slater-Bohr-Kramers and analogous the-
ories.
Einstein. New derivations of Plan k's law and appli ations of statisti s to
quanta.
Heisenberg. Adaptation of the foundations of dynami s to the quantum

29
theory.
Another report, by the ommittee's se retary Vers haelt,30 adds, on erning point 5 : `(at least, if Mr Kramers judges that it is still useful)'; it
a In the original: `Quantenverzoening <{zeer} antipathik<e>> <tendentie>
<retrograde> <{bedenkelijk<e>}> <idee> <loszinnige> [? titel wekt
misverstand'. Many thanks to Mark van Atten for help with this passage.
a To Bohr in June 1926: `.... the oni t between the lassi al theories and the
quantum theory .... '; to S hrdinger in January 1927: ` .... the ontrast between
the urrent and the earlier on eptions [Auassungen and the attempts at
development of a new me hani s'.28
11
further lists a few alternative speakers: Compton or Debye for 2 , Einstein

or Ehrenfest for 6 , and Heisenberg or S hrdinger for 7.a
Thus, the fth Solvay onferen e, as originally planned, was to fo us
mainly on the theory of radiation and on light quanta, in luding only
one report on the new quantum theory of matter. The shift in fo us
between 1926 and 1927 was learly due to major theoreti al advan es (for
example by S hrdinger and Dira ) and new experimental results (su h
as the Davisson-Germer experiments), and it an be partly followed as
the planning of the onferen e progressed.
S hrdinger's wave me hani s was one of the major theoreti al developments of the year 1926. Einstein, who had been alerted to S hrdinger's rst paper by Plan k ( f. Przibram 1967, p. 23), suggested to
Lorentz that S hrdinger should talk at the onferen e instead of himself,
on the basis of his new `theory of quantum states', whi h he des ribed
as a development of genius of de Broglie's ideas.31 While it is un lear
whether Lorentz knew of S hrdinger's papers by the time of the April
meeting,a S hrdinger was listed a week later as a possible substitute
for Heisenberg, and Lorentz himself was assuring Einstein at the end
of April that S hrdinger was already being onsidered, spe ially as
a substitute for the report on the new foundations of dynami s rather
than for the report on quantum statisti s.b
Lorentz losely followed the development of wave me hani s, indeed
ontributing some essential ritique in his orresponden e with S hrdinger from this period, for the most part translated in Przibram (1967)
(see hapter 4, espe ially se tions 4.3 and 4.4, for some more details
on this orresponden e). Lorentz also gave a number of olloquia and
le tures on wave me hani s (and on matrix me hani s) in the period
leading up to the Solvay onferen e, in Leiden, Itha a and Pasadena.32
In Pasadena he also had the opportunity of dis ussing with S hrdinger
the possibility that S hrdinger may also give a report at the onferen e,
as in fa t he did.c S hrdinger's wave me hani s had also made a great
impression on Einstein, although he repeatedly expressed his unease to
Lorentz at the use of wave fun tions on onguration spa e (òbs ure',34
`harsh',35 a `Mysterium'36 ), and again during the general dis ussion
(p. 487).
a For details of the other parti ipants, see the next se tion.
a Cf. se tion 4.1.
b Note that S hrdinger (1926a) had written on Èinstein's gas theory' in a paper
that is an immediate pre ursor to his series of papers on quantisation.
Lorentz was at Cornell from September to De ember 1926, then in Pasadena until
Mar h 1927.33 On S hrdinger's Ameri an voyage, see Moore (1989, pp. 23033).
12
One sees Lorentz's involvement with the re ent developments also in

his orresponden e with Ehrenfest. In parti ular, Lorentz appears to
have been stru k by Dira 's ontributions to quantum me hani s.a In
June 1927, Lorentz invited Dira to spend the following a ademi year
in Leiden ,38 and asked Born and Heisenberg to in lude a dis ussion of
Dira 's work in their report.39 Finally, in late August, Lorentz de ided
that Dira , and also Pauli, ought to be invited to the onferen e, for
indeed:
Sin e last year, quantum me hani s, whi h will be our topi , has developed
with an unexpe ted rapidity, and some physi ists who were formerly in the
se ond tier have made extremely notable ontributions. For this reason I
would be very keen to invite also Mr Dira of Cambridge and Mr Pauli of
Copenhagen. .... Their ollaboration would be very useful to us .... I need not
onsult the s ienti ommittee be ause Mr Dira and Mr Pauli were both on
40
a list that we had drawn up last year .... .
Lorentz invited Pauli on 5 September 1927 (Pauli 1979, pp. 4089) and
Dira sometime before 13 September 1927.41
On the experimental side, some of the main a hievements of 1927 were
the experiments on matter waves. While originally de Broglie was listed
to give a report on light quanta, the work he presented was about both
light quanta and material parti les (indeed, ele trons and photons!), and
Lorentz asked him expli itly to in lude some dis ussion of the re ent
experiments speaking in favour of the notion of matter waves, spe i ally
dis ussing Elsasser's (1925) proposals, and the experimental work of
Dymond (1927) and of Davisson and Germer (1927).42 Thus, in the nal
programme of the onferen e, we nd three reports on the foundations
of a new me hani s, by de Broglie, Heisenberg (together with Born) and
S hrdinger.
The talks given by Bragg and Compton, instead, ree t at least in part
the initial orientation of the onferen e. Here is how Compton presents
the division of labour (p. 329):
Professor W. L. Bragg has just dis ussed a whole series of radiation phenomena
in whi h the ele tromagneti theory is onrmed. .... I have been left the task of
pleading the opposing ause to that of the ele tromagneti theory of radiation,
seen from the experimental viewpoint.
Bragg fo usses in parti ular on the te hnique of X-ray analysis, as the

`most dire t way of analysing atomi and mole ular stru ture' (p. 284),
a This orresponden e in ludes for instan e a 15-page ommentary by Lorentz on
Dira (1927a).37
13
the development of whi h, as he had mentioned to Lorentz, was the

`line in whi h [he had been espe ially interested'.43 This in ludes in
parti ular the investigation of the ele troni harge distribution. At
Lorentz's request, he had also in luded a dis ussion of the refra tion
of X-rays (se tion 8 of his report), whi h is dire tly relevant to the
dis ussion after Compton's report.44 As des ribed by Lorentz in June
1927, Bragg was to report òn phenomena that still somehow allow a
lassi al des ription'.45 A few more aspe ts of Bragg's report are of
immediate relevan e for the rest of the onferen e, espe ially to the
dis ussion of S hrdinger's interpretation of the wave fun tion in terms
of an ele tri harge density (pp. 307, 312, se tion 4.4), and so are some
of the issues taken up further in the dis ussion (Hartree approximation,
problems with waves in three dimensions), but it is fair to say that the
report provides a rather distant ba kground for what followed it.
Compton's talk overs the topi s of points 2 and 3 listed above. The
expli it fo us of his report is the three-way omparison between the
photon hypothesis, the Bohr-Kramers-Slater (BKS) theory of radiation,
and the lassi al theory of radiation. Note, however, that Compton
introdu es many of the topi s of later dis ussions. For instan e, he
dis usses the problem of how to explain atomi radiation (se tion on
`The emission of radiation', p. 333), whi h is inexpli able from the point
of view of the lassi al theory, given that the òrbital frequen ies' in the
atom do not orrespond to the emission frequen ies. This problem was
one of S hrdinger's main on erns and one of the main points of oni t
between S hrdinger and, for instan e, Heisenberg (see in parti ular the
dis ussion after S hrdinger's report and, below, se tions 4.4 and 4.6).
Compton's dis ussion of the photon hypothesis relates to the question of
`guiding elds' (pp. 331 and 354) and of the lo alisation of parti les or
energy quanta within a wave (pp. 339 and 348). These in turn are losely
onne ted with some of de Broglie's and Einstein's ideas (see below
hapter 7, espe ially se tion 7.2, and hapter 9 ), and with de Broglie's
report on pilot-wave theory and Einstein's remark about lo ality in the
general dis ussion (p. 487).
Bohr had been a noted s epti of the photon hypothesis, and in 1924
Bohr, Kramers and Slater had developed a theory that was able to maintain a wave pi ture of radiation, by introdu ing a des ription of the atom
based on `virtual os illators' with frequen ies equal to the frequen ies of
emission (Bohr, Kramers and Slater 1924a,b).a A stationary state of
a As Darrigol (1992, p. 257) emphasises, while the free virtual elds obey the
14
an atom, say the nth, is asso iated with a state of ex itation of the
os illators with frequen ies orresponding to transitions from the energy
En . Su h os illators produ e a lassi al radiation eld (a `virtual' one),
whi h in turn determines the probabilities for spontaneous emission in
the atom, that is, for the emission of energy from the atom and the
jump to a stationary state of lower energy. The virtual eld of one atom
also intera ts with the virtual os illators in other atoms (whi h in turn
produ e se ondary virtual radiation) and inuen es the probabilities
for indu ed emission and absorption in the other atoms. While the
theory provides a me hanism for radiation onsistent with the pi ture
of stationary states ( f. Compton's remarks, p. 335), it violates energy
and momentum onservation for single events, in that an emission in one
atom is not onne ted dire tly to an absorption in another atom, but
only indire tly through the virtual radiation eld. Energy and momentum onservation hold only at a statisti al level. The BKS proposal was
short-lived, be ause the Bothe-Geiger and Compton-Simon experiments
established the onservation laws for individual pro esses (as explained
in detail by Compton in his report, pp. 350 .). Thus, at the time of
the planning of the fth Solvay onferen e, the experimental eviden e
had ruled out the BKS theory (hen e the above remark: ìf Mr Kramers
judges that it is still useful').a The short note 5, indeed, dropped out of
the programme altogether.b
The des ription of the intera tion between matter and radiation, in
parti ular the Compton ee t, ontinued to be a problem for Bohr, and
ontributed to the development of his views on wave-parti le dualism and
omplementarity. In his ontribution to the dis ussion after Compton's
report (p. 359, the longest of his published ontributions in the Solvay
volumea ), Bohr sket hes the motivations behind the BKS theory, the
Maxwell equations, i.e. an be onsidered to be lassi al, the virtual os illators and
the intera tion between the elds and the os illators are non- lassi al in several
respe ts.
a Bothe and Geiger had been working on their experiments sin e June 1924 (Bothe
and Geiger 1924), and provisional results were being debated by the turn of the
year. For two diering views on the signi an e of these results for instan e see
Einstein to Lorentz, 16 De ember 1924 (the same letter in whi h he wrote to
Lorentz about de Broglie's results)46 and the ex hange of letters between Born
and Bohr in January 1925 (Bohr 1984, pp. 3026). By April 1925, Bothe and
Geiger had lear- ut results against the BKS theory (Bothe and Geiger 1925a,b;
see also the letters between Geiger and Bohr in Bohr 1984, pp. 3524).
b On the BKS theory and related matters, see also hapter 3 (espe ially se tions 3.3.1
and 3.4.2), hapter 9, Darrigol (1992, hapter 9), the ex ellent introdu tion
by Stolzenburg to Part I of Bohr (1984) and Mehra and Re henberg (1982,
se tion V.2).
a See below for the fate of his ontribution to the general dis ussion.
15
on lusions to be drawn from the Bothe-Geiger and Compton-Simon

experiments and the further development of his views.
Lorentz, in his report of the meeting to Einstein had mentioned `SlaterBohr-Kramers and analogous theories'. This may refer to the further
developments (independent of the validity of the BKS theory) that led in
parti ular to Kramer's (1924) dispersion theory (and from there towards
matrix me hani s), or to Slater's original ideas, whi h were roughly along
the lines of guiding elds for the photons (even though the photons were
dropped from the nal BKS proposal).a Note that Einstein at this time
was also thinking about guiding elds (in three dimensions). Pais (1982,
pp. 44041) writes that, a ording to Wigner, Einstein did not publish
these ideas be ause they also led to problems with the onservation
laws.b
Einstein was asked by Lorentz to ontribute a report on `New derivations of Plan k's law and appli ations of statisti s to quanta' (point 6 ),
learly referring to the work by Bose (1924) on Plan k's law, hampioned
by Einstein and applied by him to the theory of the ideal gas (Einstein
1924, 1925a,b). The se ond of these papers is also where Einstein famously endorses de Broglie's idea of matter waves. Einstein thought that his
work on the subje t was already too well-known, but he a epted after
Lorentz repeated his invitation.47 On 17 June 1927, however, at about
the time when Lorentz was sending detailed requests to the speakers,
Einstein informed him in the following terms that he would not, after
all, present a report:
I re all having ommitted myself to you to give a report on quantum statisti s
at the Solvay onferen e. After mu h ree tion ba k and forth, I ome to the
onvi tion that I am not ompetent [to give su h a report in a way that really
orresponds to the state of things. The reason is that I have not been able
to parti ipate as intensively in the modern development of quantum theory
as would be ne essary for this purpose. This is in part be ause I have on the
whole too little re eptive talent for fully following the stormy developments,
in part also be ause I do not approve of the purely statisti al way of thinking
on whi h the new theories are founded .... Up until now, I kept hoping to be
able to ontribute something of value in Brussels; I have now given up that
hope. I beg you not to be angry with me be ause of that; I did not take this
lightly but tried with all my strength .... (Quoted in Pais 1982, pp. 4312)
Einstein's withdrawal may be related to the following ir umstan es. On

a Cf. Slater (1924) and Mehra and Re henberg (1982, pp. 5436). See also Pauli's
remark during the dis ussion of de Broglie's report (p. 400).
b Cf. Einstein's ontribution to the general dis ussion (p. 485) and the dis ussion
below in hapter 9.
16
5 May 1927, during a meeting of the Prussian A ademy of S ien es in

Berlin, Einstein had read a paper on the question: `Does S hrdinger's
wave me hani s determine the motion of a system ompletely or only in
the sense of statisti s?'48 As dis ussed in detail by Belousek (1996), the
paper attempts to dene deterministi parti le motions from S hrdinger's wave fun tions, but was also suddenly withdrawn on 21 May.a
The plans for the talks were nalised by Lorentz around June 1927.
An extra t from his letter to S hrdinger on the subje t reads as follows:
[We hope to have the following reports [Referate (I give them in the order
in whi h we might dis uss them):
1. From Mr W. L. Bragg on phenomena that still somehow allow a lassi al
des ription (reexion of X-rays by rystals, dira tion and total ree tion of
X-rays).
2. From Mr Compton on the ee t dis overed by him and what relates to it.
3. From Mr de Broglie on his theory. I am asking him also to take into a ount
the appli ation of his ideas to free ele trons (Elsasser, quantum me hani s of
free ele trons; Dymond, Davisson and Germer, s attering of ele trons).
4. From Dr Heisenberg
or
Prof. Born (the hoi e is left to them) on matrix
me hani s, in luding Dira 's theory.

5. Your report [on wave me hani s.
Maybe another one or two short ommuni ations [Beri hte on spe ial topi s
50
will be added.
This was, indeed, the nal programme of the onferen e, with Born
and Heisenberg de iding to ontribute a joint report.a

In 192627 the s ienti ommittee and the administrative ommission
of the Solvay institute were omposed as follows.
a The news of Einstein's ommuni ation prompted an ex hange of letters between
Heisenberg and Einstein, of whi h Heisenberg's letters, of 19 May and 10 June,
survive.49 The se ond of these is parti ularly interesting, be ause Heisenberg
presents in some detail his view of theories that in lude parti le traje tories. Both
Einstein's hidden-variables proposal and Heisenberg's rea tion will be des ribed
in se tion 11.3.
a See se tion 3.2. Note that, as we shall see below, while Bohr ontributed
signi antly to the general dis ussion and reported the views he had developed
in Como (Bohr 1949, p. 216, 1985, pp. 357), he was unable to prepare an edited
version of his omments in time and therefore suggested that a translation of his
Como le ture, in the version for Naturwissens haften (Bohr 1928), be in luded in
the volume instead. This has given rise to a ommon belief that Bohr gave a report
on a par with the other reports, and that the general dis ussion at the onferen e
was the dis ussion following it. See for instan e Mehra (1975, p. 152), and Mehra
and Re henberg (2000, pp. 246 and 249), who appear further to believe that Bohr
did not parti ipate in the o ial dis ussion.
17
S ienti ommittee: Lorentz (Leiden) as president, Knudsen (Copenhagen) as se retary, W. H. Bragg (until May 1927, London),a Marie
Curie (Paris), Einstein (sin e April 1926, Berlin), Charles-Eugne Guye
(Geneva),b Langevin (Paris), Ri hardson (London), Edm. van Aubel
(Gent).
Administrative ommission: Armand Solvay, Jules Bordet (ULB), Mauri e Bourquin (ULB), mile Henriot (ULB); the administrative se retary
sin e 1922, and thus main orrespondent of Lorentz and others, was
Charles Lefbure.
The se retary of the meeting was Jules-mile. Vers haelt (Gent),
who had a ted as se retary sin e the third Solvay onferen e.52
The rst provisional list of possible parti ipants (in addition to Ehrenfest) appears in Lorentz's letter to Ehrenfest of 29 Mar h 1926:
Einstein, Bohr, Kramers, Born, Heisenberg (Jordan surely more mathemati ian), Pauli, Ladenburg (?), Slater, the young Bragg (be ause of the ` orresponden e' to the lassi al theory that his work has often resulted in),
J. J. Thomson, another one or two Englishmen (Darwin? Fowler?), Lon
Brillouin (do not know whether he has worked on this, he has also already
been there a number of times),
Louis
de Broglie (light quanta), one or two
who have on erned themselves with dira tion of X-rays (Bergen Davis?,
53
Compton, Debye, Dira (?)).
Lorentz asked for further suggestions and omments, whi h Ehrenfest

sent in a letter dated `Leiden 30 Mar h 1926. Late at night':
Langevin, Fowler, Dira , J. Fran[ k (already for the experiments he devised
by Hanle on the destru tion of resonan e polarisation through Larmor rotation
a
and for the work he proposed by Hund on the Ramsauer ee t undisturbed
passage of slow ele trons through atoms and so on), Fermi (for interesting
ontinuation of the experiments by Hanle), Oseen (possibly a wrong attempt
at explanation of needle radiation and as sharpwitted riti ), S hrdinger (was
perhaps the rst to give quantum interpretation of the Doppler ee t, thus
lose to Compton ee t), Bothe (for Bothe-Geiger experiment on orrelation
of Compton quantum and ele tron, whi h destroys Bohr-Slater theory, altogether a ne brain!) (Bothe should be onsidered perhaps
before S hrdinger),
Darwin, Smekal (is indeed a very deserving onnoisseur of quantum nesses,

only he writes so
frightfully
mu h).
Lon Brillouin has published something re ently on matrix physi s, but I

54
have not read it yet.
a W. H. Bragg resigned due to over ommitment and was later repla ed by Cabrera
(Madrid).51
b In 1909 the university of Geneva onferred on Einstein his rst honorary degree.
A ording to Pais (1982, p. 152), this was probably due to Guye. Coin identally,
Ernest Solvay was honoured at the same time.
a For more on the spe ial interest of the Ramsauer ee t, see se tion 3.4.2 below.
18
At the April meeting (as listed in the report by Vers haelt55 ) it

was then de ided to invite: Bohr, Kramers, Ehrenfest, two among Born,
Heisenberg and Pauli, Plan k, Fowler, W. L. Bragg, C. T. R. Wilson,
L. de Broglie, L. Brillouin, Deslandres, Compton, S hrdinger and Debye. Possible substitutes were listed as: M. de Broglie or Thibaud for
Bragg, Dira for Brillouin, Fabry for Deslandres, Kapitza for Wilson,
Darwin or Dira for Fowler, Bergen Davis for Compton, and Thirring for
S hrdinger.a The members of the s ienti ommitee would all take part
ex o io, and invitations would be sent to the professors of physi s at
ULB, that is, to Pi ard, Henriot and De Donder57 (the latter apparently
somewhat to Lefbure's hagrin, who, just before the onferen e started,
felt obliged to remind Lorentz that De Donder was à paradoxi al mind,
loud [en ombrant and always ready to seize the word, often with great
maladroitness'58).
Both the number of a tual parti ipants and of observers was to be
kept limited,59 partly explaining why it was thought that one should
invite only two among Born, Heisenberg and Pauli. The hoi e initially
fell on Born and Heisenberg (although Fran k was also onsidered as an
alternative).60 Eventually, as noted above, Pauli was also in luded, as
was Dira .61 Lorentz was also keen to invite Millikan and possibly
Hall , when he heard that Millikan would be in Europe anyway for the
Como meeting (Einstein and Ri hardson agreed).62 However, nothing
ame of this plan.
When Einstein eventually withdrew as a speaker, he suggested Fermi
or Langevin as possible substitutes (Pais 1982, p. 432). For a while it was
not lear whether Langevin (who was anyway a member of the s ienti
ommittee) would be able to ome, sin e he was in Argentina over the
summer and due to go on to Pasadena from there. Ehrenfest suggested
F. Perrin instead, in rather admiring tones. Langevin was needed in Paris
in O tober, however, and was able to ome to the onferen e.63 Finally,
the week before the onferen e started, Lorentz extended the invitation
to Irving Langmuir,64 who would happen to be in Brussels at the time
of the onferen e.a
Lorentz sent most of the informal invitations around January 1927.65
In May 1927, he sent to Lefbure the list of all the people he had `provisionally invited',66 in luding all the members of the s ienti ommittee
and the prospe tive invitees as listed above by Vers haelt (that is,
a A few days later, Guye suggested also Auger as a possible substitute for Wilson.56
a To Langmuir we owe a fas inating `home movie' of the onferen e; see the report
in the AIP Bulletin of Physi s News, number 724 (2005).
19
as yet without Pauli and Dira ). All had already replied and a epted, ex ept Deslandres (who eventually replied mu h later de lining the
invitation67 ). Around early July, Lorentz invited the physi ists from the
university,68 and presumably sent a new invitation to W. H. Bragg, who
thanked him but de lined.69 Formal letters of onrmation were sent
out by Lefbure shortly before the onferen e.70
Around June 1927, Lorentz wrote to the planned speakers inviting
them in the name of the s ienti ommittee to ontribute written
reports, to rea h him preferably by 1 September. The general guidelines
were: to fo us on one's own work, without mathemati al details, but
rather so that `the prin iples are highlighted as learly as possible, and
the open questions as well as the onne tions [Zusammenhnge and
ontrasts are laried'. The material in the reports did not have to be
unpublished, and a bibliography would be wel ome.71 Compton wrote
that he would aim to deliver his manus ipt by 20 August, de Broglie
easily before the end of August, Bragg, as well as Born and Heisenberg,
by 1 September, and S hrdinger presumably only in the se ond half of
September.72 (For further details of the orresponden e between some
of the authors and Lorentz, see the relevant hapters below.)
The written reports were to be sent to all parti ipants in advan e of
the onferen e.a De Broglie's, whi h had been written dire tly in Fren h,
was sent by Lorentz to the publishers, Gauthier-Villars in Paris, before
he left for the Como meeting. They hoped to send 35 proofs to Lorentz
by the end of September. In the meantime, Vers haelt and Lorentz's
son had the remaining reports mimeographed by the `Holland Typing
O e' in Amsterdam, and Vers haelt with the help of a student added
in the formulas by hand, managing to mail on time to the parti ipants at
least Compton's and Born and Heisenberg's reports, if not all of them.74
Lorentz had further written to all speakers (ex ept Compton) to ask
them to bring reprints of their papers.75
Late during planning, a slight problem emerged, namely an unfortunate overlap of the Brussels onferen e with the festivities for the entenary
of Fresnel in Paris, to be o ially opened Thursday 27 O tober. Lorentz
informed Lefbure of the lash writing from Naples after the Como
meeting: neither the date of the onferen e ould be hanged nor that of
the Fresnel elebrations, whi h had been xed by the Fren h President.
The problem was ompounded by the fa t that de Broglie had a epted
a Mimeographed opies of Bragg's, Born and Heisenberg's and S hrdinger's reports
are to be found in the Ri hardson Colle tion, Harry Ransom Humanities Resear h
Center, University of Texas at Austin.73
20
to give a le ture to the So it de Physique on the o asion.a Lorentz

suggested the ompromise solution of a general invitation to attend the
elebrations. Those who wished to parti ipate ould travel to Paris on
27 O tober, returning to Brussels the next day, when sessions would
be resumed in the afternoon. This was the solution that was indeed
adopted.77

The fth Solvay onferen e took pla e from 24 to 29 O tober 1927 in
Brussels. As on previous o asions, the parti ipants stayed at the Htel
Britannique, where a dinner invitation from Armand Solvay awaited
them.78 Other meals were going to be taken at the institute, whi h was
housed in the building of the Institute of Physiology in the Par Lopold;
atering for 5055 people had been arranged.79 The parti ipants were
guests of the administrative ommission and all travel expenses within
Europe were met.80 From the evening of 23 O tober onwards, three seats
were reserved in a box at the Thatre de la Monnaie.81
The rst session of the onferen e started at 10:00 on Monday 24
O tober. A tentative re onstru tion of the s hedule of the onferen e is
as follows.82 We assume that the talks were given in the order they were
des ribed in the plans and printed in the volume, and that the re eption
by the university on the Tuesday ontinued throughout the morning. It
is lear that the general dis ussion extended over at least two days, from
the fa t that Dira in his main ontribution (p. 490) refers expli itly to
Bohr's omments of the day before.a
Monday 24 O tober, morning: W. L. Bragg's report, followed by dis ussion.
Monday 24 O tober, afternoon: A. H. Compton's report, followed by
dis ussion.
a In Lorentz's letter, the date of de Broglie's le ture is mentioned as 28 O tober, but
the o ial invitations state that it was Zeeman who le tured then, and de Broglie
the next evening, after the end of the Solvay onferen e. A report on de Broglie's
le ture, whi h was entitled `Fresnel's uvre and the urrent development of
physi s', was published by Guye in the Journal de Genve of 16 and 18 April
1928.76
a In a letter to Vers haelt, Kramers refers to `the general dis ussion of Thursday',
but that in fa t was the day of the Fresnel elebrations in Paris. The photograph
of Lorentz in luded in the volume, a ording to the aption, was also taken on that
day. Sin e the elebrations opened only at 8:30pm, it is on eivable that there was
a rst dis ussion session on Thursday morning. Pelseneer states, however, that
sessions were suspended for the whole day.83
21
Tuesday 25 O tober, starting 9:00 a.m.: re eption oered by the ULB.

Tuesday 25 O tober, afternoon: L. de Broglie's report, followed by
dis ussion.
Wednesday 26 O tober, morning: M. Born and W. Heisenberg's report, followed by dis ussion.
Wednesday 26 O tober, afternoon: E. S hrdinger's report, followed
by dis ussion.
Thursday 27 O tober, all day: travel to Paris and entenary of Fresnel.a
Friday 28 O tober, morning: return to Brussels.
Friday 28 O tober, afternoon: general dis ussion.
Saturday 29 O tober, morning: general dis ussion,b followed by lun h
with the King and Queen of the Belgians.
Saturday 29 O tober, evening: dinner oered by Armand Solvay.
The languages used were presumably English, German and Fren h.

S hrdinger had volunteered to give his talk in English,c while Born had
suggested that he and Heisenberg ould provide additional explanations
in English (while he thought that neither of them knew Fren h).86 The
phrasing used by Born referred to who should èxplain orally the ontents
of the report', suggesting that the speakers did not present the exa t or
full text of the reports as printed.
Multipli ity of languages had long been a hara teristi of the Solvay onferen es. A well-known letter by Ehrenfest87 tells us of `[poor
Lorentz as interpreter between the British and the Fren h who were
absolutely unable to understand ea h other. Summarising Bohr. And
Bohr responding with polite despair' (as quoted in Bohr 1985, p. 38).d
On the last day of the onferen e, Ehrenfest went to the bla kboard
and evoked the image of the tower of Babel (presumably in a more
a Most of the parti ipants at the Solvay onferen e, with the ex eption of Knudsen,
Dira , Ehrenfest, Plan k, S hrdinger, Henriot, Pi ard and Herzen, travelled to
Paris to attend the inauguration of the elebrations, in the grand amphithatre of
the Sorbonne.84
b The nal session of the onferen e also in luded a homage to Ernest Solvay's
widow.85
Note that S hrdinger was uent in English from hildhood, his mother and aunts
being half-English (Moore 1989, hapter 1).
d Both W. H. Bragg and Plan k deplored in letters to Lorentz that they were very
poor linguists. Indeed, in a letter explaining in more detail why he would not
parti ipate in the onferen e, W. H. Bragg wrote: Ì nd it impossible to follow
the dis ussions even though you so often try to make it easy for us', and Plan k
was in doubt about oming, parti ularly be ause of the language di ulties.88
22
metaphori al sense than the mere multipli ity of spoken languages),

writing:
And they said one to another: .... Go to, let us build us .... a tower, whose top
may rea h unto heaven; and
let us make us a name
.... And the Lord said: ....
Go to, let us go down, and there onfound their language, that they may not
understand one another's spee h. (Genesis 11: 37, reported by Pelseneer, his
a
emphasis )
Informal dis ussions at the onferen e must have been plentiful, but
information about them has to be gathered from other sour es. Famously, Einstein and Bohr engaged in dis ussions that were des ribed in detail
in later re olle tions by Bohr (1949), and vividly re alled by Ehrenfest
within days of the onferen e in the well-known letter quoted above (see
also hapter 12).
Little known, if at all, is another referen e by Ehrenfest to the dis ussions between Bohr and Einstein, whi h appears to relate more dire tly
to the issues raised by Einstein in the general dis ussion:
Bohr had given a very pretty argument in a onversation with Einstein,
that one ould not hope ever to master many-parti le problems with threedimensional S hrdinger ma hinery. He said something like the following (more or less!!!!!!): a wave pa ket an never simultaneously determine EXACTLY
the position and the velo ity of a parti le. Thus if one has for instan e TWO
parti les, then they annot possibly be represented in
three-dimensional
spa e
su h that one an simultaneously represent exa tly their kineti energy and the
potential energy OF THEIR INTERACTION. Therefore......... (What omes
after this therefore I already annot reprodu e properly.) In the multidimensional representation instead the potential energy of the intera tion appears
totally sharp in the relevant oe ients of the wave equation and one does [?
90
not get to see the kineti energy at all.

The editing of the pro eedings of the fth Solvay onferen e was largely
arried out by Vers haelt, who reported regularly to Lorentz and to
Lefbure. During the last months of 1927 Lorentz was busy writing up
the le ture he had given at the Como meeting in September.91 He then
died suddenly on 4 February, before the editing work was omplete.
The translation of the reports into Fren h was arried out after the
onferen e, ex ept for de Broglie's report whi h, as mentioned, was
a This may have been Ehrenfest's own emphasis. Note that, if not ne essarily present
at the sessions, Pelseneer had some onne tion with the onferen e, having taken
Lorentz's photograph reprodu ed in the pro eedings.89
23
written dire tly in Fren h. From Vers haelt's letters we gather that
by 6 January 1928 all the reports had been translated, Bragg's and
Compton's had been sent to the publishers, and Born and Heisenberg's
and S hrdinger's were to be sent on that day or the next. Several proofs
were ba k by the beginning of Mar h.92
Lorentz had envisaged preparing with Vers haelt an edited version of
the dis ussions from notes taken during the onferen e, and sending the
edited version to the speakers at proof stage.a In fa t, stenographed notes
appear to have been taken, typed up and sent to the speakers, who for the
most part used them to prepare drafts of their ontributions. From these,
Vers haelt then edited the nal version, with some help from Kramers
(who spe i ally ompleted two of Lorentz's ontributions).93 A opy of
the galley proofs of the general dis ussion, dated 1 June 1928, survives
in the Bohr ar hives in Copenhagen,94 and in ludes some ontributions
that appear to have been still largely unedited at that time.b
By January, the editing of the dis ussions was pro eeding well, and
at the beginning of Mar h it was almost ompleted. Some ontributions,
however, were still missing, most notably Bohr's. The notes sent by Vers haelt had many gaps; Bohr wanted Kramers's advi e and help with
the dis ussion ontributions, and travelled to Utre ht for this purpose
at the beginning of Mar h.95 At the end of Mar h, Kramers sent Vers haelt the edited version of Bohr's ontributions to the dis ussion after
Compton's report (pp. 359 and 369) and after Born and Heisenberg's
report (p. 443), remarking that these were all of Bohr's ontributions to
the dis ussions during the rst three days of the onferen e.c
In ontrast, material on Bohr's ontributions to the general dis ussion
survives only in the form of notes in the Bohr ar hives.d (Some notes
by Ri hardson also relate to Bohr's ontributions.97 ) A translation of
version of the Como le ture for Naturwissens haften (Bohr 1928) was
in luded instead, reprinted on a par with the other reports, and a ompanied by the following footnote (p. 215 of the published pro eedings):
This arti le, whi h is the translation of a note published very re ently in
Naturwissens haften,
vol. 16, 1928, p. 245, has been added at the author's
a A ording to D. Devriese, urator of the IIPCS ar hives, the original notes have
not survived.
b We have reprodu ed some of this material in the endnotes to the general dis ussion.
Note that Bohr also asked some brief questions after S hrdinger's report
(p. 468). Kramers further writes that Bohr suggested to òmit the whole nal
Born-Heisenberg dis ussion (Nr. 1823) and equally Fowler's remark 9'. Again,
thanks to Mark van Atten for help with this letter.96
d This material is not mi rolmed in AHQP. See also Bohr (1985, pp. 357, 100 and
4789).
24
request to repla e the exposition of his ideas that he gave in the ourse of the
following general dis ussion. It is essentially the reprodu tion of a talk on the
urrent state of quantum theory that was given in Como on 16 September
1927, on the o asion of the jubilee festivities in honour of Volta.
The last remaining material was sent to Gauthier-Villars sometime in

September 1928, and the volume was nally published in early De ember
of that year.98
1.7 Con lusion

The fth Solvay onferen e was by any standards an important and
memorable event. On this point all parti ipants presumably agreed,
as shown by numerous letters, su h as Ehrenfest's letter quoted above
(reprodu ed in Bohr 1985), Heisenberg's letter to Lefbure at the head of
this hapter, or various other letters of thanks addressed to the organisers
after the onferen e:99
I would like to take this opportunity of thanking you for your kind hospitality,
and telling you how mu h I enjoyed this parti ular Conferen e. I think it has
been the most memorable one whi h I have attended for the subje t whi h
was dis ussed was of su h vital interest and I learned so mu h. (W. L. Bragg
to Armand Solvay, 3 November 1927)
It was the most stimulating s ienti meeting I have ever taken part in.
(Max Born to Charles Lefbure, 8 November 1927)
Per eptions of the signi an e of the onferen e diered from ea h other,

however. In the o ial history, the fth Solvay onferen e went down
(perhaps together with the Como meeting) as the o asion on whi h
the interpretational issues were nally laried. This was presumably
a genuine sentiment on the part of Bohr, Heisenberg and the other
physi ists of the Copenhagen-Gttingen s hool. We nd it expli itly as
early as 1929:
In relating the development of the quantum theory, one must in parti ular
not forget the dis ussions at the Solvay onferen e in Brussels in 1927, haired
by Lorentz. Through the possibility of ex hange [Ausspra he between the
representatives of dierent lines of resear h, this onferen e has ontributed
extraordinarily to the lari ation of the physi al foundations of the quantum
theory; it forms so to speak the outward ompletion of the quantum theory .... .
(Heisenberg 1929, p. 495)
On the other hand, the onferen e was also des ribed (by Langevin)
as the one where `the onfusion of ideas rea hed its peak'.100 From
a distan e of almost 80 years, the beginnings of a more dispassionate
1.7 Con lusion
25
evaluation should be possible. In the following hapters, we shall revisit

the fth Solvay onferen e, fo ussing in parti ular on the ba kground
and ontributions relating to the three main `lines of resear h' into
quantum theory represented there: de Broglie's pilot-wave theory, Born
and Heisenberg's quantum me hani s and S hrdinger's wave me hani s.
26
Notes to pp. 39
Ar hival notes
1 Handwritten remark (by Ehrenfest) in the margin of Lorentz to
Ehrenfest, 29 Mar h 1926, AHQP-EHR-23 (in Dut h).
2 Heisenberg to Lefbure, 19 De ember [1927, IIPCS 2685 (in German).
3 AHQP-LTZ-12, talk X 23, Èrnest Solvay', dated 28 September 1914 (in
English). Cf. also the Fren h version of the same, X 10. (The two pages of
the latter are in separate pla es on the mi rolm.)
Historique des Instituts Internationaux de Physique et de

Chimie Solvay depuis leur fondation jusqu' la deuxime guerre mondiale,
4 Pelseneer, J.,
[1962, 103 pp., AHQP-58, se tion 1 (hereafter referred to simply as

`Pelseneer').
5 Plan k to Nernst, 11 June 1910, Pelseneer, p. 7 (in German).
6 Lorentz to Solvay, 4 January 1912, Pelseneer, pp. 2026 (in Fren h). See
also the reply, Solvay to Lorentz, 10 January 1912, AHQP-LTZ-12 (in
Fren h).
7 Lorentz to Solvay, 6 Mar h 1912, Pelseneer, p. 27 (in Fren h).
8 Hger to Lorentz, 16 February 1912, AHQP-LTZ-11 (in Fren h).
9 Einstein to Lorentz, 16 August 1923, AHQP-LTZ-7 (in German).
10 Van Aubel to Lorentz, 16 April, 16 May and 19 July 1923,
AHQP-LTZ-11 (in Fren h).
11 Lorentz to Solvay, 10 January 1919, Pelseneer, p. 37 (in Fren h).
12 Lorentz to Einstein, 14 Mar h 1926, AHQP-86 (in German).
13 Lefbure to Lorentz, 12 February 1926, AHQP-LTZ-12 (in Fren h).
14 Cf. Lefbure to Lorentz, 12 February 1926, AHQP-LTZ-12 (in Fren h).
15 Letter by Lorentz, 7 January 1919, Pelseneer, pp. 356 (in Fren h), two
letters from Plan k to Lorentz, 1915, AHQP-LTZ-12 (in German), Plan k
to Lorentz, Mar h 1916, Pelseneer, pp. 345 (in German), and Plan k to
Lorentz, 5 De ember 1923, AHQP-LTZ-9 (in German).
16 From Lefbure [possibly a opy for Lorentz, 1 April 1926, AHQP-LTZ-12
(in Fren h). Obituary of Tassel,
L'ventail,
15 O tober 1922,

17 Cf. Lefbure to Lorentz, 6 April 1926, AHQP-LTZ-12 (in Fren h).
18 Lorentz to Einstein, 6 April 1926, AHQP-86 (in German).
19 Einstein to Lorentz, 14 April 1926 and 1 May 1927, AHQP-LTZ-11 (in
German).
21 Compare the invitation of Lorentz to Bohr, 7 June 1926, AHQP-BSC-13
(in English), and the replies of Bohr to Lorentz, 24 June 1926,
AHQP-LTZ-11 (in English), Plan k to Lorentz, 13 June 1926,
AHQP-LTZ-8 (in German), Born to Lorentz, 19 June 1926,
AHQP-LTZ-11 (in German), and Heisenberg to Lorentz, 4 July 1926,
AHQP-LTZ-12 (in German).
22 Van Aubel to Lefbure, 6 O tober 1927, IIPCS 2545 (in Fren h), with
Lefbure's handwritten omment.
23 Lefbure to Lorentz, 14 O tober 1927, IIPCS 2534, and 15 O tober 1927,
IIPCS 2536, telegramme Lorentz to Lefbure, 15 O tober 1927, IIPCS
2535, and Lefbure to the King, 19 O tober 1927, IIPCS 2622 (all in
Fren h).
24 Lorentz to Solvay, 10 January 1919, Pelseneer, p. 37 (in Fren h).
25 There is a large amount of relevant orresponden e in AHQP-LTZ.
Notes to pp. 918
27
26 Lorentz to Ehrenfest, 29 Mar h 1926, AHQP-EHR-23 (in Dut h), with

Ehrenfest's handwritten omments.
27 Ehrenfest to Lorentz, 30 Mar h 1926, AHQP-LTZ-11 (in German).
28 Lorentz to Bohr, 7 June 1926, AHQP-BSC-13, se tion 3 (in English),
Lorentz to S hrdinger, 21 January 1927, AHQP-41, se tion 9 (in
German).
29 Lorentz to Einstein, 6 April 1926, AHQP-86 (in German and Fren h).
30 Vers haelt to Lefbure, 8 April 1926, IIPCS 2573 (in Fren h).
31 Einstein to Lorentz, 12 April 1926, AHQP-LTZ-11 (in German).
32 See for instan e the atalogue of Lorentz's manus ripts in AHQP-LTZ-11.
33 See also Lorentz to S hrdinger, 21 January and 17 June 1927,
AHQP-41, se tion 9 (in German).
34 Einstein to Lorentz, 1 May 1926, AHQP-LTZ-11 (in German).
35 Einstein to Lorentz, 22 June 1926, AHQP-LTZ-8 (in German).
36 Einstein to Lorentz, 16 February 1927, AHQP-LTZ-11 (in German).
37 En losed with Lorentz to Ehrenfest, 3 June [1927, erroneously amended
to 1925, AHQP-EHR-23 (in Dut h).
38 Lorentz to Dira , 9 June 1927, AHQP-LTZ-8 (in English).
39 Cf. Born to Lorentz, 23 June 1927, AHQP-LTZ-11 (in German).
40 Lorentz to Lefbure, 27 August 1927, IIPCS 2532A/B (in Fren h). [There
appears to be a further item also numbered 2532.
41 Dira to Lorentz, 13 September 1927, AHQP-LTZ-11 (in English).
42 Cf. de Broglie to Lorentz, 22 June 1927, AHQP-LTZ-11 (in Fren h), and
Ehrenfest to Lorentz, 14 June 1927, AHQP-EHR-23 (in Dut h and
German).
43 W. L. Bragg to Lorentz, 7 February 1927, AHQP-LTZ-11 (in English).
44 Cf. also W. L. Bragg to Lorentz, 27 June 1927, AHQP-LTZ-11 (in
English).
45 Lorentz to S hrdinger, 17 June 1927, AHQP-41, se tion 9 (in German).
46 Einstein to Lorentz, 16 De ember 1924, AHQP-LTZ-7 (in German).
47 Einstein to Lorentz, 12 April 1926, AHQP-LTZ-11, Lorentz to Einstein,
28 April 1926, AHQP-86, and Einstein to Lorentz, 1 May 1926,
AHQP-LTZ-11 (all in German).
48 `Bestimmt S hrdingers Wellenme hanik die Bewegung eines Systems
vollstndig oder nur im Sinne der Statistik?', AEA 2-100.00 (in German),
available on-line at
http://www.alberteinstein.info/db/ViewDetails.do?Do umentID=34338 .
49 Heisenberg to Einstein, 19 May 1927, AEA 12-173.00, and 10 June 1927,
AEA 12-174.00 (both in German).
50 Lorentz to S hrdinger, 17 June 1927, AHQP-41, se tion 9 (in German).
51 W. H. Bragg to Lorentz, 11 May 1927, AHQP-LTZ-11 (in English).
52 Tassel to Vers haelt, 24 February 1921, AHQP-LTZ-13 (in Fren h).
53 Lorentz to Ehrenfest, 29 Mar h 1926, AHQP-EHR-23 (in Dut h).
56 Guye to Lorentz, 14 April 1926, AHQP-LTZ-8 (in Fren h).
57 Cf. for instan e Lorentz to S hrdinger, 21 January 1927, AHQP-41,
se tion 9 (in German).
58 Lefbure to Lorentz, 22 O tober 1927, AHQP-LTZ-12 (in Fren h).
59 Lorentz to Einstein, 28 April 1926, AHQP-86 (in German), Lorentz to
Lefbure, 9 O tober 1927, IIPCS, 2530A/B (in Fren h).
28
Notes to pp. 1819
60 Lorentz to Einstein, 28 April 1926, AHQP-86 (in German), Einstein to

Lorentz, 1 May 1926, AHQP-LTZ-11 (in German).
61 For the latter, f. also Brillouin to Lorentz, 20 August 1927,
62 Lorentz to Einstein, 30 January 1927, AHQP-86 (in German), Einstein to
Lorentz, 16 February 1927, AHQP-LTZ-11 (in German), Ri hardson to
Lorentz, 19 February 1927, AHQP-LTZ-12 (in English).
63 Brillouin to Lorentz, 20 August 1927, AHQP-LTZ-11 (in Fren h),
Ehrenfest to Lorentz 18 August 1927, AHQP-EHR-23 (in German).
64 Telegramme Lorentz to Lefbure, 19 O tober 1927, IIPCS 2541 (in
Fren h), with Lefbure's note: Òui'.
65 Lorentz to Brillouin, 15 De ember 1926, AHQP-LTZ-12 (in Fren h),
Lorentz to Ehrenfest, 18 January 1927, AHQP-EHR-23 (in Dut h),
Lorentz to S hrdinger, 21 January 1927, AHQP-41, se tion 9 (in
German); ompare various other replies to Lorentz's invitation, in
AHQP-LTZ-11, AHQP-LTZ-12 and AHQP-LTZ-13: Brillouin, 8 January
1927 (in Fren h), de Broglie, 8 January 1927 (in Fren h), W. L. Bragg, 7
February 1927 (in English), Wilson, 11 February 1927 (in English),
Kramers, 14 February 1927 (in German), Debye, 24 February 1927 (in
Dut h), Compton, 3 April 1927 (in English) [late be ause he had been
away for two months ìn the Orient'.
66 Lorentz to Lefbure, 21 May 1927, IIPCS 2521A (in Fren h).
67 Deslandres to Lorentz, 19 July 1927, AHQP-LTZ-11 (in Fren h).
68 See the letters of a eptan e to Lorentz: De Donder, 8 July 1927,
AHQP-LTZ-11, Henriot, 10 July 1927, AHQP-LTZ-12, Pi ard, 2
O tober 1927, AHQP-LTZ-12 (all in Fren h).
69 W. H. Bragg to Lorentz, 12 and 17 July 1927, AHQP-LTZ-11 (in
English).
70 Copies in AHQP-LTZ-12 and IIPCS 2543. Various replies: IIPCS
254451, 25536, 2558, 25603.
71 Cf. Lorentz to S hrdinger, 17 June 1927, AHQP-41, se tion 9 (in
German).
72 Compare the answers by Bragg, 27 June 1927, and by Compton, 7 July
1927, both AHQP-LTZ-11 (in English), and the detailed ones by de
Broglie, 27 June 1927, AHQP-LTZ-11 (in Fren h) and by Born, 23 June
1927, AHQP-LTZ-11 (in German); also S hrdinger to Lorentz, 23 June
1927, AHQP-LTZ-13 (original with S hrdinger's orre tions), and
AHQP-41 se tion 9 ( arbon opy) (in German).
73 Mi rolmed in AHQP-RDN, do uments M-0059 (Bragg, atalogued as
ùnidentied author'), M-0309 (Born and Heisenberg, with seven pages of
notes by Ri hardson) and M-1354 (S hrdinger).
74 See in parti ular the already quoted Lorentz to Lefbure, 27 August
1927, IIPCS 2532A/B (in Fren h), as well as Lorentz to Lefbure, 23
September 1927, IIPCS 2523A/B, and 4 O tober 1927, IIPCS 2528 (both
in Fren h), Gauthiers-Villars to Lefbure, 16 September 1927, IIPCS
2755 (in Fren h), Vers haelt to Lorentz, 6 O tober 1927, AHQP-LTZ-13
(in Dut h), de Broglie to Lorentz, 29 August 1927, AHQP-LTZ-8, and 11
O tober 1927, AHQP-LTZ-11 (both in Fren h), and Vers haelt to
Lefbure, 15 O tober 1927, IIPCS 2756 (in Fren h).
75 Lorentz to Ehrenfest, 13 O tober 1927, AHQP-EHR-23 (in Dut h). See
also Born to Lorentz, 11 O tober 1927 AHQP-LTZ-11 (in German), de
Notes to pp. 1924
29
Broglie to Lorentz, 11 O tober 1927, AHQP-LTZ-11 (in Fren h), Plan k

to Lefbure, 17 O tober 1927, IIPCS 2558, and Ri hardson to Lefbure,
IIPCS 2561.
76 Ùne rise dans la physique moderne I & II', IIPCS 275051 (in Fren h).
77 Lorentz to Lefbure, 29 September 1927, IIPCS 2523A/B, Brillouin to
Lorentz, 11 O tober 1927, AHQP-LTZ-11, Fondation Solvay to Lefbure,
IIPCS 2582. Invitation: IIPCS 2615 and 2619, AHPQ-LTZ-8. Replies:
IIPCS 2617 and 2618. (All do uments in Fren h.)
78 IIPCS 2530A/B, 2537.
79 IIPCS 2533, 2586A/B/C/D/E (the proposed menus from the Taverne
Royale), 2587A/B.
80 Lorentz to S hrdinger, 21 January 1927, AHQP-41, se tion 9 (in
German).
81 IIPCS 2340.
82 For the time and pla e of the rst session, see IIPCS 2523A/B. For the
re eption at ULB, see IIPCS 2540 and 2629. There is a seating plan for
the lun h with the royal ouple, IIPCS 2627. For the dinner with Armand
Solvay, see IIPCS 2533, 2624 and 2625. See also Pelseneer, pp. 4950.
83 See Kramers to Vers haelt, 23 Mar h 1928, AHQP-28 (in Dut h), and
Pelseneer, p. 50.
84 IIPCS 2621.
85 See AHQP-LTZ-12 (draft), and presumably IIPCS 2667.
86 S hrdinger to Lorentz, 23 June 1927, AHQP-LTZ-13 and AHQP-41,
se tion 9 (in German), Lorentz to S hrdinger, 8 July 1927, AHPQ-41,
se tion 9 (in German), Born to Lorentz, 23 June 1927, AHQP-LTZ-11 (in
German).
87 Ehrenfest to Goudsmit, Uhlenbe k and Dieke, 3 November 1927,
AHQP-61 (in German).
88 W. H. Bragg to Lorentz, 17 July [1927, AHQP-LTZ-11 (in English),
Plan k to Lorentz, 2 February 1927, AHQP-LTZ-8 (in German).
89 See Pelseneer, pp. 5051.
90 Ehrenfest to Kramers, 6 November 1927, AHQP-9, se tion 10 (in
German). The words `bekommt man' [? are very faint.
91 Lorentz to Lefbure, 30 De ember 1927, IIPCS 2670A/B (in Fren h).
92 Vers haelt to Lefbure, 6 January 1928, IIPCS 2609, 2 Mar h 1928,
IIPCS 2610 (both in Fren h).
93 Lefbure to Lorentz, [after 27 August 1927, IIPCS 2524 (in Fren h),
Bohr to Kramers, 17 February 1928, AHQP-BSC-13 (in Danish),
Brillouin to Lorentz, 31 De ember 1927, AHQP-LTZ-8 (in Fren h), and
Kramers to Vers haelt, 28 Mar h 1928, AHQP-28, se tion 4 (in Dut h).
94 Mi rolmed in AHQP-BMSS-11, se tion 5.
95 Bohr to Kramers, 17 and 27 February 1928, AHQP-BSC-13 (both in
Danish).
96 Kramers to Vers haelt, 28 Mar h 1928, AHQP-28, se tion 4 (in Dut h).
97 In luded with the opy of Born and Heisenberg's report in AHQP-RDN,
do ument M-0309.
98 Gauthier-Villars to Vers haelt, 6 De ember 1928, IIPCS 2762 (in
Fren h), and Vers haelt to Lefbure, 11 [De ember 1928, IIPCS 2761
(in Fren h).
99 IIPCS 2671 (in English), IIPCS 2672 (in German).
100 Quoted in Pelseneer, p. 50.
2
2.1 Ba kground
At a time when no single known fa t supported this
theory, Louis de Broglie asserted that a stream of
ele trons whi h passed through a very small hole in an
opaque s reen must exhibit the same phenomena as a
light ray under the same onditions.
Prof. C. W. Oseen, Chairman of the Nobel Committee for
Physi s, presentation spee h, 12 De ember 1929 (Oseen
1999)
In September 1923, Prin e Louis de Brogliea made one of the most astonishing predi tions in the history of theoreti al physi s: that material
bodies would exhibit the wave-like phenomena of dira tion and interferen e upon passing through su iently narrow slits. Like Einstein's
predi tion of the dee tion of light by the sun, whi h was based on a
reinterpretation of gravitational for e in terms of geometry, de Broglie's
predi tion of the dee tion of ele tron paths by narrow slits was made
on the basis of a fundamental reappraisal of the nature of for es and of
dynami s. De Broglie had proposed that Newton's rst law of motion be
abandoned, and repla ed by a new postulate, a ording to whi h a freely
moving body follows a traje tory that is orthogonal to the surfa es of
equal phase of an asso iated guiding wave. The resulting `de Broglian dynami s' or pilot-wave theory as de Broglie later alled it was a new
approa h to the theory of motion, as radi al as Einstein's interpretation
a The de Broglies had ome to Fran e from Italy in the seventeenth entury, the
original name `Broglia' eventually being hanged to de Broglie. On his father's
side, de Broglie's an estors in luded dukes, prin es, ambassadors, and marshals of
Fran e. Nye (1997) onsiders the oni t between de Broglie's pursuit of s ien e
and the expe tations of his aristo rati family. For a biography of de Broglie, see
Lo hak (1992).
30
2.1 Ba kground
31
of the traje tories of falling bodies as geodesi s of a urved spa etime,

and as far-rea hing in its impli ations. In 1929 de Broglie re eived the
Nobel Prize, `for his dis overy of the wave nature of ele trons'.
Strangely enough, however, even though de Broglie's predi tion was
onrmed experimentally a few years later, for most of the twentieth
entury single-parti le dira tion and interferen e were routinely ited
as eviden e against de Broglie's ideas: even today, some textbooks on
quantum me hani s assert that su h interferen e demonstrates that parti le traje tories annot exist in the quantum domain (see se tion 6.1).
It is as if the dee tion of light by the sun had ome to be widely
regarded as eviden e against Einstein's general theory of relativity. This
remarkable misunderstanding illustrates the extent to whi h de Broglie's
work in the 1920s has been underestimated, misrepresented, and indeed
largely ignored, not only by physi ists but also by historians.
De Broglie's PhD thesis of 1924 is of ourse re ognised as a landmark
in the history of quantum theory. But what is usually remembered about
that thesis is the proposed extension of Einstein's wave-parti le duality
from light to matter, with the formulas E = h and p = h/ (relating
energy and momentum to frequen y and wavelength) being applied to
ele trons or `matter waves'. Usually, little attention is paid to the fa t
that a entral theme of de Broglie's thesis was the onstru tion of a new
form of dynami s, in whi h lassi al (Newtonian or Einsteinian) laws
are abandoned, and repla ed by new laws a ording to whi h parti le
velo ities are determined by guiding waves, in a spe i manner that
unies the variational prin iples of Maupertuis and Fermat. Nor, indeed,
have historians paid mu h attention to de Broglie's later and more
omplete form of pilot-wave dynami s, whi h he arrived at in a paper
published in May 1927 in Journal de Physique, and whi h he then
presented in O tober 1927 at the fth Solvay onferen e.
Unlike the other main ontributors to quantum theory, de Broglie
worked in relative isolation, having little onta t with the prin ipal resear h entres in Berlin, Copenhagen, Gttingen, Cambridge and Muni h.
While Bohr, Heisenberg, Born, S hrdinger, Pauli and others visited
ea h other frequently and orresponded regularly, de Broglie worked
essentially alone in Paris.a In Fran e at the time, while pure mathemati s
a In his typed `Replies to Mr Kuhn's questions' in the Ar hive for the History of
Quantum Physi s,1 de Broglie writes (p. 7): `Between 1919 and 1928, I worked
very mu h in isolation ' (emphasis in the original). Regarding his Ph.D. thesis de
Broglie re alled (p. 9): Ì worked very mu h alone and almost without any ex hange
of ideas'.
32
was well represented, there was very little a tivity in theoreti al physi s.
In addition, after the rst world war, s ienti relations with Germany
and Austria were interrupted.a All this seems to have suited de Broglie's
rather solitary temperament. De Broglie's isolation, and the fa t that
Fran e was outside the mainstream of theoreti al physi s, may a ount
in part for why so mu h of de Broglie's work went relatively unnoti ed
at the time, and has remained largely ignored even to the present day.
For some seventy years, the physi s ommunity tended to believe
either that `hidden-variables' theories like de Broglie's were impossible,
or that su h theories had been disproven experimentally. The situation
hanged onsiderably in the 1990s, with the publi ation of textbooks
presenting quantum me hani s in the pilot-wave formulation (Bohm
and Hiley 1993; Holland 1993). Pilot-wave theory as originated by
de Broglie in 1927, and elaborated by Bohm 25 years later (Bohm
1952a,b) is now a epted as an alternative (if little used) formulation
of quantum theory.
Fo ussing for simpli ity on the nonrelativisti quantum theory of a system of N (spinless) parti les with 3-ve tor positions xi (i = 1, 2, ..., N ),
it is now generally agreed that, with appropriate initial onditions, quantum
physi s may be a ounted for by the deterministi dynami s dened by
two dierential equations, the S hrdinger equation
N
X ~2 2
+V
=
i~
t
2mi i
i=1
(1)
for a `pilot wave' (x1 , x2 , ..., xN , t) in onguration spa e, and the de

Broglie guidan e equation
mi
dxi
= i S
dt
(2)
for parti le traje tories xi (t), where the phase S(x1 , x2 , ..., xN , t) is lo ally dened by S = ~ Im ln (so that = || e(i/~)S ).
This, as we shall see, is how de Broglie presented his dynami s in
1927. Bohm's presentation of 1952 was somewhat dierent. If one takes
the time derivative of (2), then using (1) one obtains Newton's law of
motion for a eleration
mi x
i = i (V + Q) ,
(3)
a For more details on de Broglie's situation in Fran e at the time, see Mehra and
Re henberg (1982a, pp. 57884).
2.1 Ba kground
33
where
Q
X 2 2 ||
i
2m
||
i
i
(4)
is the `quantum potential'. Bohm regarded (3) as the law of motion, with
(2) added as a onstraint on the initial momenta, a onstraint that Bohm
thought ould be relaxed (see se tion 11.1). For de Broglie, in ontrast,
the law of motion (2) for velo ity had a fundamental status, and for him
represented the uni ation of the prin iples of Maupertuis and Fermat.
One should then distinguish between de Broglie's rst-order (velo itybased) dynami s of 1927, and Bohm's se ond-order (a eleration-based)
dynami s of 1952.
In this hapter, we shall be on erned with the histori al origins of
de Broglie's 1927 dynami s dened by (1) and (2). Some authors have
referred to this dynami s as `Bohmian me hani s'. Su h terminology is
misleading: it disregards de Broglie's priority, and misses de Broglie's
physi al motivations for re asting dynami s in terms of velo ity; it also
misrepresents Bohm's 1952 formulation, whi h was based on (1) and (3).
These and other histori al mis on eptions on erning de Broglie-Bohm
theory will be addressed in se tion 11.1.
The two equations (1), (2) dene a deterministi (de Broglian or pilotwave) dynami s for a single multiparti le system: given an initial wave
fun tion (x1 , x2 , ..., xN , 0) at t = 0, (1) determines (x1 , x2 , ..., xN , t)
at all times t; and given an initial onguration (x1 (0), x2 (0), ..., xN (0)),
(2) then determines the traje tory (x1 (t), x2 (t), ..., xN (t)). For an ensemble of systems with the same initial wave fun tion (x1 , x2 , ..., xN , 0),
and with initial ongurations (x1 (0), x2 (0), ..., xN (0)) distributed a ording to the Born rule
2
P (x1 , x2 , ..., xN , 0) = |(x1 , x2 , ..., xN , 0)| ,
(5)
the statisti al distribution of out omes of quantum measurements will

agree with the predi tions of standard quantum theory. This is shown by
treating the measuring apparatus, together with the system being measured, as a single multiparti le system obeying de Broglian dynami s,
so that (x1 , x2 , ..., xN ) denes the `pointer position' of the apparatus
as well as the onguration of the measured system. Given the initial
ondition (5) for any multiparti le system, the statisti al distribution
of parti le positions at later times will also agree with the Born rule
2
P = || . Thus, the statisti al distribution of pointer positions in any
experiment will agree with the predi tions of quantum theory, yielding
34
the orre t statisti al distribution of out omes for standard quantum

measurements.
In his 1927 Solvay report, de Broglie gave some simple appli ations of
pilot-wave theory, with the assumed initial ondition (5). He applied the
theory to single-photon interferen e, to atomi transitions, and to the
s attering (or dira tion) of ele trons by a rystal latti e. But a detailed
demonstration of equivalen e to quantum theory, and in parti ular a
pilot-wave a ount of the general quantum theory of measurement, was
not provided until the work of Bohm in 1952.
How did de Broglie ome to propose this theory in 1927? In this
hapter, we tra e de Broglie's work in this dire tion, from his early
work leading to his do toral thesis of 1924 (de Broglie 1924e, 1925), to
his ru ial paper of 1927 published in Journal de Physique (de Broglie
1927b), and ulminating in his presentation of pilot-wave theory at the
fth Solvay onferen e. We examine in detail how de Broglie arrived at
this new form of parti le dynami s, and what his attitude towards it
was. Later, in hapter 10, we shall onsider some of the dis ussions of
de Broglie's theory that took pla e at the onferen e, in parti ular the
famous (and widely misunderstood) lash between de Broglie and Pauli.
De Broglie's dynami s has the striking feature that ele trons and photons are regarded as both parti les and waves. Like many s ienti ideas,
this mingling of parti le-like and wave-like aspe ts had pre ursors. In
Newton's Opti ks (rst published in 1704), both wave-like and parti lelike properties are attributed to light. Newton's so- alled ` orpus ular'
theory was formulated on the basis of extensive and detailed experiments
( arried out by Grimaldi, Hooke, and Newton himself) involving what
we would now all interferen e and dira tion. A ording to Newton,
light orpus les or light `Rays' as he alled thema generate `Waves of Vibrations' in an Àethereal Medium', mu h as a stone thrown
into water generates water waves (Newton 1730; reprint, pp. 3479). In
addition, Newton supposed that the waves in turn ae t the motion
of the orpus les, whi h `may be alternately a elerated and retarded
by the Vibrations' (p. 348). In parti ular, Newton thought that the
ee t of the medium on the motion of the orpus les was responsible for
the phenomena of interferen e and dira tion. He writes, for example
(p. 350):
And doth not the gradual ondensation of this Medium extend to some di-
a The opening denition of the Opti ks denes `Rays' of light as ìts least Parts'.
2.1 Ba kground
35
stan e from the Bodies, and thereby ause the Inexions of the Rays of Light,
whi h pass by the edges of dense Bodies, at some distan e from the Bodies?
Newton understood that, for dira tion to o ur, the motion of the light
orpus les would have to be ae ted at a distan e by the dira ting body
`Do not Bodies a t upon Light at a distan e, and by their a tion
bend its Rays .... ?' (p. 339) and his proposed me hanism involved
waves in an inhomogeneous ether. Further, a ording to Newton, to
a ount for the oloured fringes that had been observed by Grimaldi
in the dira tion of white light by opaque bodies, the orpus les would
have to exe ute an os illatory motion `like that of an Eel' (p. 339):
Are not the Rays of Light in passing by the edges and sides of Bodies, bent
several times ba kwards and forwards, with a motion like that of an Eel? And
do not the three Fringes of olour'd Light above-mention'd arise from three
su h bendings?
For Newton, of ourse, su h non-re tilinear motion ould be aused only

by a for e emanating from the dira ting body.
It is interesting to note that, in the general dis ussion at the fth
Solvay onferen e (p. 509), de Broglie ommented on this very point,
with referen e to the èmission' (or orpus ular) theory, and pointed out
that if pilot-wave dynami s were written in terms of a eleration (as
done later by Bohm) then just su h for es appeared:
In the orpus ular on eption of light, the existen e of dira tion phenomena
o uring at the edge of a s reen requires us to assume that, in this ase, the
traje tory of the photons is urved. The supporters of the emission theory said
that the edge of the s reen exerts a for e on the orpus le. Now, if in the new
me hani s as I develop it, one writes the Lagrange equations for the photon,
one sees appear on the right-hand side of these equations a term .... [that ....
represents a sort of for e of a new kind, whi h exists only .... where there is
interferen e. It is this for e that will urve the traje tory of the photon when
its wave
is dira ted by the edge of a s reen.
The striking similarity between Newton's qualitative ideas and pilotwave theory has also been noted by Berry, who remarks that during interferen e or dira tion the de Broglie-Bohm traje tories indeed `wriggle
like an eel' (Berry 1997, p. 42), in some sense vindi ating Newton.
A mathemati al pre ursor to de Broglian dynami s is found in the
early nineteenth entury, in Hamilton's formulation of geometri al opti s
and parti le me hani s. As de Broglie points out in his Solvay report
(pp. 376 and 382), Hamilton's theory is in fa t the short-wavelength limit
of pilot-wave dynami s: for in that limit, the phase of the wave fun tion
36
obeys the Hamilton-Ja obi equation, and de Broglie's traje tories redu e
to those of lassi al me hani s.
A physi al theory of light as both parti les and waves in ee t a
revival of Newton's views emerged again with Einstein in 1905. It is
less well known that, after 1905, Einstein tried to onstru t theories of
lo alised light quanta oupled to ve tor elds in 3-spa e. As we shall see
in hapter 9, Einstein's ideas in this vein show some resemblan e to de
Broglie's but also dier from them.
It should also be mentioned that, in the autumn of 1923 (the same year
in whi h de Broglie rst elaborated his ideas), Slater tried to develop a
theory in whi h the motion of photons was guided by the ele tromagneti
eld. It appears that Slater rst attempted to onstru t a deterministi
theory, but had trouble dening an appropriate velo ity ve tor; he then
ame to the on lusion that photons and the ele tromagneti eld were
related only statisti ally, with the photon probability density being given
by the intensity of the eld. After dis ussing his ideas with Bohr and
Kramers in 1924, the photons were removed from the theory, apparently
against Slater's wishes (Mehra and Re henberg 1982a, pp. 5426). Note
that, while de Broglie applied his theory to photons, he made it lear
(for example in the general dis ussion, p. 508) that in his theory the
guiding ` -wave' was distin t from the ele tromagneti eld.
In the ase of light, then, the idea of ombining both parti le-like and
wave-like aspe ts was an old one, going ba k indeed to Newton. In the
ase of ordinary matter, however, de Broglie seems to have been the rst
to develop a physi al theory of this form.
It is sometimes laimed that, for the ase of ele trons, ideas similar
to de Broglie's were put forward by Madelung in 1926. What Madelung
proposed, however, was to regard an ele tron with mass m and wave
fun tion not as a pointlike parti le within the wave, but as a ontinuous
uid spread over spa e with mass density m ||2 (Madelung 1926a,b). In
this `hydrodynami al' interpretation, mathemati ally the uid velo ity
oin ides with de Broglie's velo ity eld; but physi ally, Madelung's
theory seems more akin to S hrdinger's theory than to de Broglie's.
Finally, before we examine de Broglie's work, we note what appears
to be a re urring histori al opposition to dualisti physi al theories
ontaining both waves and parti les. In 180103, Thomas Young, who by
his own a ount regarded his theory as a development of Newton's ideas,a
removed the orpus les from Newton's theory and produ ed a purely
a See, for example, Bernard Cohen's prefa e to Newton's Opti ks (1730, reprint).
2.2 A new approa h to parti le dynami s: 192324
37
undulatory a ount of light. In 1905, Einstein's dualist view of light

was not taken seriously, and did not win widespread support until the
dis overy of the Compton ee t in 1923. In 1924, Bohr and Kramers, who
regarded the Bohr-Kramers-Slater theory as a development of Slater's
original idea, insisted on removing the photons from Slater's theory
of radiation.a And in 1926, S hrdinger, who regarded his work as a
development of de Broglie's ideas, removed the traje tories from de
Broglie's theory and produ ed a purely undulatory `wave me hani s'.b

In this se tion we show how de Broglie took his rst steps towards a
new form of dynami s.a His aim was to explain the quantum phenomena
known at the time in parti ular the Bohr-Sommerfeld quantisation of
atomi energy levels, and the apparently dual nature of radiation by
unifying the physi s of parti les with the physi s of waves. To a omplish
this, de Broglie began by extending Einstein's wave-parti le duality for
light to all material bodies, by introdu ing a `phase wave' a ompanying every material parti le. Then, inspired by the opti al-me hani al
analogy,b de Broglie proposed that Newton's rst law of motion should
be abandoned, and repla ed by a new prin iple that unied Maupertuis'
variational prin iple for me hani s with Fermat's variational prin iple for
opti s. The result was a new form of dynami s in whi h the velo ity v of a
parti le is determined by the gradient of the phase of an a ompanying
wave in ontrast with lassi al me hani s, where a elerations are
determined by for es. (Note that de Broglie's phase has a sign opposite
to the phase S as we would normally dene it now.)
This new approa h to dynami s enabled de Broglie to obtain a wavelike explanation for the quantisation of atomi energy levels, to explain
the observed interferen e of single photons, and to predi t for the rst
time the new and unexpe ted phenomenon of the dira tion and interferen e of ele trons.
As we shall see, the theory proposed by de Broglie in 192324 was,
a In 1925, Born and Jordan attempted to restore the photons, proposing a sto hasti
theory reminis ent of Slater's original ideas; it appears that they were dissuaded
from publi ation by Bohr. See Darrigol (1992, p. 253) and se tion 3.4.2.
b Cf. se tion 4.5.
a An insightful and general a ount of de Broglie's early work, up to 1924, has been
given by Darrigol (1993).
b Possibly, de Broglie was also inuen ed by the philosopher Henri Bergson's writings
on erning time, ontinuity and motion (Feuer 1974, pp. 20614); though this is
denied by Lo hak (1992).
38
in fa t, a simple form of pilot-wave dynami s, for the spe ial ase of

independent parti les guided by waves in 3-spa e, and without a spe i
wave equation.
2.2.1 First papers on pilot-wave theory (1923)

De Broglie's earliest experien e of physi s was losely tied to experiment.
During the rst world war he worked on wireless telegraphy, and after the
war his rst papers on erned X-ray spe tros opy. In 1922 he published
a paper treating bla kbody radiation as a gas of light quanta (de Broglie
1922). In this paper, de Broglie made the unusual assumption that
photons had a very small but non-zero rest mass m0 . He was therefore
now applying Einstein's relations E = h and p = h/ (relating energy
and momentum to frequen y and wavelength) to massive parti les, even
if these were still only photons. It seems that de Broglie made the
assumption m0 6= 0 so that light quanta ould be treated in the same
way as ordinary material parti les. It appears that this paper was the
seed from whi h de Broglie's subsequent work grew.a
A ording to de Broglie's later re olle tions (L. de Broglie, AHQP
interview, 7 January 1963, p. 1),2 his rst ideas on erning a pilot-wave
theory of massive parti les arose as follows. During onversations on
the subje t of X-rays with his older brother Mauri e de Broglie,b he
be ame onvin ed that X-rays were both parti les and waves. Then, in
the summer of 1923, de Broglie had the idea of extending this duality
to ordinary matter, in parti ular to ele trons. He was drawn in this
dire tion by onsideration of the opti al-me hani al analogy; further,
the presen e of whole numbers in quantisation onditions suggested to
him that waves must be involved.
This last motivation was re alled by de Broglie (1999) in his Nobel
le ture of 1929:
.... the determination of the stable motions of the ele trons in the atom involves
whole numbers, and so far the only phenomena in whi h whole numbers were
involved in physi s were those of interferen e and of eigenvibrations. That
suggested the idea to me that ele trons themselves ould not be represented
a In a olle tion of papers by de Broglie and Brillouin, published in 1928, a footnote

added to de Broglie's 1922 paper on bla kbody radiation and light quanta remarks:
`This paper .... was the origin of the ideas of the author on wave me hani s' (de
Broglie and Brillouin 1928, p. 1).
b Mauri e, the sixth du de Broglie, was a distinguished experimental physi ist,
having done important work on the photoele tri ee t with X-rays experiments
that were arried out in his private laboratory in Paris.
39
as simple orpus les either, but that a periodi ity had also to be assigned to
them too.
De Broglie rst presented his new ideas in three notes published (in
Fren h) in the Comptes Rendus of the A ademy of S ien es in Paris (de
Broglie 1923a,b, ), and also in two papers published in English one
in Nature (de Broglie 1923d), the other in the Philosophi al Magazine
(de Broglie 1924a).a
The ideas in these papers formed the basis for de Broglie's do toral
thesis. The paper in the Philosophi al Magazine reads, in fa t, like a
summary of mu h of the material in the thesis. Sin e the thesis provides
a more systemati presentation, we shall give a detailed summary of it
in the next subse tion; here, we give only a brief a ount of the earlier
papers, ex ept for the ru ial se ond paper, whose on eptual ontent
warrants more detailed ommentary.b
The rst ommuni ation (de Broglie 1923a), entitled `Waves and quanta', proposes that an ìnternal periodi phenomenon' should be asso iated with any massive parti le (in luding light quanta). In the rest frame
of a parti le with rest mass m0 , the periodi phenomenon is assumed to
have a frequen y 0 = m0 c2 /h. In a frame where the parti le has uniform
v , de Broglie onsiders the two frequen ies and 1 , where
velo ity p
2
= 0 / 1 v 2 /c2 is the frequen y
p with the
p = mc /h asso iated
2
2
relativisti mass in rease m = m0 / 1 v /c and 1 = 0 1 v 2 /c2
is the time-dilated frequen y. De Broglie shows that, be ause 1 = (1
v 2 /c2 ), a ` titious' wave of frequen y and phase velo ity vph = c2 /v
(propagating in the same dire tion as the parti le) will remain in phase
with the internal os illation of frequen y 1 . De Broglie then onsiders
an atomi ele tron moving uniformly on a ir ular orbit. He proposes
that orbits are stable only if the titious wave remains in phase with
the internal os illation of the ele tron. From this ondition, de Broglie
derives the Bohr-Sommerfeld quantisation ondition.
The se ond ommuni ation (de Broglie 1923b), entitled `Light quanta,
dira tion and interferen e', has a more on eptual tone. De Broglie
begins by re alling his previous result, that a moving body must be asso iated with à non-material sinusoidal wave'. He adds that the parti le
velo ity v is equal to the group velo ity of the wave, whi h de Broglie here
alls `the phase wave' be ause its phase at the lo ation of the parti le is
equal to the phase of the internal os illation of the parti le. De Broglie
a It seems possible that Comptes Rendus was not widely read by physi ists outside
Fran e, but this ertainly was not true of Nature or the Philosophi al Magazine.
b We do not always keep to de Broglie's original notation.
40
then goes on to make some very signi ant observations about dira tion
and the nature of the new dynami s that he is proposing.
De Broglie asserts that dira tion phenomena prove that light quanta
annot always propagate in a straight line, even in what would normally
be alled empty spa e. He draws the bold on lusion that Newton's rst
law of motion (the `prin iple of inertia') must be abandoned (p. 549):
The light quanta [atomes de lumire whose existen e we assume do not always
propagate in a straight line, as proved by the phenomena of dira tion. It then
seems
ne essary
to modify the prin iple of inertia.
De Broglie then suggests repla ing Newton's rst law with a new postulate (p. 549):
We propose to adopt the following postulate as the basis of the dynami s of
the free material point: Àt ea h point of its traje tory, a free moving body
follows in a uniform motion the
ray
of its phase wave, that is (in an isotropi
medium), the normal to the surfa es of equal phase'.
The dira tion of light quanta is then explained sin e, as de Broglie notes, ìf the moving body must pass through an opening whose dimensions
are small ompared to the wavelength of the phase wave, in general its
traje tory will urve like the ray of the dira ted wave'.
In retrospe t, de Broglie's postulate for free parti les may be seen as a
simplied form of the law of motion of what we now know as pilot-wave
dynami s ex ept for the statement that the motion along a ray be
ùniform' (that is, have onstant speed), whi h in pilot-wave theory is
true only in spe ial ases.a De Broglie notes that his postulate respe ts
onservation of energy but not of momentum. And indeed, in pilot-wave
theory the momentum of a `free' parti le is generally not onserved: in
ee t (from the standpoint of Bohm's Newtonian formulation), the pilot
wave or quantum potential a ts like an èxternal sour e' of momentum
(and in general of energy too).b The abandonment of something as
a From Bohm's se ond-order equation (3) applied to a single parti le, for timeindependent V it follows that d( 12 mv2 + V + Q)/dt = Q/t (the usual energy
onservation formula with a time-dependent ontribution Q to the potential). In
free spa e (V = 0), the speed v is onstant if and only if dQ/dt = Q/t or
v Q = 0 (so that the `quantum for e' does no work), whi h is true only in
spe ial ases.
b Again from (3), in free spa e the rate of hange of momentum p = mv (where
v = S/m) is dp/dt = Q, whi h is generally non-zero. Further, in general (3)
implies d( 12 mv2 + V )/dt = v Q, so that the standard ( lassi al) expression
for energy is onserved if and only if v Q = 0. If, on the other hand, one denes
1
mv2 + V + Q as the ènergy', it will be onserved if and only if Q/t = 0, whi h
2
is true only for spe ial ases (in parti ular for stationary states, sin e for these ||
is time-independent).
41
elementary as momentum onservation is ertainly a radi al step by

any standards. On the other hand, if one is willing as de Broglie was
to propose a fundamentally new approa h to the theory of motion,
then the loss of lassi al onservation laws is not surprising, as these are
really properties of lassi al equations of motion.
De Broglie then makes a remarkable predi tion, that any moving body
(not just light quanta) an undergo dira tion:
.... any moving body ould in ertain ases be dira ted. A stream of ele trons
passing through a small enough opening will show dira tion phenomena. It
is in this dire tion that one should perhaps look for experimental onrmation
of our ideas.
Next, de Broglie puts his proposals in a general on eptual and histori al perspe tive. Con erning the role of the phase wave, he writes
(p. 549):
We therefore on eive of the phase wave as guiding the movements of energy,
and this is what an allow the synthesis of waves and quanta.
Here, for the rst time, de Broglie hara terises the phase wave as a
`guiding' wave. De Broglie then remarks that, histori ally speaking, the
theory of waves `went too far' by denying the dis ontinuous stru ture of
radiation and `not far enough' by not playing a role in dynami s. For de
Broglie, his proposal has a lear histori al signi an e (p. 549, itali s in
the original):
The new dynami s of the free material point is to the old dynami s (in luding
that of Einstein) what wave opti s is to geometri al opti s. Upon ree tion
one will see that the proposed synthesis appears as the logi al ulmination of
the omparative development of dynami s and of opti s sin e the seventeenth
entury.
In the se ond part of this note, de Broglie onsiders the explanation of

opti al interferen e fringes. He assumes that the probability for an atom
to absorb or emit a light quantum is determined by `the resultant of one
of the ve tors of the phase waves rossing ea h other there [se roisant sur
lui' (pp. 54950). In Young's interferen e experiment, the light quanta
passing through the two holes are dira ted, and the probability of them
being dete ted behind the s reen will vary from point to point, depending
on the `state of interferen e' of the phase waves. De Broglie on ludes
that there will be bright and dark fringes as predi ted by the wave
theories, no matter how feeble the in ident light.
This approa h to opti al interferen e in whi h interfering phase
42
waves determine the probability for intera tion between photons and
the atoms in the dete tion apparatus is elaborated in de Broglie's
thesis (see below). Soon after ompleting his thesis (apparently), de
Broglie abandoned this idea in favour of a simpler approa h, in whi h
the interfering phase waves determine the number density of photon
traje tories (see se tion 2.2.3).
In de Broglie's third ommuni ation (de Broglie 1923 ), entitled `Quanta, the kineti theory of gases and Fermat's prin iple', part 1 onsiders
the statisti al treatment of a gas of parti les a ompanied by phase
waves. De Broglie makes the following assumption:
The state of the gas will then be stable only if the waves orresponding to all
of the atoms form a system of stationary waves.
In other words, de Broglie onsiders the stationary modes, or standing

waves, asso iated with a given spatial volume. He assumes that ea h
mode ` an transport zero, one, two or several atoms', with probabilities
determined by the Boltzmann fa tor.a A ording to de Broglie, for a
gas of nonrelativisti atoms his method yields the Maxwell distribution,
while for a gas of photons it yields the Plan k distribution.b
In part 2 of the same note, de Broglie shows how his new dynami al
postulate amounts to a uni ation of Maupertuis' prin iple of least
a tion with Fermat's prin iple of least time in opti s. Let us re all that,
in the me hani al prin iple of Maupertuis for parti le traje tories,
Z b
mv dx = 0 ,
(6)
a
the ondition of stationarity determines the parti le paths. (In (6) the
energy is xed on the varied paths; at the end points, x = 0 but t
need not be zero.) While in the opti al prin iple of Fermat for light rays,
Z b
d = 0 ,
(7)
a
the stationary line integral for the phase hange the stationary òpti al
path length' provides a ondition that determines the path of a ray
onne ting two points, in spa e (for the time-independent ase) or in
spa etime. Now, a ording to de Broglie's basi postulate: `The rays
a As remarked by Pais (1982, pp. 4356), in this paper de Broglie èvaluated
independently of Bose (and published before him) the density of radiation states
in terms of parti le (photon) language'.
b As shown by Darrigol (1993), de Broglie made some errors in his appli ation of
the methods of statisti al me hani s.
43
of the phase waves oin ide with the dynami ally possible traje tories'
(p. 632). The rays are des ribed by Fermat's prin iple (for the ase
of a dispersive medium), whi h de Broglie shows oin ides with Maupertuis' prin iple, as follows: writing the element of phase hange as
d = 2dl/vph , where dl is an element of path and vph = c2 /v is the
phase velo ity, and using the relation = E/h = mc2 /h, the element
of phase hange may be rewritten as (2/h)mvdl, so that (7) oin ides
with (6). As de Broglie puts it (p. 632):
In this way the fundamental link that unites the two great prin iples of
geometri al opti s and of dynami s is brought fully to light.
De Broglie remarks that some of the dynami ally possible traje tories
will be ìn resonan e with the phase
wave', and that these orrespond to
R
Bohr's stable orbits, for whi h dl/vph is a whole number.
Soon afterwards, de Broglie introdu es a ovariant 4-ve tor formulation of his basi dynami al postulate (de Broglie 1924a,b). He denes
a 4-ve tor w = (/c, (/vph )
n), where n
is a unit ve tor in the
dire tion of a ray of the phase wave, and assumes it to be related to the
energy-momentum 4-ve tor p = (E/c, p) by p = hw . De Broglie
notes that the identity of the prin iples of Maupertuis and Fermat then
follows immediately. We shall dis uss this in more detail in the next
subse tion.
2.2.2 Thesis (1924)

He has lifted a orner of the great veil.
Einstein, ommenting on de Broglie's thesis
De Broglie's do toral thesis (de Broglie 1924e) was mostly based on

the above papers. It seems to have been ompleted in the summer of
1924, and was defended at the Sorbonne in November. The thesis was
published early in 1925 in the Annales de Physique (de Broglie 1925).b
When writing his thesis, de Broglie was well aware that his theory had
gaps. As he put it (p. 30):c
a Letter to Langevin, 16 De ember 1924 (quoted in Darrigol 1993, p. 355).
b An English translation of extra ts from de Broglie's thesis appears in Ludwig
(1968). A omplete translation has been done by A. F. Kra klauer ( urrently online
at http://www.ensmp.fr/ab/LDB-oeuvres/De_Broglie_Kra klauer.htm ). All
translations here are ours.
Here and below, page referen es for de Broglie's thesis orrespond to the published
version in Annales de Physique (de Broglie 1925).
44
.... the main aim of the present thesis is to present a more omplete a ount
of the new ideas that we have proposed, of the su esses to whi h they have
led, and also of the many gaps they ontain.
De Broglie begins his thesis with a histori al introdu tion. Newtonian

me hani s, he notes, was eventually formulated in terms of the prin iple
of least a tion, whi h was rst given by Maupertuis and then later in
another form by Hamilton. As for the s ien e of light and opti s, the
laws of geometri al opti s were eventually summarised by Fermat in
terms of a prin iple whose form is reminis ent of the prin iple of least
a tion. Newton tried to explain some of the phenomena of wave opti s
in terms of his orpus ular theory, but the work of Young and Fresnel
led to the rise of the wave theory of light, in parti ular the su essful
wave explanation of the re tilinear propagation of light (whi h had been
so lear in the orpus ular or èmission' theory). On this, de Broglie
omments (p. 25):
When two theories, based on ideas that seem entirely dierent, a ount for
the same experimental fa t with equal elegan e, one an always wonder if the
opposition between the two points of view is truly real and is not due solely
to an inadequa y of our eorts at synthesis.
This remark is, of ourse, a hint that the aim of the thesis is to ee t just
su h a synthesis. De Broglie then turns to the rise of ele trodynami s, relativity, and the theory of energy quanta. He notes that Einstein's theory
of the photoele tri ee t amounts to a revival of Newton's orpus ular
theory. De Broglie then sket hes Bohr's 1913 theory of the atom, and
goes on to point out that observations of the photoele tri ee t for
X- and -rays seem to onrm the orpus ular hara ter of radiation.
At the same time, the wave aspe t ontinues to be onrmed by the
observed interferen e and dira tion of X-rays. Finally, de Broglie notes
the very re ent orpus ular interpretation of Compton s attering. De
Broglie on ludes his histori al introdu tion with a mention of his own
re ent work (p. 30):
.... the moment seemed to have arrived to make an eort towards unifying the
orpus ular and wave points of view and to go a bit more deeply into the true
meaning of the quanta. That is what we have done re ently ....
De Broglie learly regarded his own work as a synthesis of earlier theories of dynami s and opti s, a synthesis in reasingly for ed upon us by
a umulating experimental eviden e.
Chapter 1 of the thesis is entitled `The phase wave'. De Broglie begins
45
by re alling the equivalen e of mass and energy implied by the theory

of relativity. Turning to the problem of quanta, he remarks (pp. 323):
It seems to us that the fundamental idea of the quantum theory is the impossibility of onsidering an isolated quantity of energy without asso iating a
ertain frequen y with it. This onne tion is expressed by what I shall all the
quantum relation:
energy = h frequency
where
is Plan k's onstant.
To make sense of the quantum relation, de Broglie proposes that (p. 33)
.... to ea h energy fragment of proper mass
phenomenon of frequen y
m0
there is atta hed a periodi
su h that one has:
h0 = m0 c2
0
being measured, of ourse, in the system tied to the energy fragment.
De Broglie asks if the periodi phenomenon must be assumed to be

lo alised inside the energy fragment. He asserts that this is not at all
ne essary, and that it will be seen to be `without doubt spread over an
extensive region of spa e' (p. 34).
De Broglie goes on to onsider
p the apparent ontradi tion between the
2
=
mc
/h
=
/
1 v 2 /c2 and the time-dilated frequen y
frequen y
0
p
2
2
1 = 0 1 v /c . He proposes that the ontradi tion is resolved by
the following `theorem of phase harmony' (p. 35): in a frame where
the moving body has velo ity v , the periodi phenomenon tied to the
moving body and with frequen y 1 is always in phase with a wave
of frequen y propagating in the same dire tion as the moving body
with phase velo ity vph = c2 /v . This is shown by applying the Lorentz
transformation to a rest-frame wave sin (0 t0 ), yielding a wave
i
h
p
sin 0 t vx/c2 / 1 v 2 /c2
(8)
p
of frequen y = 0 / 1 v 2 /c2 and phase velo ity c2 /v . Regarding the
nature of this wave de Broglie says that, be ause its velo ity is greater
than c, it annot be a wave transporting energy: rather, ìt represents the
spatial distribution of the phases of a phenomenon; it is a phase wave '
(p. 36). De Broglie shows that the group velo ity of the phase wave is
equal to the velo ity of the parti le. In the nal se tion of hapter 1
(`The phase wave in spa etime'), he dis usses the appearan e of surfa es
of onstant phase for dierently moving observers, from a spa etime
perspe tive.
46
Chapter 2 is entitled `Maupertuis' prin iple and Fermat's prin iple'.

The aim is to generalise the results of the rst hapter to non-uniform,
non-re tilinear motion. In the introdu tion to hapter 2 de Broglie writes
(p. 45):
Guided by the idea of a deep unity between the prin iple of least a tion and
that of Fermat, from the beginning of my investigations on this subje t I was
led to
assume
that, for a given value of the total energy of the moving body
and therefore of the frequen y of its phase wave, the dynami ally possible
traje tories of the one oin ided with the possible rays of the other.
De Broglie dis usses the prin iple of least a tion, in the dierent forms
given by Hamilton and by Maupertuis, and also for relativisti parti les
in an external ele tromagneti eld. He writes Hamilton's prin iple as
Z Q
p dx = 0
(9)
P
( = 0, 1, 2, 3, with dx = cdt), where P , Q are points in spa etime and

p is the anoni al energy-momentum 4-ve tor, and notes that if p0 is
onstant the prin iple be omes
Z B
(10)
pi dxi = 0
0
(i = 1, 2, 3), where A, B are the orresponding points in spa e that

is, Hamilton's prin iple redu es to Maupertuis' prin iple.
De Broglie then dis usses wave propagation and Fermat's prin iple
from a spa etime perspe tive. He onsiders a sinusoidal fun tion sin ,
where the phase has a spa etime-dependent dierential d, and writes
the variational prin iple for the ray in spa etime in the Hamiltonian
form
Z Q
d = 0 .
(11)
P
De Broglie then introdu es a 4-ve tor eld w on spa etime, dened by

d = 2w dx ,
(12)
where the w are generally fun tions on spa etime. (Of ourse, this
implies that 2w = , though de Broglie does not write this expli itly.) De Broglie also notes that d = 2(dt (/vph )dl) and w =
is a unit ve tor in the dire tion of propaga(/c, (/vph )
n), where n
tion; and that if is onstant, the prin iple in the Hamiltonian form
Z Q
w dx = 0
(13)
P

redu es to the prin iple in the Maupertuisian form
Z B
wi dxi = 0 ,
47
(14)
or
B
A
dl = 0 ,
vph
(15)
whi h is Fermat's prin iple.

De Broglie then dis usses an èxtension of the quantum relation' (that
is, an extension of E = h ). He states that the two 4-ve tors p and
w play perfe tly symmetri al roles in the motion of a parti le and in
the propagation of a wave. Writing the `quantum relation' E = h as
w0 = h1 p0 , de Broglie proposes the generalisation
w =
1
p ,
h
(16)
so that
d = 2w dx =
2
p dx .
h
Fermat's prin iple then be omes

Z B
pi dxi = 0 ,
(17)
(18)
whi h is the same as Maupertuis' prin iple. Thus, de Broglie arrives at

the following statement (p. 56):
Fermat's prin iple applied to the phase wave is identi al to Maupertuis' prin iple applied to the moving body; the dynami ally possible traje tories of the
moving body are identi al to the possible rays of the wave.
He adds that (p. 56):

We think that this idea of a deep relationship between the two great prin iples
of Geometri al Opti s and Dynami s ould be a valuable guide in realising the
synthesis of waves and quanta.
De Broglie then dis usses some parti ular ases: the free parti le,
a parti le in an ele trostati eld, and a parti le in a general ele tromagneti eld. He al ulates the phase velo ity, whi h depends on the
ele tromagneti potentials. He notes that the propagation of a phase
wave in an external eld depends on the harge and mass of the moving
body. And he shows that the group velo ity along a ray is still equal to
the velo ity of the moving body.
48
For the ase of an ele tron of harge e and velo ity v in an ele trostati
potential , de Broglie writes down the following expressions for the
frequen y and phase velo ity vph of the phase wave (p. 57):
= (mc2 + e)/h , vph = (mc2 + e)/mv
(19)
p
(where again m = m0 / 1 v 2 /c2 ). He shows that vph may be rewritten
as the free value c2 /v multiplied by a fa tor h/(h e) that depends
on the potential . The expressions (19) formed the starting point for
S hrdinger's work on the wave equation for de Broglie's phase waves
(as re onstru ted by Mehra and Re henberg (1987, pp. 4235), see se tion 2.3).
While de Broglie does not expli itly say so in his thesis, note that from
the denition (12) of w , the generalised quantum relation (16) may be
written in the form
p = .
(20)
This is what we would now all a relativisti guidan e equation, giving

the velo ity of a parti le in terms of the gradient of the phase of a pilot
wave (where here de Broglie denes the phase to be dimensionless).
In other words, the extended quantum relation is a rst-order equation
of motion. In the presen e of an ele tromagneti eld, the anoni al
momentum p ontains the 4-ve tor potential. For a free parti le, with
p = (E/c, p), the guidan e equation has omponents
p = ,
E = ,
where the spatial omponents may also be written as

m0 v
mv = p
= .
1 v 2 /c2
(21)
(22)
For a plane wave of phase = t k x, we have

E = , p = k .
(23)
Thus, de Broglie's uni ation of the prin iples of Maupertuis and

Fermat amounts to a new dynami al law, (16) or (20), in whi h the
phase of a guiding wave determines the parti le velo ity. This new law
of motion is the essen e of de Broglie's new, rst-order dynami s.
Chapter 3 of de Broglie's thesis is entitled `The quantum onditions for
the stability of orbits'. De Broglie reviews Bohr's ondition for ir ular
orbits, a ording to whi h the angular
R 2momentum of the ele tron must
be a multiple of , or equivalently 0 p d = nh (p onjugate to ).

49
H
He also reviews Sommerfeld's generalisation,
H P3 pi dqi = ni h (integral
ni ) and Einstein's invariant formulation
i=1 pi dqi = nh (integral
n). De Broglie then provides an explanation for Einstein's ondition.
The traje tory of the moving body oin ides with one of the rays of
its phase wave, and the phase wave moves along the traje tory with
a onstant frequen y (be ause the total energy is onstant) and with
a variable speed whose value has been al ulated. To have a stable
orbit, laims de Broglie, the length l of the orbit must be in `resonan e'
with
the wave: thus l = n in the ase of onstant wavelength, and
H
(/vph )dl = n (n integral) generally. De Broglie notes that this is
pre isely the integral appearing in Fermat's prin iple, whi h has been
shown to be equal to the integral giving the Maupertuisian a tion divided
by h. The resonan e ondition is then identi al to the required stability
ondition. For the simple ase of ir ular orbits in the Bohr atom de
Broglie shows, using
H vph = and h/ =2 m0 v , that the resonan e
ondition be omes m0 vdl = nh or m0 R = n (with v = R), as
originally given by Bohr.a (Note that the simple argument ommonly
found in textbooks, about the tting of whole numbers of wavelengths
along a Bohr orbit, originates in this work of de Broglie's.)
De Broglie thought that his explanation of the stability or quantisation
onditions onstituted important eviden e for his ideas. As he puts it
(p. 65):
This beautiful result, whose demonstration is so immediate when one has
a epted the ideas of the pre eding hapter, is the best justi ation we an
give for our way of atta king the problem of quanta.
Certainly, de Broglie had a hieved a on rete realisation of his initial

intuition that quantisation onditions for atomi energy levels ould arise
from the properties of waves.
In his hapter 4, de Broglie onsiders the two-body problem, in parti ular the hydrogen atom. He expresses on ern over how to dene the
proper masses, taking into a ount the intera tion energy. He dis usses
the quantisation onditions for hydrogen from a two-body point of view:
he has two phase waves, one for the ele tron and one for the nu leus.
The subje t of hapter 5 is light quanta. De Broglie suggests that the
lassi al (ele tromagneti ) wave distribution in spa e is some sort of time
average over the true distribution of phase waves. His light quantum is
a De Broglie also laims to generalise his results from losed orbits to quasi-periodi
(or multi-periodi ) motion: however, as shown by Darrigol (1993), de Broglie's
derivation is faulty.
50
assigned a very small proper mass: the velo ity v of the quantum, and
the phase velo ity c2 /v of the a ompanying phase wave, are then both
very lose to c.
De Broglie points out that radiation is sometimes observed to violate
re tilinear propagation: a light wave striking the edge of a s reen dira ts
into the geometri al shadow, and rays passing lose to the s reen deviate
from a straight line. De Broglie notes the two histori al explanations for
this phenomenon on the one hand the explanation for dira tion given
by the wave theory, and on the other the explanation given by Newton in
his emission theory: `Newton assumed [the existen e of a for e exerted
by the edge of the s reen on the orpus le' (p. 80). De Broglie asserts
that he an now give a unied explanation for dira tion, by abandoning
Newton's rst law of motion (p. 80):
.... the ray of the wave would urve as predi ted by the theory of waves, and
the moving body, for whi h the prin iple of inertia would no longer be valid,
would suer the same deviation as the ray with whi h its motion is bound up
[solidaire ....
De Broglie's words here deserve emphasis. As is also very lear in his

se ond paper of the pre eding year (see se tion 2.2.1), de Broglie regards
his explanation of parti le dira tion as based on a new form of dynami s
in whi h Newton's rst law the prin iple that a free body will always
move uniformly in a straight line is abandoned. At the same time,
de Broglie re ognises that one an always adopt a lassi al-me hani al
viewpoint if one wishes (pp. 8081):
.... perhaps one ould say that the wall exerts a for e on it [the moving body
if one takes the urvature of the traje tory as a riterion for the existen e of
a for e.
Here, de Broglie re ognises that one may still think in Newtonian terms,
if one ontinues to identify a eleration as indi ative of the presen e of
a for e. Similarly, as we shall see, in 1927 de Broglie notes that his pilotwave dynami s may if one wishes be written in Newtonian form with a
quantum potential. But de Broglie's preferred approa h, throughout his
work in the period 192327, is to abandon Newton's rst law and base
his dynami s on velo ity rather than on a eleration.
After onsidering the Doppler ee t, ree tion by a moving mirror,
and radiation pressure, all from a photon viewpoint, de Broglie turns to
the phenomena of wave opti s, noting that (p. 86):
The stumbling blo k of the theory of light quanta is the explanation of the
phenomena that onstitute wave opti s.
51
Here it be omes apparent that, despite his understanding of how nonre tilinear parti le traje tories arise during dira tion and interferen e,
de Broglie is not sure of the details of how to explain the observed bright
and dark fringes in dira tion and interferen e experiments with light.
In parti ular, de Broglie did not have a pre ise theory of the assumed
statisti al relationship between his phase waves and the ele tromagneti
eld. Even so, he went on to make what he alled `vague suggestions'
(p. 87) towards a detailed theory of opti al interferen e. De Broglie's
idea was that the phase waves would determine the probability for the
light quanta to intera t with the atoms onstituting the equipment used
to observe the radiation, in su h a way as to a ount for the observed
fringes (p. 88):
.... the probability of rea tions between atoms of matter and atoms of light is
at ea h point tied to the resultant (or rather to the mean value of this) of one
of the ve tors hara terising the phase wave; where this resultant vanishes the
light is undete table; there is interferen e. One then on eives that an atom
of light traversing a region where the phase waves interfere will be able to be
absorbed by matter at ertain points and not at others. This is the still very
qualitative prin iple of an explanation of interferen e .... .
As we shall see in the next se tion, after ompleting his thesis de Broglie
arrived at a simpler explanation of opti al interferen e fringes.
The nal se tion of hapter 5 onsiders the explanation of Bohr's
frequen y ondition h = E1 E2 for the light emitted by an atomi
transition from energy state E1 to energy state E2 . De Broglie derives
this from the assumption that ea h transition involves the emission of a
single light quantum of energy E = h (together with the assumption
of energy onservation).
De Broglie's hapter 6 dis usses the s attering of X- and -rays.
In his hapter 7, de Broglie turns to statisti al me hani s, and shows
how the on ept of statisti al equilibrium is to be modied in the presen e of phase waves. If ea h parti le or atom in a gas is a ompanied
by a phase wave, then a box of gas will be ` riss- rossed in all dire tions'
(p. 110) by the waves. De Broglie nds it natural to assume that the
only stable phase waves in the box will be those that form stationary or
standing waves, and that only these will be relevant to thermodynami
equilibrium. He illustrates his idea with a simple example of mole ules
moving in one dimension, onned to an interval of length l. In the
nonrelativisti limit, ea h phase wave has a wavelength = h/m0 v and
the `resonan e ondition' is l = n with n integral. Writing v0 = h/m0 l,
one then has v = nv0 . As de Broglie notes (p. 112): `The speed will then
52
be able to take only values equal to integer multiples of v0 '. (This is, of
ourse, the well-known quantisation of momentum for parti les onned
to a box.) De Broglie then argues that a velo ity element v orresponds
to a number n = (m0 l/h)v of states of a mole ule ( ompatible with
the existen e of stationary phase waves), so that an element m0 xv
of phase spa e volume orresponds to a number m0 xv/h of possible
states. Generalising to three dimensions, de Broglie is led to take the
element of phase spa e volume divided by h3 as the measure of the
number of possible states of a mole ule, as assumed by Plan k.
De Broglie then turns to the photon gas, for whi h he obtains Wien's
law. He laims that, in order to get the Plan k law, the following further
hypothesis is required (p. 116):
If two or several atoms [of light have phase waves that are exa tly superposed,
of whi h one an therefore say that they are transported by the same wave,
their motions an no longer be onsidered as entirely independent and these
atoms an no longer be treated as separate units in al ulating the probabilities.
In de Broglie's approa h, the stationary phase waves play the role of the
elementary obje ts of statisti al me hani s. De Broglie denes stationary
waves as a superposition of two waves of the form
i
i
x
x
sin h
sin h
and
, (24)
2 t + + 0
2 t + 0
cos
cos
where 0 an take any value from 0 to 1 and takes one of the allowed
values. Ea h elementary wave an arry any number 0, 1, 2, ... of atoms,
and the probability of arrying n atoms is given by the Boltzmann fa tor
enh/kT . Applying this method to a gas of light quanta, de Broglie
laims to derive the Plan k distribution.a
De Broglie's thesis ends with a summary and on lusions (pp. 1258).
The seeds of the problem of quanta have been shown, he laims, to be
ontained in the histori al `parallelism of the orpus ular and wave-like
on eptions of radiation'. He has postulated a periodi phenomenon
asso iated with ea h energy fragment, and shown how relativity requires
us to asso iate a phase wave with every uniformly moving body. For the
ase of non-uniform motion, Maupertuis' prin iple and Fermat's prin iple ` ould well be two aspe ts of a single law', and this new approa h
to dynami s led to an extension of the quantum relation, giving the
speed of a phase wave in an ele tromagneti eld. The most important
a Again, as shown by Darrigol (1993), de Broglie's appli ation of statisti al
me hani s ontains some errors.
53
onsequen e is the interpretation of the quantum onditions for atomi

orbits in terms of a resonan e of the phase wave along the traje tories:
`this is the rst physi ally plausible explanation proposed for the BohrSommerfeld stability onditions'. A `qualitative theory of interferen e'
has been suggested. The phase wave has been introdu ed into statisti al
me hani s, yielding a derivation of Plan k's phase volume element, and
of the bla kbody spe trum. De Broglie has, he laims, perhaps ontributed to a uni ation of the opposing on eptions of waves and parti les,
in whi h the dynami s of the material point is understood in terms of
wave propagation. He adds that the ideas need further development: rst
of all, a new ele tromagneti theory is required, that takes into a ount
the dis ontinuous stru ture of radiation and the physi al nature of phase
waves, with Maxwell's theory emerging as a statisti al approximation.
The nal paragraph of de Broglie's thesis emphasises the in ompleteness
of his theory at the time:
I have deliberately left rather vague the denition of the phase wave, and of
the periodi phenomenon of whi h it must in some sense be the translation,
as well as that of the light quantum. The present theory should therefore be
onsidered as one whose physi al ontent is not entirely spe ied, rather than
as a onsistent and denitively onstituted do trine.
As de Broglie's on luding paragraph makes lear, his theory of 1924

was rather abstra t. There was no spe ied basis for the phase waves
(they were ertainly not regarded as `material' waves); nor was any
parti ular wave equation suggested. It should also be noted that at this
time de Broglie's waves were real-valued fun tions of spa e and time,
of the form sin(t k x), with a real os illating amplitude. They
were not omplex waves ei(kxt) of uniform amplitude. Thus, de
Broglie's `phase waves' had an os illating amplitude as well as a phase.
(De Broglie seems to have alled them `phase waves' only be ause of his
theorem of phase harmony.) Note also that, in his treatment of parti les
in a box, de Broglie superposes waves propagating in opposite dire tions,
yielding stationary waves whose amplitudes os illate in time.
In his thesis de Broglie does not expli itly dis uss dira tion or interferen e experiments with ele trons, even though in his se ond ommuni ation of 1923 (de Broglie 1923b) he had suggested ele tron dira tion
as an experimental test. A ording to de Broglie's later re olle tions (L.
de Broglie, AHQP Interview, 7 January 1963, p. 6),3 at his thesis defen e
on 25 November 1924:
Mr Jean Perrin, who haired the ommittee, asked me if my ideas ould lead to
54
experimental onrmation. I replied that yes they ould, and I mentioned the
dira tion of ele trons by rystals. Soon afterwards, I advised Mr Dauvillier ....
to try the experiment, but, absorbed by other resear h, he did not do it. I do
not know if he believed, or if he said to himself that it was perhaps very
un ertain, that he was going to go to a lot of trouble for nothing it's
possible. .... But the following year it was dis overed in Ameri a by Davisson
and Germer.
2.2.3 Opti al interferen e fringes: November 1924

On 17 November 1924, just a few days before de Broglie defended his
thesis, a further ommuni ation of de Broglie's was presented to the
A ademy of S ien es: entitled Òn the dynami s of the light quantum
and interferen e', and published in the Comptes Rendus, this short note
gave a new and improved a ount of opti al interferen e in terms of light
quanta (de Broglie 1924d).
De Broglie began his note by re alling his unsatisfa tory dis ussion of
opti al interferen e in his re ent work on the quantum theory (p. 1039):
.... I had not rea hed a truly satisfying explanation for the phenomena of
wave opti s whi h, in prin iple, all ome down to interferen e. I limited myself
to putting forward a ertain onne tion between the state of interferen e of
the waves and the probability for the absorption of light quanta by matter.
This viewpoint now seems to me a bit arti ial and I tend towards adopting
another, more in harmony with the broad outlines of my theory itself.
As we have seen, in his thesis de Broglie was unsure about how to

a ount for the bright and dark fringes observed in opti al interferen e
experiments. He did not have a theory of the ele tromagneti eld,
whi h he assumed emerged as some sort of average over his phase waves.
To a ount for opti al fringes, he had suggested that the phase waves
somehow determined the probability for intera tions between photons
and the atoms in the apparatus. Now, after ompleting his thesis, he
felt he had a better explanation, that was based purely on the spatial
distribution of the photon traje tories.
De Broglie's note ontinues by outlining his `new dynami s', in whi h
the energy-momentum 4-ve tor of every material point is proportional
to the ` hara teristi ' 4-ve tor of an asso iated wave, even when the
wave undergoes interferen e or dira tion. He then gives his new view
of interferen e fringes (p. 1040):
The rays predi ted by the wave theories would then be in every ase the
possible traje tories of the quantum. In the phenomena of interferen e, the rays
55
be ome on entrated in those regions alled `bright fringes' and be ome diluted
in those regions alled `dark fringes'. In my rst explanation of interferen e,
the dark fringes were dark be ause the a tion of fragments of light on matter
was zero there; in my urrent explanation, these fringes are dark be ause the
number of quanta passing through them is small or zero.
Here, then, de Broglie explains bright and dark fringes simply in terms of
a high or low density of photon traje tories in the orresponding regions.
When de Broglie speaks of the traje tories being on entrated and
diluted in regions of bright and dark fringes respe tively, he presumably
had in mind that the number density of parti les in an interferen e zone
should be proportional to the lassi al wave intensity, though he does
not say this expli itly. We an dis ern the essen e of the more pre ise
and omplete explanation of opti al interferen e given by de Broglie
three years later in his Solvay report: there, de Broglie has the same
photon traje tories, with a number density spe ied as proportional
to the amplitude-squared of the guiding wave (see pp. 383 f., and our
dis ussion in se tion 6.1.1).
In his note of November 1924, de Broglie goes on to illustrate his proposal for the ase of Young's interferen e experiment with two pinholes
a ting as point sour es. De Broglie ites the well-known fa ts that in
this ase the surfa es of equal phase are ellipsoids with the pinholes as
fo i, and that the rays (whi h are normal to the ellipsoidal surfa es)
are on entrated on hyperboloids of onstru tive interferen e where the
lassi al intensity has maxima. He then notes (p. 1040):
Let
let
r1 and r2 be the distan es from a point in spa e

be the fun tion 12 (r1 + r2 ), whi h is onstant on
to the two holes and

ea h surfa e of equal
phase. One easily shows that the phase velo ity of the waves along a ray is
equal to the value it would have in the ase of free propagation divided by the
derivative of
taken along the ray; as for the speed of the quantum, it will
be equal to the speed of free motion multiplied by the same derivative.
De Broglie gives no further details, but these are easily re onstru ted.
For an in ident beam of wavelength = 2/k , ea h pinhole a ts as a
sour e of a spheri al wave of wavelength , yielding a resultant wave
proportional to the real part of
ei(kr2 t)
ei(kr1 t)
+
.
r1
r2
(25)
If the pinholes have a separation d, then at large distan es (r1 , r2 >> d)
56
from the s reen the resultant wave may be approximated as

ei(krt)
2
cos
r
kd
(26)
where is the (small) angular deviation from the normal to the s reen.
The amplitude shows the well-known interferen e pattern. As for the
phase = kr t, with r = 12 (r1 + r2 ), the surfa es of equal phase are
indeed the well-known ellipsoids. Further, the phase velo ity is given by
vph =
|/t|
1
c
=
=
,
||
k |r|
|r|
(27)
while de Broglie's parti le velo ity given by (22) has magnitude

v=
||
||
k
=
= c2 |r| = c |r|
2
m
(/c )
(28)
(where m is the relativisti photon mass), in agreement with de Broglie's

assertions.
At the end of his note, de Broglie omments that this method may be
applied to the study of s attering.
In November 1924, then, de Broglie understood how interfering phase
waves would ae t photon traje tories, ausing them to bun h together
in regions oin iding with the observed bright fringes.
Note that in this paper de Broglie treats his phase waves as if they were
a dire t representation of the ele tromagneti eld. In his dis ussion of
Young's interferen e experiment, he has phase waves emerging from the
two holes and interfering, and he identies the interferen e fringes of his
phase waves with opti al interferen e fringes. However, he seems quite
aware that this is a simpli ation,a remarking that `the whole theory will
be ome truly lear only if one manages to dene the stru ture of the light
wave'. De Broglie is still not sure about the pre ise relationship between
his phase waves and light waves, a situation that persists even until
the fth Solvay onferen e: there, while he gives (pp. 383 f.) a pre ise
a ount of opti al interferen e in his report, he points out (p. 508) that
the onne tion between his guiding wave and the ele tromagneti eld
is still unknown.
a De Broglie may have thought of this as analogous to s alar wave opti s, whi h
predi ts the orre t opti al interferen e fringes by treating the ele tromagneti
eld simply as a s alar wave.
2.3 Towards a omplete pilot-wave dynami s: 192527
57

On 16 De ember 1924, Einstein wrote to Lorentz:a
A younger brother of the de Broglie known to us [Mauri e de Broglie has made
a very interesting attempt to interpret the Bohr-Sommerfeld quantization rules
(Paris Dissertation, 1924). I believe that it is the rst feeble ray of light
to illuminate this, the worst of our physi al riddles. I have also dis overed
something that supports his onstru tion.
What Einstein had dis overed, in support of de Broglie's ideas, appeared

in the se ond of his famous papers on the quantum theory of the ideal
gas (Einstein 1925a). Einstein showed that the u tuations asso iated
with the new Bose-Einstein statisti s ontained two distin t terms that
ould be interpreted as parti le-like and wave-like ontributions just
as Einstein had shown, many years earlier, for bla kbody radiation.
Einstein argued that the wave-like ontribution should be interpreted
in terms of de Broglie's matter waves, and he ited de Broglie's thesis.
It was largely through this paper by Einstein that de Broglie's work
be ame known outside Fran e.
In the same paper, Einstein suggested that a mole ular beam would
undergo dira tion through a su iently small aperture. De Broglie had
already made a similar suggestion for ele trons, in his se ond ommuni ation to the Comptes Rendus (de Broglie 1923b). Even so, in
their report at the fth Solvay onferen e, Born and Heisenberg state
that in his gas theory paper Einstein `dedu ed from de Broglie's daring
theory the possibility of dira tion of material parti les' (p. 425),
giving the in orre t impression that Einstein had been the rst to see
this onsequen e of de Broglie's theory. It seems likely that Born and
Heisenberg did not noti e de Broglie's early papers in the Comptes
Rendus.
In 1925 Elsasser a student of Born's in Gttingen read de
Broglie's thesis. Like most others outside Fran e, Elsasser had heard
about de Broglie's thesis through Einstein's gas theory papers. Elsasser
suspe ted that two observed experimental anomalies ould be explained
by de Broglie's new dynami s. First, the Ramsauer ee t the surprisingly large mean free path of low-velo ity ele trons in gases whi h
Elsasser thought ould be explained by ele tron interferen e. Se ond, the
intensity maxima observed by Davisson and Kunsman at ertain angles
of ree tion of ele trons from metal surfa es, whi h had been assumed
to be aused by atomi shell stru ture, and whi h Elsasser thought were
a Quoted in Mehra and Re henberg (1982a, p. 604).
58
aused by ele tron dira tion. Elsasser published a short note sket hing
these ideas in Die Naturwissens haften (Elsasser 1925). Elsasser then
tried to design an experiment to test the ideas further, with low-velo ity
ele trons, but never arried it out. A ording to Heisenberg's later re olle tion, Elsasser's supervisor Born was s epti al about the reality of
matter waves, be ause they seemed in oni t with the observed parti le
tra ks in loud hambers.a
On 3 November 1925, S hrdinger wrote to Einstein: À few days
ago I read with the greatest interest the ingenious thesis of Louis de
Broglie .... '.a S hrdinger too had be ome interested in de Broglie's
thesis by reading Einstein's gas theory papers, and he set about trying
to nd the wave equation for de Broglie's phase waves. As we have seen
(se tion 2.2.2), in his thesis de Broglie had shown that, in an ele trostati
potential , the phase wave of an ele tron of harge e and velo ity v
would have (see equation (19)) frequen y = p
(mc2 + V )/h and phase
2
velo ity vph = (mc + V )/mv , where m = m0 / 1 v 2 /c2 and V = e
is the potential energy. These expressions for and vph , given by de
Broglie, formed the starting point for S hrdinger's work on the wave
equation.
S hrdinger took de Broglie's formulas for and vph and applied them
to the hydrogen atom, with a Coulomb eld = e/r.b Using the
formula for to eliminate v , S hrdinger rewrote the expression for the
phase velo ity vph purely in terms of the frequen y and the ele tronproton distan e r:
vph =
h
q
m0 c
1
2
(29)
(h/m0 c2 V /m0 c2 ) 1
(where V = e2 /r). Then, writing de Broglie's phase wave as =

(x, t), he took the equation for to be the usual wave equation
2 =
1 2
2 t2
vph
(30)
a For this and further details on erning Elsasser, see Mehra and Re henberg (1982a,
pp. 6247). See also the dis ussion in se tion 3.4.2.
a Quoted in Mehra and Re henberg (1987, p. 412).
b Here we follow the analysis by Mehra and Re henberg (1987, pp. 4235) of
what they all S hrdinger's èarliest preserved [unpublished manus ript on wave
me hani s'. Similar reasoning is found in a letter from Pauli to Jordan of 12 April
1926 (Pauli 1979, pp. 31520).
59
with phase velo ity vph . Assuming to have a time dependen e

e2it , S hrdinger then obtained the time-independent equation
2 =
4 2 2
2
vph
(31)
with vph given by (29). This was S hrdinger's original (relativisti )

equation for the energy states of the hydrogen atom.
As is well known, S hrdinger found that the energy levels predi ted
by (31) that is, the eigenvalues h disagreed with experiment.
He then adopted a nonrelativisti approximation, and found that this
yielded the orre t energy levels for the low-energy limit.
It is in fa t straightforward to obtain the orre t nonrelativisti limit.
Writing E = h m0 c2 , in the low-energy limit |E| /m0 c2 << 1 and
|V | /m0 c2 << 1, we have
2m0
2
2 = h2 (E V ) ,
vph
(32)
so that (31) redu es to what we now know as the nonrelativisti timeindependent S hrdinger equation for a single parti le in a potential V :
2 2
+ V = E .
2m0
(33)
Histori ally speaking, then, S hrdinger's original equation (31) for

stationary states amounted to a mathemati al trans ription into the
language of the standard wave equation (30) of de Broglie's expressions for the frequen y and phase velo ity of an ele tron wave. By studying
the eigenvalue problem of this equation in its nonrelativisti form (33),
S hrdinger was able to show that the eigenvalues agreed remarkably
well with the observed features of the hydrogen spe trum. Thus, by
adopting the formalism of a wave equation, S hrdinger transformed de
Broglie's elementary derivation of the quantisation of energy levels into
a rigorous and powerful te hnique.
The time-dependent S hrdinger equation
i
2 2
+V ,
=
t
2m0
(34)
with a single time derivative, was eventually obtained by S hrdinger

in his fourth paper on wave me hani s, ompleted in June 1926 (S hrdinger 1926g). The path taken by S hrdinger in these four published
papers was rather tortuous. In the fourth paper, he a tually began by
onsidering a wave equation that was of se ond order in time and of
60
fourth order in the spatial derivatives. However, he eventually settled

on (34), deriving it by the following argument: for a wave with time
dependen e e(i/)Et , one may write the term E in (33) as i/t
and thus obtain (34), whi h must be valid for any E and therefore for
any that an be expanded as a Fourier time series.a The pri e paid
for having an equation that was of only rst order in time was that
had to be omplex.
In retrospe t, the time-dependent S hrdinger equation for a free
parti le ((34) with V = 0) may be immediately derived as the simplest
wave equation obeyed by a omplex plane wave ei(kxt) with the
nonrelativisti dispersion relation = (k)2 /2m (that is, E = p2 /2m
ombined with de Broglie's relations E = and p = k ). This derivation is, in fa t, often found in textbooks.
Not only did S hrdinger adopt de Broglie's idea that quantised energy
levels ould be explained in terms of waves, he also took up de Broglie's
onvi tion that lassi al me hani s was merely the short-wavelength
limit of a broader theory. Thus in his se ond paper on wave me hani s
S hrdinger (1926 , p. 497) wrote:
Maybe our lassi al me hani s is the
omplete
analogue of geometri al opti s
and as su h is wrong, not in agreement with reality; it fails as soon as the radii
of urvature and the dimensions of the traje tory are no longer large ompared
to a ertain wavelength, whi h has a real meaning in the q -spa e. Then one has
a
to look for an ùndulatory me hani s' and the most natural path towards
this is surely the wave-theoreti al elaboration of the Hamiltonian pi ture.
S hrdinger used the opti al-me hani al analogy as a guide in the onstru tion of wave equations for atomi systems, parti ularly in his se ond
paper. S hrdinger tended to think of the opti al-me hani al analogy in
terms of the Hamilton-Ja obi equation and the equation of geometri al
opti s, whereas de Broglie tended to think of it in terms of the prin iples
of Maupertuis and Fermat. These are of ourse two dierent ways of
drawing the same analogy.
While S hrdinger took up and developed many of de Broglie's ideas,
he did not a ept de Broglie's view that the parti le was lo alised within
an extended wave. In ee t, S hrdinger removed the parti le traje a A tually, S hrdinger onsidered the time dependen e e(i/)Et , so that E =
i/t, leading to two possible wave equations diering by the sign of i. He
wrote: `We shall require that the omplex wave fun tion satisfy one of these two
equations. Sin e at the same time the omplex onjugate fun tion satises the
other equation, one may onsider the real part of as a real wave fun tion (if one
needs it)' (S hrdinger 1926g, p. 112, original itali s).
a Here S hrdinger adds a footnote: `Cf. also A. Einstein, Berl. Ber., pp. 9 . (1925)'.
61
tories from de Broglie's theory, and worked only with the extended
(non-singular) waves. (For a detailed dis ussion of S hrdinger's work,
see hapter 4.)
In the period 192526, then, many of the ideas in de Broglie's thesis
were taken up and developed by other workers, espe ially S hrdinger.
But what of de Broglie himself? Unlike in his earlier work, during this
period de Broglie onsidered spe i equations for his waves. Like S hrdinger, he took standard relativisti wave equations as his starting point.
Unlike S hrdinger, however, de Broglie was guided by the following two
ideas. First, that parti les are really small singular regions, of very large
wave amplitude, within an extended wave. Se ond, that the motion of the
parti les or singularities must in some sense satisfy the ondition
(35)
Maupertuis Fermat ,
so as to bring about a synthesis of the dynami s of parti les with the

theory of waves, along the lines already sket hed in de Broglie's thesis.
This work ame to a head in a remarkable paper published in May 1927,
to whi h we now turn.
2.3.1 `Stru ture': Journal de Physique, May 1927

In the last number of the Journal de Physique, a paper
by de Broglie has appeared .... de Broglie attempts here
to re on ile the full determinism of physi al pro esses
with the dualism between waves and orpus les .... even if
this paper by de Broglie is o the mark (and I hope that
a tually), still it is very ri h in ideas and very sharp, and
on a mu h higher level than the hildish papers by
S hrdinger, who even today still thinks he may ....
abolish material points.
Pauli, letter to Bohr, 6 August 1927 (Pauli 1979,
pp. 4045)
What we now know as pilot-wave theory rst appears in a paper by de

Broglie (1927b) entitled `Wave me hani s and the atomi stru ture of
matter and of radiation', whi h was published in Journal de Physique
in May 1927, and whi h we shall dis uss in detail in this se tion.a We
shall refer to this ru ial paper as `Stru ture' for short.
In `Stru ture', de Broglie presents a theory of parti les as moving
a Again, we do not always follow de Broglie's notation. Note also that de Broglie
moves ba k and forth between solutions of the form f cos and solutions `written
in omplex form' f ei .
62
singularities. It is argued, on the basis of ertain assumptions, that the

equations of what we now all pilot-wave dynami s will emerge from this
theory. At the end of the paper de Broglie proposes, as a provisional
theory, simply taking the equations of pilot-wave dynami s as given,
without trying to derive them from something deeper. It is this last,
provisional theory that de Broglie presents a few months later at the
fth Solvay onferen e.
For histori al ompleteness we should point out that, a ording to a
footnote to the introdu tion, `Stru ture' was `the development of two
notes' published earlier in Comptes Rendus (de Broglie 1926, 1927a).
The rst, of 28 August 1926, onsiders a model of photons as moving
singularities: arguments are given (similar to those found in `Stru ture')
leading to the usual velo ity formula, and to the on lusion that the
probability density of photons is proportional to the lassi al wave intensity. The se ond note, of 31 January 1927, sket hes analogous ideas
for material bodies in an external potential, with the probability density
now proportional to the amplitude-squared of the wave fun tion. Unlike
`Stru ture', neither of these notes ontains the suggestion that pilot-wave
theory may be adopted as a provisional view.
We now turn to a detailed analysis of de Broglie's `Stru ture' paper.
De Broglie rst onsiders a `material point of proper mass m0 ' moving in
free spa e with a onstant velo ity v and represented by a wave u(x, t)
satisfying what we would now all the Klein-Gordon equation
2 u
1 2u
4 2 02
=
u,
2
2
c t
c2
(36)
with 0 = m0 c2 /h. De Broglie onsiders solutions of the form

2
(x, t) ,
(37)
h
where the amplitude f is singular at the lo ation x = vt of the moving
body and the phasea

h0
v x
(x, t) = p
t 2
(38)
c
1 v 2 /c2
u(x, t) = f (x vt) cos
is equal to the lassi al Hamiltonian a tion of the parti le. A single

parti le is represented by a wave (37), where the moving singularity has
velo ity v and the phase (x, t) is extended over all spa e.
De Broglie then onsiders an ensemble (or ` loud') of similar free
a For simpli ity we ignore an arbitrary onstant in the phase, whi h was in luded
by de Broglie.
63
parti les with no mutual intera tion, all having the same velo ity v,
and with singular amplitudes f entred at dierent points in spa e.
A ording to de Broglie, this ensemble of moving singularities may be
represented by a ontinuous solution of the same wave equation (36), of
the form
2
(x, t) = a cos (x, t) ,
(39)
h
where a is a onstant and is given by (38). This ontinuous solution
is said to ` orrespond to' the singular solution (37). (De Broglie points
out that the ontinuous solutions are the same as those onsidered by
S hrdinger.)
The number density of parti les in the ensemble is taken to have a
onstant value , whi h may be written as
= Ka2
(40)
for some onstant K . Thus, de Broglie notes, (39) gives the `distribution
of phases in the loud of points' as well as the `density of the loud'; and
for a single parti le with known velo ity and unknown position, a2 d3 x
will measure the probability for the parti le to be in a volume element
d3 x.
Having dis ussed the free parti le, de Broglie moves on to the ase of
a single parti le in a stati external potential V (x). The wave u `written
in omplex form' now satises what we would all the Klein-Gordon
equation in an external potentiala

V2
2i u
1
1 2u
2 2
2
2 m0 c 2 u = 0 .
u 2 2 + 2V
(41)
c t
c
t
De Broglie assumes that the parti le begins in free spa e, represented

by a singular wave (37), and then enters a region where V 6= 0, into
whi h the free solution (37) must be extended. De Broglie writes (37) ìn
omplex form', substitutes into (41), and takes the real and imaginary
parts, yielding two oupled partial dierential equations for f and .
Writing
(x, t) = Et 1 (x)
(42)
(where E = h is onstant), one of the said partial dierential equations

be omes
f
1
2
2
(43)
= (1 ) 2 (E V )2 + m20 c2
f
c
a This follows from (36) by the substitution i/t i/t + V .
64
(where 2 (1/c2 ) 2 /t2 ). De Broglie notes that if the left-hand

side is negligible, (43) redu es to the `Ja obi equation' for relativisti
dynami s in a stati potential, so that 1 redu es to the `Ja obi fun tion'.
Deviations from lassi al me hani s o ur, de Broglie notes, when f is
non-zero.
De Broglie now remarks that, in the lassi al limit, the velo ity of
the parti le has the same dire tion as the ve tor 1 ; he then expli itly
assumes that this is still true in the general ase. Here, the identity (35)
of the prin iples of Maupertuis and Fermat is being invoked. With this
assumption, de Broglie shows that the singularity must have velo ity
v=
c2 1
,
EV
(44)
and he adds that in the nonrelativisti approximation, E V m0 c2 ,

so that
1
1 .
v=
(45)
m0
De Broglie remarks that, in the lassi al approximation, 1 be omes
the `Ja obi fun tion' and that (44) agrees with the relativisti relation
between velo ity and momentum. He adds (p. 230):
The aim of the pre eding arguments is to make it plausible that the relation
[(44) is stri tly valid in the new me hani s.
The importan e of the relation (44) for de Broglie is, of ourse, that it
embodies the identity of the prin iples of Maupertuis and Fermat.
As in the free ase, de Broglie then goes on to onsider an ensemble
of su h parti les, with ea h parti le represented by a moving singularity.
He assumes that the phase fun tion 1 is the same for all the parti les,
whose velo ities are then given by (44). The (time-independent) parti le
density (x) must then satisfy the ontinuity equation
(v) = 0 .
(46)
On e again, de Broglie introdu es a representation of the ensemble by

a ontinuous solution of (41). In the presen e of the potential V , the
solution is taken to have the form

1
2
(x, t) = a(x) cos 2 t 1 (x) .
(x, t) = a(x) cos
(47)
h
h
Again, writing ìn omplex form', substituting into (41) and taking real
65
and imaginary parts, de Broglie obtains two oupled partial dierential

equations, now in a and 1 . One of these reads
2 a
1
2
(48)
= (1 ) 2 (E V )2 + m20 c2 .
a
c
In this ase de Broglie notes that if the left-hand side is negligible, (48)
redu es to the equation of geometri al opti s asso iated with the wave
equation (41).
Comparing (43) and (48), de Broglie notes that when the left-hand
sides are negligible, one re overs both lassi al me hani s (for a moving
singularity) and geometri al opti s (for a ontinuous wave). In this limit,
the fun tions 1 and 1 are identi al, being both equal to the `Ja obi
fun tion'.
At this point de Broglie makes a ru ial assumption. He proposes to
assume, as a hypothesis, that 1 and 1 are always equal, regardless of
any approximation (p. 2312):
2
1 and 1 are still identi al when

[the left-hand sides of (43) and (48) an no longer be negle ted. Obviously,
We now make the essential hypothesis that

this requires that one have:
f
2 a
=
.
a
f
We shall refer to this postulate by the name of `prin iple of the double
solution', be ause it implies the existen e of two sinusoidal solutions of equation [(41) having the same phase fa tor, the one onsisting of a point-like
singularity and the other having, on the ontrary, a ontinuous amplitude.
This hypothesis expresses, in a more on rete form, de Broglie's earlier

idea of `phase harmony' between the internal os illation of a parti le and
the os illation of an a ompanying extended wave: the ondition that 1
and 1 should oin ide amounts to a phase harmony between the singular
u-wave representing the point-like parti le and the ontinuous -wave.
Given this ondition, de Broglie dedu es that the ratio /a2 (E V ) is
onstant along the parti le traje tories. Sin e ea h parti le is assumed
to begin in free spa e, where V = 0 and (a ording to (40)) = Ka2 ,
de Broglie dedu es that in general

V (x)
(x) = Ka2 (x) 1
(49)
,
E
where K is a onstant. The ontinuous wave then gives the ensemble
density at ea h point. In the nonrelativisti approximation, this be omes
(x) = Ka2 (x) .
(50)
66
De Broglie notes that ea h possible initial position for the parti le gives
rise to a possible traje tory, and that an ensemble of initial positions
gives rise to an ensemble of motions. Again, de Broglie takes (x)d3 x
as the probability for a single parti le to be in a volume element d3 x.
This probability is, as he notes, given in terms of by (49) or (50). De
Broglie adds that also determines the traje tories (p. 232):
The form of the traje tories is moreover equally determined by knowledge of
the ontinuous wave, sin e these traje tories are orthogonal to the surfa es of
equal phase.
Here we see the rst suggestion that itself may be regarded as determining the traje tories, an idea that de Broglie proposes more fully at
the end of the paper.
After sket hing how the above ould be used to al ulate probabilities
for ele tron s attering o a xed potential, de Broglie goes on to generalise his results to the ase of a parti le of harge e in a time-dependent
ele tromagneti eld (V(x, t), A(x, t)). He writes down the orresponding Klein-Gordon equation, and on e again onsiders solutions of the
form (37) with a moving singularity, following the same pro edure as
before. Be ause of the time-dependen e of the potentials, no longer
takes the form (42). Instead of (44), the velo ity of the singularity is
now found to be
+ ec A
v = c2
(51)
,
eV
or, in the nonrelativisti approximation,
1
e
v=
+ A .
m0
c
(52)
On e again, de Broglie then onsiders an ensemble of su h moving singularities, with the same phase fun tion (x, t). The density (x, t) now
obeys a ontinuity equation
(53)
+ (v) = 0 ,
t
with velo ity eld v given by (51). Again, de Broglie pro eeds to represent the motion of the ensemble by means of a ontinuous solution of
the wave equation, this time of the form
2
(x, t) ,
(54)
h
with a time-dependent amplitude and a phase no longer linear in t. And
again, de Broglie assumes the prin iple of the double solution, that the
(x, t) = a(x, t) cos
67
phase fun tions (x, t) and (x, t) are the same. From this, de Broglie
is onstant along parti le traje tories.

dedu es that the ratio /a2 (eV)
Using on e more the expression = Ka2 for free spa e, where V = 0
and = E0 , de Broglie argues that in general

K 2
(x, t) =
(55)
a (x, t) eV = K a2 eV ,
E0
and that in the nonrelativisti limit is still proportional to a2 . As
before, the ensemble may be regarded as omposed of all the possible
positions of a single parti le, of whi h only the initial velo ity is known,
and the probability that the parti le is in the volume element d3 x at
time t is equal to (x, t)d3 x and is given in terms of by (55).
De Broglie now shows how the above results for the motion of a parti le
in a potential V = eV (ignoring for simpli ity the ve tor potential A)
may be obtained from lassi al me hani s, with a Lagrangian of the
standard form
p
L = M0 c2 1 v 2 /c2 V ,
(56)
by assuming that the parti le has a variable proper mass

r
2 a
,
M0 (x, t) = m20 2
(57)
c a
where a is the amplitude of the ontinuous wave. Further, he onsiders
this point of view in the nonrelativisti limit. Writing M0 (x, t) = m0 +
(x, t) with small, the Lagrangian takes the approximate form
1
L = m0 c2 + m0 v 2 c2 V .
2
(58)
As de Broglie remarks (p. 237):

Everything then takes pla e as if there existed, in addition to
2
energy term c .
V,
a potential
This extra term oin ides, of ourse, with Bohm's quantum potential Q
(equation (4)), sin e to lowest order
c2 =
2 2 a
2 a
2 a
=
+
2m0 a
2m0 a
2m0 c2 a
(59)
and the se ond term is negligible in the nonrelativisti limit (c ).

As we have already repeatedly remarked, for de Broglie the guidan e
equation v for velo ity is the fundamental equation of motion,
expressing as it does the identity of the prin iples of Maupertuis and
Fermat. Even so, de Broglie points out that one an write the dynami s
68
of the parti le in lassi al (Newtonian or Einsteinian) terms, provided

one in ludes a variable proper mass or, in the nonrelativisti limit, an
additional quantum potential. It is this latter, Newtonian formulation
that Bohm proposes in 1952.
De Broglie now turns to the ase of a many-body system, onsisting
of N parti les with proper masses m1 , m2 , ..., mN . In the nonrelativisti
approximation, if the system has total (Newtonian) energy E and potential energy V (x1 , ..., xN ), then following S hrdinger one may onsider
the propagation of a wave u in the 3N -dimensional onguration spa e,
satisfying the wave equation
N
X
i=1
2 2
u + V u = Eu
2mi i
(60)
(the time-independent S hrdinger equation). De Broglie remarks that

this seems natural, be ause (60) is the obvious generalisation of the
one-body S hrdinger equation, whi h follows from the nonrelativisti
limit of de Broglie's wave equation (41). However, de Broglie riti ises
S hrdinger's interpretation, a ording to whi h a parti le is identied
with an extended, non-singular wave pa ket, having no pre ise position or traje tory. De Broglie obje ts that, without well-dened parti le
positions, the oordinates x1 , ..., xN used to onstru t the onguration
spa e would have no meaning. Further, de Broglie asserts that onguration spa e is `purely abstra t', and that a wave propagating in this spa e
annot be a physi al wave: instead, the physi al pi ture of the system
must involve N waves propagating in 3-spa e.
`What then', asks de Broglie (p. 238), ìs the true meaning of the
S hrdinger equation?' To answer this question, de Broglie onsiders,
for simpli ity, the ase of two parti les, whi h for him are singularities of
a wave-like phenomenon in 3-spa e. Negle ting the ve tor potential, the
two singular waves u1 (x, t) and u2 (x, t) satisfy the oupled equations

V2
u1
1
2i
2 m21 c2 12 u1 = 0 ,
u1 + 2 V1
c
t
c
(61)

V2
u2
1
2i
2 m22 c2 22 u2 = 0 ,
u2 + 2 V2
c
t
c
with potentials
V1 = V (|x x2 |)
, V2 = V (|x x1 |) ,
(62)
where x1 , x2 are the positions of the two parti les (or singularities). The
69
propagation of ea h wave then depends, through V , on the position of

the singularity in the other wave.
De Broglie is now fa ed with the formidable problem of solving the
simultaneous partial dierential equations (61), in order to obtain the
motions of the two singularities. Not surprisingly, de Broglie does not
attempt to arry through su h a solution. Instead he notes that, in
lassi al me hani s in the nonrelativisti limit, it is possible to nd a
`Ja obi fun tion' (x1 , x2 ) for the two-parti le system, su h that the
parti le momenta are given by
m1 v1 = 1 ,
m2 v2 = 2 .
(63)
De Broglie then asks (p. 238): `Can the new me hani s .... dene su h a
fun tion ?'
De Broglie is asking whether, if one ould solve the oupled equations
(61), the resulting motions of the singularities would (in the nonrelativisti limit) satisfy the guidan e equations (63) for some fun tion
(x1 , x2 ). He then asserts, on the basis of an in orre t argument (to
whi h we shall return in a moment), that this will indeed be the ase.
Having on luded that a fun tion (x1 , x2 ) generating the motions of
the singularities will in fa t exist, he goes on to identify (x1 , x2 ) with
the phase of a ontinuous solution (x1 , x2 , t) of the S hrdinger equation (60) for the two-parti le system. While de Broglie does not say so
expli itly, in ee t he assumes that there exists a ontinuous solution of
(60) whose phase oin ides with the fun tion (x1 , x2 ) that generates the
motions of the singularities via (63). This seems to be de Broglie's answer
to the question he raises, as to the `true meaning' of the S hrdinger
equation: solving the S hrdinger equation (60) in onguration spa e
provides an ee tive means of obtaining the motions of the singularities
without having to solve the oupled partial dierential equations (61) in
3-spa e.
De Broglie's in orre t argument for the existen e of an appropriate
fun tion , derived from the equations (61), pro eeds as follows. He rst
imagines that the motion of singularity 2 is already known, so that the
problem of motion for singularity 1 redu es to a ase that has already
been dis ussed, of a single parti le in a given time-dependent potential.
For this ase, it has already been shown that the ensemble of possible
traje tories may be represented by a ontinuous wave
a1 (x, t)e(i/)1 (x,t) .
(64)
Further, as was shown earlier, the equations of motion for the par-
70
ti le may be written in Lagrangian form with an additional potential

1 (x1 , t)c2 . Similarly, onsidering the motion of singularity 1 as known,
the motion of singularity 2 an be des ribed by a Lagrangian with an
additional potential 2 (x2 , t)c2 . The two sets of ( lassi al) equations of
motion an be derived from a single Lagrangian, de Broglie argues, only
if 1 and 2 redu e to a single fun tion (|x1 x2 |) of the interparti le
separation. Ìf we assume this', says de Broglie, the total Lagrangian for
the system will be
L=
1
1
m1 v12 + m2 v22 V (|x1 x2 |) (|x1 x2 |)c2 ,
2
2
(65)
and one will be able to dedu e, in the usual fashion, the existen e of a
fun tion (x1 , x2 ) satisfying the equations (63). It is not entirely lear
from the text if de Broglie is really onvin ed that 1 and 2 will indeed
redu e to a single fun tion (|x1 x2 |). With hindsight one sees, in fa t,
that this will not usually be the ase. For it amounts to requiring that
the quantum potential Q(x1 , x2 ) for two parti les should take the form
Q(x1 , x2 ) = Q(|x1 x2 |), whi h is generally false.a
After giving his in orre t argument for the existen e of , de Broglie points out that, sin e the parti les have denite traje tories, it is
meaningful to onsider their six-dimensional onguration spa e, and to
represent the two parti les by a single point in this spa e, with velo ity
omponents given by (63). If the initial velo ity omponents are given,
but the initial positions of the two parti les are not, one may onsider
an ensemble of representative points (x1 , x2 ), whose density (x1 , x2 ) in
onguration spa e satises the ontinuity equation,
(v) = 0 ,
(66)
where v is the (six-dimensional) velo ity eld given by (63).

De Broglie then onsiders, as we have said, a ontinuous solution of
the S hrdinger equation (60), of the form
(x1 , x2 , t) = A(x1 , x2 )e(i/) ,
(67)
with the ta it assumption that the phase is the same fun tion whose
existen e has been (apparently) established by the above (in orre t)
argument. He then shows, from (60), that A2 satises the ongurationspa e ontinuity equation (66). De Broglie on ludes that A2 d3 x1 d3 x2
a One might also question the meaning of de Broglie's wave fun tion (64) for parti le
1, in the ontext of a system of two intera ting parti les, whi h must have an
entangled wave fun tion (x1 , x2 , t).
71
is the probability for the representative point (x1 , x2 ) to be present in a

volume element d3 x1 d3 x2 of onguration spa e.
Here, then, de Broglie has arrived at a higher-dimensional analogue of
his previous results: the wave fun tion in onguration spa e determines the ensemble probability density through its amplitude, as well as
the motion of a single system through its phase.
De Broglie ends this se tion by remarking that it seems di ult to nd
an equation playing a role similar to (60) outside of the nonrelativisti
approximation, and that Fermi's al ulation on erning the s attering
of ele trons by a rotator may be regarded as illustrating the above.
From the point of view of the history of pilot-wave theory, the most
important se tion of de Broglie's paper now follows. Entitled `The pilot
wave' (`L'onde pilote'), it begins by re alling the results obtained in the
ase of a single parti le in a time-dependent potential, whi h de Broglie
states may be summarised by `the two fundamental formulas' (51) and
(55) for the parti le velo ity and probability density respe tively. De
Broglie notes that he arrived at the velo ity formula (51) by invoking
the prin iple of the double solution a prin iple that is valid in free
spa e but whi h `remains a hypothesis in the general ase' (p. 241). At
this point, de Broglie makes a remarkable suggestion: that instead of
trying to derive the velo ity eld (51) from an underlying theory, one
ould simply take it as a postulate, and regard as a physi ally-real
`pilot wave' guiding the motion of the parti le. To quote de Broglie
(p. 241):
But if one does not wish to invoke the prin iple of the double solution,
it is a eptable to adopt the following point of view: one will assume the
existen e, as distin t realities, of the material point and of the ontinuous
wave represented by the fun tion
and one will take it as a postulate that
the motion of the point is determined as a fun tion of the phase of the wave
by the equation [(51). One then on eives the ontinuous wave as guiding the
motion of the parti le. It is a pilot wave.
This is the rst appearan e in the literature of what we now know as

pilot-wave or de Broglian dynami s (albeit stated expli itly for only a
single parti le). The pilot wave satisfying the S hrdinger equation,
and the material point, are regarded as `distin t realities', with the
former guiding the motion of the latter a ording to the velo ity law
(51).
De Broglie made it lear, however, that he thought su h a dynami s
ould be only provisional (p. 241):
72
By thus taking [(51) as a postulate, one avoids having to justify it by the

prin iple of the double solution; but this an only be, I believe, a provisional
attitude. No doubt one will indeed have to
rein orporate
the parti le into the
wave-like phenomenon, and one will probably be led ba k to ideas analogous

to those that have been developed above.
As we shall dis uss in the next se tion, de Broglie's proposal of pilotwave theory as a provisional measure has striking analogues in the early
history of Newtonian gravity and of Maxwellian ele tromagneti theory.
De Broglie goes on to point out two important appli ations of the
formulas (51) and (55). First, applying the theory to light, a ording
to (55) `the density of the photons is proportional to the square of
the amplitude' of the guiding wave , yielding agreement with the
predi tions of wave opti s. Se ond, an ensemble of hydrogen atoms with
a denite state will have a mean ele troni harge density proportional
2
to || , as assumed by S hrdinger.
De Broglie's `Stru ture' paper ends with some remarks on the pressure
exerted by parti les on a wall (for example of a box of gas). He notes
that a ording to (51), be ause of the interferen e between the in ident
and ree ted waves, the parti les will not a tually strike the wall, raising
the question of how the pressure is produ ed. De Broglie laims that the
pressure omes from stresses in the interferen e zone, as appear in an
expression obtained by S hrdinger (1927 ) for the stress-energy tensor
asso iated with the wave .
2.3.2 Signi an e of de Broglie's `Stru ture' paper

The ex eptional quality of de Broglie's `Stru ture' paper in Journal de
Physique was noted by Pauli, in a letter to Bohr dated 6 August 1927,
already quoted as the epigraph to the last subse tion: `.... it is very ri h
in ideas and very sharp, and on a mu h higher level than the hildish
papers by S hrdinger ....' (Pauli 1979, pp. 4045). In the same letter,
Pauli suggested that Bohr would have to refer to de Broglie's paper in
his Como le ture (in whi h Bohr developed the idea of omplementarity
between waves and parti les); and in fa t, in what was to remain an
unpublished version of the Como le ture, Bohr did take an expli it stand
against de Broglie's ideas (Bohr 1985, pp. 8998). Bohr hara terised de
Broglie's work as attempting `to re on ile the two apparently ontradi tory sides of the phenomena by regarding the individual parti les or
light quanta as singularities in the wave eld'; further, Bohr suggested
that de Broglie's view rested upon `the on epts of lassi al physi s' and
73
was therefore not `suited to help us over the fundamental di ulties'
(Bohr 1985, p. 92). On the whole, though, this paper by de Broglie has
been essentially ignored, by both physi ists and historians.
De Broglie's `Stru ture' paper may be summed up as follows. De
Broglie has a model of parti les as singularities of 3-spa e waves ui , in
whi h the motion of the individual parti les, as well as the ensemble distribution of the parti les, are determined by a ontinuous wave fun tion
. (We emphasise that de Broglie has both the wave fun tion and the
singular u-waves.) The result is a rst-order theory of motion, based on
the velo ity law mi vi = i , in whi h the prin iples of Maupertuis and
Fermat are unied. Then, at the end of the paper, de Broglie re ognises
that his singularity model of parti les an be dropped, and that the
results he has obtained the formulas for velo ity and probability
density in terms of an be simply postulated, yielding pilot-wave
theory as a provisional measure.
A few months later, in O tober 1927, de Broglie presented this pilotwave theory at the fth Solvay onferen e, for a nonrelativisti system of
N parti les guided by a wave fun tion in onguration spa e. Before
dis ussing de Broglie's Solvay report, however, it is worth pausing to
onsider the role played by the `Stru ture' paper in de Broglie's thinking.
From a histori al point of view, the signi an e of de Broglie's `Stru ture' paper inevitably depends on the signi an e one as ribes to the
provisional pilot-wave theory arrived at in that paper. Given what we
know today that pilot-wave theory provides a onsistent a ount
of quantum phenomena de Broglie's `Stru ture' paper now seems
onsiderably more signi ant than it has seemed in the past.
From the point of view of pilot-wave theory as we know it today, the
singular u-waves played a similar role for de Broglie as the material
ether did for Maxwell in ele tromagneti theory. In both ases, there
was a on eptual s aolding that was used to build a new theory, and
that ould be dropped on e the results had been arrived at. De Broglie
re ognised at the end of his paper that, if one took pilot-wave dynami s
as a provisional theory, then the s aolding he had used to onstru t
this theory ould be dropped. At the same time, de Broglie insisted that
taking su h a step was indeed only provisional, and that an underlying
theory was still needed, probably along the lines he had been pursuing.
Similar situations have arisen before in the history of s ien e. Abstra ting away the details of a model based on an older theory sometimes
results in a new theory in its own right, involving new on epts, where,
however, the author regards the new theory as only a provisional measu-
74
re, and expe ts that the model based on the older theory will eventually
provide a proper basis for the provisional theory. Thus, for example,
Newton tried to explain gravitation on the basis of a tion by onta t,
involving a material medium lling spa e the same Àethereal Medium', in fa t, as he thought responsible for the interferen e and dira tion
of light (Newton 1730; reprint, pp. 35053). Newton regarded his theory
of gravitation, with a tion at a distan e, as merely a provisional and
phenomenologi al theory, that would later nd an explanation in terms
of ontiguous a tion. Eventually, however, the on ept of `gravitational
a tion at a distan e' be ame widely a epted in its own right (though it
was later to be overthrown, of ourse, by general relativity). Similarly,
Maxwell used me hani al models of an ether to develop his theory of the
ele tromagneti eld. Maxwell himself may or may not have re ognised
that this s aolding ould be dropped (Hendry 1986). But ertainly,
many of his immediate followers did not: they regarded working only
with Maxwell's equations as provisional and phenomenologi al, pending
the development of an underlying me hani al model of an ether. Again,
the on ept of èle tromagneti eld' eventually be ame widely a epted
as a physi al entity in its own right.
In the ase of de Broglie in 1927, in his `Stru ture' paper (and subsequently at the fth Solvay onferen e), he arrived at the new on ept
of `pilot wave in onguration spa e', an entirely new kind of physi al
entity that, a ording to de Broglie's (provisional) theory, guides the
motion of material systems.
2.4 1927 Solvay report: the new dynami s of quanta

We now turn to a (brief) summary of and ommentary on de Broglie's
report at the fth Solvay onferen e.a As we have noted, the theory
presented in this report is pilot-wave theory as we know it today, for a
nonrelativisti many-body system, with a guiding wave in onguration
spa e that determines the parti le velo ities a ording to de Broglie's
basi law of motion. De Broglie's ideas about parti les as singularities of
3-spa e waves are mentioned only briey. The theory de Broglie presents
in Brussels in O tober 1927 is, indeed, just the provisional theory he
proposed a few months earlier at the end of his `Stru ture' paper (though
now expli itly applied to many-body systems as well).
De Broglie begins part I of his report by reviewing the results obtained
a Again, our notation sometimes departs from de Broglie's.
75
in his do toral thesis. The energy E and momentum p of a parti le are

determined by the phase of an asso iated wave:a
E = /t ,
p = .
(68)
(These are the relativisti guidan e equations of de Broglie's early pilotwave theory of 192324.) It follows that, as de Broglie remarks, `the
prin iples of least a tion and of Fermat are identi al' (p. 375). Quantisation onditions appear in a natural way, and there are far-rea hing
impli ations for statisti al me hani s. De Broglie ends the review of his
early work with some general remarks. He points out (p. 376) that he `has
always assumed that the material point o upies a well-dened position
in spa e', and he asserts that as a result the wave amplitude must be
singular, or take very large values in a small region, somewhere within the
extended wave. Signi antly, de Broglie adds: `But, in fa t, the form of
the amplitude plays no role in the results reviewed above'. This seems to
be a rst hint that the a tual results, su h as (68), do not depend on the
details of any underlying model of the parti les as moving singularities.
De Broglie then outlines the work of S hrdinger. He notes that S hrdinger's wave equation is onstru ted in order that the phase of the
wave fun tion
2
= a cos
(69)
h
be a solution of the Hamilton-Ja obi equation in the geometri al-opti s
limit. He points out that S hrdinger identies parti les with lo alised
wave pa kets instead of with a small on entration within an extended
wave, and that for a many-body system S hrdinger has a wave propagating in onguration spa e. For both the one-body and many-body
ases, de Broglie writes down the time-independent S hrdinger equation
only, with a stati potential energy fun tion. As we shall see, de Broglie in
fa t onsiders non-stationary wave fun tions as well. For the one-body
ase, de Broglie also writes down a relativisti , and time-dependent,
equation what we now know as the Klein-Gordon equation in an
external ele tromagneti eld.
De Broglie then raises two on eptual di ulties with S hrdinger's
work (similar in spirit to the obje tions he raises in `Stru ture'): (1)
He questions how one an meaningfully onstru t a onguration spa e
without a real onguration, asserting that it `seems a little paradoxi al
to onstru t a onguration spa e with the oordinates of points that do
a Again, de Broglie's phase has a sign opposite to the phase S as we would normally
dene it now.
76
not exist' (p. 379). (2) He laims that the physi al meaning of the wave
annot be ompared with that of an ordinary wave in 3-spa e, be ause
the number of dimensions of the abstra t spa e on whi h is dened
is determined by the number of degrees of freedom of the system. (This
se ond point seems indeed a very ee tive way to make lear that is
quite dierent from a onventional wave or eld on 3-spa e.)
Part I of the report ends with some remarks on Born's statisti al
interpretation, whi h de Broglie asserts is analogous to his own.
In part II of his report, de Broglie presents his own interpretation
of the wave fun tion. Pilot-wave theory is learly formulated, rst for a
single parti le, and then for a system of N parti les. Several appli ations
are outlined. It is interesting to see how de Broglie motivates his theory
and ompares it with the ontenders.
De Broglie begins by asking what the relationship is between parti les
and the wave . He rst onsiders a single relativisti parti le in an
external ele tromagneti eld with potentials (V , A). De Broglie notes
that, in the lassi al limit, the phase of obeys the Hamilton-Ja obi
equation, and the velo ity of the parti le is given by
v = c2
+ ec A
eV
(70)
(the same formula (51) dis ussed in `Stru ture'). De Broglie then proposes that this velo ity formula is valid even outside the lassi al limit
(p. 382, itali s in the original):
We propose to assume by indu tion that this formula is still valid when the old
Me hani s is no longer su ient, that is to say when [ is no longer a solution
of the Ja obi equation. If one a epts this hypothesis, whi h appears justied
by its onsequen es, the formula [(70) ompletely determines the motion of
the orpus le
as soon as one is given its position at an initial instant.
In
other words, the fun tion [, just like the Ja obi fun tion of whi h it is the
generalisation, determines a whole lass of motions, and to know whi h of
these motions is a tually des ribed it su es to know the initial position.
We emphasise that there is no appeal to singular u-waves anywhere

in de Broglie's report. No use is made of the prin iple of the double
solution. As he had suggested at the end of his `Stru ture' paper a
few months earlier, de Broglie simply postulates the basi equations
of pilot-wave dynami s, without trying to derive them from anything
else. To motivate the guidan e equation, de Broglie simply generalises
the lassi al Hamilton-Ja obi velo ity formula to the non- lassi al domain: he assumes `by indu tion' that the formula holds even outside
77
the lassi al limit. And de Broglie is quite expli it that he is proposing a

rst-order theory of motion, based on velo ities: given the wave fun tion,
the initial position alone determines the traje tory.
So far, then, for a single parti le de Broglie has a (relativisti ) wave
fun tion = a cos 2
h whose phase determines the parti le velo ity
by the guidan e equation (70). De Broglie now goes on to point out
that, for an ensemble of parti les guided by the velo ity eld (70), the
distribution

Ka2 eV
(71)
is preserved in time (that is, equivariant). He on ludes that, if the initial

position of the parti le is ignored, then the probability for the parti le
to be present (at time t) in a spatial volume d is

d = Ka2 eV d .
(72)
(This is the same probability formula (55) arrived at in `Stru ture'.)

De Broglie's expression `probabilit de prsen e' makes it lear that
we have to do with a probability for the ele tron being somewhere, and
not merely with a probability for an experimenter nding the ele tron
somewhere f. Bell (1990, p. 29). De Broglie simply assumes that the
equivariant distribution is the orre t probability measure for a parti le
of unknown position. In fa t, this distribution is only an equilibrium
distribution, analogous to thermal equilibrium in lassi al statisti al
me hani s (see se tion 8.2).
De Broglie sums up his results so far (p. 382):
In brief, in our hypotheses, ea h wave
determines a ` lass of motions', and
ea h one of these motions is governed by equation [(70) when one knows the
initial position of the orpus le. If one ignores this initial position, the formula
[(72) gives the probability for the presen e of the orpus le in the element of
volume
at the instant
t.
then appears
probability wave.
The wave
(Fhrungsfeld of Mr Born) and a
as both a
pilot wave
There are, de Broglie adds, no grounds for abandoning determinism, and

in this his theory diers from that of Born.
De Broglie adds that, in the nonrelativisti approximation (where
eV m0 c2 ), the guidan e and probability formulas (70) and (72) redu e
to
e
1
+ A
v=
(73)
m0
c
78
and
= const a2 .
(74)
These are the standard pilot-wave equations for a single parti le (in an
external ele tromagneti eld).
De Broglie then remarks on how the above formulas may be applied
to the s attering of a single parti le by a xed potential. The ensemble
of in ident parti les may be represented by a plane wave with a uniform
probability distribution. Upon entering a region of non-zero eld, the
behaviour of the wave fun tion may be al ulated using Born's perturbation theory (for example, for the Rutherford s attering of an ele tron by
an atomi nu leus). De Broglie draws an analogy between the s attering
of the wave fun tion and the lassi al s attering of light by a refra ting
medium.
As in his `Stru ture' paper, de Broglie remarks in passing that, for a
relativisti parti le governed by the guidan e equation (70), one may
write down the equations of lassi al dynami s with a variable rest
mass M0 given by equation (57). As we have already noted, in the
nonrelativisti approximation this yields an additional potential energy
term, whi h is pre isely Bohm's `quantum potential'. And, on e again,
we emphasise that for de Broglie the equation v for velo ity is the
fundamental equation of motion, expressing the identity of the prin iples
of Maupertuis and Fermat: de Broglie merely points out that, if one
wishes, the dynami s an be written in lassi al terms (as Bohm did in
1952).
De Broglie then turns to the interpretation of interferen e and diffra tion, for the ase of photons. Here, the guiding wave is similar to
but not the same as a light wave.a De Broglie onsiders s attering by
xed obsta les, in whi h ase the guiding wave may be taken to have a
onstant frequen y . The relativisti equations (70), (72) then be ome
c2
, = const a2 .
(75)
h
De Broglie points out that the se ond equation predi ts the well-known
interferen e and dira tion patterns. He argues that the results will be
the same, whether the experiment is done with an intense beam over a
short time or with a feeble beam over a long time.
De Broglie remarks, as he did at the end of `Stru ture', on the stresses
v=
a De Broglie did not identify the photoni pilot wave with the ele tromagneti eld.
In the general dis ussion, on p. 508, de Broglie expli itly states that in his theory
the wave for the ase of photons is distin t from the ele tromagneti eld.
79
appearing in S hrdinger's expression for the stress-energy tensor of the

wave . A ording to de Broglie, these stresses provide an explanation
for the pressure exerted by light ree ting on a mirror despite the fa t
that, a ording to (70), the photons never a tually strike the surfa e of
the mirror (as shown expli itly by Brillouin in the dis ussion after de
Broglie's report).
De Broglie then turns to the generalisation of the above dynami s
to a (nonrelativisti ) many-body system. Remarkably, many historians
and ommentators have not noti ed this proposal by de Broglie of a
many-body dynami s in onguration spa e, a proposal that is usually
attributed to Bohm ( f. se tion 11.1.)
De Broglie begins by pointing out how the two di ulties he has
raised against S hrdinger's wave me hani s might be solved. First, if a
real onguration exists at all times, one an meaningfully onstru t the
onguration spa e. As for the se ond di ulty, regarding the meaning
of a wave in the abstra t onguration spa e, de Broglie makes the
following preliminary remark (p. 385):
It appears to us ertain that if one wants to
of a system of
in spa e, ea h of the
N 1
physi ally
represent the evolution
orpus les, one must onsider the propagation of
waves
propagations being determined by the a tion of the
orpus les onne ted to the other waves.
This seems a lear referen e to the theory of intera ting singular u-waves
dis ussed a few months earlier in de Broglie's `Stru ture' paper.a (The
propagation of ea h wave must depend on the motion of the other N 1
parti les, of ourse, in order to a ount for intera tions between the
parti les.) As we saw above (se tion 2.3.1), in `Stru ture' de Broglie
expli itly onsidered the ase of two parti les, whi h were represented
by two singular waves u1 (x, t) and u2 (x, t) satisfying a pair of oupled
partial dierential equations (61), where the equation for ea h wave
ontained a potential depending on the position of the singularity in the
other wave. Instead of trying to solve the equations, de Broglie assumed
a It might be thought that here de Broglie has in mind the fa t that N moving
parti les may be asso iated with N velo ity elds in 3-spa e, whi h may be
asso iated with N (non-singular) guiding waves in 3-spa e. From a pilot-wave
perspe tive, su h N guiding waves may be identied with the N ` onditional' wave
fun tions i (xi , t) (X1 , X2 , ..., xi , ..., XN , t), where xi (i = 1, 2, ..., N ) ranges
over all positions in 3-spa e and Xj is the a tual position of the j th parti le. Ea h
i (xi , t) denes a wave in 3-spa e that determines the velo ity of the ith parti le
(through the gradient of its phase); and ea h i (xi , t) depends on the positions
of the other N 1 parti les. However, given the ontext (in parti ular the re ent
publi ation of `Stru ture'), it seems lear that here de Broglie's N waves are the
singular u-waves of his double-solution theory.
80
that the resulting velo ities of the moving singularities ould be written
as the gradient of a fun tion (x1 , x2 ), whi h ould be identied with the
phase of a solution (x1 , x2 , t) of the (time-independent) S hrdinger
equation. In `Stru ture', then, the wave fun tion in onguration spa e
appeared as an ee tive des ription of the motions of the parti les (or
singularities). An e ho of this view is dis ernible in the Solvay report,
whi h ontinues with the following justi ation for introdu ing a guiding
wave in onguration spa e (p. 385):
Nevertheless, if one fo usses one's attention only on the orpus les, one an
represent their states by a point in onguration spa e, and one an try to
relate the motion of this representative point to the propagation of a titious
wave
in onguration spa e.
De Broglie seems to be saying that, if one is on erned only with a su in t mathemati al a ount of parti le motion (as opposed to a physi al
representation), then this an be obtained by introdu ing a guiding wave
in onguration spa e. Certainly, the dynami s is mu h simpler with a
single, autonomous wave in onguration spa e. As in his `Stru ture'
paper, this seems to be de Broglie's explanation for S hrdinger's otherwise mysterious onguration-spa e wave.
When de Broglie states that the wave is ` titious', he presumably
means that it has only mathemati al, and not physi al, signi an e.a It
provides a onvenient mathemati al a ount of parti le motion, pending
a full physi al des ription by a more detailed theory. As we dis ussed in
se tion 2.3.2, the provisional introdu tion of a `mathemati al' des ription, pending the development of a proper `physi al' model, has distinguished pre edents in the history of Newtonian gravity and Maxwellian
ele trodynami s.
Having motivated the introdu tion of a guiding wave in onguration spa e, de Broglie suggests (p. 385) that
.... plays for the representative point of the system in onguration spa e the
same role of pilot wave and of probability wave that the wave
plays in
ordinary spa e in the ase of a single material point.
Thus, in the nonrelativisti approximation, de Broglie onsiders N para This is somewhat in ontrast with de Broglie's introdu tion of the pilot wave at
the end of `Stru ture', where he refers to the parti le and the guiding wave as
`distin t realities'. However, there de Broglie expli itly proposed pilot-wave theory
for a single parti le only, and it is likely that while he was omfortable with the
idea of a physi ally real pilot wave in 3-spa e (for the one-body ase), he ould
not regard a pilot wave in onguration spa e as having more than mathemati al
signi an e.
81
ti les with positions x1 , x2 , ..., xN . He states that the wave determines

the velo ity of the representative point in onguration spa e by the
formula
1
vk =
(76)
k
mk
(where mk is the mass of the k th parti le). As in the ase of a single
parti le, notes de Broglie, the probability for the system to be present
in a volume element d of onguration spa e is
d = const a2 d .
(77)
These are the standard pilot-wave equations for a many-body system,

repla ing the single-parti le formulas (73) and (74). De Broglie remarks
that (77) seems to agree with Born's results for ele tron s attering by
an atom and with Fermi's for s attering by a rotator.
Finally, at the end of his se tion on the many-body ase, de Broglie
notes that it seems di ult to onstru t a wave that an generate
the motion of a relativisti many-body system (in ontrast with the
relativisti one-body ase), a point that de Broglie had already noted
in `Stru ture'. From the very beginning, then, it was re ognised that it
would be di ult to formulate a fundamentally Lorentz-invariant pilotwave theory for a many-body system, a situation that persists to this
day.
Thus, in his report at the fth Solvay onferen e, de Broglie arrived
at pilot-wave theory for a many-body system, with a guiding wave in
onguration spa e. Judging by de Broglie's omments about a physi al
representation requiring N waves in 3-spa e, and his hara terisation
of as a ` titious' wave, it seems lear that he still regarded his new
dynami s as an ee tive, mathemati al theory only (just as he had a few
months earlier in his `Stru ture' paper).
De Broglie's report then moves on to sket h some appli ations to
atomi theory. For stationary states of hydrogen, de Broglie notes that
the ele tron motion is ir ular, ex ept in the ase of magneti quantum
number m = 0 for whi h the ele tron is at rest. (Note that de Broglie
expresses no on ern that the ele tron is predi ted to be motionless in
the ground state of hydrogen. This result may seem puzzling lassi ally,
just as Bohr's quantised atomi orbits seem puzzling lassi ally: but both
are natural onsequen es of de Broglie's non- lassi al dynami s.)
De Broglie also points out that one an al ulate the ele tron velo ity
during an atomi transition i j , so that su h transitions an be
visualised. In this example, de Broglie's guiding wave is a solution of the
82
time-dependent S hrdinger equation, with a time-dependent amplitude

as well as phase. (The atomi wave fun tion is taken to have the form
= ci i + cj j , where i , j are eigenfun tions orresponding to the
atomi states i, j , and ci , cj are fun tions of time.) Clearly, then, de
Broglie applied his pilot-wave dynami s not just to stationary states,
but to quite general wave fun tions as he had also done (in the ase
of a single parti le) in `Stru ture'.
De Broglie then outlines how, in his theory, one an obtain expressions
for the mean harge and urrent density for an ensemble of atoms. These
expressions are the same as those used by S hrdinger and others. De
Broglie remarks that, `denoting by the wave written in omplex form,
and by [ the onjugate fun tion', in the nonrelativisti limit the
harge density is proportional to ||2 : the ele tri dipole moment then
ontains the orre t transition frequen ies. By using these expressions
as sour es in Maxwell's equations, says de Broglie, one an orre tly
predi t the mean energy radiated by an atom. This is just semi lassi al
radiation theory, whi h is still widely used today in quantum opti s.a
Part II of de Broglie's report ends with some general remarks. First
and foremost, de Broglie makes it lear that he regards pilot-wave theory
as only provisional (p. 389):
So far we have onsidered the orpus les as èxterior' to the wave
their
motion being only determined by the propagation of the wave. This is, no
doubt, only a provisional point of view: a true theory of the atomi stru ture
of matter and radiation should, it seems to us,
in orporate
the orpus les in
the wave phenomenon by onsidering singular solutions of the wave equations.
De Broglie goes on to suggest that in a deeper theory one ould `show

that there exists a orresponden e between the singular waves and the
waves , su h that the motion of the singularities is onne ted to the
propagation of the waves ', just as he had suggested in `Stru ture'.
Part II ends by noting the in omplete state of the theory with respe t
to the ele tromagneti eld and ele tron spin.
De Broglie's nal part III ontains a lengthy dis ussion and review of
re ent experiments involving the dira tion, interferen e and s attering
of ele trons (as had been requested by Lorentz4 ). He regards the results
as eviden e for his `new Dynami s'. For the ase of the dira tion of
ele trons by a rystal latti e, de Broglie points out that the s attered
wave fun tion has maxima in ertain dire tions, and notes that a a Cf. se tion 4.4.
83
ording to his theory the ele trons should be preferentially s attered in

these dire tions (p. 391):
Be ause of the role of pilot wave played by the wave
, one must then observe
a sele tive s attering of the ele trons in these dire tions.
What de Broglie is (briey) des ribing here is the separation of the in ident wave fun tion into distin t (non-overlapping) emerging beams,
with ea h outgoing ele tron o upying one beam, and with an ensemble
of ele trons being distributed among the emerging beams a ording to
the Born rule. This is relevant to a proper understanding of the de
Broglie-Pauli en ounter, dis ussed in se tion 10.2.
Con erning the s attering maxima observed re ently by Davisson and
Germer, for ele trons in ident on a rystal, de Broglie remarks (p. 394):
There is dire t numeri al onrmation of the formulas of the new Dynami s .... .
De Broglie also dis usses the inelasti s attering of ele trons by atoms.
A ording to Born's al ulations, one should observe maxima and minima in the angular dependen e of the dierential s attering ross se tion.
A ording to de Broglie, it is premature to speak of an agreement with
experiment. The results of Dymond, for the inelasti s attering of ele trons by helium atoms, do however show maxima in the ross se tion,
in qualitative agreement with the predi tions. De Broglie makes it quite
lear that, at the time, theory was far behind experiment.
The dis ussion following de Broglie's report was extensive, detailed,
and varied. A number of parti ipants raised queries, and de Broglie
replied to most of them. Some of this dis ussion will be onsidered in
detail in hapters 10 and 11. Here, we limit ourselves to a brief summary
of the questions raised.
Lorentz asked how, in the simple pilot-wave theory of 1924, de Broglie
derived quantisation onditions for the ase of multiperiodi atomi
orbits. Born questioned the validity of de Broglie's guidan e equation
for an elasti ollision, while Pauli suggested that the key idea behind de
Broglie's theory was the asso iation of parti le traje tories with a lo ally
onserved urrent. S hrdinger raised the question of an alternative
velo ity eld dierent from that assumed by de Broglie, while Kramers raised the question of how the Maxwell energy-momentum tensor
ould arise from independently moving photons. Lorentz, Ehrenfest and
S hrdinger asked about the properties of ele tron orbits in hydrogen.
Brillouin dis ussed at length the simple example of a photon olliding
84
with a mirror: in his Fig. 2, the in ident and emergent photon traje tories
are lo ated inside pa kets of limited extent (a point relevant for the
de Broglie-Pauli en ounter in the general dis ussion). Finally, Lorentz
onsidered, in lassi al Maxwell theory, the near-eld attra tive stress
between two prisms, and laimed that this `negative pressure' ould not
be produ ed by the motion of orpus les (photons).

In his papers and thesis of 192324, de Broglie proposed a simple form of
a new, non- lassi al dynami s of parti les with velo ities determined by
the phase gradient of abstra t waves in 3-spa e. De Broglie onstru ted
the dynami s in su h a way as to unify the me hani al prin iple of
Maupertuis with the opti al prin iple of Fermat. He showed that it gave
an a ount of some simple quantum phenomena, in luding single-parti le
interferen e and quantised atomi energy levels.
As we have seen, the s ope and ambition of de Broglie's do toral thesis
went far beyond a mere extension of the relations E = h , p = h/ from
photons to other parti les. Yet, the thesis is usually remembered solely
for this idea. The depth and inner logi of de Broglie's thinking is not
usually appre iated, neither by physi ists nor by historians. An ex eption
is Darrigol (1993), who on this very point writes (pp. 3034):
For one who only knows of de Broglie's relation
= h/p,
two explanations of
his originality oer themselves. The rst has him as a lu ky dreamer who
hit upon a great idea amidst a foolish play with analogies and formulas.
The se ond has him a providential deep thinker, who dedu ed unsuspe ted
onne tions by rationally ombining distant on epts. The rst explanation is
the most popular, though rarely expressed in print. .... The se ond explanation
of Louis de Broglie's originality, though also extreme, is ertainly loser to the
truth. Anyone who has read de Broglie's thesis annot help admiring the unity
and inner onsisten y of his views, the inspired use of general prin iples, and
a ne essary reserve.
In 1926, starting from de Broglie's expressions for the frequen y and

phase velo ity of an ele tron wave in an external potential, S hrdinger
found the (nonrelativisti ) wave equation for de Broglie's waves. It was
de Broglie's work, beginning in 1923, that initiated the notion of a
wave fun tion for material parti les. And it was de Broglie's view of
the signi an e of the opti al-me hani al analogy, and of the role waves
ould play in bringing about the existen e of integer-valued quantum
85
numbers, that formed the basis for S hrdinger's development of the

wave equation and the asso iated eigenvalue problem.
In 1927, de Broglie proposed what we now know as pilot-wave or de
Broglian dynami s for a many-body system, with a guiding wave in
onguration spa e. De Broglie regarded this theory as provisional: he
thought it should emerge as an ee tive theory, from a more fundamental
theory in whi h parti les are represented by singularities of 3-spa e
waves. Even so, he did propose the theory, in the rst-order form most
ommonly used today. Further, as we have seen, de Broglie's view of the
pilot wave as merely phenomenologi al is strikingly reminis ent of (for
example) late-nineteenth- entury views of the ele tromagneti eld. In
retrospe t, one may regard de Broglie as having unwittingly arrived at
a new and fundamental on ept, that of a pilot wave in onguration
spa e.
De Broglie was unable to show that his new dynami s a ounted for all
quantum phenomena. In parti ular, as we shall dis uss in hapters 10 and
11, de Broglie did not understand how to des ribe the pro ess of measurement of arbitrary quantum observables in pilot-wave theory: as shown
in detail by Bohm in 1952, this requires an appli ation of de Broglie's
dynami s to the measuring devi e itself. De Broglie did, however, possess
the fundamental dynami s in omplete form. Furthermore, de Broglie
did understand how his theory a ounted for single-parti le interferen e,
for the dire ted s attering asso iated with rystal dira tion, and for
ele tron s attering by atoms (for the latter, see se tion 10.2). Clearly, in
1927, many appli ations of pilot-wave theory remained to be developed;
just as, indeed, many appli ations of quantum theory in luding the
general quantum theory of measurement remained to be developed
and laried.
In retrospe t, de Broglian dynami s seems as radi al as and indeed
somewhat reminis ent of Einstein's theory of gravity. A ording to
Einstein, there is no gravitational for e, and a freely falling body follows
the straightest path in a urved spa etime. A ording to de Broglie, a
massive body undergoing dira tion and following a urvilinear path is
not a ted upon by a Newtonian for e: it is following the ray of a guiding
wave. De Broglie's abandonment of Newton's rst law of motion in 1923,
and the adoption of a dynami s based on velo ity rather than a eleration, amounts to a far-rea hing departure from lassi al me hani s and
(arguably) from lassi al kinemati s too with impli ations for the
stru ture of spa etime that have perhaps not been understood (Valentini
1997). Certainly, the extent to whi h de Broglie's dynami s departs from
86
lassi al ideas was unfortunately obs ured by Bohm's presentation of it,

in 1952, in terms of a eleration and a pseudo-Newtonian quantum potential, a formulation that today seems arti ial and inelegant ompared
with de Broglie's (mu h as the rewriting of general relativity as a eld
theory on at spa etime seems unnatural and hardly illuminating). The
fundamentally se ond-order nature of lassi al physi s is today embodied
in the formalism of Hamiltonian dynami s in phase spa e. In ontrast,
de Broglie's rst-order approa h to the theory of motion seems more
naturally ast in terms of a dynami s in onguration spa e.
Regardless of how one may wish to interpret it, by any standards
de Broglie's work from 1923 to 1927 shows a remarkable progression
of thought, beginning from early intuitions and simple models of the
relationship between parti les and waves, and ending (with S hrdinger's
help) with a omplete and new form of dynami s for nonrelativisti
systems a deterministi dynami s that was later shown by Bohm to be
empiri ally equivalent to quantum theory (given a Born-rule distribution
of initial parti le positions). The inner logi of de Broglie's work in this
period, his drive to unite the physi s of parti les with the physi s of
waves by unifying the prin iples of Maupertuis and Fermat, his wave-like
explanation for quantised energy levels, his predi tion of ele tron diffra tion, his explanation for single-parti le interferen e, his attempts to
onstru t a eld-theoreti al pi ture of parti les as moving singularities,
and his eventual proposal of pilot-wave dynami s as a provisional theory
all this ompels admiration, all the more for being largely unknown
and unappre iated.
Today, pilot-wave theory is often hara terised as simply adding parti le traje tories to the S hrdinger equation. An understanding of de
Broglie's thought from 1923 to 1927, and of the role it played in S hrdinger's work, shows the gross ina ura y of this hara terisation: after
all, it was a tually S hrdinger who removed the traje tories from de
Broglie's theory ( f. se tion 11.1). It is di ult to avoid the on lusion
that de Broglie's stature as a major ontributor to quantum theory
has suered unduly from the ir umstan e that, for most of the twentieth entury, the theory proposed by him was in orre tly regarded as
untenable or in onsistent with experiment. Regardless of whether or
not de Broglie's pilot-wave theory is loser to the truth than other
interpretations, the fa t that it is a onsistent and viable approa h to
quantum physi s whi h has no measurement problem, whi h shows
that obje tivity and determinism are not in ompatible with quantum
physi s, and whi h stimulated Bell to develop his famous inequalities
87
ne essarily entails a reappraisal of de Broglie's pla e in the history of

twentieth- entury physi s.
88
Notes to pp. 3182
Ar hival notes
1 AHQP-OHI, Louis de Broglie, `Replies to Mr Kuhn's questions'.
2 AHQP-OHI, Louis de Broglie, session 1 (tape 43a), 7 January 1963,
Paris; 0.75 h, in Fren h, 13 pp.; by T. S. Kuhn, T. Kahan and A. George.
Ibid.
4 Cf. de Broglie to Lorentz, 27 June 1927, AHQP-LTZ-11 (in Fren h).
3
From matrix me hani s to quantum
me hani s
The report by Born and Heisenberg on `quantum me hani s' may seem
surprisingly di ult to the modern reader. This is partly be ause Born
and Heisenberg are des ribing various stages of development of the theory that are quite dierent from today's quantum me hani s. Among
these, it should be noted in parti ular that the theory developed by
Heisenberg, Born and Jordan in the years 192526 and known today
as matrix me hani s (Heisenberg 1925b [1, Born and Jordan 1925 [2,
Born, Heisenberg and Jordan 1926 [4)a diers from standard quantum
me hani s in several important respe ts. At the same time, the interpretation of the theory (the topi of se tion II of the report) also appears
to have undergone important modi ations, in parti ular regarding the
notion of the state of a system. Initially, Born and Heisenberg insist
on the notion that a system is always in a stationary state (performing
quantum jumps between dierent stationary states). Then the notion
of the wave fun tion is introdu ed and related to probabilities for the
stationary states. At a later stage probabilisti notions (in parti ular,
what one now alls transition probabilities) are extended to arbitrary
observables, but it remains somewhat un lear whether the wave fun tion
itself should be regarded as a fundamental entity or merely as an ee tive
one. This may ree t the dierent routes followed by Born and by
Heisenberg in the development of their ideas. The ommon position
presented by Born and Heisenberg emphasises the probabilisti aspe t
of the theory as fundamental, and the on lusion of the report expresses
strong onden e in the resulting pi ture.
The two main se tions of this hapter, se tion 3.3 and se tion 3.4,
will be devoted, respe tively, to providing more details on the various
a Throughout this hapter, numbers in square bra kets refer to entries in the original
bibliography at the end of Born and Heisenberg's report.
89
90
From matrix to quantum me hani s
stages of development of the theory, and to disentangling various threads

of interpretation that appear to be present in Born and Heisenberg's
report. Before that, we provide a summary (se tion 3.1) and a few
remarks on the authorship and writing of the various se tions of the
report (se tion 3.2).
3.1 Summary of Born and Heisenberg's report

Born and Heisenberg's report has four se tions (together with an introdu tion and on lusion): I on formalism, II on interpretation, III on
axiomati formulations and on un ertainty, and IV on appli ations. The
formalism that is des ribed is initially that of matrix me hani s, whi h
is then extended beyond the original framework (among other things, in
order to make the onne tion with S hrdinger's wave me hani s). Then,
further developments of matrix me hani s are sket hed. These allow one
to in orporate a `statisti al' interpretation. After a brief dis ussion of
Jordan's (1927b, [39) axiomati formulation, the un ertainty relations
are used to justify the statisti al element of the interpretation. A few
appli ations of spe ial interest and some brief nal remarks on lude the
report. As we dis uss below, Born drafted se tions I and II, and Heisenberg drafted the introdu tion, se tions III and IV and the on lusion.
Born and Heisenberg's introdu tion stresses the ontinuity of quantum
me hani s with the old quantum theory of Plan k, Einstein and Bohr,
and tou hes briey on su h themes as dis ontinuity, observability in
prin iple, Àns hauli hkeit' (for whi h see se tion 4.6 below) and the
statisti al element in quantum me hani s.
Se tion I, `The mathemati al methods of quantum me hani s', rst
sket hes matrix me hani s roughly as developed in the `three-man paper'
by Born, Heisenberg and Jordan (1926 [4): the basi framework of
position and momentum matri es and of the anoni al equations, the
perturbation theory, and the onne tion with the theory of quadrati
forms (the latter leading to both dis rete and ontinuous spe tra).a Next,
Born and Heisenberg des ribe two generalisations of matrix me hani s:
Dira 's (1926a [7) q-number theory and what they refer to as Born and
Wiener's (1926a [21, 1926b) operator al ulus. As a matter of fa t, they
already sket h von Neumann's representation of physi al quantities by
a Note that Born and Heisenberg use the term `quantum me hani s' to refer also
to matrix me hani s, in keeping with the terminology of the original papers. See
also Mehra and Re henberg (1982 , fn. 72 on pp. 612). Heisenberg expresses his
dislike for the term `matrix physi s' in Heisenberg to Pauli, 16 November 1925
(Pauli, 1979, pp. 2556).
3.1 Summary of Born and Heisenberg's report
91
operators in Hilbert spa e. This is identied as a mathemati ally rigorous

version of the transformation theory of Dira (1927a [38) and Jordan
(1927b, [39).a Born and Heisenberg note that, if one takes as the Hilbert spa e the appropriate fun tion spa e, the problem of diagonalising
the Hamiltonian operator leads to the time-independent S hrdinger
equation. Thus, in this (formal) sense, the S hrdinger theory is a spe ial
ase of the operator version of quantum me hani s. The eigenfun tions
of the Hamiltonian an be asso iated with the unitary transformation
that diagonalises it. Matrix me hani s, too, is a spe ial ase of this more
general formalism, if one takes the (dis rete) energy eigenstates as the
basis for the matrix representation.
Se tion II on `Physi al interpretation' is probably the most striking.
It begins by stating that matrix me hani s des ribes neither the a tual
state of a system, nor when hanges in the a tual state take pla e, suggesting that this is related to the idea that matrix me hani s des ribes only
losed systems. In order to obtain some des ription of hange, one must
onsider open systems. Two methods for doing this are introdu ed. First
of all, following Heisenberg (1926b [35) and Jordan (1927a [36), one an
onsider the matrix me hani al des ription of two oupled systems that
are in resonan e. One an show that the resulting des ription an be
interpreted in terms of quantum jumps between the energy levels of the
two systems, with an expli it expression for the transition probabilities.
The se ond method uses the generalised formalism of se tion I, in whi h
time dependen e an be introdu ed expli itly via the time-dependent
S hrdinger equation. Again, expressions an be found that an be interpreted as transition probabilities, and similarly the squared modulus
of the wave fun tion's oe ients in the energy basis is interpreted
as the probability for the o urren e of the orresponding stationary
state. Born and Heisenberg introdu e the notion of interferen e of probabilities. Only at this stage are experiments mentioned as playing a
on eptually ru ial role, namely in an argument why su h interferen e
`does not represent a ontradi tion with the rules of the probability
al ulus' (p. 423). Interferen e is then related to the wave theory of de
Broglie and S hrdinger, and the interpretation of the squared modulus
of the wave fun tion as a probability density for position is introdu ed.
a See below, se tions 3.3.3 and 3.4.1, for some dis ussion of Born and Wiener's work,
and se tion 3.4.5 for the transformation theory (mainly Dira 's). Note that von
Neumann had published a series of papers on the Hilbert-spa e formulation of
quantum me hani s before his well-known treatise of 1932. It is to two of these
that Born and Heisenberg refer, Hilbert, von Neumann and Nordheim (1928 [41,
submitted April 1927) and von Neumann (1927 [42).
92
Born and Heisenberg also dene the `relative' (i.e. onditional) position
density given an energy value, as the squared modulus of the amplitude
of a stationary state. Finally, in the ontext of Dira 's and Jordan's
transformation theory, the notions of transition probability (for a single
quantity) and of onditional probability (for a pair of quantities) are
generalised to arbitrary physi al quantities.
Se tion III opens with a on ise exposition of Jordan's (1927b, [39)
axioms for quantum me hani s, whi h take the notion of probability
amplitude as primary. Born and Heisenberg point out some formal drawba ks, su h as the use of -fun tions and the presen e of unobservable
phases in the probability amplitudes. They note that su h drawba ks
have been over ome by the formulation of von Neumann (whi h they
do not go on to dis uss further). Born and Heisenberg then pro eed to
dis uss in parti ular whether the statisti al element in the theory an be
re on iled with ma ros opi determinism. To this ee t they rst justify
the ne essity of using probabilisti notions by appeal to the notion of
un ertainty (dis ussed using the example of dira tion of light by a single
slit); then they point out that, while for instan e in ases of dira tion
the laws of propagation of the probabilities in quantum me hani s are
very dierent from the lassi al evolution of a probability density, there
are ases where the two oin ide to a very good approximation; they
write that this justies the lassi al treatment of - and -parti les in a
loud hamber.
Se tion IV dis usses briey some appli ations, hosen for their ` lose
relation to questions of prin iple' (p. 432). Mostly, these appli ations
highlight the importan e of the generalised formalism (introdu ed in
se tion I), by going beyond matrix me hani s proper or wave me hani s
proper. The rst example is that of spin, whi h, requiring nite matri es,
is taken to be a problemati on ept for wave me hani s (but not for
matrix me hani s). The main example is given by the dis ussion of
identi al parti les, the Pauli prin iple and quantum statisti s. Born and
Heisenberg note that the hoi e of whether the wave fun tion should
be symmetri or antisymmetri appears to be arbitrary. They note, moreover, that the appropriate hoi e arises naturally if one quantises the
normal modes of a bla k body, leading to Bose-Einstein statisti s, or, if
one adopts Jordan's (1927d [54) quantisation pro edure, to Fermi-Dira
statisti s.a Finally, Born and Heisenberg omment on Dira 's quantum
a The general dis ussion returns to these issues in more detail; see for instan e
Dira 's riti al remarks on Jordan's pro edure, p. 501.
3.2 Writing of the report
93
ele trodynami s (Dira 1927b, [51,52), noting in parti ular that it

yields the transition probabilities for spontaneous emission.
The on lusion dis usses quantum me hani s as a ` losed theory' (see
se tion 3.4.7 below) and addresses the question of whether indeterminism in quantum me hani s is fundamental. In parti ular, Born and
Heisenberg state that the assumption of indeterminism agrees with experien e and that the treatment of ele trodynami s will not modify this
state of aairs. They on lude by noting that the existing problems
on ern rather the development of a fully relativisti theory of quantum
ele trodynami s, but that there is progress also in this dire tion.
The dis ussion is omparatively brief. Dira des ribes the analogy between the matrix method and the Hamilton-Ja obi theory. Then Lorentz
makes a few remarks, in parti ular emphasising that the freedom to
hoose the phases in the matri es q and p, (2) and (5) below, is not
limited to a dierent hoi e of the time origin, a possibility already noted
by Heisenberg (1925b [1), but extends to arbitrary phase fa tors of the
form ei(m n ) , as noted by Born and Jordan (1925 [2).a A ording to
Lorentz, this fa t suggests that the `true os illators' are asso iated with
the dierent stationary states rather than with the spe tral frequen ies.1
The last few and brief ontributions address the question of the phases.b
3.2 Writing of the report

The respe tive ontributions by Born and by Heisenberg to the report
be ome lear from their orresponden e with Lorentz. As mentioned in
hapter 1, Lorentz had originally planned to have a report on quantum
me hani s from either Born or Heisenberg, leaving to them the hoi e
of who was to write it. In reply, Born and Heisenberg suggested that
they would provide a joint report.2 Heisenberg was going to visit Born
in Gttingen in early July 1927, on whi h o asion they would de ide
on the stru ture of the report and divide up the work, planning to meet
again in August to nish it. The report (as requested by Lorentz) would
in lude the results by Dira , who was in Gttingen at the time.c As Born
a This orresponds of ourse to the hoi e of a phase fa tor for ea h energy state.
b As mentioned in footnote on p. 23, Bohr suggested to omit this entire dis ussion
from the published pro eedings.
Dira was in Gttingen from February 1927 to the end of June, after having
spent September to February in Copenhagen. Dira 's paper on transformation
theory (1927a [38) and his paper on emission and absorption (Dira 1927b [51)
were written in Copenhagen, while the paper on dispersion (Dira 1927 [52) was
written in Gttingen. Compare Kragh (1990, h. 2. pp. 37 .).
94
explains, the report would àbove all emphasise the viewpoints that are
presumably taken less into a ount by S hrdinger, namely the statisti al on eption of quantum me hani s'. Either author, or both, ould
orally present the report at the onferen e (Heisenberg, in a parallel
letter, suggested that Born should be the appropriate hoi e, sin e he
was the senior s ientist3 ); Born and Heisenberg ould also give additional
explanations in English. Born did not speak Fren h, nor (Born thought)
did Heisenberg.
In a later letter,4 Born reports that Heisenberg had sent him at the end
of July the draft of introdu tion, on lusion and se tions III and IV, upon
whi h Born had written se tions I and II, reworked Heisenberg's draft,
and sent everything ba k to Heisenberg. Heisenberg in turn had made
some further small hanges. Due to illness and a small operation, Born
had been unable to go to Muni h in mid-August, so that he had not seen
the hanges. He trusted, however, that he would agree with the details
of Heisenberg's nal version.a Born also mentions that Heisenberg and
himself would like to go through the text again, at least at proof stage.b
The presentation would be split between Born (introdu tion and se tions
I and II) and Heisenberg (se tions III and IV and on lusion). Born
on ludes with the following words: Ìt would be parti ularly important
for us to ome to an agreement with S hrdinger regarding the physi al
interpretation of the quantum formalisms'.
3.3 Formalism
3.3.1 Before matrix me hani s

In the old Bohr-Sommerfeld theory, ele tron orbits are des ribed as
lassi al Kepler orbits subje t to additional onstraints (the `quantum
onditions'), yielding dis rete stationary states. Su h a pro edure, in
Born and Heisenberg's introdu tion, is riti ised as arti ial.c An atom is
assumed to `jump' between its various stationary states, and energy differen es between stationary states are related to the spe tral frequen ies
a Heisenberg had sent the nal text to Lorentz on 24 August.5
b The published version and the types ript show only minor dis repan ies, and many
of these are learly mistakes in the published version (detailed in our endnotes to
the report). This and the fa t that Born and Heisenberg appear not to have spoken
Fren h suggests that there was in fa t no proofreading on their part.
The remark is rather brief, but note that Born and Heisenberg seem to onsider
the introdu tion of photons into the lassi al ele tromagneti theory (a orpus ular
dis ontinuity) to be as arti ial as the introdu tion of dis rete stationary states
into lassi al me hani s.
3.3 Formalism
95
via Bohr's frequen y ondition. Spe tral frequen ies and orbital frequen ies, however, appear to be quite unrelated. This is the ru ial `radiation
problem' of the old quantum theory (see also below, se tion 4.4).
The BKS theory of radiationa in ludes a rather dierent pi ture of
the atom, and arguably provides a solution to the radiation problem
just mentioned. In the BKS theory, one asso iates to ea h atom a set of
`virtual' os illators, with frequen ies equal to the spe tral frequen ies.
Spe i ally, when the atom is in the stationary state n, the os illators
that are ex ited are those with frequen ies orresponding to transitions from the energy En to both lower and higher energy levels. These
os illators produ e a virtual radiation eld, whi h propagates ( lassi ally) a ording to Maxwell's equations and intera ts (non- lassi ally)
with the virtual os illators, in parti ular inuen ing the probabilities
for indu ed emission and absorption in other atoms and determining
the probabilities for spontaneous emission in the atom itself. The a tual
emission or absorption of energy is asso iated with the orresponding
transition between stationary states. Note that an emission in one atom
is not onne ted dire tly to an absorption in another, so that energy and
momentum are onserved only statisti ally.
Several important results derived by developing orresponden e arguments into translation rules from lassi al to quantum expressions
were formulated within the ontext of the BKS theory, in parti ular
Kramers' (1924) dispersion theory. Other examples were Born's (1924)
perturbation formula, Kramers and Heisenberg's (1925) joint paper on
dispersion and Heisenberg's (1925a) paper on polarisation of uores en e
radiation.b The same arguments also led to a natural reformulation of
the quantum onditions independent of the old ` lassi al models' (Kuhn
1925, Thomas 1925). As rst argued by Pauli (1925), su h results were
entirely independent of the presumed me hanism of radiation.c
Indeed, after the demise of the BKS theory following the experimental
veri ation of the onservation laws in the Bothe-Geiger experiments,
these results and the orresponding te hniques of symboli translation
formed the basis for Heisenberg's formulation of matrix me hani s in
his famous paper of 1925 (Heisenberg 1925b [1). Heisenberg's stroke of
genius was to give up altogether the kinemati al des ription in terms of
spatial oordinates, while retaining the lassi al form of the equations
a See also the des ription in se tion 1.3.
b See Darrigol (1992, pp. 22446) and Mehra and Re henberg (1982b, se tions III.4
and III.5).
Cf. Darrigol (1992, pp. 2445). Cf. also Jordan (1927e [63), among others.
96
of motion, to be solved under suitably reformulated quantum onditions

(those of Kuhn and Thomas). While Heisenberg had dropped both the
me hanism of radiation and the older ele troni orbits, there were points
of ontinuity with the BKS theory and with the old quantum theory;
in parti ular, as we shall see, the new variables were at least formally
related to the virtual os illators, and the pi ture of quantum jumps was
retained, at least for the time being.
3.3.2 Matrix me hani s

Heisenberg's original paper is rather dierent from the more denitive
presentations of matrix me hani s: the equations are given in Newtonian
(not Hamiltonian) form, energy onservation is not established to all
orders, and, as is well known, Heisenberg at rst had not re ognised
that his theory used the mathemati al ma hinery of matri es. Like Born
and Heisenberg in their report, we shall a ordingly follow in this se tion
mainly the papers by Born and Jordan (1925 [2) and Born, Heisenberg
and Jordan (1926 [4), as well as the book by Born (1926d,e [58),
supplementing the report with more details when useful.
Kinemati ally, the des ription of `motion' in matrix me hani s generalises that given by the set of the omponents of a lassi al Fourier
series,
qn e2int
(1)
(and thus generalises the idea of a periodi motion). As frequen ies, one
hooses, instead of (n) = n , the transition frequen ies of the system
under onsideration, whi h are required to obey Ritz's ombination prin iple, ij = Ti Tj (relating the frequen ies to the spe tros opi terms
Ti = Ei /h, or, in Born and Heisenberg's notation, Ti = Wi /h). The
omponents thus dened form a doubly innite array:
q11 e2i11 t q12 e2i12 t . . .

2i21 t
q22 e2i22 t . . .
q = q21 e
(2)
.
..
..
..
.
.
.
From the ombination prin iple for the ij , it follows that ij = ji .
Further, it is required of the (generally omplex) amplitudes that
qji = qij
,
(3)
by analogy with a lassi al Fourier series, so that the array is in fa t a

Hermitian matrix.
3.3 Formalism
97
In modern notation, the above matrix onsists of the elements of the

position observable (in Heisenberg pi ture) in the energy basis. For a
losed system with Hamiltonian H , these indeed take the form
hEi | Q(t) |Ej i = hEi | e(i/~)Ht Q(0)e(i/~)Ht |Ej i
= hEi | Q(0) |Ej i e(i/~)(Ei Ej )t .
(4)
Along with the position matrix q , one onsiders also a momentum

matrix:
p11 e2i11 t p12 e2i12 t . . .

2i21 t
p22 e2i22 t . . .
p = p21 e
(5)
,
..
..
..
.
.
.
as well as all other matrix quantities that an be obtained as fun tions of

q and p, dened as polynomials or as power series (leaving aside questions
of onvergen e). Again from the ombination prin iple, it follows that
under multipli ation of two su h matri es, one obtains another matrix
of the same form (i.e. with the same time-dependent phases). Thus one
justies taking these two-dimensional arrays as matri es. The most general physi al quantity in matrix me hani s is a matrix whose elements (in
modern terminology) are the elements of an arbitrary Heisenberg-pi ture
observable in the energy basis.
The values of these matri es (in parti ular the amplitudes qij and the
frequen ies ij ) will have to be determined by solving the equations
of motion under some suitable quantum onditions. Spe i ally, the
equations of motion are postulated to be the analogues of the lassi al
Hamiltonian equations:
H
dq
=
,
dt
p
dp
H
=
,
dt
q
(6)
where H is a suitable matrix fun tion of q and p. The quantum onditions

are formulated as
h
pq qp =
(7)
1 ,
2i
where 1 is the identity matrix. In the limit of large quantum numbers,
(7) orresponds to the òld' quantum onditions.
Of ourse, dierentiation with respe t to t and with respe t to matrix
arguments need to be suitably dened. The former is dened elementwise
as
(8)
fij = 2iij fij ,
98
whi h an be written equivalently as

2i
(9)
(W f f W ) ,
f =
h
where W is the diagonal matrix with Wij = Wi ij . Dierentiation with
respe t to matrix arguments is dened (in slightly modernised form) as
df
f (x + 1) f (x)
(10)
= lim
,
dx 0
where the limit is also understood elementwise.a

On e the pre eding s heme has been set up, the next question is how
to go about solving the equations of motion. First, for the spe ial ase
of dierentiation with respe t to q or p, one shows that
f q qf =
h f
2i p
(11)
pf f p =
h f
2i q
(12)
and
(by indu tion on the form of f and using (7)). From this and from (9),
one an easily show that the equations of motion are equivalent to
W q qW = Hq qH
(13)
W p pW = Hp pH .
(14)
and
Thus, W H ommutes with both q and p, and therefore also with H .
It follows that
W H HW = 0 ,
(15)
H = 0 ,
(16)
whi h by (9) means that

that is, energy is onserved. In the non-degenerate ase, it follows that
H is a diagonal matrix. Denoting its diagonal terms by Hi , (13) yields
further the Bohr frequen y ondition,
Hi Hj = Wi Wj = hij ,
(17)
and xing an arbitrary onstant one an set H = W . The above proof

shows also that onversely, (16), (17) and the ommutation relations
a This denition was agreed upon after some debate. See Mehra and Re henberg
(1982 , pp. 6971 and 97100).
3.3 Formalism
99
imply the equations of motion. Therefore, the problem of solving the

equations of motion essentially redu es to that of diagonalising the energy matrix.
This dis ussion leads also to the notion of a ` anoni al transformation'
(a transformation that leaves the equations of motion invariant) as a
transformation that preserves the ommutation relations. Obviously any
transformation of the form
Q = SqS 1 ,
P = SpS 1
(18)
(with S invertible) will be su h a transformation, and it was onje tured

that these were the most general anoni al transformations. At this
stage, S is required only to be invertible; indeed, the notion of unitarity
has not been introdu ed yet. If H is a suitably symmetrised fun tion of
q and p, however, it will also be a Hermitian matrix, and the problem
of diagonalising it an be solved using the theory of quadrati forms
of innitely many variables (assuming that the theory as known at the
time extends also to unbounded quadrati forms). In parti ular, it will
be possible to diagonalise H using a unitary S , whi h also guarantees
that the new oordinates Q and P will be Hermitian.
What does su h a solution to the equations of motion yield? In the rst
pla e, one obtains the values of the frequen ies ij and of the energies Wi
(along with the values of other onserved quantities). In the se ond pla e,
based on orresponden e arguments, one an identify the diagonal elements of a matrix with `time average' in the respe tive stationary states,a
and relate the squared amplitudes |qij |2 to the intensities of spontaneous
emission, or equivalently to the orresponding transition probabilities
between stationary states.b This, however, is not an argument based on
rst prin iples. Indeed, as Born and Heisenberg repeatedly stress, matrix
me hani s in the above form des ribes a losed, onservative system,
in whi h no hange should take pla e. As Born and Heisenberg note in
se tion II, a tual transfer of energy is to be expe ted only when the atom
is oupled to some other system, spe i ally the radiation eld. Born
and Jordan's paper (1925 [2) and Born, Heisenberg and Jordan's paper
(1926 [4) in lude a rst attempt at treating the radiation eld quantum
me hani ally and at justifying the assumed interpretation of the squared
amplitudes. A satisfa tory treatment of spontaneous radiation was later
a The on ept of `time averages' an also be thought of as related to the BKS theory,
as remarked by Heisenberg himself in a letter to Einstein of February 1926.6
b The determination of intensities had been a pressing problem sin e the work by
Born (1924) and the Ùtre ht observations' referred to in the report (p. 411). See
Darrigol (1992, pp. 2345).
100
given by Dira (1927b [51), as also mentioned in the report.a In the third
pla e, by setting up relations between the amplitudes qij and the matrix
elements of other onserved quantities su h as angular momentum, one
is able to identify whi h transitions are possible between states with
the orresponding quantum numbers (derivation of `sele tion rules').
Expe tation values, other than in the form of time averages for the
stationary states, are la king.
It should be lear that matrix me hani s in its histori al formulation
is not to be identied with today's quantum me hani s in the Heisenberg pi ture. The matri es allowed as solutions in matrix me hani s
are basis-dependent. Moreover, quite apart from the fa t that solutions
to the equations of motion are hard to nd,a the results obtained are
relatively modest. In parti ular, there are no general expe tation values
for physi al quantities. In dis ussing both the mathemati al formalism
and the physi al interpretation, Born and Heisenberg suggest that the
original matrix theory is inadequate and in need of extension.
3.3.3 Formal extensions of matrix me hani s

As presented in the report, the hief formal rather than interpretational
di ulty for matrix me hani s is the failure to des ribe aperiodi quantities. As we have seen, the matri es were understood as a generalisation
of lassi al Fourier series. Stri tly speaking, if one follows this analogy, a
periodi quantity is represented by a dis rete (if doubly innite) matrix,
and one an envisage representing aperiodi quantities by ontinuous
matri es analogously to the representation of lassi al quantities by
Fourier integrals. Indeed, already Heisenberg (1925b [1) points out that
his quadrati arrays would have in general both a periodi part (dis rete)
and an aperiodi part ( ontinuous). The onne tion between matri es
and quadrati forms showed in fa t that matri es in general had also
a ontinous spe trum (if the theory extended to the unbounded ase).
Still, ontinuous matri es are unwieldy, and in ertain ases the matrix
elements (whi h are always evaluated in the energy basis) will be ome
singular, as when trying to des ribe a free parti le. (This is not at all
surprising, onsidering that also in the lassi al analogy the Fourier
integral of the fun tion at + b fails to onverge.)
a Note that by this time Dira (1927a [38) had already introdu ed a probabilisti
interpretation based on the transformation theory.
a In parti ular, one was unable to introdu e a tion-angle variables to solve the
Hamiltonian equations as in lassi al me hani s; f. Born and Heisenberg's remarks
on àperiodi motions'.
3.3 Formalism
101
This problem was addressed by generalising the notion of a matrix

to that of a q-number (Dira 1926a [7) and to that of an operator
(Born and Wiener 1926a [21, 1926b). Dira 's q-numbers are abstra t
obje ts that are assumed to form a non ommutative algebra, while Born
and Wiener's operators are hara terised not by their elements in some
basis, but by their a tion on a spa e of fun tions. In both approa hes,
one an deal with aperiodi quantities and quantities with a singular
matrix representation.
Born and Wiener's approa h is the lesser-known of the two, and shall
therefore be briey sket hed.a The operators in question are linear operators a ting on a spa e of fun tions x(t) (whi h are not given any spe i
physi al interpretation). These fun tions are understood as generalising
the fun tions having the form
x(t) =
xn e
2i
h Wn t
(19)
n=1
That is, they generalise the fun tions that an be written in a Fourier-like
series with frequen ies equal to the spe tral frequen ies. Therefore, the
operators generalise the matri es that a t on innite-dimensional ve tors
indexed by the energy values, i.e. they generalise the matri es in the
energy basis. By using these operators, Born and Wiener free themselves
from the need to operate with the matrix elements, and they are able to
solve the equations of motion expli itly for systems more general than
those treated in matrix me hani s until then. Their main example is the
(aperiodi ) one-dimensional free parti le.
Instead of presenting this formalism, however, in the report Born
and Heisenberg present dire tly von Neumann's formalism of operators
on Hilbert spa e (evidently but ta itly onsidering Born and Wiener's
formalism as its natural pre ursor). They note that in this formalism it
is possible to onsider matri es in arbitrary, even ontinuous bases, and
they use this fa t to make the onne tion with S hrdinger's theory.
In parti ular, they point out that solving the (time-independent) S hrdinger equation is equivalent to diagonalising the Hamiltonian quadrati
form, in the sense that the set of S hrdinger eigenfun tions (q, W )
(wave fun tions in q indexed by W ) yields the transformation matrix S
from the position basis to the energy basis. Thus one an use the familiar
methods of partial dierential equations to diagonalise H .b
a Born and Wiener published two very similar papers on their operator theory, one
in German, referred to in the report (1926a [21), and one in English (1926b).
b Cf. below, se tion 3.4.5. The onne tion between matrix and wave me hani s was
102
Despite providing very useful formal extensions of matrix me hani s,

the q-numbers and the operators (at least as presented in se tion I of the
report) do not provide further insights into the physi al interpretation
of the theory. In the following se tion II, on `Physi al interpretation',
Born and Heisenberg dis uss the problem of des ribing a tual states
and pro esses in matrix me hani s, suitably extended, and the surprising
rami ations of this problem.
3.4 Interpretation
Born and Heisenberg's se tion II begins with the following statement
(p. 419):
The most noti eable defe t of the original matrix me hani s onsists in the
fa t that at rst it appears to give information not about a tual phenomena,
but rather only about possible states and pro esses. It allows one to al ulate
the possible stationary states of a system; further it makes a statement about
the nature of the harmoni os illation that an manifest itself as a light wave
in a quantum jump. But it says nothing about when a given state is present, or
when a hange is to be expe ted. The reason for this is lear: matrix me hani s
deals only with losed periodi systems, and in these there are indeed no
hanges. In order to have true pro esses, as long as one remains in the domain
of matrix me hani s, one must dire t one's attention to a
part
of the system;
this is no longer losed and enters into intera tion with the rest of the system.
The question is what matrix me hani s an tell us about this.
(This is again one of the se tions originally drafted by Born.) As raised

here, the question to be addressed is how to in orporate into matrix
me hani s the (a tual) state of a system, and the time development
of su h a state. The dis ussion given in the report may give rise to
some onfusion, be ause it arguably ontains at least two, if not three,
disparate approa hes to what is a state in quantum me hani s. The rst,
ree ted in the above quotation, is the idea that a state of a system is
always a stationary state, whi h stems from Bohr's quantum theory and
whi h appears to have lived on through the BKS phase until well into the
development of matrix me hani s, indeed at least as late as Heisenberg's
dis overed independently by S hrdinger (1926d), E kart (1926 [22) and Pauli
(Pauli to Jordan, 12 April 1926, in Pauli 1979, pp. 31520). For a des ription of
the various ontributions, in luding the work of Lan zos (1926 [23), see Mehra
and Re henberg (1987, pp. 63684). Muller (1997) argues that matrix me hani s
and wave me hani s, as formulated and understood at the time, were nevertheless
inequivalent theories. For on rete examples in whi h the two theories were indeed
onsidered to yield dierent predi tions, see below p. 143 (in luding the footnote)
and the dis ussion on p. 469.
3.4 Interpretation
103
paper on resonan e (Heisenberg 1926b [35). The se ond is the idea that
the state of a system is given by its wave fun tion, but it is bound up
with the question of whether the latter should be seen as a `spread-out'
entity, a `guiding eld', a `statisti al state' or something else.a Born's
papers on ollisions (Born 1926a,b [30) an be said to ontain elements
of both these approa hes. Yet a third approa h may well be present in
the report, an approa h in whi h the notion of state would be purely an
ee tive one. Some pronoun ements by Heisenberg, in orresponden e
and in the un ertainty paper (Heisenberg 1927 [46), may support this
further (tentative) suggestion.
It seems to us that Born and Heisenberg's statements be ome learer
if one is aware of the dierent ba kgrounds to their dis ussion. A ordingly, we shall dis uss in turn, briey, various developments that appear
to have fed into their on eption of quantum me hani s, in parti ular
previous work by Born and Wiener on generalising matrix me hani s
(se tion 3.4.1), by Born on guiding elds (se tion 3.4.2), by Bohr as
well as famously by Born on ollision pro esses (again se tion 3.4.2 and
se tion 3.4.3), by Heisenberg on atoms in resonan e (se tion 3.4.4) and
by Dira on the transformation theory (se tion 3.4.5). We shall then
dis uss the treatment of interpretational issues given in the report in
se tions 3.4.6 and 3.4.7. The latter in ludes some brief omments on the
notion of a ` losed theory', whi h is prominent in some of Heisenberg's
later writings (e.g. Heisenberg 1948), and whi h appears to be used here
for the rst time. We do not laim to have settled the interpretational
issues, and will return to several of them in Part II. Overall, the `physi al
interpretation' of the report requires areful reading and assessment.
3.4.1 Matrix me hani s, Born and Wiener

As we have seen, in the old quantum theory the only states allowed for
atomi systems are stationary states, understood in terms of lassi al
orbits subje t to the quantum onditions of Bohr and Sommerfeld. The
same is true in the BKS theory, where stationary states be ome more
abstra t and are represented by the olle tion of virtual os illators orresponding to the transitions of the atom from a given energy level. In
both theories, dis ontinuous transitions between the stationary states,
so- alled quantum jumps, are assumed to o ur. Although in today's
quantum theory one usually still talks about dis ontinuous transitions,
a Cf. the next hapter.
104
these are asso iated with the ollapse postulate and with the on ept of a
measurement. The states performing these transitions are not ne essarily
stationary states, i.e. eigenstates of energy (whether one des ribes them
dynami ally in the S hrdinger pi ture or stati ally in the Heisenberg
pi ture).
This modern notion of state is strikingly absent also in matrix me hani s, as we have seen it formulated in the original papers. Indeed, the
interpretation of the theory is still ostensibly in terms of stationary states
and quantum jumps, but the formalism itself ontains only matri es,
whi h an at most be seen as a olle tive representation of all stationary
states, in the following sense. The matrix (2) in orporates the os illations
orresponding to all possible transitions of the system. Ea h matrix
element is formally analogous to a virtual os illator with frequen y ij =
(Ei Ej )/h ( orresponding to an atomi transition Ei Ej ). Therefore,
ea h row (or olumn) ontains all the frequen ies orresponding to the
transitions from (or to) a given energy level, mu h like a stationary
state in the BKS theory. As in the above quotation, the matrix an thus
be seen as the olle tive representation of all the stationary states of a
losed system.a
This analysis is further supported by examining Born and Wiener's
work, in whi h the pi ture of stationary states as the rows of the position
matrix (now the position operator) be omes even more expli it. While
the fun tions x(t) remain uninterpreted, Born and Wiener appear to
asso iate the `rows' or, rather, the ` olumns' of their operators with
(stationary) states of motion. This is lear from the following. Born and
Wiener introdu e a notion of ` olumn sum', that is, a generalisation of
the sum of the elements in the olumn of a matrix. In dis ussing their
main example, the free parti le, they then show that, for the position
operator, this generalised olumn sum q(t, W ) takes the form
s
2W
+ F (W ) ,
q(t, W ) = t
(20)
where is the mass of the parti le and F (W ) is a omplex-valued

expression independent of t. They expli itly draw the on lusion that at
least the real part of the generalised olumn sum represents a lassi al
inertial motion with the energy W .
a A somewhat similar idea appears to be expressed by Dira at the beginning
of the dis ussion after the report (p. 441), where he emphasises the parallel
between matrix me hani s and lassi al Hamiton-Ja obi theory, with the latter
also des ribing not single traje tories but whole families of traje tories.
3.4 Interpretation
105
Note that this indi ates not only that Born and Wiener asso iate the
state of a system with a olumn of the position operator. It also suggests
that, in their view, at least a limited spatio-temporal pi ture of parti le
traje tories in the absen e of intera tions is possible (analogously to the
earlier limited use of spatio-temporal pi tures in des ribing atomi states
in the absen e of transitions).
3.4.2 Born and Jordan on guiding elds, Bohr on ollisions

The early history of the guiding eld idea, in onne tion with Einstein
and with the BKS theory (in the ase of photons), and in onne tion
with de Broglie (in the ase of both photons and material parti les),
is dis ussed mainly in hapters 9 and 2, respe tively. Slater's original
intention was in fa t to have the virtual elds to be guiding elds for the
photons, whi h were to arry energy and momentum, but this aspe t
was not in orporated in the BKS theory. However, after this theory
was reje ted pre isely be ause of the results on energy and momentum
onservation in individual pro esses (detailed, for instan e, in Compton's
report), Slater's original idea was eetingly revived.
On 24 April 1925, after learning from Fran k that Bohr had in fa t
given up the BKS theory, Born sent Bohr the des ription of su h a
proposal (whi h, as he wrote, he had been working on for some weeks
with Jordan). A manus ript by Born and Jordan followed, entitled `Zur
Strahlungstheorie', whi h is found today in the Bohr ar hives. a The
proposal ombines the BKS idea of emission of waves while the atom
is in a stationary state with the emission of a light quantum during
an instantaneous quantum jump, in order to give a spa etime pi ture of
radiation. The light quantum thus follows the rear end of the wave. It an
be s attered or absorbed by other atoms (both pro esses depending on
the dipole moment of the appropriate virtual os illators in the atoms),
in the latter ase leaving a `dead' wave that has no further physi al
ee ts. Born and Jordan had applied this pi ture with some su ess
to a few simple examples, and were intending to publish the idea in
Naturwissens haften, provided Bohr or Kramers did not nd fault with
it (Bohr 1984, pp. 845 and 30810).
Bohr replied on 1 May, after re eipt of both the letter and the manus ript, riti ising the proposal, on the grounds, rst, that the proposed
me hanism did not guarantee that the traje tories of the light quanta
a Cf. Darrigol (1992, p. 253). We are espe ially grateful to Olivier Darrigol for helpful
orresponden e on this matter.
106
would oin ide with the propagation of the wave, and se ond, that the
ross se tion for the absorption ought to be a onstant in order for the
parti le number density to be proportional to the intensity of the wave.
Bohr reiterated the beliefs he had expressed to Fran k: that the oupling
between state transitions in dierent atoms ex luded a des ription that
used ans hauli h pi tures, and that he suspe ted the same on lusion
to be likely in the ase of ollision phenomena (Bohr 1984, pp. 85 and
31011).a
Bohr's work on ollisions, to whi h he alluded here, was an extension of
the BKS idea of merely statisti al onservation laws (Bohr 1925).a The
idea, as paraphrased by Born, was `to regard the eld of the parti le
passing by in the same way as the eld of a light wave; thus, it only
produ es a probability for the absorption of energy, and this [absorption
only o urs when the ollision lasts su iently long (the parti le passes
slowly)' (Born to Bohr, 15 January 1925, quoted in Bohr 1984, p. 73).b
However, the Ramsauer ee t the anomalously low ross se tion of
atoms of ertain gases for slow ele trons ould not be a ommodated
in Bohr's s heme. Bohr therefore was developing doubts about statisti al onservation laws (and further, about the feasibility altogether
of a spa etime pi ture of ollisionsc), even as he was submitting his
paper on ollisions,d that is, even before the results of the Bothe-Geiger
experiments were onrmed in April 1925. Bohr's paper was published
only after Bohr in luded an addendum in July 1925, whi h draws the
onsequen es from both the Bothe-Geiger experiments and the di ulties with the Ramsauer ee t. In the same month of July, an explanation
a See Bohr to Fran k, 21 April 1925, in Bohr (1984, pp. 35051).
a Cf. also Darrigol (1992, pp. 24951) and pp. 8993 in Stolzenburg's introdu tion
to Part I of Bohr (1984).
b Another olourful paraphrase is in Bohr to Fran k, 30 Mar h 1925: Ìf two atoms
have the possibility of settling their mutual a ount, it is, of ourse, simplest that
they do so. However, when the invoi es annot be submitted simultaneously, they
must be satised with a running a ount' (quoted in Bohr 1984, p. 74).
One of the most striking expressions of this is a passage in Bohr to Heisenberg, 18
April 1925: `Stimulated espe ially by talks with Pauli, I am for ing myself these
days with all my strength to familiarise myself with the mysti ism of nature and am
attempting to prepare myself for all eventualities, indeed even for the assumption
of a oupling of quantum pro esses in separated atoms. However, the osts of
this assumption are so great that they annot be estimated within the ordinary
spa etime des ription' (Bohr 1984, pp. 36061). For other qualms about su h
`quantum nonlo ality', f. also Jordan's habilitation le ture (1927f [62), in whi h
Jordan onsiders the idea of mi ros opi indeterminism to be omprehensible only
if the elementary random events are independent. (This le ture was translated into
English by Oppenheimer and published in Nature as Jordan (1927g).)
d The paper was re eived by Zeits hrift fr Physik on 30 Mar h 1925; on that very
day Bohr was expressing his doubts in a letter to Fran k (Bohr 1984, pp. 34850).
3.4 Interpretation
107
for the Ramsauer ee t was suggested by Elsasser (1925) in Gttingen,

on the basis of de Broglie's matter waves.
3.4.3 Born's ollision papers

The above provides a useful ba kdrop for dis ussing Born's own work
on ollisions (Born 1926a,b [30),a whi h treats ollision problems on the
basis of S hrdinger's wave me hani s.b
This work also ree ts the idea that the states of a system are stationary states undergoing transitions. Indeed, Born presents the problem
as that of in luding in matrix me hani s a des ription of the transitions
between stationary states. The ase of ollisions, say between an atom
and an ele tron, is hosen as the simplest for treating this problem (while
still leading to interesting predi tions), be ause it is natural to expe t
that in this ase the ombined system is asymptoti ally in a stationary
state for the atom and a state of uniform translational motion for the
ele tron. (Note that this is onne ted to the treatment of the free parti le
by Born and Wiener.) If one an manage to des ribe the asymptoti
behaviour of the ombined system mathemati ally, this will give on rete
indi ations as to the transitions between the initial and nal asymptoti
states. Born managed to nd the solution spe i ally by wave me hani al
methods.c
Note that Born onsiders two on eptually distin t obje ts, the wave
fun tion on the one hand and the stationary states of the atom and the
ele tron on the other, the onne tion between them being that the wave
fun tion denes a probability distribution over the stationary states.
(Note also that he reserves the word `state' only for the stationary
states.)
Born solves by perturbation methods the time-independent S hrdinger equation for the ombined system of atom and ele tron under
the ondition that asymptoti ally for z , the solution has the form
2
n (q)e z (a produ t of the n-th eigenstate of the atom with a plane
wave oming from the z -dire tion), with energy . Born's solution has
a For a modern dis ussion of ollisions, f. se tion 10.1.
b The Ramsauer ee t, however, is ex luded from Born's dis ussion (1926b [30,
footnote on p. 824). A few passages might be seen to refer to Born's ex hange
with Bohr, in parti ular the remark that, at the pri e of dropping ausality, the
usual spa etime pi ture an be maintained (1926b [30, p. 826).
Cf. Born to S hrdinger, 16 May 1927: `the simple possibility of treating with it
aperiodi pro esses ( ollisions) made me rst believe that your on eption was
superior' (quoted in Mehra and Re henberg 2000, p. 135).
108
the form:
XZ
m
x+y+z>0
nm (, , )m (q)eknm (x+y+z+) d ,
(21)
equals
where the energy orresponding to the wave number knm
Enm
= hnm + ,
(22)
the nm being the transition frequen ies of the atom. The omponents of
the superposition an thus be asso iated with various, generally inelasti ,
ollisions in whi h energy is onserved, and Born interprets |nm (, , )|2
as the probability for the atom to be in the stationary state m and the
ele tron to be s attered in the dire tion (, , ).a
But now, ru ially, sin e the initial wave fun tion orresponds to a
fully determined stationary state and inertial motion, this probability is
also the probability for a quantum jump from the given initial state to
the given nal state, i.e. a transition probability.
This idea is linked to that of a guiding eld. The link is made expli itly
at the beginning of Born's se ond paper (whi h in ludes the details of the
derivation and some quantitative predi tions). While Born judges that in
the ontext of opti s one ought to wait until the development of a proper
quantum ele trodynami s, in the ontext of the quantum me hani s of
material parti les the guiding eld idea an be applied already, using
the de Broglie-S hrdinger waves as guiding elds. The traje tories of
material parti les, however, are determined by the guiding eld merely
probabilisti ally (1926b [30, pp. 8034). In his on lusion, Born regards
the pi ture of the guiding eld as fundamentally indeterministi . A
deterministi ompletion, if possible, would not have any pra ti al use.
Born also expresses the hope that the `laws of motion for light quanta'
will nd a similar treatment to the one given for ele trons, and refers
to the di ulties `so far' of pursuing a guiding eld approa h in opti s
(pp. 8267).b
a A statisti al interpretation for the modulus squared of the oe ients of the wave
fun tion in the energy basis was introdu ed also by Dira (1926 [37), at the
same time as and presumably independently of Born ( f. Darrigol 1992, p. 333).
Note further that, even though it may be tempting to assume that ea h traje tory
pro eeds from the s attering entre, stri tly speaking to ea h stationary state of
the free parti le orresponds a whole family of inertial traje tories. (Similarly, in
the ase treated by Born and Wiener, ea h generalised olumn sum asso iated with
a stationary state orresponds to two dierent inertial traje tories, depending on
the sign of the square root in (20).)
b Born's views on quantum me hani s from this period are also presented in Born
(1927), an expanded version (published Mar h 1927) of a talk given by Born in
August 1926.
3.4 Interpretation
109
3.4.4 Heisenberg on energy u tuations

As dis ussed in the next hapter, Heisenberg in parti ular among matrix
physi ists was opposed to S hrdinger's attempt to re ast and reinterpret quantum theory on the basis of ontinuous wave fun tions. S hrdinger's wave fun tions were meant from the start as des riptions of
individual states of a physi al system. Even though in general they
are abstra t fun tions (on onguration spa e), they an provide a pi ture of Bohr's stationary states. Furthermore, the solution of the timedependent S hrdinger equation appears to provide a generalisation of
the state of a system as it evolves in time.
Heisenberg appears to have been disturbed initially also by Born's
use of S hrdinger's theory in the treatment of ollisions, an attitude
ree ted in parti ular in his orresponden e with Pauli.a In this onne tion, the fa t that Pauli in his letter of 19 O tober (Pauli 1979,
pp. 3409) was in ee t able to sket h how one ould reinterpret
Born's results in terms of matrix elements, must have been of parti ular
signi an e: `Your al ulations have given me again great hope, be ause
they show that Born's somewhat dogmati viewpoint of the probability
waves is only one of many possible s hemes'.b
A few days later, Heisenberg sent Pauli the manus ript of his paper on
u tuation phenomena (Heisenberg 1926b [35), in whi h he developed
onsiderations similar to Pauli's in the ontext of a hara teristi example, that of two atoms in resonan e. By fo ussing on the subsystems of
a losed system, Heisenberg was able to derive expressions for (transition) probabilities within matrix me hani s proper, without having to
introdu e the wave fun tion as an external aid. A very similar result was
derived at the same time by Jordan (1927 [36), using two systems with
a single energy dieren e in ommon.
We shall now sket h Heisenberg's reasoning. We adapt the presentation of the argument given by Heisenberg in The Physi al Prin iples
of the Quantum Theory (Heisenberg 1930b, pp. 1427),c whi h is more
general than the one given in the paper and learer than the one given
in Born and Heisenberg's report.
Take two systems, 1 and 2, that are in resonan e. Consider, to begin with, that the frequen y of the transition n1 m1 in system 1
a Òne senten e [of Born's paper reminded me vividly of a hapter in the Christian
reed: An ele tron is a plane wave... ' (Heisenberg to Pauli 28 July 1926, in Pauli
1979, p. 338, original emphasis).
b Heisenberg to Pauli, 28 O tober 1926, in Pauli (1979, p. 350).
This very remarkable book is an expanded English edition of Heisenberg (1930a).
110
orresponds to exa tly one transition frequen y in system 2, say

(1) (n1 m1 ) = (2) (n2 m2 ) .
(23)
If the systems are un oupled, the ombined system has the degenerate
eigenvalue of energy
(1)
(2)
.
+ Wn(2)
= Wm
+ Wm
Wn(1)
2
1
2
1
(24)
If we ouple weakly the two systems, the degenera y will be lifted. Let
us label the new eigenstates of the ombined system by a and b. We
an now onsider the matrix S that transforms the basis of eigenstates
of energy of the oupled system to the (produ t) basis of eigenstates
of energy of the un oupled systems. In parti ular we an onsider the
submatrix

Sn1 m2 ,a Sn1 m2 ,b
.
(25)
Sn2 m1 ,a Sn2 m1 ,b
Choose one of the stationary states of the ombined system, say a.
If the ombined system is in the state a, what an one say about the
energy of the subsystems, for instan e H (1) ? Heisenberg's answer, in the
terminology and notation of his book (1930b), is that in the state a,
the time average of H (1) (whi h is no longer a diagonal matrix, thus no
longer time-independent), or of any fun tion f (H (1) ), is
(1)
)|Sm1 n2 ,a |2 .
)|Sn1 m2 ,a |2 + f (Wm
f (H (1) ) = f (Wn(1)
1
1
(26)
Sin e f is arbitrary, Heisenberg on ludes that |Sn1 m2 ,a |2 is the probability that the state n1 has remained the same (and the state m2 has
remained the same), and |Sm1 n2 ,a |2 is the probability that the state n1
has jumped to the state m1 (and m2 has jumped to n2 ).
The asso iated transfer of energy between the systems appears to be
instantaneous, in that a quantum jump in one system (from a higher
to a lower energy level) is a ompanied by a orresponding jump in the
other system (from a lower to a higher energy level). The paper merely
mentions, without elaborating further, that a light quantum (or better
a `sound quantum') is ex hanged over and over again between the two
systems. The pi ture thus avoids the non- onservation of energy of the
BKS theory, at the pri e of what appears to be an expli it orrelation
at a distan e.
In modern terms, Heisenberg has al ulated the expe tation value of
3.4 Interpretation
111
the observable f (H (1) ) 1 in the state a:

ha|f (H (1) ) 1|ai =
ha |n1 m2 ihn1 m2 | f (H (1) ) 1 |n1 m2 ihn1 m2 | ai+
ha |m1 n2 ihm1 n2 | f (H (1) ) 1 |m1 n2 ihm1 n2 | ai =
(27)
hn1 |f (H (1) )|n1 i|hn1 m2 |ai|2 +
hm1 |f (H (1) )|m1 i|hm1 n2 |ai|2 .
Note, however, that rather than fo ussing on the idea that

|Sn1 m2 ,a |2 = |hn1 m2 |ai|2
(28)
is a onditional probability (in fa t what we would today all a transition

probability), Heisenberg is still fo ussing on the transitions between the
stationary states of the subsystems, as in Born's work on ollisions.
3.4.5 Transformation theory

Less than three weeks after ompleting his draft on u tuation phenomena, we nd Heisenberg reporting to Pauli about Dira 's transformation
theory, whi h generalises pre isely the formal expression of a onditional
probability given by Heisenberg in terms of the transformation matrix:
`Here [in Copenhagen we have also been thinking more about the question of the meaning of the transformation fun tion S and Dira has
a hieved an extraordinarily broad generalisation of this assumption from
my note on u tuations' (Pauli 1979, p. 357, original emphasis).
Dira indeed presents his results in his paper, signi antly titled `The
physi al interpretation of quantum dynami s' (1927a [38), as a generalisation of Heisenberg's approa h. The main goal of the paper is the
following.a
Take any pair of onjugate matrix quantities and , and any ` onstant of integration' g(, ).b Given a value of as a -number, nd
the fra tion of -spa e for whi h g lies between any two numeri al
values. If is assumed to be distributed uniformly,c this result will
yield the frequen y of the given values of g in an ensemble of systems.
a In our presentation we shall partly follow the analysis by Darrigol (1992, pp. 33745).
b This term is meant to in lude any value of a dynami al quantity at a spe ied
time t = t0 (Dira 1927 [38, p. 623, footnote).
This assumption may sound strange, espe ially sin e is assumed to have a denite
value and and are anoni ally onjugate. Cf. however Dira 's remarks on
interpretation below.
112
Equivalently, we an state Dira 's goal as that of nding the expe tation
value (or more pre isely, the -average) of any xed-time observable g ,
given a ertain value of .
The main part of the paper is devoted to developing a `transformation
theory' that will allow Dira to write the quantity g not in the usual
energy representation but in an arbitrary -representation. Dira then
suggests taking the -averaged value of g , for taking the -number
value , as given by the diagonal element g of g in this representation. This is in fa t a natural if èxtremely broad' generalisation, to
an arbitrary pair of onjugate quantities (, ), of the assumption that
the diagonal elements of a matrix in the energy representation (su h
as Heisenberg's f (H (1) ) from the previous se tion) are time averages,
although the justi ation Dira gives for this assumption is merely that
`the diagonal elements .... ertainly would [determine the average values
in the limiting ase of large quantum numbers' (Dira 1927 [38, p. 637).
Dira writes the elements of a general transformation matrix between
two omplete sets of variables and (whether dis rete or ontinuous)
as ( / ) (what we would now write h | i), so that the matrix elements
transform as
Z
g = ( / )g ( / )d d
(29)
(Dira 's notation is meant to in lude the possibility of dis rete sums).
His main analyti tool is the manipulation of -fun tions and their
derivatives. Dira shows in parti ular that for the quantity itself,
= ( ) ,
(30)
and that for the quantity anoni ally onjugate to ,

= i~ ( ) .
(31)
In a mixed representation, one has

= ( / ) ,
(32)
and
= i~

( / ) ,

(33)
from whi h follows

g = g , i~ ( / )
for arbitrary g = g(, ).
(34)
3.4 Interpretation
113
Choosing su h that g is diagonal in the -representation, one has

g = g ( / ) ,
(35)
where the g are the eigenvalues of g . Therefore, (34) be omes a dierential equation for the ( / ) (seen as fun tions of ), whi h generalises
the time-independent S hrdinger equation.a
On e this equation is solved, one ould obtain the desired g from
the g by the appropriate transformation. The way Dira states his
nal result, however, is by onsidering the matrix (gg ). The numeri al
fun tion (g g ), when integrated,
Z b
(36)
(g g )dg ,
a
yields the hara teristi fun tion of the set a < g < b. Therefore, in
Dira 's proposed interpretation, the diagonal elements of the orresponding matrix in the -representation yield, for ea h value = , the
fra tion of the -spa e for whi h a < g < b. That is, if is assumed to
be distributed uniformly, these diagonal elements yield the onditional
probability for a < g < b given = . Thus, the diagonal elements
of the matrix (g g ) yield the orresponding onditional probability
density for g given = . But now, e.g. sin e for any fun tion f (g) one
has
Z
f (g) = ( /g )f (g )(g / )dg ,
(37)
one has in parti ular
Z
(g g ) = ( /g )(g g )(g / )dg = ( /g )(g / ) .
(38)
Therefore, the onditional probability density for g given = is equal

to |(g / )|2 , a result that Dira illustrates by dis ussing Heisenberg's
example of transition probabilities in resonant atoms and Born's ollision
problem.
In parallel with Dira 's development of transformation theory, Jordan
(1927b, [39) also arrived at a similar theory, following on dire tly from
his paper on quantum jumps (1927a [36) and from his earlier work
on anoni al transformations (1926a,b), to whi h Dira also makes an
expli it onne tion. Although Born and Heisenberg state in the report
(p. 429) that the two methods are equivalent, Darrigol (1992, pp. 3434)
a Dira also gives a generalisation of the time-dependent S hrdinger equation.
114
points to some subtle dieren es, whi h are also related to Dira 's riti ism in the general dis ussion of Jordan's introdu tion of anti ommuting
elds (p. 501). Dira also notes that his theory generalises the work by
Lan zos (1926 [23). The development of transformation theory from
the idea of anoni al transformations led further towards the realisation
that quantum me hani al operators a t on a Hilbert spa e and that the
natural transformations are in fa t unitary.a
Dira on ludes his paper with an intriguing suggestion of
.... a point of view for regarding quantum phenomena rather dierent from
the usual ones. One an suppose that the initial state of a system determines
denitely the state of the system at any subsequent time. If, however, one
des ribes the state of the system at an arbitrary time by giving numeri al
values to the o-ordinates and momenta, then one annot a tually set up a
one-one orresponden e between the values of these o-ordinates and momenta
initially and their values at a subsequent time. All the same one an obtain
a good deal of information (of the nature of averages) about the values at
the subsequent time onsidered as fun tions of the initial values. The notion
of probabilities does not enter into the ultimate des ription of me hani al
pro esses: only when one is given some information that involves a probability
(e.g., that all points in
-spa e
are equally probable for representing the
system) an one dedu e results that involve probabilities. (Dira 1927a [38,
p. 641)
Here Dira does not impute indeterminism to nature itself (the matrix
equations are after all deterministi ), but instead apparently identies
the sour e of the statisti al element in the hoi e of probabilisti initial
data.a
A ording to Heisenberg, however, and despite the generality of the
results and the èxtraordinary progress' obtained (Pauli 1979, p. 358),
Dira 's transformation theory did not resolve the question of the meaning of quantum me hani s. As Darrigol (1992, p. 344) emphasises, there
is no notion of state ve tor in Dira 's paper (the well-known bras and kets
do not appear yet). C-number values, and probability distributions over
a See La ki (2004), who gives details also of London's (1926a,b) ontributions to
this development. Note also the onne tion between transformation theory and
the work on the èquivalen e' between matrix me hani s and wave me hani s.
a This should be ompared to the general dis ussion, in whi h Dira (a) talks
of àn irrevo able hoi e of nature' (p. 494) in relation to the out omes of an
experiment, (b) uses expli itly (perhaps for the rst time) the notion of the state
ve tor, when he arms that `[a ording to quantum me hani s the state of the
world at any time is des ribable by a wave fun tion , whi h normally varies
a ording to a ausal law, so that its initial value determines its value at any later
time' (p. 493), and in whi h ( ) he des ribes the initial data taken for quantum
me hani al al ulations as des ribing à ts of freewill', namely `the disturban es
that an experimenter applies to a system to observe it' (p. 493); see also se tion 8.2.
3.4 Interpretation
115
-number values, now refer to arbitrary quantities or pairs of quantities.

As Heisenberg wrote: `there are too many -numbers in all our utteran es
used to des ribe a fa t' (Pauli 1979, p. 359). Cru ially, however, the
energy variable and the stationary states no longer played a privileged
role.
3.4.6 Development of the `statisti al view' in the report

In the report, Born and Heisenberg appear to understand Born's ollision
papers (1926a,b [30) on the one hand and the papers by Heisenberg
(1926b [35) and Jordan (1927a [36) on the other broadly in the same
way, as seeking to obtain ìnformation .... about a tual phenomena', by
`dire t[ing one's attention to a part of the system' (p. 419 of the report).
And indeed, by onsidering oupled systems all of these papers manage
to derive quantitative expressions for the probabilities of quantum jumps
between energy eigenstates.
Heisenberg's setting is the one hosen in the report, and sin e Heisenberg's treatment of intera ting systems does not use the formalism
of wave me hani s, this hoi e may be intended to make the point that
matrix me hani s an indeed a ount for time-dependent phenomena
without the aid of wave me hani s.
The form of Heisenberg's result (26) as given in the report is in terms
of the expe ted deviation of the value of energy from a given initial value,
for instan e n1 :a
)=
fn1 =f (H (1) ) f (Wn(1)
1
)}|Sn1 m2 ,a |2 +
) f (Wn(1)
{f (Wn(1)
1
1
(1)
)
{f (Wm
1
)}|Sm1 n2 ,a |2
f (Wn(1)
1
(39)
If we write n1 m1 for the probability of the transition n1 m1 in

system 1, this be omes equation (20) of the report, ex ept that Born
and Heisenberg label the matrix elements Sn1 m2 ,a and Sm1 n2 ,a , respe tively, by the transitions n1 m2 n1 m2 and n1 m2 m1 n2 , that is,
as Sn1 m2 ,n1 m2 and Sn1 m2 ,m1 n2 , omitting referen e to the state a.b One
further dieren e between our des ription above and the one given in
the report is that Born and Heisenberg are treating the ase in whi h
the transition n1 m1 in system 1 may resonate with more than one
a Note that |Sn1 m2 ,a |2 + |Sm1 n2 ,a |2 = 1, be ause a is normalised.
b In Heisenberg's paper, the matrix (25) orresponds to a 45-degree rotation, so the
probabilities are independent of the hoi e of a or b.
116
transition in system 2. In this ase, the total transition probability n1 m1

is no longer equal to |Sn1 m2 ,m1 n2 |2 but to
X
|Sn1 m2 ,m1 n2 |2 .
n1 m1 =
(40)
m2 n2
Thus we also have equation (21) of the report.

It is only after this matrix me hani al dis ussion that Born and Heisenberg introdu e the time-dependent S hrdinger equation, as a more
` onvenient' formalism for `thinking of the system under onsideration
as oupled to another one and negle ting the rea tion on the latter'
(p. 421). This suggests that Born and Heisenberg may onsider the timedependent S hrdinger equation only as an ee tive des ription. This
impression is reinfor ed by omparing with Heisenberg's book (1930b,
pp. 14850) where, after the above derivation, Heisenberg ontinues with
a more general derivation of time-dependent probabilities, whi h he then
relates to the usual time-dependent S hrdinger equation. Nevertheless,
Born and Heisenberg use the wave fun tion throughout the ensuing
dis ussion of probabilities, noting that this formalism `leads to a further
development of the statisti al view', by whi h they mean in parti ular
the idea of interferen e of probabilities.a
First of all Born and Heisenberg relate the wave fun tion to probabilities. They take the time-dependent transformation matrix S(t) given
by the unitary evolution. For the oe ients of the wave fun tion in the
energy basis one has (equation (25) in the report):
X
cn (t) =
(41)
Snm (t)cm (0) .
m
If now all cm (0) ex ept one (say, ck (0)) are zero, from the assumption
that a system is always in a stationary state it is natural to on lude
that the |Snk (t)|2 are the probabilities for transitions to the respe tive
energy states (`transition probabilties'), and the |cn (t)|2 are the resulting
probabilities for the stationary states (`state probabilities'). In support
of this interpretation (whi h is the same as in Born's ollision papers),
the report quotes in parti ular Born's paper on the adiabati prin iple
(1926 [34), that is, Born's proof that in the adiabati ase the transition
probabilities between dierent states tend to zero, in a ordan e with
Ehrenfest's (1917) prin iple.b
a On these matters f. also se tion 8.1.
b Ehrenfest had the idea that sin e quantised variables annot hange by arbitrarily
small amounts, they should remain onstant under adiabati perturbations. This
3.4 Interpretation
117
Then Born and Heisenberg ome to dis ussing interferen e.a This is
also the rst time in the presentation of the physi al interpretation of
the theory that measurements enter the pi ture. Born and Heisenberg
note that if ck is not the only non-zero oe ient at t = 0, then (41)
does not imply
X
|cn (t)|2 =
(42)
|Snm (t)|2 |cm (0)|2 ,
m
but that instead one has
|cn (t)|2 = |
Snm (t)cm (0)|2 .
(43)
The passage immediately following this is both remarkable and, in our

opinion, very signi ant (p. 423):a
it should be noted that this ìnterferen e' does not represent a ontradi tion
with the rules of the probability al ulus, that is, with the assumption that
2
the |Snk | are quite usual probabilities. In fa t, the omposition rule [(42)
follows from the on ept of probability for the problem treated here when and
|cn |2 of the atoms in
only when the relative number, that is, the probability

the state
phases
n,
has been
established
beforehand
experimentally.
In this ase the
are unknown in prin iple, so that [(43) then naturally goes over to
[(42) .... .
(The passage ends with a referen e to Heisenberg's un ertainty paper.)

How do Born and Heisenberg propose to resolve this apparent ontradi tion?
It would make sense to say that the |Snk (t)|2 annot be taken in
general as probabilities for quantum jumps, be ause the derivation of
|Snk (t)|2 as a transition probability works only in a spe ial ase (that is,
presumably, if the energy at t = 0 has in fa t been measured). On this
reading, there might on eivably exist some quite dierent transition
probabilities, whi h lie outside the s ope of quantum me hani s and are
led him to formulate the prin iple stating that the lassi al variables to be
quantised are the adiabati invariants of the system. Cf. Born (1969, pp. 113 .).
a Born and Heisenberg give redit to Pauli for the notion of interferen e of
probabilities (p. 423). Note that Pauli ontributed signi antly to the development
of the `statisti al view', albeit mainly in orresponden e and dis ussion. Contrary
to what is ommonly assumed, the idea of a probability density for position is
not ontained in Born's ollision papers, but appears in fa t in Pauli's letter to
Heisenberg of 19 O tober 1926 (Pauli 1979, p. 3409), together with the idea of
the orresponding momentum density, and in print in a footnote of Pauli's paper
on gas degenera y and paramagnetism (1927 [44). (See also Heisenberg to Pauli,
28 O tober 1926, in Pauli 1979, pp. 34052.) Jordan (1927b [39), in his se ond
paper on the transformation theory, even gives redit to Pauli for the introdu tion
of arbitrary transition probabilities and amplitudes.
a For further dis ussion, see se tion 6.1.2.
118
presumbly of no pra ti al value (like a deterministi ompletion in the

ase of ollision pro esses).a
However, this reading does not seem to t what Born and Heisenberg a tually say. Their suggestion seems to be that the |Snk (t)|2 are
indeed always transition probabilities, but that the |cn |2 are not always
state probabilities: the |cn |2 will be state probabilities if and only if the
energies have been measured (non-sele tively). This seems analogous to
Heisenberg's (1927 [46, p. 197) idea in the un ertainty paper that the
`law of ausality' is inappli able be ause it is impossible in prin iple to
know the present with su ient a ura y (i.e. the ante edent of the law
of ausality fails).a
Born and Heisenberg swiftly move on to generalising the dis ussion to
the ase of arbitrary observables, on the basis of Dira 's and Jordan's
transformation theory (Dira 1927a [38, Jordan 1927b, [39). They
introdu e the interpretation of ||2 as a position density, and onsider
in parti ular the density |(q , W )|2 dened by the stationary state
(q , W ) with the energy W , or in modern notation,
|(q , W )|2 = |W (q )|2 = |hq |W i|2 .
(44)
This is immediately generalised to arbitrary pairs of observables q and

Q with values q and Q :
|(q , Q )|2 = |hq |Q i|2 .
(45)
Born and Heisenberg all this a `relative state probability', reserving the
term `transition probability' for the ase of a single observable evolving
in time (or depending on some external parameter), always with the proviso that in general one should expe t interferen e of the orresponding
`transition amplitudes'.
Note that the physi al interpretation of this generalisation makes sense
only if one takes over from the above the idea that probabilities su h as
||2 are well-dened only upon measurement, or more pre isely, that a a That su h probabilities an be dened (albeit non-uniquely), leading to
well-dened sto hasti pro esses for the quantum jumps, was shown expli itly
by Bell (1984).
a It appears not to be well known that the last 20 pages of the original types ript
of Heisenberg's un ertainty paper are ontained in AHQP, mis atalogued as
an 'in omplete and unpublished paper (pp. 12-31)' by Kramers.7 The types ript
ontains slight textual variants (as ompared with the published version) and
manus ript orre tions in what appears to be Heisenberg's hand, but does not
in lude the famous addendum in proof in response to Bohr's riti ism ( f. p. 146
below). On Heisenberg's treatment of the `law of ausality', see also Beller (1999,
pp. 11013).
3.4 Interpretation
119
tual frequen ies upon measurement will be given by the expression ||2 .
Born and Heisenberg's terminology, however, is somewhat ambiguous.a
A major on eptual shift appears to be taking pla e, whi h may
be easy to miss. Do quantum jumps still o ur whenever two systems
intera t, or do they now o ur only between measurements? Indeed, are
systems always in stationary states, as has been expli itly assumed until
now, or only when the energy is measured? Heisenberg's un ertainty
paper (on pp. 1901), as well as the orresponden e with Pauli (Heisenberg to Pauli, 23 February 1927, in Pauli 1979, pp. 37682) both
mention expli itly the loss of a privileged status for stationary states.
It seems that, even though we are not expli itly told so, the pi ture of
quantum jumps (that is, of probabilisti transitions between possessed
values of energy) is shifting to that of probabilisti transitions from one
measurement to the next.
The idea that frequen ies are well-dened only upon measurement
appears to play the same role as von Neumann's proje tion postulate.
As dis ussed in more detail in hapter 6, however, it is far from lear
whether that is what Born and Heisenberg have in mind. A similar notion
is introdu ed in Heisenberg's un ertainty paper (1927 [46, p. 186), but
again in terms that are `somewhat mysti al' (Pauli to Bohr, 17 O tober
1927, in Pauli 1979, p. 411). It appears expli itly in the pro eedings
only in the general dis ussion, in Born's main ontribution (p. 482) and
in the intriguing ex hange between Dira and Heisenberg (pp. 492 .).
In hapter 6 and se tion 8.1, we shall return to Born and Heisenberg's
view of interferen e and to the question of whether, a ording to them,
the ollapse of the wave fun tion and the time-dependent S hrdinger
evolution are at all fundamental pro esses.
3.4.7 Justi ation and overall on lusions

The following se tion III (drafted by Heisenberg) presents Jordan's axiomati formulation of quantum me hani s (Jordan 1927b, [39),a and
justies the ne essity of a statisti al view in the ontext of Heisenberg's
notion of un ertainty (Heisenberg 1927 [46). Born and Heisenberg argue
as follows. Even in lassi al me hani s, if ertain quantities (for instan e
a In one paragraph they refer to `the probability that for given energy W the
oordinate q is in some given element dq ', whereas in the next they refer to `the
probability, given q , to nd the value of Q in dQ ' (p. 426; itali s added).
a This formulation, whi h is how Jordan presents his transformation theory, was
expli itly intended as a generalisation of the formalisms of matrix me hani s, wave
me hani s, q-number theory and of Born and Wiener's original operator formalism.
120
the phases of the motion) were known only with a ertain impre ision,
the future evolution of the system would be only statisti ally onstrained. Now, the un ertainty relations prevent one from determining
the values of all physi al quantities, providing a fundamental limit of
pre ision. In addition, quantum me hani s pres ibes dierent laws for
the time evolution of the statisti al onstraints. Impre ise initial onditions an be des ribed by hoosing ertain `probability fun tions' (this
is the losest Born and Heisenberg ome to dis ussing the `redu tion
of the wave pa ket' as presented in the un ertainty paper), and `the
quantum me hani al laws determine the hange (wave-like propagation)
of these probability fun tions' (p. 431). Born and Heisenberg laim
that dis ussion of the ases in whi h these laws oin ide to a very
good approximation with the lassi al evolution of a probability density
justies the lassi al treatment of - and -parti le traje tories in a
loud hamber.a They thus maintain that the statisti al element in the
theory an be re on iled with ma ros opi determinism.b
Se tion III arguably addresses the dual task set in the introdu tion of
ensuring that quantum me hani s is ` onsistent in itself' and of showing
that quantum me hani s an be taken to `predi t unambiguously the
results for all experiments on eivable in its domain' (p. 408). This task
appears to be related to two on eptual desiderata, that the theory be
intuitive (ans hauli h) and losed (abges hlossen). These are tou hed
upon briey in the report, espe ially in the introdu tion and on lusion,
but are important both in the debate with S hrdinger (see se tion 4.6)
and in some of Heisenberg's later writings (in parti ular, Heisenberg
1948). The report appears to be the rst instan e in whi h Heisenberg
uses the on ept of a losed theory.a
As dened in the report, a losed theory is one that has a hieved
a Born and Heisenberg's remarks about dierent laws of propagation of the
probabilities may refer to the onditions under whi h (43) redu es to (42), whi h
would arguably be an early example of de oheren e onsiderations. However, the
remark is too brief, and Born and Heisenberg may be merely omparing the
spreading of the quantum probabilities with that of a Liouville distribution. For a
modern treatment of the latter omparison, see Ballentine (2003).
b For further dis ussion of these issues see se tions 6.2.1 and 6.4.
a Compare also S heibe (1993) on the on ept of losed theories in Heisenberg's
thought. The origin of this on ept has also been tra ed to two earlier papers
(Heisenberg 1926a [28 and 1926 [60); see for instan e Chevalley (1988). However,
in the rst paper there is no mention of losed theories, only of losed systems of
terms (symmetri and antisymmetri ), a point also repeated in the se ond paper.
In the latter, Heisenberg mentions the need to introdu e equations for the matrix
variables in order to obtain a ` losed theory', but this does not seem to be the
same use of the term as in the report. We wish to thank also Gregor S hiemann
for orresponden e and referen es on this topi .
3.4 Interpretation
121
a denitive form, and is no longer liable to modi ation, either in its

mathemati al formulation or in its physi al meaning. This is made more
pre ise in later presentations (e.g. Heisenberg 1948), in whi h Heisenberg
in ludes the appli ability of the on epts of a theory in the analysis. All
losed theories possess a spe i domain of appli ation, within whi h
they are and will always remain orre t. Indeed, their on epts always
remain part of the s ienti language and are onstitutive of our physi al
understanding of the world. In the report, quantum me hani s (without
the in lusion of ele trodynami s) is indeed taken to be a losed theory,a
so that dierent assumptions about the physi al meaning of quantum
me hani s (su h as S hrdinger's idea of taking ||2 to be a harge
densityb ), would lead to ontradi tions with experien e. Thus, the report
ends on a note of utmost onden e.
a Heisenberg (1948) lists Newtonian me hani s, Maxwellian ele trodynami s

and spe ial-relativisti physi s, thermodynami s, and nonrelativisti quantum
me hani s as the four main examples of su h theories.
b Des ribed in more detail in se tion 4.4.
122
Notes to pp. 93118
Ar hival notes
1 Cf. also Lorentz to Ehrenfest, 4 July 1927, AHQP-EHR-23 (in Dut h).
2 This and the following remarks are based on Born to Lorentz, 23 June
1927, AHQP-LTZ-11 (in German).
3 Heisenberg to Lorentz, 23 June 1927, AHQP-LTZ-12 (in German).
4 Born to Lorentz, 29 August 1927, AHQP-LTZ-11 (in German).
5 Heisenberg to Lorentz, 24 August 1927, AHQP-LTZ-12 (in German).
6 Heisenberg to Einstein, 18 February 1926, AEA 12-172.00 (in German).
7 AHQP-28 (H. A. Kramers, notes and drafts 192652), se tion 6.
4
S hrdinger's work on wave me hani s in 1926 appears to have been driven by the idea that one ould give a purely wave-theoreti al
des ription of matter. Key elements in this pi ture were the idea of
parti les as wave pa kets (se tion 4.3) and the possible impli ations for
the problem of radiation (se tion 4.4). This pure wave theory, in ontrast
to de Broglie's work, did away with the idea of point parti les altogether
(se tion 4.5). The main oni t, however, was between S hrdinger and
the proponents of quantum me hani s (se tion 4.6), both in its form
at the time of S hrdinger's papers and in its further developments as
sket hed in the previous hapter.
For referen e, we provide a brief hronology of S hrdinger's writings
relating to wave me hani s up to the Solvay onferen e:
Paper on Einstein's gas theory, submitted 15 De ember 1925, published 1 Mar h 1926 (S hrdinger 1926a).
First paper on quantisation, submitted 27 January 1926, addendum
in proof 28 February 1926, published 13 Mar h 1926 (S hrdinger
1926b).
Se ond paper on quantisation, submitted 23 February 1926, published
6 April 1926 (S hrdinger 1926 ).
Paper on the relation between wave and matrix me hani s (èquivalen e paper'), submitted 18 Mar h 1926, published 4 May 1926
(S hrdinger 1926d).
Paper on mi ro- and ma rome hani s ( oherent states for the harmoni os illator), published 9 July 1926 (S hrdinger 1926e).
Third paper on quantisation, submitted 10 May 1926, published 13
July 1926 (S hrdinger 1926f).
123
124
Fourth paper on quantisation, submitted 21 June 1926, published 5

September 1926 (S hrdinger 1926g).
Review paper in English for the Physi al Review, submitted 3 September 1926, published De ember 1926 (S hrdinger 1926h).
Prefa e to the rst edition of Abhandlungen zur Wellenme hanik, dated November 1926 (S hrdinger 1926i).
Paper on the Compton ee t in wave me hani s, submitted 30 November 1926, published 10 January 1927 (S hrdinger 1927a).
Paper on the energy-momentum tensor, submitted 10 De ember 1926,
published 10 January 1927 (S hrdinger 1927b).
Paper on energy ex hange in wave me hani s, submitted 10 June 1927,
published 9 August 1927 (S hrdinger 1927 ).
We shall now dis uss the above points in turn, after a brief dis ussion
of the planning of S hrdinger's report for the onferen e (se tion 4.1)
and a summary of the report itself (se tion 4.2).
4.1 Planning of S hrdinger's report

As reported in hapter 1, the s ienti ommittee of the Solvay institute
met in Brussels on 1 and 2 April 1926 to plan the fth Solvay onferen e.
Lorentz had asked Ehrenfest to suggest some further names of possible
parti ipants, and it is in Ehrenfest's letter of 30 Mar h1 that S hrdinger's name is rst mentioned in onne tion with the onferen e.a On
this o asion, Ehrenfest suggested S hrdinger on the basis of a paper in
whi h S hrdinger proposed an expression for the broadening of spe tral
lines due to the Doppler ee t, and whi h applied the onservation
laws to phenomena involving single light quanta (S hrdinger 1922).b
Evidently, neither Lorentz nor Ehrenfest were yet aware of S hrdinger's
work on wave me hani s.
In the meantime, S hrdinger sent to Lorentz the proof sheets of
his rst two papers on quantisation, also on 30 Mar h,2 thus initiating his well-known orresponden e with Lorentz on wave me hani s.c
a S hrdinger had already been a parti ipant in the fourth Solvay onferen e, though
not a speaker; f. Moore (1989, pp. 1578).
b As S hrdinger points out, the al ulated broadening of the spe tral lines is small
ompared to that expe ted on the basis of the thermal agitation of the radiating
gas, otherwise the ee t ould be used as a test of the light quantum hypothesis.
Most of this orresponden e is translated in Przibram (1967). In the letter of
30 Mar h, not in luded there, S hrdinger suggests reading the se ond paper on
quantisation before the rst, whi h should be seen rather an as example of an
appli ation. Also, he writes that the variational prin iple of the rst paper is given
4.1 Planning of S hrdinger's report
125
A ordingly, already in a report of 8 April 1926 to the administrative

ommission,3 S hrdinger is listed as a possible substitute for Heisenberg for a le ture on the àdaptation of the foundations of dynami s to
the quantum theory'. (It is unlikely, however, that the papers rea hed
Lorentz before the meeting in Brussels.a ) In January 1927 then, as most
of the other parti ipants, S hrdinger was invited to the fth Solvay
onferen e.5
A few weeks later, S hrdinger had the opportunity to dis uss personally with Lorentz the plans for the report ùnder the beautiful palms of
Pasadena', as he re alls in a letter of June 1927. In the same letter, we
nd a useful sket h of the theme and fo us of S hrdinger's report; we also gather that S hrdinger was wary of the potential for a onfrontation
in Brussels:6
.... I nurtured the quiet hope you would yet return to your rst plan and entrust
only Messrs [de Broglie and Heisenberg with reports on the new me hani s.
But now you have de ided otherwise and I will of ourse happily perform my
duty.
Yet I fear that the `matri ians' (as Mr Ehrenfest used to say) will feel
disadvantaged. Should it ome to dierent views, whi h might after all urge on
the ommittee the wish to limit the reports to two, you know, dear Professor,
that I shall always happily remit my harge into your hands.
A ording to S hrdinger's sket h, the report is to stress points of prin iple, rather than the (by then many) appli ations of the theory. First of
all, one has to distinguish learly between two wave-me hani al theories:
a theory of waves in spa e and time (whi h however runs into di ulties
espe ially with the many-ele tron problem), and the highly su essful
theory of waves in onguration spa e (whi h however is not relativisti ).
A di ulty of prin iple to be dis ussed in the ontext of the spa etime theory is the possibility of developing an intera ting theory, whi h
seems to require distinguishing between the elds generated by dierent
parti les, and whether this an be done in a sensible way, or perhaps
be avoided.a In the ontext of the onguration-spa e theory, the main
question is how to interpret the wave fun tion. S hrdinger mentions
a sensible formulation only in the addendum in proof. Finally, he mentions the
paper on Einstein's gas theory in the Physikalis he Zeits hrift (1926a) as a kind
of preparatory work. Note that Lorentz on 27 May thanks S hrdinger for the proof
sheets of three papers rather than the two mentioned in S hrdinger's letter. This
third paper is learly the equivalen e paper (1926d), and was presumably sent
separately ( f. Przibram 1967, pp. 43 and 556).
a Lorentz had written to Ehrenfest on 29 Mar h from Paris, where he had another
meeting, and appears to have travelled to Brussels dire tly from there.4
a Cf. p. 460 of the report.
126
the widespread view that the wave des ribes only ensembles, as well as
his own `preferred interpretation as a real des ription of the individual
system, whi h thereby be omes a kind of mollus '. In the letter (but
not in the report), he is expli it about some of his misgivings about the
ensemble view (as well as about the di ulties with his own preferred
understanding, namely the `failure of the ele trons to stay together
and similar'). Indeed, he points out that the S hrdinger equation is
time-symmetri (if one in ludes omplex onjugation), while experien e
tea hes us that the statisti al behaviour of ensembles annot be des ribed time-symmetri ally. Also, insisten e on a statisti al interpretation
leads to `mysti al' al ulations with amplitudes and thus to problems
with the laws of probability.7
4.2 Summary of the report

The eventual form of S hrdinger's report follows roughly the sket h
given above, with an introdu tion, followed by three main se tions,
respe tively on the onguration-spa e theory, on the spa etime theory
and on the many-ele tron problem.
Introdu tion. S hrdinger draws the distin tion between the spa etime
theory (four-dimensional) and the onguration-spa e theory (multi-dimensional). He states that the use of onguration spa e is a mathemati al way for des ribing what are in fa t events in spa e and time.
However, it is the multi-dimensional theory that is the most su essful
and has proved to be a powerful analyti tool in relation to Heisenberg
and Born's matrix me hani s. The multi-dimensional point of view has
not been re on iled yet with the four-dimensional one.a
I. Multi-dimensional theory. S hrdinger sket hes a derivation of his
time-independent wave equation, noting that it reprodu es or improves
on the results of Bohr's quantum theory. He also notes that the stationary states allow one to al ulate the transition probabilities en ountered
in matrix me hani s. If one wishes to onsistently develop a formalism in whi h there are only dis ontinuous transitions, he suggests one
should take seriously the idea that the transitions do not o ur against
a ontinuous time ba kground; the appearen e of a ontinuous time
a As be omes lear in the dis ussion, S hrdinger thinks that the multi-dimensional
theory may prove indispensable, so that one should a ept the notion of a fun tion on onguration spa e and try instead to understand its physi al meaning
in terms of its manifestation in spa e and time. (See the dis ussion in se tion 4.4
below.)
4.2 Summary of the report
127
parameter would be purely statisti al, so to speak.a As an alternative,

he suggests interpreting the time-independent equation as arising from
a time-dependent one from whi h the time variable is eliminated by
assuming a stationary solution. He thus arrives to the des ription of a
quantum system in terms of a time-dependent wave fun tion on onguration spa e. He then asks what the meaning of this wave fun tion is:
`how does the system des ribed by it really look like in three dimensions?'
(p. 453, S hrdinger's emphasis). He briey mentions the view that the
-fun tion des ribes an ensemble of systems, whi h Born and Heisenberg
are going to dis uss. S hrdinger instead nds it useful (if perhaps à
bit naive') to imagine an individual system as ontinuously lling the
whole of spa e somehow weighted by ||2 (as he further laries in the
dis ussion). S hrdinger then arefully spells out that the spatial density
resulting from the onguration spa e density is not a lassi al harge
density, in the sense that the a tion on the parti les by external elds
and the intera tion between the parti les are already des ribed by the
potentials in the wave equation, and that it is in onsistent to assume that
this spatial density is also a ted upon in the manner of a lassi al harge
density. Instead, it is possible to interpret it as a harge density (with
some quali ations, some of whi h are spelled out only in the dis ussion)
for the purpose of al ulating the ( lassi al) radiation eld, thus yielding
a partial vindi ation of the idea of spatial densities. This, however, must
be an approximation, sin e the observation of su h emitted radiation is
itself an intera tion between the emitting atom and some other absorbing
atom or mole ule, to be des ribed again by the appropriate potentials
in the wave equation.
II. Four-dimensional theory. S hrdinger shows that the time-dependent wave equation for a single parti le is a nonrelativisti approximation
(with subtra tion of a rest frequen y) to the wave equation for the
de Broglie phase wave of the parti le. The latter an be made manifestly
relativisti by in luding ve tor potentials. (In modern terminology, this
is the Klein-Gordon equation.) If one ouples the Maxwell eld to it,
the same spatial densities dis ussed in se tion I appear as harge (and
urrent) densities. However, in the appli ation to the ele tron in the
hydrogen atom, it be omes apparent that adding the self-eld of the
ele tron to the (external) eld of the nu leus yields the wrong results.a
Thus S hrdinger argues that if one hopes to develop a spa etime theory
of intera ting parti les, it will be ne essary to onsider not just the
a Cf. the dis ussion in h. 8.1.
a Cf. S hrdinger (1927b), as mentioned in se tion 4.6 below.
128
overall eld generated by the parti les, but to distinguish between the
(spatially overlapping) elds generated by ea h individual parti le, ea h
eld a ting only on the other parti les of the system. Finally, he notes
that the Klein-Gordon equation needs to be modied in order to des ribe
spin ee ts, and that it may be possible to do so by onsidering a
ve torial instead of a s alar .
III. The many-ele tron problem. S hrdinger returns to the multidimensional theory and its treatment, by approximation, of the manyele tron atom. His interest in this spe i example relates to the question
of whether this multi-dimensional system an be understood in spa etime terms. The treatment rst negle ts the intera tion potentials between
the ele trons, and as a rst approximation takes produ ts k1 . . . kn of
the single-ele tron wave fun tions as solutions. One then expands the
solution of the full equation in terms of the produ t wave fun tions. The
time-dependent oe ients ak1 ...kn (t) in this expansion an then be al ulated approximately if the intera tion between the ele trons is small.
S hrdinger shows that before any approximation the oe ients
in the equations for the ak1 ...kn (t) depend only on potentials al ulated
from the spatial harge densities asso iated with the ki . Thus, although
the solution to the full equation is a fun tion on onguration spa e, it is
determined by purely spatial harge densities. A ording to S hrdinger,
this reinfor es the hope of providing a spatial interpretation of the wave
fun tion. A sket h of the approximation method on ludes this se tion
and the report.
4.3 Parti les as wave pa kets

The idea of parti les as wave pa kets is ru ial to the development of
S hrdinger's ideas and appears to provide one of the main motivations,
at least initially, behind the idea of a des ription of matter purely in
terms of waves.
S hrdinger's earliest spe ulation about wave pa kets (for both material parti les and light quanta) is found in his paper on Einstein's
gas statisti s (1926a), in se tion 5, Òn the possibility of representing
mole ules or light quanta through interferen e of phase waves'. As he
explains, S hrdinger nds it un anny that in de Broglie's theory one
should onsider the phase waves of the orpus les to be plane waves, sin e
it is lear that by appropriate superposition of dierent plane waves one
an onstru t a `signal', whi h following Debye (1909) and von Laue
129
(1914, se tion 2) an be onstrained to a small spatial volume. He then

ontinues:
On the other hand, it is of ourse
not
to be a hieved by the lassi al wave
laws, that the onstru ted `model of a light quantum' whi h by the way
extends indeed for
many
wavelengths in every dire tion also
permanently
stays together. Rather, it spreads itself out [zerstreut si h over ever larger
volumes after passing through a fo al point.
If one ould avoid this last on lusion by a quantum theoreti al modi ation
of the lassi al wave laws, then a way to deliveran e from the light quantum
dilemma would appear to be paved [angebahnt. (1926a, p. 101)
Wave pa kets are rst dis ussed at length in the se ond paper on
quantisation (1926 ): after des ribing the opti al-me hani al analogy,
S hrdinger dis usses how one an onstru t wave pa kets (using the
analogues of the opti al onstru tions by Debye and von Laue), and
then shows that the entroid of su h a wave pa ket follows the lassi al
equations of motion. S hrdinger onje tures that material points are in
fa t des ribed by wave pa kets of small dimensions. He notes also that,
for systems moving along very small orbits, the pa ket will be spread out,
so that the idea of the traje tory or of the position of the ele tron inside
the atom loses its meaning. In fa t, the main problem that S hrdinger
was to fa e with regard to wave pa kets turned out to be that spreading
of wave pa kets is a mu h more generi feature than he imagined at rst.
As mentioned, S hrdinger sent to Lorentz the proofs of his rst
two papers on quantisation (1926b, ) on 30 Mar h 1926. In his reply,
among many other things, Lorentz dis ussed expli itly the idea of wave
pa kets, indeed doubting that they would stay together (Przibram 1967,
pp. 478).a S hrdinger ommented on this both in his reply of 6 June
(Przibram 1967, pp. 5566) and in an earlier letter of 31 May to Plan k
(Przibram 1967, pp. 811). In the latter, he admits that there will always
exist spread-out states, be ause of linearity, but still hopes it will be
possible to onstru t pa kets that stay together for hydrogen orbits
of high quantum number. As he notes in the reply to Lorentz, this
would imply that there is no general identi ation between hydrogen
eigenstates and Bohr orbits, sin e a Bohr orbit of high quantum number
would be represented by a wave pa ket rather than a stationary wave.
(For an ele tron in a hydrogen orbit of low quantum number, S hrdinger
a As Lorentz remarks, the alternative would be `to dissolve the ele tron ompletely
.... and to repla e it by a system of waves' (p. 48), whi h, however, would make it
di ult to understand phenomena su h as the photoele tri ee t. The latter was
in fa t one of the riti isms levelled at S hrdinger's theory by Heisenberg (see
below, se tion 4.6).
130
did not envisage an orbiting pa ket but indeed a spread-out ele tron.)
With the reply to Lorentz, S hrdinger further sent his paper on mi roand ma rome hani s (1926e), in whi h he showed that for the harmoni
os illator, wave pa kets do stay together. He hoped that the result would
generalise to all quasi-periodi motions (admitting that maybe there
would be `dissolution' for a free ele tron). However, on 19 June Lorentz
sent S hrdinger a al ulation showing that wave pa kets on a high
hydrogen orbit would indeed spread out ( f. Przibram 1967, pp. 6971,
where the details of the al ulation, however, are omitted).
The relevan e of high hydrogen orbits is, of ourse, the role they
play in Bohr's orresponden e prin iple. The fa t that wave pa kets
along su h orbits do not stay together was thus a blow for any hopes
S hrdinger might have had of explaining the transition from mi ro- to
ma rome hani s along the lines of the orresponden e prin iple.
A further blow to the idea of wave pa kets must have ome with Born's
papers on ollision theory during the summer of 1926 (Born 1926a,b; f.
se tion 3.4.3 above). In the equivalen e paper, S hrdinger had in luded
a remark about s attering, for whi h, he wrote, it is ìndeed ne essary to
understand learly the ontinuous transition between the ma ros opi
ans hauli h me hani s and the mi rome hani s of the atom' (1926d,
p. 753; see below, se tion 4.6, for the notion of Ans hauli hkeit). Given
that at the time S hrdinger thought that ele trons on high quantum
orbits should be des ribed by wave pa kets, this remark may indi ate
that S hrdinger also thought that s attering should involve a dee tion
of wave pa kets, whi h would move asymptoti ally in straight lines.a
Born's work on ollisions in the summer of 1926 made essential use
of wave me hani s, but suggested a very dierent pi ture of s attering. Born expli itly understood his work as providing an alternative
interpretation both to his earlier views on matrix me hani s and to
S hrdinger's views. In turn, S hrdinger appeared to be s epti al of
Born's suggestions, writing on 25 August to Wilhelm Wien:
Zeits h. f. Phys. I know more or
waves must be stri tly ausally determined
through eld laws, the wavefun tions on the other hand have only the meaning
of probabilities for the a tual motions of light- or material-parti les. I believe
From an oprint of Born's last work in the
less how he thinks of things: the
that Born thereby overlooks that .... it would depend on the taste of the
observer whi h he now wishes to regard as real, the parti le or the guiding
eld. (Quoted in Moore 1989, p. 225)
a If thus was indeed S hrdinger's intuition, it may seem quite remarkable. On the
other hand, so is the fa t that S hrdinger does not seem to dis uss dira tion of
material parti les, whi h seems equally problemati for the idea of wave pa kets.
131
A few days later, S hrdinger submitted a review paper on ùndulatory

me hani s' to the Physi al Review (1926h). In it, he qualied rather
strongly the idea that `material points onsist of, or are nothing but,
wave systems' (p. 1049). Indeed, he ontinued (pp. 104950):
This extreme on eption may be wrong, indeed it does not oer as yet the
slightest explanation of why su h wave-systems seem to be realized in nature
as orrespond to mass-points of denite mass and harge. .... a thorough
orrelation of all features of physi al phenomena an probably be aorded
only by a harmonious union of these two extremes.
During the general dis ussion at the Solvay onferen e, S hrdinger

summarised the situation with the following words (this volume, p. 518):
The original pi ture was this, that what moves is in reality not a point but
a domain of ex itation of nite dimensions .... . One has sin e found that
the naive identi ation of an ele tron, moving on a ma ros opi orbit, with
a wave pa ket en ounters di ulties and so annot be a epted to the letter.
The main di ulty is this, that with ertainty the wave pa ket spreads in
all dire tions when it strikes an obsta le, an atom for example. We know
today, from the interferen e experiments with athode rays by Davisson and
Germer, that this is part of the truth, while on the other hand the Wilson
loud hamber experiments have shown that there must be something that
ontinues to des ribe a well-dened traje tory after the ollision with the
obsta le. I regard the ompromise proposed from dierent sides, whi h onsists
of assuming a ombination of waves and point ele trons, as simply a provisional
manner of resolving the di ulty.
The problem of the relation between mi ro- and ma rophysi s is onne ted of ourse to the linearity of the wave equation, whi h appears to
lead dire tly to highly non lassi al states (witness S hrdinger's famous
` at' example, S hrdinger 1935).a One might ask whether S hrdinger himself onsidered the idea of a nonlinear wave equation. In this
onne tion, a few remarks by S hrdinger may be worth investigating
further. One expli it, if early, referen e to nonlinearity is ontained in the
letter of 31 May to Plan k, where, after noti ing that linearity for es the
existen e of non- lassi al states, S hrdinger indeed spe ulates that the
equations might be only approximately linear (Przibram 1967, p. 10). In
the orresponden e with Lorentz, the question of nonlinearity arises in
other ontexts. In the ontext of the problem of radiation, it appears at
rst to be ne essary for ombination tones to arise (pp. 4950). In the
a Note, on the other hand, that S hulman (1997) shows there are ` lassi al' solutions
to the linear equation in quite realisti models of oupling between mi ro- and
ma rosystems (measurements), provided one requires them to satify appropriate
boundary onditions at both the initial and the nal time.
132
ontext of radiation rea tion, S hrdinger writes that the ex hange with
Lorentz has onvin ed him of the ne essity of nonlinear terms (p. 62). In
print, S hrdinger mentioned the possibility of a nonlinear term in order
to in lude radiaton rea tion in his fourth paper on quantisation (1926g,
p. 130).a Finally, a few years later, in his se ond paper on entanglement
(1936, pp. 4512), S hrdinger mentioned the possibility of spontaneous
de ay of entanglement at spatial separation, whi h would have meant a
yet untested violation of the S hrdinger equation.
Modern ollapse theories, su h as those by Ghirardi, Rimini and Weber (1986) or by Pearle (1976, 1979, 1989), modify the S hrdinger
equation sto hasti ally, in a way that su essfully ountera ts spreading
with in reasing s ale of the system. S hrdinger's strategy based on wave
pa kets thus appears to be viable at least if one a epts sto hasti modi ations to S hrdinger's equation (as S hrdinger was not ne essarily
likely to do).
Note that while ru ial, the failure of the straightforward idea of wave
pa kets for representing the ma ros opi , or lassi al, regime of the theory is distin t from the question of whether S hrdinger's wave pi ture
ould adequately des ribe what appeared to be other examples or lear
indi ations of parti ulate or `dis ontinuous' behaviour, su h as en ountered in the photoele tri ee t. This question was parti ularly important
in the dialogue with Heisenberg and Bohr (se tion 4.6). We shall also
see that S hrdinger ontinued to explore how ontinuous waves might
provide des riptions of apparently parti ulate or dis ontinuous quantum
phenomena, su h as the Compton ee t, quantum jumps, bla kbody
radiation and even the photoele tri ee t (se tion 4.6.3).
a An expli it proposal for in luding radiation rea tion through a nonlinear term is
due to Fermi (1927). Classi ally, if one in ludes radiation rea tion, one has the
third-order nonrelativisti Abraham-Lorentz equation
...
x
x = F/m ,
(1)
where F is the externally applied for e and = (2/3)e2 /mc3 (e is the ele tron
harge, m is the mass). Fermi (1927) proposed modifying the S hrdinger equation
as follows:
i
m x d
= H
t
dt3
d3 x x .
(2)
d
That is, he added an extra `potential' m x dt
3 hxi. Rederiving the Ehrenfest
theorem, one nds that hxi then obeys the above Abraham-Lorentz equation.
4.4 The problem of radiation
133

In Bohr's theory of the atom, the frequen y of emitted light orresponded
not to the frequen y of os illation of an ele tron on a Bohr orbit, but
to the term dieren e between two Bohr orbits. No known me hanism
ould explain the dieren e between the frequen y of os illation and
the frequen y of emission. Bohr's theory simply postulated quantum
jumps Ei Ej between the stationary states of energy Ei and Ej ,
a ompanied by emission (or absorption) of light of the orresponding
E E
frequen y ij = i h j . In this respe t, the Bohr-Kramers-Slater (BKS)
theory had the advantage of postulating a olle tion of virtual os illators
with the observed frequen ies. This was also, in a sense, that aspe t of
the BKS theory that survived into Heisenberg's matrix me hani s.a
The idea that wave me hani s ould provide a ontinuous des ription
of the radiation pro ess (as opposed to the pi ture of quantum jumps)
appears to have been also one of the main bones of ontention between
S hrdinger and the Copenhagen-Gttingen physi ists. As S hrdinger
wrote to Lorentz on 6 June 1926:
The frequen y dis repan y in the Bohr model, on the other hand, seems to
me, (and has indeed seemed to me sin e 1914), to be something so
monstrous,
that I should like to hara terize the ex itation of light in this way as really
a
almost in on eivable. (Przibram 1967, p. 61)
The rea tion by S hrdinger to the BKS theory instead was quite enthusiasti . In part, S hrdinger was well predisposed towards the possibility
that energy and momentum onservation be only statisti ally valid.b In
part (as argued by de Regt, 1997), this was pre isely be ause the BKS
theory provided a me hanism for radiation, in fa t a very ans hauli h
me hanism, albeit `virtual'. S hrdinger published a paper on the BKS
theory, ontaining an estimate of its energy u tuations (1924b).c
In 1926, with the development of wave me hani s, S hrdinger saw a
new possibility of on eiving a me hanism for radiation: the superposia See above se tion 1.3 (p. 13) and hapter 3, espe ially se tion 3.3.1.
a The emphasis here is strong, but f. the ontext of this passage in S hrdinger's
letter.
b Cf. S hrdinger to Pauli, 8 November 1922, in Pauli (1979, pp. 6971). The idea
of abandoning exa t onservation laws seems to be derived from S hrdinger's
tea her Exner (ibid., and Moore 1989, pp. 1524). S hrdinger publi ly stated
this view and allegian e in his inaugural le ture at the University of Zri h (9
De ember 1922), published as S hrdinger (1929a).
Cf. Darrigol (1992, pp. 2478) on the problem of the indenite growth of the
energy u tuations with time (as raised in parti ular by Einstein). A ording to
S hrdinger, an isolated system would behave in this way, but the problem would
disappear for a system oupled to an innite thermal bath.
134
tion of two waves would involve two frequen ies, and emitted radiation
ould be understood as some kind of `dieren e tone'. In his rst paper
on quantisation (1926b), S hrdinger states that this pi ture would be
`mu h more pleasing [um vieles sympathis her' than the one of quantum
jumps (p. 375), but the idea is rather sket hy: S hrdinger spe ulates
that the energies of the dierent eigenstates all share a large onstant
term, and that if the square of the frequen y is proportional to mc2 + E ,
then the frequen y dieren es (and therefore the beat frequen ies) are
approximately given by the hydrogen term dieren es. The se ond paper
(1926 ) refers to radiation only in passing.
Commenting on these papers in his letter to S hrdinger of 27 May,
Lorentz pointed out that while beats would arise if the time-dependent
wave equation (whi h S hrdinger did not have yet) were linear, they
would still not produ e radiation by any known me hanism. Combination tones would arise if the wave equation were nonlinear (Przibram 1967,
pp. 4950). As be omes lear in the following letters between S hrdinger
and Lorentz, on e the harge density of a parti le is asso iated with a
quadrati fun tion of , su h as ||2 , `dieren e tones' in the os illating
harge density arise regardless of whether the wave equation is linear or
nonlinear.
This idea is still the basis of today's semi lassi al radiation theory
(often used in quantum opti s), that is, the determination of lassi al
ele tromagneti radiation from the urrent asso iated with a harge density proportional to ||2 (for a non-stationary ).a S hrdinger arrived
at this result through his work onne ting wave me hani s and matrix
me hani s (his èquivalen e paper', 1926d). In fa t, S hrdinger showed
how to express the elements of Heisenberg's matri es wave me hani ally, in parti ular the elements of the dipole moment matrix, whi h
(by orresponden e arguments) were interpreted as proportional to the
radiation intensities. S hrdinger now suggested it might be possible `to
give an extraordinarily ans hauli h interpretation of the intensity and
polarisation of radiation' (1926d, p. 755), by dening an appropriate
harge density in terms of . In this paper, he suggested as yet, for a
single ele tron, to use (the real part of) the quadrati fun tion
t .
The third paper on quantisation (1926f) is on erned with perturbation theory and its appli ation to the Stark ee t, but S hrdinger
notes in an addendum in proof (footnote on p. 476) that the orre t
a This method is tou hed on previously. Pauli uses it (for the ase of one parti le)
in al ulating the s attered radiation in the Compton ee t during the dis ussion
after Compton's talk (p. 364).
135
harge density is given by ||2 . In the letter to Lorentz of 6 June 1926,

he explains in detail how this gives rise to a sensible notion of harge
density also for several parti les, ea h ontribution being obtained by
integrating over the other parti les (Przibram 1967, p. 56). This idea is
then used and dis ussed in print (as a `heuristi hypothesis') in the fourth
paper on quantisation (1926g, p. 118, S hrdinger's itali s), where the
wave fun tion is also expli itly interpreted in terms of a superposition
of all lassi alR ongurations of a system weighted by ||2 , and the time
onstan y of ||2 is derived (pp. 1356).a
This is also the pi ture of the wave fun tion given by S hrdinger in
the Solvay report (pp. 453 .). S hrdinger dis reetly skips dis ussing the
view of the wave fun tion as des ribing only an ensemble. S hrdinger's
on ern in interpreting the wave fun tion is to understand its manifestation in spa etime. This on ern also motivates S hrdinger's dis ussion
of the many-ele tron atoma , where he stresses that the spatial harge
distributions of the single (non-intera ting) ele trons already determine
the wave fun tion of the intera ting ele trons.
In the report, S hrdinger starts by rephrasing the question of the
meaning of the wave fun tion as that of how a system des ribed by a
ertain (multi-parti le) wave fun tion looks like in three dimensions. He
des ribes this as taking all possible ongurations of the lassi al system
simultaneously and weighting them a ording to ||2 . To this pi ture of
the system as a `snapshot' (as he alls it in the report) or as a `mollus '
(as he had written to Lorentz) S hrdinger then asso iates the orresponding ele tri harge density in 3-spa e. He is areful to state, however,
that this is not an ele tri harge in the usual sense. For one thing, the
ele tromagneti eld does not exert for es on it. While this harge does
des ribe the sour es of the eld in the semi- lassi al theory, this oupling
of the eld to the harges is des ribed as `provisional', rst, be ause of
the problem of radiation rea tion (whi h is not taken into a ount in the
S hrdinger equation), and se ond, be ause within a losed system it is
in onsistent to model the intera tion between dierent harged parti les
using the semi- lassi al eld. In parti ular, the observation of emitted
radiation is in prin iple again a quantum me hani al intera tion, and
should be des ribed by orresponding potentials in the equation for the
total system.
a The latter question had also been raised by Lorentz (Przibram 1967, p. 71) and
answered in S hrdinger's next letter, not in luded in Przibram's olle tion.8
a Previously unpublished but deriving from methods used in S hrdinger's paper on
energy ex hange (1927 ).
136
The above pi ture of the wave fun tion and the question of regarding
S hrdinger's formal harge density as a sour e of lassi al ele tromagneti radiation evidently raised many questions. (S hrdinger introdu es
the dis ussion after his report by the remark: Ìt would seem that my
des ription in terms of a snapshot was not very fortunate, sin e it has been misunderstood', p. 468.) It also eli ited the most dis ussion following
S hrdinger's report.a
A di ulty that is spelled out only in the dis ussion ( ontributions
by Bohr and by S hrdinger, p. 468), is that using S hrdinger's dipole
moment (equation (13) in S hrdinger's talk) to al ulate the radiation
does not dire tly yield the orre t intensities. As pointed out in the
P
talk, if one evaluates the dipole moment for aR superposition k ck k ,
one obtains terms ontaining the integrals e qk l d . These are the
matrix elements of the dipole moment matrix in matrix me hani s, and
they an be used within ertain limits to al ulate the emitted radiation
(in parti ular, as S hrdinger points out, vanishing of the integral implies
vanishing of the orresponding spe tral line). However, these integrals
appear with the oe ients ck cl , whereas both a ording to Bohr's old
quantum theory and to experiment, the intensity of radiation should not
depend on the oe ient of the `nal state', say l . Thus, the use of (13)
as a lassi al dipole moment in the al ulation of emitted radiation does
not in general yield the orre t intensities.
Bohr also drew attention to the fa t that by the time of the Solvay
onferen e, Dira (1927b, ) had already published his treatment of the
intera tion of the (S hrdinger) ele tron with the quantised ele tromagneti eld. S hrdinger replied that he was aware of Dira 's work,
but had the same misgivings with q-numbers as he had with matri es:
the la k of `physi al meaning', without whi h he thought the further
development of a relativisti theory would be di ult.a
a The rest of the dis ussion in ludes a few te hni al questions and omments
( ontributions by Fowler and De Donder, Born's report on some numeri al work
on perturbation theory), and some dis ussion of the `three-dimensionality' of the
many-ele tron atom. Further dis ussion of the meaning of S hrdinger's harge
densities (and of the meaning of the S hrdinger wave in the ontext of the
transformation theory of matrix me hani s) took pla e in the general dis ussion,
espe ially in ontributions by Dira and by Kramers (pp. 491 and 495).
a Note that in his treatment of the emission of radiation in The Physi al Prin iples
of the Quantum Theory (Heisenberg 1930b, pp. 824), Heisenberg seems to follow
and expand on the dis ussion of S hrdinger's report. Indeed, Heisenberg rst
des ribes two related methods for al ulating the radiation, based respe tively on
al ulating the matrix element of the dipole moment of the atom (justied via the
orresponden e prin iple), and on al ulating the dipole moment of S hrdinger's
`virtual harges' (as he alls them). He then explains pre isely the di ulty with
4.5 S hrdinger and de Broglie
137
If the interpretation of the wave fun tion as a harge density raises

problems of prin iple (as Born and Heisenberg also stress in their report,
p. 435), what is the point of suggesting su h an interpretation? One
possible way of understanding S hrdinger's intentions is to say that he
is proposing an ans hauli h image of the wave fun tion in terms of the
spatial density it denes, without in general spe ifying the dynami al role
played by this density. Under ertain ir umstan es, then, this spatial
density takes on the dynami al role of a harge density, in parti ular
as a sour e of radiation. This point of view does not resolve the other
problems onne ted with the wave on eption of matter (spreading of
the wave pa ket, S hrdinger's at, mi ro-ma ro question), but it oers
a platform from whi h to work towards their possible resolution. This
of ourse may be an overinterpretation of S hrdinger's position, but it
ts approximately with later developments that take S hrdinger's wave
on eption seriously, e.g. Bell's (1990) idea of ||2 as `density of stu'
(in onguration spa e).
4.5 S hrdinger and de Broglie

Thus opens S hrdinger's review paper on wave me hani s for the Physi al Review (1926h, p. 1049, referen es omitted):
The theory whi h is reported in the following pages is based on the very
interesting and fundamental resear hes of L. de Broglie on what he alled
`phase waves' (òndes de phase') and thought to be asso iated with the motion
a
of material points, espe ially with the motion of an ele tron or [photon. The
point of view taken here, whi h was rst published in a series of German
papers, is rather that material points onsist of, or are nothing but, wavesystems.
This passage both illustrates the well-known fa t that S hrdinger arrived at his wave me hani s by developing further the ideas of de Broglie
starting with his paper on gas statisti s (S hrdinger 1926a) and
emphasises the main on eptual dieren e between de Broglie's and
S hrdinger's approa hes.
While some details of the relation between S hrdinger's and de Broglie's work are dis ussed in se tion 2.3, in this se tion we wish to raise
the question of why S hrdinger should have developed su h a dierent
the latter dis ussed here by Bohr and S hrdinger, and goes on to sket h (a variant
of) Dira 's treatment of the problem.
a The text here reads `proton', whi h is very likely a misprint, sin e de Broglie's
work indeed fo ussed on ele trons and photons.
138
pi ture of wave me hani s. Indeed, although S hrdinger quotes de Broglie as the redis overer of Hamilton's opti al-me hani al analogy (1926h,
footnote 3 on p. 1052), the two authors apply the analogy in opposite
dire tions: de Broglie treats even the photon as a material parti le with
a traje tory, while S hrdinger treats even the ele tron as a pure wave.a
As S hrdinger writes in his rst paper on quantisation, it was in
parti ular `ree tion on the spatial distribution' of de Broglie's phase
waves that gave the impulse for the development of his own theory of
wave me hani s (1926b, p. 372). We gain some insight as to what this
refers to from a letter from S hrdinger to Land of 16 November 1925:
I have tried in vain to make for myself a pi ture of the phase wave of the ele tron in the Kepler orbit. Closely neighbouring Kepler ellipses are onsidered
as `rays'. This, however, gives horrible ` austi s' or the like for the wave fronts.
(Quoted in Moore 1989, p. 192)
Thus, one reason for S hrdinger to abandon the idea of traje tories
in favour of the pure wave theory might have been that well-behaved
traje tories seemed to be in ompatible with well-behaved waves. And indeed, in his presentations of the opti al-me hani al analogy, S hrdinger
states that outside of the geometri limit the notion of `ray' be omes
meaningless (1926b, pp. 495 and 507, 1926h, pp. 10523). This also seems
to be at issue in the ex hange between S hrdinger and Lorentz during
the dis ussion of de Broglie's report, where S hrdinger points out that
in ases of degenera y, any arbitrary linear ombination of solutions is
allowed, and de Broglie's theory would predi t very ompli ated orbits
(p. 401).
There are other aspe ts that ould on eivably provide further reasons
for S hrdinger's denite abandoning of the traje tories. One possibility
is that S hrdinger pi ked up from de Broglie spe i ally the idea of
parti les as singularities of the wave,a and was happy to relax it to the
a As is lear from de Broglie's thesis and as mentioned by de Broglie himself in
his report (p. 376), de Broglie had always assumed the pi ture of traje tories.
The above quotation in any ase makes it lear that this was S hrdinger's own
reading of de Broglie. Already as early as January 1926, S hrdinger writes in his
gas theory paper about `the de Broglie-Einstein undulatory theory of orpus les
in motion, a ording to whi h the latter are no more than a kind of foam rest
[S hammkaum on the wave radiation that onstitutes the world ba kground
[Weltgrund' (1926a, p. 95). Other indi ations that this was well-known are
Lorentz's omments to S hrdinger on the onstru tion of ele tron orbits in his
letter of 19 June 1926 oming ` lose to de Broglie's arguments' (Przibram 1967,
p. 74) and Pauli's referen e to Einstein's and de Broglie's `moving point masses'
in his letter to Jordan of 12 April 1926 (Pauli 1979, p. 316).
a Cf. S hrdinger (1926a, p. 99): `The universal radiation, as signals or perhaps
singularities of whi h the parti les are meant to o ur, is thus something quite
4.6 The oni t with matrix me hani s
139
idea of wave pa kets. Indeed, as we have seen above, S hrdinger pla ed

great emphasis on the notion of a wave pa ket, and if it had been an
adequate notion, there would have been no need for a separate notion
of a orpus le in order to explain the parti ulate aspe ts of matter.
We have also seen that S hrdinger was a utely aware of the radiation
problem, namely of the dis repan y between the orbital frequen y of the
ele tron and the frequen y of the emitted radiation. While de Broglie in
his thesis was able to derive the Bohr orbits from wave onsiderations,
this would in no way seem to alleviate the problem: the frequen y of
revolution of the ele tron in de Broglie's theory was the same as in the
Bohr model, and would lead to the wrong frequen y of radiation if the
ele tron was treated as a lassi al sour e. (In the ase of degenera y
noted above, the situation would be even more ompli ated.)a
Finally, S hrdinger appeared to be riti al of proposals ombining
waves and parti les, for instan e as appeared to be done by Born in
his ollision papers (see se tion 4.3 above). Su h misgivings ould easily
have applied also to de Broglie's theory.

In the early dis ussions on the meaning of quantum theory, the notion of
Àns hauli hkeit' resurfa es time and again. The verb àns hauen' means
`to look at', and àns hauli h', whi h means ` lear', `vivid' or ìntuitive',
has visual onnotations that the English word ìntuitive' la ks. Ans hauli hkeit of a physi al theory is thus a quality of ready omprehensibility
that may (but need notb ) in lude a strong omponent of literal pi turability. For S hrdinger, at any rate, it seems that the possibility of
grasping a theory through some kind of spatio-temporal intuition was a
key omponent of physi al understanding.c
It is lear, however, that the oni t between wave me hani s and
matrix or quantum me hani s was not, or not only, a philosophi al issue,
say about the validity of the on epts of spa etime and ausality,d and
a
b

essentially more ompli ated than for instan e the wave radiation of Maxwell's
theory .... '
Note, however, that de Broglie (1924 ) had dis ussed the solution of this problem
in the orresponden e limit (i.e. for the ase of high quantum numbers).
Cf. below Heisenberg's use of the term in the un ertainty paper.
A good dis ussion of the role the notion of Ans hauli hkeit played for S hrdinger
is given by de Regt (1997, 2001), who argues that S hrdinger's requirement
of Ans hauli hkeit is derived indire tly from Boltzmann, and is essentially a
methodologi al requirement (as opposed to being a ommitment to realism
f. also the introdu tion by Bitbol in S hrdinger 1995, p. 4 fn. 10).
Note that Kant's on eption of spa e and time is formulated in terms of what he
140
that, in the minds of the parties involved, these issues were onne ted
with spe i ally s ienti questions.a
The list of these questions is extensive. In his èquivalen e' paper
(1926d), S hrdinger states that mathemati al equivalen e is not the
same as physi al equivalen e, in the sense that two theories an oer
quite dierent possibilities for generalisation and further development.a
He thinks of two problems in parti ular: rst, the problem of s attering
(as mentioned in se tion 4.3 above), se ond, the problem of radiation
(dis ussed in se tion 4.4), in onne tion with whi h he then des ribes
the idea of the vibrating harge density. The radiation problem in turn
links to further issues at the heart of the debate with Heisenberg and
Bohr in parti ular, on whether there are quantum jumps or whether
the pro ess of radiation and the atomi transitions an be des ribed as
ontinuous pro esses in spa e and time. In a sense, neither the issue of
s attering nor that of radiation resolved the debate in favour of either
theory: both theories were modied or reinterpreted in the ourse of
these developments, even though the result (statisti al interpretation,
Copenhagen interpretation) was not to S hrdinger's taste.
We shall now follow how the oni t evolved, from the beginnings to
the time of the Solvay onferen e, sin e both sides developed onsiderably during the ru ial period between S hrdinger's rst papers and the
onferen e.
4.6.1 Early days

At the time of S hrdinger's rst two papers on quantisation, as we
have seen in the previous hapter, matrix me hani s was a theory that
reje ted the notion of ele tron orbits, indeed the very possibility of a
spa etime des ription, substituting the lassi al kinemati al quantities
with matrix quantities; but it kept the postulate of stationary states
and of quantum jumps between these states. Matrix me hani s did not
des ribe the stationary states individually, only olle tively, and allowed
one to al ulate only transition probabilities for the jumps.
S hrdinger's rst paper on quantisation (1926b) ontains only a few
omments on the possible ontinuous pi ture of atomi transitions, as
alls the Àns hauungsformen' (the `forms of intuition'), so that the dis ussion has
indeed strong philosophi al overtones.
a This point has been re ently argued also by Perovi (2006).
a Similarly, in the dis ussion after his own report, S hrdinger insists that nding a
physi al interpretation of the theory is ìndispensable for the further development
of the theory' (p. 472).
141
opposed to quantum jumps. (But of ourse, at this stage, he has very

little to say about the problem of radiation, in parti ular nothing about
intensities.) The rst time he omments expli itly on the dieren es
between wave me hani s and quantum me hani s is in his se ond paper.
There he writes that wave me hani s oers a way to interpret
rst: one should
phase of the ele tron motions in the atom; se ond:
the onvi tion, more and more oming to the fore today, that
deny real signi an e to the
that one may not even laim that the ele tron at a ertain time is lo ated
on one parti ular

onditions; third:
of the quantum traje tories distinguished by the quantum

the true laws of quantum me hani s onsist not in de-
nite pres riptions for the
individual traje tory,
rather these true laws relate
through equations the elements of the whole manifold of traje tories of a

system through, so that apparently a ertain intera tion between the dierent
traje tories obtains. (1926 , pp. 508)
As S hrdinger pro eeds to say, these laims are in ontradi tion with
the ideas of ele tron position and ele tron orbit, but should not be
taken as for ing a omplete surrender of spatio-temporal ideas. At the
time, the mathemati al relation between wave me hani s and matrix
me hani s was not yet laried, but S hrdinger hopes there would be
a well-dened mathemati al relation between the two, whi h ould then
omplement ea h other. A ording to S hrdinger, Heisenberg's theory
yields the line intensities, his own oers the possibility of bridging the
mi ro-ma ro gap.a Personally, he nds the on eption of emitted frequen ies as `beats' parti ularly attra tive and believes it will provide an
ans hauli h understanding of the intensity formulas (1926 , pp. 51314).
Within weeks, the relation between the two theories was laried
independently by S hrdinger (1926d), by E kart (1926) and by Pauli,
in his remarkable letter to Jordan of 12 April 1926 (Pauli 1979, pp. 315
20). This is one of the rst do umented rea tions to S hrdinger's new
work from a physi ist of the Copenhagen-Gttingen s hool.b In it, Pauli
emphasises in fa t that in S hrdinger's theory there are no ele tron
orbits, sin e traje tories are a on ept belonging to the geometri limit
of the theory. Pauli seems to say that insofar as S hrdinger provides a
a S hrdinger's olle ted works (1984) reprodu e this and other papers from Erwin
and Anny S hrdinger's own opy of the se ond edition of Abhandlungen zur
Wellenme hanik (S hrdinger 1928). In this opy of the book, the passage about
the mi ro-ma ro bridge is underlined.
b Ehrenfest informed Lorentz of the èquivalen e' result (and of Klein's (1926)
theory) on 5 May 1926, when Kramers reported them in the olloquium at Leiden.
Ehrenfest also mentioned the relation of this result to Lan zos's (1927) work.9 As
noted above (footnote on p. 125), by 27 May Lorentz had re eived a opy of the
paper dire tly from S hrdinger.
142
des ription of individual stationary states, his theory is not in oni t

with matrix me hani s, sin e it also does not ontain the on ept of
ele tron orbits.a
The tone of S hrdinger's remarks and of the omments they eli it
hanges with the equivalen e paper. In a mu h-quoted footnote, S hrdinger says he had known of Heisenberg's theory but was `s ared away,
not to say repelled', by the ompli ated algebrai methods and the la k
of Ans hauli hkeit (1926d, p. 735). On 8 June 1926, Heisenberg sends
Pauli an equally notorious (but often mis-quoted) omment:
The more I ree t on the physi al part of S hrdinger's theory, the more
disgusting [abs heuli h I nd it. Imagine the rotating ele tron, whose harge
is distributed over the whole spa e with its axis in a fourth and fth dimension.
What S hrdinger writes of the Ans hauli hkeit of his theory would
[be
an appropriate...
s ar ely
in other words I nd it poppy o k [Mist. (Pauli 1979,
p. 328)
It seems lear that, a ording to Heisenberg, it is S hrdinger's laim

of Ans hauli hkeit for his theory that is ludi rous, presumably partly
be ause of the spread-out ele tron and partly be ause the waves are in
onguration spa e.a
4.6.2 From Muni h to Copenhagen

In the summer of 1926, S hrdinger gave a series of talks on wave
me hani s in various German universities, in parti ular, on 21 and 23
July, he talked in Muni h at the invitation of Sommerfeld and of Wien.b
The des ription of a me hanism for radiation eli ited enthusiasti omments by Wien, but riti ism from Sommerfeld and from Heisenberg.
The dis ussion was apparently heated, and eventually identied a ru ial experiment that would de ide between the idea of the ontinuous
a Pauli knew very early about S hrdinger's paper, having been informed by
Sommerfeld, to whom at S hrdinger's request a opy of the paper had been
forwarded by Wien, the editor of Annalen der Physik ( f. Pauli 1979, pp. 278
and 293). Pauli's analysis in the letter of 12 April appears to be based only on
S hrdinger's rst paper on quantisation, although at least the o ial publi ation
date of the se ond paper was 6 April 1926. See also the letters between Pauli and
S hrdinger reprodu ed in Pauli (1979).
a Alternatively, with a `fourth and fth dimension', Heisenberg might on eivably
be referring to Klein's ve-dimensional extension of S hrdinger's equation (Klein
1926). As one example, here is how Moore (1989, p. 221) quotes the passage: `The
more I think of the physi al part of the S hrdinger theory, the more abominable I
nd it. What S hrdinger writes about Ans hauli hkeit makes s ar ely any sense,
in other words I think it is bullshit [Mist'.
b Here we follow mainly the a ount given by Heisenberg (1946).
143
me hanism of radiation envisaged in wave me hani s and the idea of

quantum jumps. This was in oherent s attering, i.e. the Raman ee t (at
that time neither observed nor thus named). A ording to the quantum
predi tion (Smekal 1923, Kramers and Heisenberg 1925), in oherent
s attering would exist also for atoms in the ground state, be ause the
atom ould be ex ited by the in ident light. A ording to S hrdinger
instead, the ee t was due to indu ed vibrations for the ase in whi h at
least two atomi frequen ies were already present, and so would not
o ur in the ground state.a Apparently, Sommerfeld and Heisenberg
were `prepared to enter a bet for its existen e, while the experimental
physi ists were against it and S hrdinger took a more wait-and-see
attitude' (Heisenberg 1946, p. 5).
In a letter to Pauli of 28 July, Heisenberg gives other spe i riti isms
of S hrdinger, for throwing overboard èverything quantum theoreti al: namely photoele tri ee t, Fran k ollisions, Stern-Gerla h ee t
et .' (Pauli 1979, p. 338). As a matter of fa t, in the letter to Wien
of 25 August 1926, quoted above (se tion 4.3), S hrdinger admitted
that he had great on eptual di ulties with the photoele tri ee t
(but see below the dis ussion of S hrdinger, 1929b). Heisenberg further
mentioned to Pauli that, together with S hrdinger and Wien, he had
dis ussed Wien's experiments on the de ay of lumines en e (Wien 1923,
h. XX). This was another point where S hrdinger thought wave me hani s proved superior to matrix me hani s, and Heisenberg en ouraged
Pauli to al ulate and publish the damping oe ients for the hydrogen
spe trum.a
Heisenberg mentions similar riti isms in his talk on `Quantum me hani s', given at the 89th meeting of German S ientists and Physi ians in
Dsseldorf on 23 September 1926 and published in the issue of 5 November of Naturwissens haften (Heisenberg 1926 ). This talk ould be seen
as a publi response to the laims that the return to a ` ontinuum theory'
was possible. In it, Heisenberg addresses in parti ular the problem of
Ans hauli hkeit: a ording to Heisenberg, the usual notions of spa e and
time, and in parti ular their appli ation to physi s with the idea that
a Note the implied inequivalen e of wave and matrix me hani s (despite the
re ent èquivalen e proofs'). Another su h possibility of experimental inequivalen e
is mentioned in the dis ussion after S hrdinger's report, with regard to the
quadrupole radiation of the atom (p. 469). Cf. also Muller (1997) on the
inequivalen e of the two theories.
a Cf. Born and Heisenberg's report (pp. 432 and 435), where spin is onsidered to
be problemati for wave me hani s, and where it is expli itly stated that Dira
(1927 ) provides an explanation for the de ay experiments.
144
spa e and matter are in prin iple ontinuously divisible, turn out to be
mistaken, rst of all due to the Unans hauli hkeit of the orpus ular
nature of matter, then through the theoreti al and experimental onsiderations leading to the idea of stationary states and quantum jumps
(Bohr, Fran k-Hertz, Stern-Gerla h), and nally through onsideration
of radiation phenomena (Plan k's radiation formula, Einstein's light
quantum, the Compton ee t and the Bothe-Geiger experiments). This
issue, it is laimed, also relates losely to the question of the degree of
`reality' to be as ribed to material parti les or light quanta. Quantum
me hani s in its development had thus rst of all to free itself from
notions of Ans hauli hkeit, in order to set up a new kind of kinemati s
and me hani s. In dis ussing the wave theory, Heisenberg onsiders rst
de Broglie and Einstein as having developed wave-parti le dualism for
matter and having suggested the possibility of interferen e for an ensemble [S har of parti les. He then explains that S hrdinger found a
dierential equation for the matter waves that reprodu es the eigenvalue
problem of quantum (i.e. matrix) me hani s. However, a ording to
Heisenberg, the S hrdinger theory fails to provide the link with de
Broglie's ideas, that is, it fails to provide the analogy with light waves in
ordinary spa e, be ause of the need to onsider waves in onguration
spa e; the latter therefore have only a formal signi an e. Heisenberg
refers to the laim that on the basis of S hrdinger's theory one may
be able to return to à purely ontinuous des ription of the quantum
theoreti al phenomena', and ontinues (Heisenberg 1926 , p. 992):
In developing onsequently this point of view one leaves in fa t the ground of
de Broglie's theory, thus of Q.M. and indeed of all quantum theory and arrives
in my opinion at a omplete ontradi tion with experien e (law of bla kbody
radiation; dispersion theory). This route is thus not viable. The a tual reality
of the de Broglie waves lies rather in the interferen e phenomena mentioned
above, whi h defy any interpretation on the basis of lassi al on epts. The
extraordinary physi al signi an e of S hrdinger's results lies in the realisation that an ans hauli h interpretation of the quantum me hani al formulas
ontains both typi al features of a orpus ular theory and typi al features of
a wave theory.
After dis ussing quantum statisti s (`whi h in any ase represents a

very bizarre further limiting of the reality of the orpus les', p. 992),
Heisenberg on ludes (p. 994):
In our ans hauli h interpretation of the physi al pro esses and mathemati al
formulas there is a dualism between wave theory and orpus ular theory su h
that many phenomena are des ribed most naturally by a wave theory of light
145
as well as of matter, in parti ular interferen e and dira tion phenomena,

while other phenomena in turn an be interpreted only on the basis of the
orpus ular theory. .... The ontradi tions of the ans hauli h interpretations
of dierent phenomena ontained in the urrent s heme are ompletely unsatisfa tory. For a ontradi tion-free ans hauli h interpretation of the experiments,
whi h in themselves are indeed ontradi tion-free, some essential trait in our
pi ture of the stru ture of matter is urrently still missing.
After the summer, S hrdinger visited Copenhagen for an intense

round of dis ussions. A ording to Heisenberg's re onstru tion in Der
Teil und das Ganze (Heisenberg 1969), S hrdinger argued with Bohr
pre isely about the ne essity of nding a me hanism for radiation, while
Bohr insisted that quantum jumps were ne essary for the derivation of
Plan k's radiation law, as well as being dire tly observable in experiments. Heisenberg (1946, p. 6) states that the dis ussion ended with the
re ognition `that an interpretation of wave me hani s without quantum
jumps was impossible and that the mentioned ru ial experiment [i.e.
the Raman ee t in any ase would turn out in favour of the quantum
jumps'.a
S hrdinger in turn admitted his di ulties. In a letter to Wien of
21 O tober 1926 (quoted in Pauli 1979, p. 339), he wrote: Ìt is quite
ertain that the position of ans hauli h images, whi h de Broglie and I
take, has not nearly been developed far enough to a ount even just for
the most important fa ts. And it is downright probable that here and
there a wrong path has been taken that needs to be abandoned'. And,
ommenting in his prefa e (dated November 1926) to the rst edition of
Abhandlungen zur Wellenme hanik (1926i) on the fa t that the papers
were being reprinted un hanged, he invoked `the impossibility at the
urrent stage of giving an essentially more satisfa tory or even denitive
new presentation'.
In the meantime, both Bohr and Heisenberg were working to nd their
own satisfa tory interpretation of the theory.b In parti ular as regards
Heisenberg, the orresponden e with Pauli is very telling.c First of all
one nds the dis ussion of Born's ollision papers (Born 1926a,b), whi h
led to Heisenberg's paper on u tuation phenomena (Heisenberg 1926b).
As already mentioned in se tion 3.4.5, Heisenberg then reports to Pauli
a The well-known quotation Ìf this damned quantum jumping is indeed to stay,
then I regret having worked on this subje t at all' is reported both in Heisenberg
(1946) and Heisenberg (1969).
b See also se tion 3.4 above.
As noted in Pauli (1979, p. 339 and fn. 3 on p. 340), however, most of Pauli's
letters to Heisenberg from this period appear to have been destroyed in the war.
146
about Dira 's transformation theory (Dira 1927a), whi h, however, is

seen only as an extraordinary formal development. The next letters
from Heisenberg to Pauli are from February 1927 and in lude on 23
February the sket h of the ideas of the un ertainty paper (Pauli 1979,
pp. 37681). This, as was Bohr's simultaneous development of the idea of
omplementarity, was meant to provide the ans hauli h pi ture that was
still missing. Indeed, in the un ertainty paper, Heisenberg (1927, p. 172)
formulates Ans hauli hkeit as the possibility of arriving at qualitative
predi tions in simple ases together with formal onsisten y. With the
formulation and appli ation of the un ertainty relations, he was then
satised to have found su h an interpretation.
In a sense, however, the un ertainty paper was also the end of the
original notion of `quantum jumps' as transitions between stationary
states of a system. Indeed, as had to be the ase given the generalisation
of transition probabilities to arbitrary pairs of observables (and as was
implied in the Heisenberg-Pauli orresponden e), the privileged role of
stationary states had to give way. It is somewhat ironi that Heisenberg
was to give up quantum jumps only a few months after the dis ussions
with S hrdinger in Copenhagen; nevertheless, quantum me hani s thus
wedded to probabilisti transitions between measurements was just as
dis ontinuous as the pi ture of quantum jumps between stationary states
that it repla ed, and, for S hrdinger, it was equally unsatisfa tory.
Note that, for Bohr at least, wave aspe ts played a ru ial role in
the resulting `Copenhagen' interpretation. Yet, just as in the ase of
Born's (1926a,b) use of wave me hani s in the dis ussion of ollisions
(see se tion 3.4.4), Heisenberg appears to have been onvin ed that the
apparent wave aspe ts ould be interpreted entirely in matrix terms.
This dieren e of opinion was ree ted in the sometimes tense dis ussions between Heisenberg and Bohr at the time, in parti ular on the topi
of Heisenberg's treatment of the -ray mi ros ope in the un ertainty
paper and the orresponding addendum in proof.a
4.6.3 Continuity and dis ontinuity

Between late 1926 and the time of the Solvay onferen e, S hrdinger
ontinued to work along lines that brought out the attra tive features
and sometimes the limitations of the wave pi ture. In late 1926
and early 1927, S hrdinger fo ussed on the `four-dimensional' theory,
a For more on this issue, see Beller (1999, pp. 714 and 13841) and Camilleri (2006).
For Heisenberg's way out of the di ulty, see Heisenberg (1929, pp. 4945).
147
with his papers on the Compton ee t (1927a), and on the energymomentum tensor (1927b), while he returned to the `many-dimensional'
theory shortly before the Solvay onferen e, with his work on energy ex hange (1927 ) and the treatment of the many-ele tron atom presented
in his report.
The Compton ee t is of ourse a paradigmati example of a `dis ontinuous' phenomenon, but S hrdinger (1927a) gives it a wave me hani al
treatment, by analogy with the lassi al ase of ree tion of a light wave
when it en ounters a sound wave (Brillouin 1922).a As he remarks in the
paper, and as remarked in the dis ussion after Compton's report, this
treatment relies on the onsideration of stationary waves and does not
dire tly des ribe an individual Compton ollision.b
The paper on the energy-momentum tensor (1927b), following Gordon
(1926), takes the Lagrangian approa h to deriving the Klein-Gordon
equation, and varies also the ele tromagneti potentials, thus deriving
the Maxwell equations as well. S hrdinger then onsiders in parti ular the energy-momentum tensor of the ombined Maxwell and KleinGordon elds. As he remarks (S hrdinger 1928, p. x), this is a beautiful
formal development of the theory, but it heightens starkly the di ulties
with the four-dimensional view, sin e one annot insert the ele tromagneti potentials thus obtained ba k into the Klein-Gordon equation. For
instan e, in luding the self-eld of the ele tron in the treatment of the
hydrogen atom would yield the wrong results, as S hrdinger also mentions in se tion II of his report. In this onne tion, S hrdinger states that:
`The ex hange of energy and momentum between the ele tromagneti
eld and matter does not in reality take pla e in a ontinuous way,
as the [given eldlike expression suggests' (1927b, p. 271).c A ording
to Heisenberg (1929), the further development of the four-dimensional
theory was indeed purely formal, but provided a basis for the later
development of quantum eld theory.d
The paper on energy ex hange (1927 ), S hrdinger's last paper before
a S hrdinger had given a derivation of Brillouin's result by assuming that energy
and momentum were ex hanged in the form of quanta (1924a). In the paper on
the Compton ee t he now omments on how, in a sense, he is reversing his own
earlier reasoning.
b See in parti ular the remarks by Pauli (p. 363) and the dis ussion following
S hrdinger's ontribution (p. 369). Cf. also the losing paragraph of Pauli's letter
to S hrdinger of 12 De ember 1927 (Pauli 1979, p. 366).
See again the ex hange of letters between Pauli and S hrdinger in De ember 1926
(Pauli 1979, pp. 3648).
d Heisenberg (1930b) in orporates into his view of quantum theory both the
multi-dimensional theory and the four-dimensional theory of S hrdinger's report.
The latter is interpreted as a lassi al wave theory of matter that forms the
148
the Solvay onferen e, marks an attempt to meet the riti ism that wave
me hani s annot a ount for ru ial phenomena involving `quantum
jumps'. Following Dira (1926 ), S hrdinger sket hes the method of
the variation of onstants, whi h he is to use also in the dis ussion
of the many-ele tron atom in his Solvay report. He then applies it to
the system of two atoms in resonan e dis ussed by Heisenberg (1926b)
(and by Jordan (1927a)). As dis ussed in se tion 3.4, Heisenberg uses
this example to show how in matrix me hani s one an indeed des ribe
hange starting from rst prin iples, in parti ular how one an determine the probabilities for quantum jumps. Now S hrdinger turns the
tables around and argues that the treatment of atoms in resonan e using
wave me hani s shows how one an eliminate quantum jumps from the
pi ture. He argues that the two atoms ex hange energies as if they
were ex hanging denite quanta. Indeed, he goes further and suggests
that the idea of quantised energy itself should be reinterpreted in terms
of wave frequen y,a and that resonan e phenomena are indeed the key
to the `quantum postulates'. S hrdinger then pro eeds to formulate
statisti al onsiderations about the distributions of the amplitudes of
the two systems in resonan e, leading to the idea of the squares of the
amplitudes as measures of the strength of ex itation of an eigenvalue. He
then returns to resonan e onsiderations in the ase of a system oupled
to a heat bath, whi h he onsiders would su e in prin iple for the
derivation of Plan k's radiation formula and of all the results of the òld
quantum statisti s'.
In the aftermath of the Solvay onferen e, although in pla es one
nds S hrdinger at least temporarily espousing views mu h loser to
those of the Copenhagen-Gttingen s hool ( f. Moore 1989, pp. 25051),
S hrdinger ontinued to explore the possibilities of the wave pi ture.a
The following is a telling example. In a short paper in Naturwissens haften, S hrdinger (1929b) proposes to illustrate `how the quantum
theory in its newest phase again makes use of ontinuous spa etime
fun tions, indeed of properties of their form, to des ribe the state and
behaviour of a system ....' (p. 487). S hrdinger quotes the example of
ba kground for a se ond quantisation, and in ludes the ba krea tion of the
self-eld via in lusion in the potential ( f. also above, fn. on p. 132).
a S hrdinger had expressed the idea of energy as frequen y already in his letter to
Wien of 25 August 1926: `What we all the energy of an individual ele tron is its
frequen y. Basi ally it does not move with a ertain speed be ause it has re eived
a ertain shove, but be ause a dispersion law holds for the waves of whi h it
onsists, as a onsequen e of whi h a wave pa ket of this frequen y has exa tly
this speed of propagation' (as quoted in Moore 1989, p. 225).
a Cf. also Bitbol's introdu tion in S hrdinger (1995), p. 5.
149
how photo hemi al and photoele tri phenomena depend on the form
of an impinging wave (i.e. on its Fourier de omposition), rather than
on lo al properties of the wave at the point where it impinges on the
relevant system. He argues that a ontinuous pi ture an be retained, but
introdu es the idea (whi h, as he remarks, generalises without di ulty
to the ase of many-parti le wave fun tions) that the ru ial properties
of wave fun tions are in fa t properties pertaining to the entire wave.
S hrdinger may thus have been the rst to introdu e the idea of nonlo alisable properties, as part of the pri e to pay in order to pursue a
wave pi ture of matter.
150
Notes to pp. 124141
Ar hival notes
2 S hrdinger to Lorentz, 30 Mar h 1926, AHQP-LTZ-8 (in German).
5 Lorentz to S hrdinger, 21 January 1927, AHQP-41, se tion 9 (in
German).
6 S hrdinger to Lorentz, 23 June 1927, AHQP-LTZ-13 (original with
S hrdinger's orre tions) and AHQP-41, se tion 9 ( arbon opy) (in
German).
7 See also S hrdinger to Lorentz, 16 July 1927, AHQP-LTZ-13 (in
German).
8 S hrdinger to Lorentz, 23 June 1926, AHQP-LTZ-8 (in German).
9 Ehrenfest to Lorentz, 5 May 1926, AHQP-EHR-23 (in German).
Part II
Quantum foundations and the 1927 Solvay
onferen e
5
Quantum theory and the measurement
problem
5.1 What is quantum theory?

For mu h of the twentieth entury, it was widely believed that the
interpretation of quantum theory had been essentially settled by Bohr
and Heisenberg in 1927. But not only were the `dissenters' of 1927 in
parti ular de Broglie, Einstein, and S hrdinger un onvin ed at the
time: similar dissenting points of view are not un ommon even today.
What Popper alled `the s hism in physi s' (Popper 1982) never really
healed. Soon after 1927 it be ame standard to assert that matters of
interpretation had been dealt with, but the sense of puzzlement and
paradox surrounding quantum theory never disappeared.
As the entury wore on, many of the on erns and alternative viewpoints expressed in 1927 slowly but surely revived. In 1952, Bohm revived and extended de Broglie's theory (Bohm 1952a,b), and in 1993
the de Broglie-Bohm theory nally re eived textbook treatment as an
alternative formulation of quantum theory (Bohm and Hiley 1993; Holland 1993). In 1957, Everett (1957) revived S hrdinger's view that the
wave fun tion, and the wave fun tion alone, is real (albeit in a very
novel sense), and the resulting Èverett' or `many-worlds' interpretation (DeWitt and Graham 1973) gradually won widespread support,
espe ially among physi ists interested in quantum gravity and quantum
osmology. Theories even loser to S hrdinger's ideas ollapse theories, with ma ros opi obje ts regarded as wave pa kets whose spreading
is prevented by sto hasti ollapse were developed from the 1970s
onwards (Pearle 1976, 1979; Ghirardi, Rimini and Weber 1986). As for
Einstein's on erns in 1927 about the nonlo ality of quantum theory (see
hapter 7), re-expressed in the famous EPR paper of 1935, matters ame
to a head in 1964 with the publi ation of Bell's theorem (Bell 1964). In
153
154
the losing de ades of the twentieth entury, after many stringent experimental tests showed that Bell's inequality was violated by entangled
quantum states, nonlo ality ame to be widely regarded as a entral fa t
of the quantum world.
Other on erns, voi ed by S hrdinger just a few years after the
fth Solvay onferen e, also eventually played a entral role after de ades of obs urity. S hrdinger's ` at paradox' of 1935 ame to dominate
dis ussions about the meaning of quantum theory. And the pe uliar
èntanglement' that S hrdinger had highlighted as a key dieren e
between lassi al and quantum physi s (S hrdinger 1935) eventually
found its pla e as a entral on ept in quantum information theory: as
well as being a matter of `philosophi al' on ern, entanglement ame to
be seen as a physi al resour e to be exploited for te hnologi al purposes,
and as a entral feature of quantum physi s that had been strangely
under-appre iated for most of the twentieth entury.
The interpretation of quantum theory is probably as ontroversial
now as it ever has been. Many workers now re ognise that standard
quantum theory entred as it is around the notion of `measurement'
requires a lassi al ba kground ( ontaining ma ros opi measuring
devi es), whi h an never be sharply dened, and whi h in prin iple does
not even exist. Even so, the operational approa h to the interpretation of
quantum physi s is still being pursued by some, in terms of new axioms
that onstrain the stru ture of quantum theory (Hardy 2001, 2002;
Clifton, Bub and Halvorson 2003). On the other hand, those who do
regard the ba kground problem as ru ial tend to assert that everything
in the universe mi ros opi systems, ma ros opi equipment, and even
human experimenters should in prin iple be des ribed in a unied
manner, and that `measurement' pro esses must be regarded as physi al
pro esses like any other. Approa hes of this type in lude: the Everett
interpretation (Everett 1957), whi h is being subje ted to in reasing
s rutiny at a foundational level (Saunders 1995, 1998; Deuts h 1999;
Walla e 2003a,b); the pilot-wave theory of de Broglie and Bohm (de
Broglie 1928; Bohm 1952a,b; Bohm and Hiley 1993; Holland 1993), whi h
is being pursued and developed more than ever before (Cushing, Fine
and Goldstein 1996; Pearle and Valentini 2006; Valentini 2007); ollapse
models (Pearle 1976, 1979, 1989; Ghirardi, Rimini and Weber 1986),
whi h are being subje ted to ever more stringent experimental tests
(Pearle and Valentini 2006); and theories of ` onsistent' or `de oherent'
histories (Griths 1984, 2002; Gell-Mann and Hartle 1990; Hartle 1995;
Omns 1992, 1994).
5.2 The measurement problem today
155
Today, it is simply untenable to regard the views of Bohr and Heisenberg (whi h in any ase diered onsiderably from ea h other) as in any
sense standard or anoni al. The meaning of quantum theory is today
an open question, arguably as mu h as it was in O tober 1927.

The problem of measurement and the observer is the
problem of where the measurement begins and ends, and
where the observer begins and ends. .... I think, that
when you analyse this language that the physi ists have
fallen into, that physi s is about the results of
observations you nd that on analysis it evaporates,
and nothing very lear is being said.
J. S. Bell (1986, p. 48)
The re urring puzzlement over the meaning of quantum theory often

entres around a group of related on eptual questions that usually ome
under the general heading of the `measurement problem'.
5.2.1 A fundamental ambiguity

As normally presented in textbooks, quantum theory des ribes experiments in a way that is ertainly pra ti ally su essful, but seemingly fundamentally ill-dened. For it is usually impli itly or expli itly assumed
that there is a lear boundary between mi ros opi quantum systems
and ma ros opi lassi al apparatus, or that there is a lear dividing
line between `mi ros opi indeniteness' and the denite states of our
lassi al ma ros opi realm. Yet, su h distin tions defy sharp and pre ise
formulation.
That quantum theory is therefore fundamentally ambiguous was argued
with parti ular larity by Bell. For example (Bell 1986, p. 54):
The formulations of quantum me hani s that you nd in the books involve
dividing the world into an observer and an observed, and you are not told
where that division omes on whi h side of my spe ta les it omes, for
example or at whi h end of my opti nerve.
The problem being pointed to here is the la k of a pre ise boundary

between the quantum system and the rest of the world (in luding the
apparatus and the experimenter).
A losely-related aspe t of the `measurement problem' is the need to
explain what happens to the denite states of the everyday ma ros opi
156
domain as one goes to smaller s ales. Where does ma ros opi deniteness give way to mi ros opi indeniteness? Does the transition o ur
somewhere between pollen grains and ma romole ules, and if so, where?
On whi h side of the line is a virus?
Nor an quantum ìndeniteness' or `fuzziness' be easily onned to
the atomi level. For ma ros opi obje ts are made of atoms, and so
inevitably one is led to doubt whether ro ks, trees, or even the Moon,
have denite ma ros opi states, espe ially when observers are not present. And this in the fa e of remarkable developments in twentieth entury astrophysi s and osmology, whi h have tra ed the origins of
stars, galaxies, helium and the other elements, to times long before
human observers existed.
The notion of a `real state of aairs' is familiar from everyday experien e: for example, the lo ation and number of ma ros opi bodies in
a laboratory. S ien e has shown that there is more to the real state of
things than is immediately obvious (for example, the ele tromagneti
eld). Further, it has been shown that the hara ter of the real state
of things hanges with s ale: on large s ales we nd planets, stars and
galaxies, while on small s ales we nd pollen grains, viruses, mole ules,
and atoms. Nevertheless, at least outside of the quantum domain, the
notion of `real state' remains. The ambiguity emphasised by Bell onsists
of the la k of a sharp boundary between the ` lassi al' domain, in whi h
`real state' is a valid on ept, and the `quantum domain', in whi h `real
state' is not a valid on ept.
Despite de ades of eort, this ambiguity remains unresolved within
standard textbook quantum theory, and many riti s have been led
to argue that the notion of real state should be extended, in some
appropriate way, into the quantum domain. Thus, for example, Bell
(1987, pp. 2930):
Theoreti al physi ists live in a lassi al world, looking out into a quantumme hani al world. The latter we des ribe only subje tively, in terms of pro edures and results in our lassi al domain. This subje tive des ription is ee ted
by means of quantum-me hani al state fun tions
.... . The lassi al world
of ourse is des ribed quite dire tly às it is'. .... Now nobody knows just
where the boundary between the lassi al and quantum domain is situated. ....
A possibility is that we nd exa tly where the boundary lies. More plausible to
me is that we will nd that there is no boundary. It is hard for me to envisage
intelligible dis ourse about a world with no lassi al part no base of given
events .... to be orrelated. On the other hand, it is easy to imagine that the
lassi al domain ould be extended to over the whole.
157
While Bell goes on to argue in favour of adding extra (`hidden')

parameters to the quantum formalism, for our purposes the key point
being made here is the need to extend the notion of real state into
the mi ros opi domain. This might indeed be a hieved by introdu ing
hidden variables, or by other means (for example, the Everett approa h).
Whatever form the theory may take, the real ma ros opi states onsidered in the rest of s ien e should be part of a unied des ription
of mi ros opi and ma ros opi phenomena what Bell (1987, p. 30)
alled à homogeneous a ount of the world'.
There are in fa t, as we have mentioned, several well-developed proposals for su h a unied or homogeneous (or `realist') a ount of the
world: the pilot-wave theory of de Broglie (1928) and Bohm (1952a,b);
theories of dynami al wave-fun tion ollapse (Pearle 1976, 1979, 1989;
Ghirardi, Rimini and Weber 1986); and the many-worlds interpretation
of Everett (1957). The available proposals that have broad s ope assume
that the wave fun tion is a real obje t that is part of the stru ture of
an individual system. At the time of writing, it is not known if realist
theories may be onstru ted without this feature.a
5.2.2 Measurement as a physi al pro ess: quantum theory

`without observers'
Another losely-related aspe t of the measurement problem is the question of how quantum theory may be applied to the pro ess of measurement
itself. For it seems ines apable that it should be possible (in prin iple)
to treat apparatus and observers as physi al systems, and to dis uss the
pro ess of measurement in purely quantum-theoreti al terms. However,
attempts to do so are notoriously ontroversial and apt to result in
paradox and onfusion.
For example, in the paradox of `Wigner's friend' (Wigner 1961), an
experimenter A (Wigner) possesses a box ontaining an experimenter B
(his friend) and a mi ros opi system S. Suppose S is initially in, for
example, a superposition of energy states
1
|0 i = (|E1 i + |E2 i)
2
a For example, a ording to the sto hasti hidden-variables theory of Fnyes (1952)
and Nelson (1966), the wave fun tion merely provides an emergent des ription of
probabilities. However, despite appearan es, it seems that for te hni al reasons this
theory is awed and does not really reprodu e quantum theory: the S hrdinger
equation is obtained only for ex eptional (nodeless) wave fun tions (Wallstrom
1994; Pearle and Valentini 2006).
158
and that the whole box is initially in a state |0 i = |B0 i|0 i (idealising
B as initially in a pure state |B0 i). Let B perform an ideal energy
measurement on S. If A does not arry out any measurement, then from
the point of view of A the quantum state of the whole box evolves ontinuously (a ording to the S hrdinger equation) into a superposition
of states
1
|(t)i = (|B1 i |E1 i + |B2 i |E2 i)
(1)
2
(where |Bi i is a state su h that B has found the energy value Ei ).
Now, if A wished to, ould he (in prin iple) at later times, by appropriate experiments on the whole box, observe interferen e ee ts
involving both bran hes of the superposition in (1)? If so, ould this
be onsistent with the point of view of B, a ording to whi h the energy
measurement had a denite result?
One may well question whether the above s enario is realisti , even
in prin iple (given the resour es in our universe). For example, one
might question whether a box ontaining a human observer ould ever
be su iently isolated for environmental de oheren e to be negligible.
However, if the above èxperimenter B' were repla ed by an automati
devi e or ma hine, the s enario may indeed be ome realisti , depending
on the possibility of isolating the box to su ient a ura y.a
Most s ientists agree that ma ros opi equipment is subje t to the
laws of physi s, just like any other system, and that it should be possible
to des ribe the operation of su h equipment purely in terms of the most
fundamental theory available. There is somewhat less onsensus over the
status of human experimenters as physi al systems. Some physi ists have
suggested, in the ontext of quantum physi s, that human beings annot
be treated as just another physi al system, and that human ons iousness plays a spe ial role. For example, Wigner (1961) on luded from his
paradox that `the being with a ons iousness must have a dierent role
in quantum me hani s than the inanimate measuring devi e', and that
for a system ontaining a ons ious observer `the quantum me hani al
equations of motion annot be linear'.
Wigner's on lusion, that living beings violate quantum laws, seems
in reasingly in redible given the impressive progress made in human
biology and neuros ien e, in whi h the human organism in luding
the brain is treated as (ultimately) a omplex ele tro- hemi al sya It is perhaps worth remarking that, even if de oheren e has a role to play here,
one has to realise what the problem is in order to understand whether and how
de oheren e might ontribute to a solution ( f. Ba iagaluppi 2005).
159
stem. There is no eviden e that human beings are able to violate, for
example, the laws of gravity, or of thermodynami s, or basi prin iples of
hemistry, and the on lusion that human beings in parti ular should be
outside the domain of quantum laws seems di ult to a ept. An alternative on lusion, of ourse, is that something is missing from orthodox
quantum theory.
Assuming, then, that human experimenters and their equipment may
in prin iple be regarded as physi al systems subje t to the usual laws,
their intera tion with mi rosystems ought to be analysable, and the
pro ess of measurement ought to be treatable as a physi al pro ess like
any other. One an then ask if, over an ensemble of similar experiments,
it would be possible in prin iple for the external experimenter A to
observe (at the statisti al level) interferen e ee ts asso iated with both
terms in (1). To deny this possibility would be to laim (with Wigner)
that a box ontaining a human being violates the laws of quantum
theory. To a ept the possibility would seem to imply that, at time
t before experimenter A makes a measurement, there was (at least
a ording to A) no matter of fa t about the result of B's observation,
notwithstanding the expli it supposition that B had indeed arried out
an observation by time t.
It is sometimes said that Wigner's paradox may be evaded by noting
that, if the external experimenter A a tually performs an experiment on
the whole box that reveals interferen e between the two bran hes of (1),
then this operation will destroy the memory the internal experimenter
B had of obtaining a parti ular result, so that there is no ontradi tion.
But this misses the point. For while it is true that B will then not have
any memory of having obtained a parti ular experimental result, the
ontradi tion remains with there being a purported matter of fa t (at
time t) as to the result of B's observation, regardless of whether or not
B has subsequently forgotten it. (Note that in this dis ussion we are
talking about matters of fa t, not for mi rosystems, but for ma ros opi
experimental results.)
Let us examine the reasoning behind Wigner's paradox more losely.
We take it that experimenter B agreed beforehand to enter the box
and perform an energy measurement on the mi ros opi system S; and
that it was further agreed that after su ient time had elapsed for B
to perform the measurement, A would de ide whether or not to arry
out an experiment showing interferen e between the two bran hes of (1).
Considering an ensemble of similar experiments, the paradox onsists of
160
a ontradi tion between the following statements on erning the physi al

state of B just before A de ides what to do:
(I) There is no denite state of B, be ause A an if he wishes perform
measurements showing the presen e of interferen e between dierent
states of B.
(II) There is a denite state of B, be ause B is a human experimenter like
any other, and be ause instead of testing for interferen e A an simply
ask B what he saw.
The argument in (I) is the familiar one from standard quantum theory,
applied to the unusual ase of a box ontaining an experimenter. The
argument in (II) is unusual: it requires omment and elaboration.
Be ause B is a human experimenter like any other, we are driven
to onsider the theoreti al possibility that, in the distant future, some
`super-experimenter' ould de ide to perform an interferen e experiment
on a `box' ontaining us and our equipment. What would happen to our
urrent (ma ros opi ally-re orded) experimental fa ts on erning for
example the out ome of a spin measurement performed in the laboratory? To be sure, our re ords of these experiments ould one day be erased,
but it would be illogi al to suppose that the fa t of these experiments
having been arried out (with denite results) ould ever be hanged.
Unless we a ept (II), we are in danger of en ountering the paradox that
fa ts about what we have done in the laboratory today might later turn
out not to be fa ts.
Further support for (II) omes from Wigner's original argument, whi h
entred around the assumed reality of other minds. (Wigner did not regard solipsism as worthy of serious onsideration.) From this assumption
Wigner inferred that, whatever the ir umstan es, if one asks a `friend'
what he saw, the answer given by the friend must have been, as Wigner
put it, àlready de ided in his mind, before I asked him'. But then, if
A de ides not to perform an interferen e experiment on the box, and
simply asks B what he saw, A is obliged to take B's answer as indi ative
of the state of B's mind before A asked the question indeed, before
A de ided on whether or not to perform an interferen e experiment. For
Wigner, a superposition of the form (1) is una eptable for a system
ontaining a human experimenter or `friend', be ause it implies that the
friend `was in a state of suspended animation before he answered my
question'.
As Wigner presented it, the argument is based on the assumption that
161
a ons ious being will always have a denite state of ons iousness. In
orthodox quantum theory, of ourse, one might dismiss as `meaningless'
the question of whether the friend's ons iousness ontained one impression or the other (B1 or B2 ) before he was asked. However, as Wigner
put it, `to deny the existen e of the ons iousness of a friend to this
extent is surely an unnatural attitude, approa hing solipsism'.
Finally, on the topi of Wigner's paradox, it is important to emphasise
the distin tion between `matters of fa t' on the one hand, and `memories'
(true or false) on the other a distin tion that is omparable to the
distin tion between fa ts and opinions, or between truth and belief, or
between reality and appearan e, distin tions that form part and par el
of the s ienti method. Thus, again, while an experimenter's memory
of having obtained a ertain result might be erased in the future, the
fa t that he on e obtained a ertain result will ne essarily remain a fa t:
to assert otherwise would be a logi al ontradi tion.
A further, more subtle motivation for treating measurement as a physi al pro ess omes from onsidering the very nature of `measurement'. As
is well known to philosophers, and to experimental physi ists, the pro ess
of measurement is `theory-laden'. That is, in order to know how to
arry out a measurement orre tly, or how to design a spe i measuring
apparatus orre tly, some prior body of theory is required: in parti ular,
one needs some understanding of how the equipment fun tions, and
how it intera ts with the system being examined. For this reason, it
is di ult to see how the pro ess of quantum measurement an be
properly understood, without some prior body of theory that des ribes
the equipment itself and its intera tion with the `system'. And sin e
the equipment usually belongs to the denite ma ros opi realm, and
the `system' often does not, a proper understanding seems to require a
`homogeneous a ount of the world' as dis ussed above, that is, a theory
in whi h an obje tive a ount is provided not only of the ma ros opi
apparatus, but also of the mi rosystem and its intera tion with the
apparatus.
A ommon on lusion, then, is that a oherent a ount of quantum
measurement requires that quantum theory be somehow extended from a
theory of mi rosystems to a universal physi al theory with an unbounded
domain of appli ation, with our everyday ma ros opi realism being
somehow extended to the mi ros opi level. Given su h a well-dened
and universal physi al theory, whose subje t matter onsists of the real
states of the world as a whole, it would be possible in prin iple to use the
theory to analyse the pro ess of measurement as a physi al pro ess like
162
any other (just as, for example, lassi al ele trodynami s may be used
to analyse the pro ess involving for es exerted by magneti elds
by whi h an ammeter measures an ele tri urrent).
Thus, for example, pilot-wave theory, or the Everett interpretation, or
ollapse models, may be applied to situations where a quantum measurement is taking pla e. If the theory provides an unambiguous a ount
of obje tive pro esses in general, it will provide an unambiguous a ount of the quantum measurement pro ess in parti ular. The result is
a quantum theory `without observers', in the sense that observers are
physi al systems obeying the same laws as all other systems, and do not
have to be added to the theory as extra-physi al elements.
Con lusions as to what is a tually happening during quantum measurements will, of ourse, depend on the details of the theory. For example, onsider again Wigner's s enario above. In the Everett interpretation, B's observation within the box has two results, and there is
no ontradi tion if the external experimenter A subsequently observes
interferen e between them. In de Broglie-Bohm theory, B's observation
has only one result sele ted by the a tual onguration, but even so
the empty wave pa ket still exists in onguration spa e, and an in
prin iple re-overlap with (and hen e interfere with) the o upied pa ket
if appropriate Hamiltonians are applied.
5.2.3 Quantum osmology

Further losely-related questions, again broadly under the heading of
the `measurement problem', on ern the des ription of the distant past
before human beings and other life forms evolved on Earth, and indeed
the des ription of the universe as a whole in epo hs before life existed.
While the basi theoreti al foundations of big-bang osmology had already been laid by 1927 (through the work of Friedmann and Lematre),
at that time any suggestion of the need to provide a quantum-theoreti al
a ount of the early universe ould easily have been dismissed as being
of no pra ti al or experimental import. By the 1980s, however, with
the development of inationary osmology (Guth 1981), the theoreti al
question be ame a pra ti al one, with observational impli ations.
A ording to our urrent understanding, the small non-uniformities of
temperature observed in the osmi mi rowave ba kground originated
from lassi al density perturbations in the early (and approximately
homogenous) universe (Padmanabhan 1993). And a ording to inationary theory, those early lassi al density perturbations originated from
163
quantum u tuations at even earlier times (Liddle and Lyth 2000). Here
we have an example of a osmologi al theory in whi h a `quantum-to lassi al transition' o urred long before life (or even galaxies) developed,
and whose details have left an imprint on the sky that an be measured
today. This is the measurement problem on a osmi s ale (Kiefer,
Polarski and Starobinsky 1998; Perez, Sahlmann and Sudarsky 2006;
Valentini 2006).
But the tension between `Copenhagen' quantum theory and the requirements of osmology was felt long before osmology matured as an
experimental s ien e. Thus, for example, in his pioneering work in the
1960s on quantum gravity, when it ame to applying the theory to a
losed universe DeWitt wrote (DeWitt 1967, p. 1131):
The Copenhagen view depends on the assumed
a priori
existen e of a lassi al
level to whi h all questions of observation may ultimately be referred. Here,

however, the whole universe is the obje t of inspe tion; there is no lassi al
vantage point, and hen e the interpretation question must be re-argued from
the beginning.
DeWitt went on to argue (pp. 11402) that, in the absen e of a lassi al

level, the Everett interpretation should be adopted. A ording to DeWitt
(p. 1141):
Everett's view of the world is a very natural one to adopt in the quantum
theory of gravity, where one is a ustomed to speak without embarrassment
of the `wave fun tion of the universe'.
While DeWitt expresses a preferen e for the Everett interpretation, for

our purposes the entral point being made is that, if the whole universe
is treated as a quantum obje t, with no denite ( lassi al) ba kground
`to whi h all questions of observation may ultimately be referred', then
the physi s be omes unintelligible unless some form of real state (or
ontology) is as ribed to the quantum obje t. The Everett interpretation
provides one way, among others, to do this.a
Everett himself, in 1957, had already ited the quantum theory of
osmology as one of his main motivations for going beyond what he alled the ` onventional or external observation formulation of quantum
me hani s' (Everett 1957, p. 454). Everett's general motivation was the
need to des ribe the quantum physi s internal to an isolated system, in
a Everett's original formulation was of ourse open to a number of riti isms,
in parti ular on erning the notion of `world' and the idea of probability.
Su h riti isms are arguably being met only through more re ent developments
(Saunders 1995, 1998; Deuts h 1999; Walla e 2003a,b).
164
parti ular one ontaining observers. A losed universe was a spe ial ase
of su h a system, and one that arguably would have to be onsidered
as the s ien e of osmology progressed (as has indeed proved to be the
ase). Thus Everett wrote (p. 455):
How is one to apply the onventional formulation of quantum me hani s to the
spa e-time geometry itself ? The issue be omes espe ially a ute in the ase of
a losed universe. There is no pla e to stand outside the system to observe it.
There is nothing outside it to produ e transitions from one state to another. ....
No way is evident to apply the onventional formulation of quantum me hani s
to a system that is not subje t to
external
observation. The whole interpretive
s heme of that formalism rests upon the notion of external observation. The
probabilities of the various possible out omes of the observation are pres ribed
ex lusively by Pro ess 1 [dis ontinuous wave fun tion ollapse.
In more re ent years, similar on erns have motivated the development

of a `generalised quantum me hani s' based on ` onsistent' or `de oherent' histories (Griths 1984, 2002; Gell-Mann and Hartle 1990; Hartle
1995; Omns 1992, 1994), an approa h that is also supposed to provide
a quantum theory `without observers', and without a presumed lassi al
ba kground, so as to be appli able to quantum osmology.
5.2.4 The measurement problem in `statisti al'

interpretations of
The measurement problem is often posed simply as the problem of how to
interpret a ma ros opi superposition of quantum states, su h as (pure)
states of S hrdinger's at. This way of posing the measurement problem
an be misleading, however, as it usually rests on the impli it assumption
(or suggestion) that the quantum wave fun tion is a real physi al
obje t identiable as a omplete des ription of an individual system. It
might be that is indeed a real obje t, but not a omplete des ription
(as in pilot-wave theory). Or, might not be a real obje t at all. Here
we fo us on the latter possibility.
The quantum wave fun tion might be merely a mathemati al tool
for al ulating and predi ting the measured frequen ies of out omes
over an ensemble of similar experiments. In whi h ase, it would be
immediately wrong to interpret a mathemati al superposition of terms
in as somehow orresponding to a physi al superposition of real states
for individual systems, and the `measurement problem' in the limited
sense just mentioned would be a pseudo-problem. This `statisti al in-
165
terpretation' of quantum theory has been hampioned in parti ular by

Ballentine (1970).
However, even in the statisti al interpretation, the `measurement problem' in the more general sense remains. For quantum theory is then
an in omplete theory that refers only to ensembles, and simply does
not fully des ribe individual quantum systems or their relation to real,
individual ma ros opi states. The statisti al interpretation gives no
a ount of what happens, for example, when an individual ele tron is
being measured: it talks only about the distribution of (ma ros opi allyregistered) measurement out omes over an ensemble of similar experiments. Nor does the statisti al interpretation provide any sharp delineation of the boundary between `ma ros opi ' obje ts with an individual
(non-ensemble) des ription and `mi ros opi ' obje ts with no su h des ription.a
In the statisti al interpretation, then, a solution of the measurement
problem in the general sense will require the development of a omplete
des ription of individual systems. This was Einstein's point of view
(Einstein 1949, pp. 6712):
The attempt to on eive the quantum-theoreti al des ription as the omplete
des ription of the individual systems leads to unnatural theoreti al interpretations, whi h be ome immediately unne essary if one a epts the interpretation
that the des ription refers to ensembles of systems and not to individual systems. .... [It appears unavoidable to look elsewhere for a omplete des ription
of the individual system .... .
Einstein was arguably the founder of the statisti al interpretation.b

It should be noted, however, that while Einstein's on lusion about the
nature of might turn out to be orre t, what seems to have been his
main argument for this on lusion now appears to be wrong, in that it
was based on what now appears to be a false premise the assumption
of lo ality. For example, in a letter to his friend Mi hele Besso, dated 8
O tober 1952, Einstein argued that the `quantum state' ould not be
a omplete hara terisation of the `real state' of an individual system,
on the following grounds:
A system
S2 ,
S12 ,
with known fun tion
whi h at time
12 ,
is omposed of subsystems
S1
and
are far away from ea h other. If one makes a ` omplete'
a There are dierent ways of onsidering a `statisti al' interpretation, depending

on one's point of view on erning the nature of probability. In any ase, the
`measurement problem' in the general sense still stands.
b Cf. Ballentine (1972).
166
S1 , this an be done in dierent ways .... . From the measuand the -fun tion 12 , one an determine .... the -fun tion
2 of the se ond system. This will take on dierent forms, a ording to the
kind of measurement applied to S1 .
measurement on
rement result
But this is in ontradi tion with assumption (1) [that the quantum state
hara terises the real state ompletely,
Then in fa t the measurement on
S2 ,
S1
if one ex ludes a tion at a distan e.
an have no inuen e on the real state of
and therefore a ording to (1) an have also no inuen e on the quantum
state of
S2
des ribed by
2 .
(Einstein and Besso 1972, pp. 4878, emphasis in
the original)
Einstein's argument hinges on the fa t that, in a lo al physi s, the

measurement made on S1 an have no ee t on the real state of S2 .a
With the development of Bell's theorem, however, it seems to be
beyond reasonable doubt that quantum physi s is not lo al (if one
assumes the absen e of ba kwards ausation or of many worlds). For if
lo ality is assumed, one may use the EPR argument to infer determinism
for the out omes of quantum measurements at widely-separated wings
of an entangled state.b Following further reasoning by Bell (1964), one
may then show that any lo al and deterministi ompletion of quantum
theory annot reprodu e quantum orrelations for all measurements
on entangled states. Therefore, lo ality ontradi ts quantum theory.c
Be ause the premise of Einstein's argument ontradi ts quantum theory,
the argument annot be used to infer anything about quantum theory
or about the nature of . Thus, Einstein's argument does not establish
the `statisti al' or ènsemble' nature of .d
a The notion of lo ality that Einstein uses here is, to be pre ise, a ombination of
the prin iples of `separability' (that widely-separated systems have lo ally-dened
real states) and of `no a tion at a distan e'. Cf. Howard (1990).
b Ìt is important to note that .... determinism .... in the EPR argument .... is not
assumed but inferred [from lo ality. .... It is remarkably di ult to get this point
a ross, that determinism is not a presupposition of the analysis' (Bell 1987, p. 143,
itali s in the original).
See, however, Fine (1999) for a dissenting view.
d Einstein's argument above has re ently been revived by Fu hs (2002) (who is,
however, not expli it about the ompleteness or in ompleteness of quantum
theory). Fu hs states (p. 9) that Einstein `was the rst person to say in absolutely
unambiguous terms why the quantum state should be viewed as information ....
. His argument was simply that a quantum-state assignment for a system an be
for ed to go one way or the other by intera ting with a part of the world that
should have no ausal onne tion with the system of interest'. Fu hs then quotes
at length (p. 10) the above letter by Einstein. Later in the same paper, Fu hs
writes (p. 39): `Re all what I viewed to be the most powerful argument for the
quantum state's subje tivity the Einsteinian argument of [the above letter.
Sin e [for entangled systems we an toggle the quantum state from a distan e, it
must not be something sitting over there, but rather something sitting over here:
It an only be our information about the far-away system'. Again, the premise
167
of this Einsteinian argument lo ality is nowadays no longer reasonable (as of

ourse it was in 1952), and so the argument annot be used to infer the subje tive
or epistemi nature of the quantum state.
6
Interferen e, superposition, and wave
pa ket ollapse
6.1 Probability and interferen e

A ording to Feynman (1965, hap. 1, p. 1), single-parti le interferen e
is `the only mystery' of quantum theory. Feynman onsidered an experiment in whi h parti les are red, one at a time, towards a s reen with
two holes labelled 1 and 2. With both holes open, the distribution P12
of parti les at the ba kstop displays an os illatory pattern of bright and
dark fringes. If P1 is the distribution with only hole 1 open, and P2 is
the distribution with only hole 2 open, then experimentally it is found
that P12 6= P1 + P2 . A ording to the argument given by Feynman (as
well as by many other authors), this result is inexpli able by ` lassi al'
reasoning.
By his presentation of the two-slit experiment (as well as by his development of the path-integral formulation of quantum theory), Feynman
popularised the idea that the usual probability al ulus breaks down in
the presen e of quantum interferen e, where it is probability amplitudes
(and not probabilities themselves) that are to be added. As pointed
out by Koopman (1955), and by Ballentine (1986), this argument is
mistaken: the probability distributions at the ba kstop P12 , P1 and P2
are onditional probabilities with three distin t onditions (both slits
open, one or other slit losed), and probability al ulus does not imply
any relationship between these. Feynman's argument notwithstanding,
standard probability al ulus is perfe tly onsistent with the two-slit
experiment.
In his inuential le tures on physi s, as well as asserting the breakdown
of probability al ulus, Feynman laimed that no theory with parti le
traje tories ould explain the two-slit experiment. This laim is still
168
169
found in many textbooks.a From a histori al point of view, it is remarkable indeed that single-parti le interferen e ame to be widely regarded
as in onsistent with any theory ontaining parti le traje tories: for as
we have seen in hapter 2, in the ase of ele trons this phenomenon was
in fa t rst predi ted (by de Broglie) on the basis of pre isely su h a
theory.
As we shall now dis uss, in his report at the fth Solvay onferen e
de Broglie gave a lear and simple explanation for single-parti le interferen e on the basis of his pilot-wave theory; and the extensive dis ussions
at the onferen e ontain no sign of any obje tion to the onsisten y of
de Broglie's position on this point.
As for S hrdinger's theory of wave me hani s, in whi h parti les were
supposed to be onstru ted out of lo alised wave pa kets, in retrospe t it
is di ult to see how single-parti le interferen e ould have been a ounted for. It is then perhaps not surprising that, in Brussels in 1927, no
spe i dis ussion of interferen e appears in S hrdinger's ontributions.
Born and Heisenberg, on the other hand, do dis uss interferen e in
their report, from the point of view of their `quantum me hani s'. And,
they do onsider the question of the appli ability of probability al ulus.
The views they present are, interestingly enough, rather dierent from
the views usually asso iated with quantum me hani s today. In parti ular, as we shall see below, a ording to Born and Heisenberg there was
(in a very spe i sense) no oni t between quantum interferen e and
the ordinary probability al ulus.
Interferen e was also onsidered in the general dis ussion, in parti ular
by Dira and Heisenberg: this latter material will be dis ussed later, in
se tion 6.3.
6.1.1 Interferen e in de Broglie's pilot-wave theory

At the fth Solvay onferen e, the subje t of interferen e was addressed
from a pilot-wave perspe tive by de Broglie in his report. In his se tion
5, `The interpretation of interferen e', de Broglie onsidered interferen e
experiments with light of a given frequen y . For a guiding wave
of phase and amplitude a, de Broglie took the photon velo ity to be
a For example, Shankar (1994) dis usses the two-slit experiment at length in his
hapter 3, and laims (p. 111) that the observed single-photon interferen e pattern
` ompletely rules out the possibility that photons move in well-dened traje tories'.
Further, a ording to Shankar (p. 112): Ìt is now widely a epted that all parti les
are des ribed by probability amplitudes (x), and that the assumption that they
move in denite traje tories is ruled out by experiment'.
170
Interferen e, superposition, and ollapse

2
c
given by v = h
, while the probability distribution was taken to be
2
= consta . As de Broglie had pointed out, the latter distribution is
preserved over time by the assumed motion of the photons. Therefore,
the usual interferen e and dira tion patterns follow immediately. To
quote de Broglie (p. 384):
the bright and dark fringes predi ted by the new theory will oin ide with
those predi ted by the old [that is, by lassi al wave opti s.
De Broglie also pointed out that his theory gave the orre t bright and
dark fringes for photon interferen e experiments, regardless of whether
the experiments were performed with an intense or a very feeble sour e.
As he put it (p. 384):
one an do an experiment of short duration with intense irradiation, or an
experiment of long duration with feeble irradiation .... if the light quanta do
not a t on ea h other the statisti al result must evidently be the same.
De Broglie's dis ussion here addresses pre isely the supposed di ulty
highlighted mu h later by Feynman. It is noteworthy that a lear and
simple answer to what Feynman thought was `the only mystery' of
quantum me hani s was published as long ago as the 1920s.
Even so, for the rest of the twentieth entury, the two-slit experiment
was widely ited as proof of the non-existen e of parti le traje tories
in the quantum domain. Su h traje tories were thought to imply the
relation P12 = P1 + P2 , whi h is violated by experiment. As Feynman
(1965, hap. 1, p. 6) put it, on the basis of this argument it should
ùndoubtedly' be on luded that: Ìt is not true that the ele trons go
either through hole 1 or hole 2'. Feynman also suggested that, by 1965,
there had been a long history of failures to explain interferen e in terms
of traje tories:
Many ideas have been on o ted to try to explain the urve for
P12
[that
is, the interferen e pattern in terms of individual ele trons going around in
ompli ated ways through the holes. None of them has su eeded. (Feynman
1965, hap. 1, p. 6)
Yet, de Broglie's onstru tion is so simple as to be almost trivial: the

quantum probability density ||2 for parti le position obeys a ontinuity
equation, with a lo al probability urrent; if the traje tories follow the
ow lines of the quantum urrent then, by onstru tion, an in ident
distribution |0 |2 of parti les will ne essarily evolve into a distribution
2
|| at the ba kstop with interferen e or dira tion, as the ase may
be, depending on the potential in whi h the wave evolves.
171
Not only did Feynman laim, wrongly, that no one had ever su eeded in explaining interferen e in terms of traje tories; he also gave an
argument to the ee t that any su h explanation was impossible:
Suppose we were to assume that inside the ele tron there is some kind of
ma hinery that determines where it is going to end up. That ma hine must
also
determine whi h hole it is going to go through on its way. But .... what
is inside the ele tron should not be dependent .... upon whether we open or
lose one of the holes. So if an ele tron, before it starts, has already made up
its mind (a) whi h hole it is going to use, and (b) where it is going to land,
we should nd
P1
for those ele trons that have hosen hole 1,
that have hosen hole 2,
and ne essarily
the sum
P1 + P2
P2
for those
for those that arrive
through the two holes. There seems to be no way around this. (Feynman 1965,
hap. 1, p. 10)
Feynman's argument assumes that the motion of the ele tron is unae ted by opening or losing one of the holes. This assumption is violated
in pilot-wave theory, where the form of the guiding wave behind the
two-slit s reen does depend on whether or not both slits are open.
A similar assumption is made in the dis ussion of the two-slit experiment by Heisenberg (1962), in hapter III of his book Physi s and
Philosophy. Heisenberg onsiders single photons in ident on a s reen
with two small holes and a photographi plate on the far side, and gives
the familiar argument that the existen e of parti le traje tories implies
the non-interfering result P1 + P2 . As Heisenberg puts it:
If [a single photon goes through the rst hole and is s attered there, its
probability for being absorbed at a ertain point of the photographi plate
annot depend upon whether the se ond hole is losed or open. (Heisenberg
1962)
This assertion is denied by pilot-wave theory, whi h provides a simple

ounterexample to Heisenberg's on lusion that `the statement that any
light quantum must have gone either through the rst or through the
se ond hole is problemati and leads to ontradi tions'.
Finally, we note that interferen e was also onsidered by Brillouin
(pp. 401 .) in the dis ussion following de Broglie's report, for the ase
of photons ree ted by a mirror. Brillouin drew a gure (p. 403), with
a sket h of a photon traje tory passing through an interferen e region.
To our knowledge, plots of traje tories in ases of interferen e did not
appear again in the literature until the pioneering numeri al work by
Philippidis, Dewdney and Hiley (1979).
172
6.1.2 Interferen e in the `quantum me hani s' of Born and

Heisenberg
The subje t of interferen e was onsidered by Born and Heisenberg, in
their report on quantum me hani s (pp. 423 f.), for the ase of an atom
initially in a superposition
X
|(0)i =
(1)
cn (0) |ni
n
of energy states |ni, with oe ients cn (0) = |cn (0)| ein and eigenvalues
En . The S hrdinger equation implies a time evolution
X
cn (t) =
(2)
Snm (t)cm (0)
m
with (in modern notation) Snm (t) = hn| U (t, 0) |mi, where U (t, 0) is the
evolution operator. In the spe ial ase where cm (0) = mk for some k , we
2
2
2
have |cn (t)| = |Snk (t)| , and Born and Heisenberg interpret |Snk (t)| as
a transition probability. They also draw the on lusion that `the |cn (t)|2
must be the state probabilities' (p. 423).
Born and Heisenberg seem to adopt a statisti al interpretation, a ording to whi h the system is always in a denite energy state, with jump
2
2
probabilities |Snk (t)| and o upation (or `state') probabilities |cn (t)|
( f. se tion 3.4.6). This is stated quite expli itly (p. 422):
From the point of view of Bohr's theory a system an always be in only
one
quantum state. .... A ording to Bohr's prin iples it makes no sense to say
a system is simultaneously in several states. The only possible interpretation
seems to be statisti al: the superposition of several eigensolutions expresses
that through the perturbation the initial state an go over to any other
quantum state .... .
Note that this is quite dierent from present-day quantum me hani s, in

whi h a system des ribed by the superposition (1) would not normally
be regarded as always o upying only one energy state.
At the same time, Born and Heisenberg re ognise a di ulty (p. 423):
Here, however, one runs into a di ulty of prin iple that is of great importan e,
as soon as one starts from an initial state for whi h not all the
cn (0)
ex ept
one vanish.
The di ulty, of ourse, is that for an initial superposition the nal
probability distribution is given by
X
2

|cn (t)|2 =
(3)
Snm (t)cm (0)
m
173
|cn (t)|2 =
(4)
as opposed to
X
m
|Snm (t)|2 |cm (0)|2
whi h, as Born and Heisenberg remark, òne might suppose from the
usual probability al ulus'. (In standard probability al ulus, of ourse,
(4) expresses |cn (t)|2 as a sum over onditional transition probabilities
2
2
|Snm (t)| weighted by the initial population probabilities |cm (0)| .)
While Born and Heisenberg refer to (3) as the `theorem of the interferen e of probabilities', they make the remarkable assertion that there is
in fa t no ontradi tion with the usual rules of probability al ulus, and
2
that the |Snm | may still be regarded as ordinary probabilities. Further,
and equally remarkably, it is laimed that in any ase where the state
probabilities |cn |2 are experimentally established, the presen e of the
unknown phases n makes the interfering expression (3) redu e to the
non-interfering expression (4) (p. 423):
.... it should be noted that this ìnterferen e' does not represent a ontradi tion
with the rules of the probability al ulus, that is, with the assumption that
2
the |Snk | are quite usual probabilities. In fa t, .... [(4) follows from the
on ept of probability .... when and only when the relative number, that is, the
|cn |2 of the atoms in the state n, has been established beforehand
probability
experimentally.
In this ase the phases
are unknown in prin iple, so that
[(3) then naturally goes over to [(4).... .
Here, Born and Heisenberg refer to Heisenberg's (re ently-published)

un ertainty paper, whi h ontains a similar laim. There, Heisenberg
onsiders a Stern-Gerla h atomi beam passing through two su essive
regions of eld inhomogeneous in the dire tion of the beam (so as to
indu e transitions between energy states without separating the beam
into omponents). If the input beam is in a denite energy state then
the beam emerging from the rst region will be in a superposition. The
probability distribution for energy emerging from the se ond region will
then ontain interferen e as in (3), where the ìnitial' superposition (1)
is now the state emerging from the rst region. Heisenberg asserts that, if
the energy of an atom is a tually measured between the two regions, then
be ause of the resulting perturbation `the phase of the atom hanges by
amounts that are in prin iple un ontrollable' (Heisenberg 1927, pp. 183
4), and averaging over the unknown phases in the nal superposition
yields a non-interfering result.
The same example, with the same phase randomisation argument, is
also given by Heisenberg (1930b) in his book The Physi al Prin iples of
174
the Quantum Theory ( hapter IV, se tion 2), whi h was based on le tures
delivered at Chi ago in 1929. Heisenberg asserts (p. 60) that an energy
measurement for an atom in the intermediate region `will ne essarily
alter the phase of the de Broglie wave of the atom in state m by an
unknown amount of order of magnitude one', so that in applying the
(analogue of the) interfering expression (3) ea h term in the sum `must
thus be multiplied by the arbitrary fa tor exp (im ) and then averaged
over all values of m '.
From a modern perspe tive, this argument seems strange and unfamiliar, and indeed quite wrong. However, the argument makes rather more
sense, if one re ognises that the `quantum me hani s' des ribed by Born
and Heisenberg is not quantum me hani s as we usually know it today.
In parti ular, the theory as they present it appears to ontain no notion
of wave pa ket ollapse (or state ve tor redu tion).
The argument given by Born and Heisenberg amounts to saying, in
modern language, that if the energies of an atomi population have
a tually been measured, then one will have a mixture of states of the
superposed form (1), with randomly-distributed phases n . Su h a mixture is indeed statisti ally equivalent to a mixture of energy states |ni
with weights |cn (0)|2 , be ause the density operators are the same:
!
X
Y 1 Z
|cn (0)| |cm (0)| ei(n m ) |ni hm| =
dk
2
n,m
k
X
2
|cn (0)| |ni hn| . (5)
n
However, from a modern point of view, if one did measure the atomi
energies and nd the value En with frequen y pn , the resulting total
ensemble would naturally be represented by a density operator =
P
n pn |ni hn|, and there would seem to be no parti ular reason to rewrite
this in terms of the alternative de omposition on the left-hand side of (5)
(with |cn (0)|2 = pn and random phases ein ); though of ourse one ould
if one wished to. What is more, an a tual in onsisten y would appear if
having measured the atomi energies one sele ted a parti ular atom
that was found to have energy Em : a subsequent and immediate energy
measurement for this parti ular atom should again yield the result Em
with ertainty, as is onsistent with the usual representation of the atom
by the state |mi, and this ertainty would be in onsistent with what
appears to be (at least in ee t) the proposed representation of the atom
P
by a state n |cn (0)| ein |ni with randomised phases n . Indeed, for any
175
subensemble omposed of the latter states, all the energy values present
in the sum will be possible out omes of subsequent and immediate energy
measurements.a
This in onsisten y arises, however, if one applies the modern notion of
state ve tor ollapse a notion that, upon lose examination, appears
to be quite absent from the theory presented by Born and Heisenberg.
Instead of applying the usual ollapse rule, Born and Heisenberg seem
to interpret the quantities |Snm |2 as `quite usual' transition probabilities
in all ir umstan es, even in the presen e of interferen e. On this view,
then, an atom that has been found to have energy Em an be represented
P
by a state n |cn (0)| ein |ni with randomised phases n , and the probability of obtaining a value En in an immediately su essive measurement
is given not by |cn (0)|2 (as would follow from the usual ollapse rule)
but by the transition probability |Snm |2 where the latter does indeed
approa h nm as the time interval between the two energy measurements
tends to zero, so that the above ontradi tion does not in fa t arise.
Considering now the whole atomi ensemble, if the energies of the
atoms have indeed been measured, then using the |Snm |2 as transition
probabilities and the |cm (0)|2 as population probabilities, appli ation of
the probability al ulus gives the non-interfering result (4). As noted by
Born and Heisenberg, on their view exa tly the same result is obtained
from the ìnterfering' expression (3), with random phases m appearing
in the oe ients cm (0) = |cm (0)| eim .
If instead the atomi energies have not been measured, then, a ording
to Born and Heisenberg, the phases m have not been randomised and
the expression (3) does show interferen e, in ontradi tion with the
non-interfering expression (4). How do Born and Heisenberg re on ile
the breakdown of (4) with their laim that the ordinary probability
al ulus still holds, with the |Snm |2 being quite ordinary probabilities?
The answer seems to be that, if the energies have not been measured,
2
then the population probabilities |cm (0)| are in some sense ill-dened,
so that the usual probability formula (4) simply annot be applied: `[(4)
follows from the on ept of probability .... when and only when .... the
probability |cn |2 .... has been established beforehand experimentally'.
Born and Heisenberg seem to take an òperational' view of the poa Of ourse, if we do not subdivide the atomi ensemble on the basis of the
measured
energies, the total ensemble will be a mixture with density operator
P
2
n |cn (0)|P|ni hn| and will indeed be indistinguishable from the proposed mixture
in |ni with randomly-distributed phases. But there is
of states
n |cn (0)| e
nothing to prevent an experimenter from sele ting atoms a ording to their
measured energies.
176
pulation probabilities, in the sense that these are to be regarded as

meaningful only when dire tly measured. And the ited argument in
Heisenberg's un ertainty paper suggests that it is operationally impossible to have simultaneously well-dened phase relations and population
probabilities in the same experiment. This impossibility was presumably
regarded as omparable to the (operational) impossibility of having
simultaneously a well-dened position and momentum for a parti le.
What seems to be at work here, then, is some form of un ertainty
relation (or omplementarity) between population probabilities and phases: measurement of the former makes the latter ill-dened, and vi e
versa. Interferen e requires denite phase relationships, whi h pre lude
a well-dened population probability, so that the ordinary probability
al ulus annot be applied.a If instead the population probability has
a tually been measured, then the phases are indenite and averaging
over them washes out any interferen e.
The resulting viewpoint is ertainly remarkable. A ording to Born
2
and Heisenberg, in the presen e of interferen e, the quantities |Snm |
ontinue to be quite ordinary (transition) probabilities, while the quanti2
ties |cm (0)| annot be regarded as population probabilities rendering
the formula (4) inappli able. On this view, ordinary probability al ulus
is not violated; it is simply wrong to assert that the |cm (0)|2 represent
state probabilities in an interfering ase.
One may well obje t to this point of view on the grounds that, even
without measuring the energies, for a given preparation of the state
2
(1) the oe ients cm (0) and hen e the values of |cm (0)| will
be known (up to an overall phase). However, presumably, Born and
2
Heisenberg would have had to assert that while the |cm (0)| always
exist as mathemati al quantities, they annot be properly interpreted as
population probabilities unless the energies have been measured dire tly.
From a modern perspe tive, Born and Heisenberg's treatment of interferen e is surprising: in modern quantum me hani s, of ourse, in ases
2
where interferen e o urs the quantities |Snm | would not normally be
interpreted as `quite usual' transition probabilities; while in ases where
interferen e does not o ur, the non-interfering result (4) (as applied
a There is an analogy here with Heisenberg's view of ausality, expressed in his
un ertainty paper, a ording to whi h ausality annot be applied be ause its
premiss is generally false: `.... in the sharp formulation of the law of ausality, If
we know the present exa tly, we an al ulate the future, it is not the onsequent
that is wrong, but the ante edent. We annot in prin iple get to know the present
in all [its determining data' (Heisenberg 1927, p. 197).
6.2 Ma ros opi superposition: Born's dis ussion of the loud hamber
177
here) would not normally be regarded as arising from the interfering
result (3) through a pro ess of phase randomisation.
6.2 Ma ros opi superposition: Born's dis ussion of the loud

hamber
Quantum theory is normally understood to allow the `superposition of
distin t physi al states'. However, while `superposition' is well-dened as
a mathemati al term, it is hard to make sense of when applied to physi al
states that is, when the omponents in a superposition are regarded
as simultaneous physi al attributes of a single system. The need to
understand su h `physi al superposition' seems parti ularly a ute when
it is onsidered at the ma ros opi level. The di ulty here is losely
related to the question of wave pa ket ollapse: how is a mathemati al
superposition of ma ros opi ally-distin t states related to the denite
ma ros opi states seen in the laboratory?
The Wilson loud hamber, as used to observe the tra ks of -parti les,
was dis ussed at length by Born in the general dis ussion. The loud
hamber illustrates the measurement problem rather well, and is a good
example of how mi ros opi superpositions an be ome transferred to
the ma ros opi domain. It also illustrates how extending the formal
quantum des ription to the environment does not by itself alleviate the
measurement problem (despite many laims to the ontrary, for example
Zurek (1991)).a Remarkably, as we shall see, Born asserts that wave
pa ket ollapse is not required to dis uss the loud hamber.
The me hanism of the loud hamber is well known. The -parti les
pass through a supersaturated vapour. The passage of the parti les
auses ionisation, and the vapour ondenses around the ions, resulting
in the formation of tiny droplets. The droplets s atter light, making the
parti le tra ks visible.
If the emission of an -parti le is undire ted, so that the emitted
wave fun tion is approximately spheri al, how does one a ount for the
approximately straight parti le tra k revealed by the loud hamber?
In the general dis ussion, Born attributes this question to Einstein, and
asserts that to answer it (p. 482)
.... one must appeal to the notion of `redu tion of the probability pa ket'
developed by Heisenberg.
a For a summary of riti isms of environmental de oheren e as a solution to the

measurement problem, see Ba iagaluppi (2005).
178
This notion appears in Heisenberg's un ertainty paper, whi h had been

published in May of 1927. In se tion 3, entitled `The transition from
mi ro- to ma rome hani s', Heisenberg had des ribed how a lassi al
ele tron orbit ` omes into being' through repeated observation of the
ele tron position, using light of wavelength . A ording to Heisenberg,
the result of ea h observation an be hara terised by a probability
pa ket of width , where the pa ket spreads freely until the next observation: Èvery determination of position redu es therefore the wave
pa ket ba k to its original size ' (Heisenberg 1927, p. 186).
In the ase of the loud hamber, the ollapse of the wave pa ket
is applied repeatedly to the -parti le alone. Upon produ ing visible
ionisation, the wave pa ket of the -parti le ollapses, and then starts
to spread again, until further visible ionisation is produ ed, whereupon
ollapse o urs again, and so on. The probability for the resulting `traje tory' is on entrated along straight lines, a ounting for the observed
tra k in the loud hamber. As Born puts it (p. 482):
The des ription of the emission by a spheri al wave is valid only for as long
as one does not observe ionisation; as soon as su h ionisation is shown by the
appearan e of loud droplets, in order to des ribe what happens afterwards
one must `redu e' the wave pa ket in the immediate vi inity of the drops. One
thus obtains a wave pa ket in the form of a ray, whi h orresponds to the
orpus ular hara ter of the phenomenon.
Here, the loud hamber itself the ionisation, and the formation of
droplets from the vapour is treated as if it were an external ` lassi al
apparatus': only the -parti le appears in the wave fun tion.
6.2.1 Quantum me hani s without wave pa ket ollapse?

Born goes on to onsider if wave pa ket redu tion an be avoided by
treating the atoms of the loud hamber, along with the -parti le,
as a single system des ribed by quantum theory, a suggestion that he
attributes to Pauli (p. 482):
Mr Pauli has asked me if it is not possible to des ribe the pro ess without the
redu tion of wave pa kets, by resorting to a multi-dimensional spa e, whose
number of dimensions is three times the number of all the parti les present
.... . This is in fa t possible .... but this does not lead us further as regards the
fundamental questions.
Remarkably, Born laims that a treatment without redu tion is ìn

fa t' possible, and goes on to illustrate how, in his opinion, this an be
6.2 Ma ros opi superposition
179
done. As we shall see, Born seems to make use of a ` lassi al' probability
redu tion only, without any redu tion for the onguration-spa e wave
pa ket.
As for Born's referen e to Pauli, around the time of the Solvay onferen e Pauli believed that wave pa ket redu tion was needed only for
des ribing subsystems. This is lear from a letter he wrote to Bohr, on
17 O tober 1927 (one week before the Solvay meeting began), in whi h
Pauli omments on wave pa ket redu tion (Pauli 1979, p. 411):
This is pre isely a point that was not quite satisfa tory in Heisenberg [that
is, in the un ertainty paper; there the `redu tion of the pa kets' seemed a bit
mysti al. Now however, it is to be stressed that at rst su h redu tions are
not ne essary if one
in ludes
in the system all means of measurement. But
in order to des ribe observational results theoreti ally at all, one has to ask
what one an say about just a
part
of the whole system. And then from the
omplete solution one sees immediately that, in many ases (of ourse not
always), leaving out the means of observation an be formally repla ed by
su h redu tions.
Thus, at that time, Pauli thought that the redu tion was a formality
asso iated with an ee tive des ription of subsystems alone, and that if
the apparatus were in luded in the system then redu tion would not be
needed at all.
Born, then, presents a multi-dimensional treatment, in whi h atoms
in the loud hamber are des ribed by quantum theory on the same
footing as the -parti le. Born onsiders the simple ase of a model loud
hamber onsisting of just two atoms in one spatial dimension. There are
two ases, one with both atoms on the same side of the origin (where the
-parti le is emitted), and the other with the atoms on opposite sides of
the origin. The two `tubes' in Born's diagram (see his gure) represent,
for the two ases, the time development of the total (lo alised) pa ket in
3-dimensional onguration spa e. The oordinate x0 of the -parti le
is perpendi ular to the page, while x1 , x2 are the oordinates of the two
atoms. In ase I, the initial state is lo alised at x0 = 0, x1 , x2 > 0; in ase
II it is lo alised at x0 = 0, x1 > 0, x2 < 0. Ea h initial pa ket separates
into two pa kets moving in opposite dire tions along the x0 -axis. In the
rst ase the two ollisions (indi ated by kinks in the traje tory of the
pa keta ) take pla e on the same side of the origin; in the se ond ase they
take pla e on opposite sides. Note that, in both ases, both bran hes of
a Note that the motion o urs in one spatial dimension; a kink in the ongurationspa e traje tory shows that the orresponding atom has undergone a small spatial
displa ement.
180
the wave pa ket moving in opposite dire tions are shown; that is,
the omplete (ùn ollapsed') pa kets are shown in the gure.
Born remarks (p. 485) that:
To the `redu tion' of the wave pa ket orresponds the hoi e of one of the
two dire tions of propagation
+x0 , x0 ,
whi h one must take as soon as it is
established that one of the two points 1 and 2 is hit, that is to say, that the
traje tory of the pa ket has re eived a kink.
Here Born seems to be saying that, instead of redu tion, what takes pla e
is a hoi e of dire tion of propagation. But propagation of what? Born
presumably does not mean a hoi e of dire tion of propagation of the
wave pa ket, for that would amount to wave pa ket redu tion, whi h
Born at the outset has laimed is unne essary in a multi-dimensional
treatment. (And indeed, his gure shows the wave pa ket propagating
in both dire tions.) Instead, Born seems to be referring to a hoi e in
the dire tion of propagation of the system (whi h we would represent
by a point in onguration spa e). The wave pa ket spreads in both
dire tions, and determines the probabilities for the dierent possible
motions of the system. On e the dire tion of motion of the system
is established, by the o urren e of ollisions, an ordinary (` lassi al')
redu tion of the probability distribution o urs while the wave pa ket
itself is un hanged. In other words, the probabilities are updated but
the wave pa ket does not ollapse. This, at least, appears to be Born's
point of view.
This may seem a pe uliar interpretation ordinary probabilisti
ollapse without wave pa ket ollapse but it is perhaps related to
the intuitive thinking behind Born's famous ollision papers of the previous year (Born 1926a,b) (papers in whi h probabilities for results of
ollisions were identied with squares of s attering amplitudes for the
wave fun tiona ). Born drew an analogy with Einstein's notion of a `ghost
eld' that determines probabilities for photons (see hapter 9). As Born
put it:
In this, I start from a remark by Einstein on the relationship between the wave
eld and light quanta; he said, for instan e, that the waves are there only to
show the orpus ular light quanta the way, and in this sense he talked of a
`ghost eld'. This determines the probability for a light quantum, the arrier
of energy and momentum, to take a parti ular path; the eld itself, however,
possesses no energy and no momentum. .... Given the perfe t analogy between
the light quantum and the ele tron .... one will think of formulating the laws
of motion of ele trons in a similar way. And here it is natural to onsider
a Cf. the dis ussion in se tion 3.4.3.
6.2 Ma ros opi superposition
181
the de Broglie-S hrdinger waves as the `ghost eld', or better, `guiding eld'
.... [whi h propagates a ording to the S hrdinger equation. Momentum and
energy, however, are transferred as if orpus les (ele trons) were a tually ying
around. The traje tories of these orpus les are determined only insofar as they
are onstrained by the onservation of energy and momentum; furthermore,
only a probability for taking a ertain path is determined by the distribution
of values of the fun tion
(Born 1926b, pp. 8034)
Here, Born seems to be suggesting that there are sto hasti traje tories for ele trons, with probabilities for paths determined by the wave
fun tion . It is not lear, though, whether the eld is to be regarded
as a physi al eld asso iated with individual systems (as the ele tromagneti eld usually is), or whether it is to be regarded merely as relating
to an ensemble. If the former, then it might make sense to apply ollapse
to the probabilities without applying ollapse to itself. For example,
this ould happen in a sto hasti version of de Broglie-Bohm theory:
ould be a physi al eld evolving at all times by the S hrdinger equation
(hen e never ollapsing), and instead of generating deterministi parti le
traje tories (as in standard de Broglie-Bohm theory) ould generate
probabilisti motions only. Further eviden e that Born was indeed thinking along su h lines omes from an unpublished manus ript by Born
and Jordan, written in 1925, in whi h they propose a sto hasti theory of
photon traje tories with probabilities determined by the ele tromagneti
eld (as originally envisaged by Slater) see Darrigol (1992, p. 253)
and se tion 3.4.2.a
Thus, in his dis ussion of the loud hamber, when Born spoke of `the
hoi e of one of the two dire tions of propagation', he may indeed have
been referring to the possible dire tions of propagation of the system
onguration, with the eld remaining in a superposition. On this
reading, wave pa ket redu tion for the -parti le would be only an
ee tive des ription, whi h properly orresponds to the bran hing of
the total wave fun tion together with a random hoi e of traje tory (in
multi-dimensional onguration spa e). Unfortunately, however, Born's
intentions are not entirely lear: whether this really is what Born had
in mind, in 1926 or 1927, is di ult to say. Certainly, in the general
dis ussion at the fth Solvay onferen e, Born did maintain that does
a In this onne tion it is interesting to note the following passage from Born's book
Atomi Physi s (Born 1969), whi h was rst published in German in 1933: À
me hani al pro ess is therefore a ompanied by a wave pro ess, the guiding wave,
des ribed by S hrdinger's equation, the signi an e of whi h is that it gives the
probability of a denite ourse of the me hani al pro ess'. Born's referen e to the
wave fun tion as `the guiding wave' shows the lingering inuen e of the ideas that
had inspired him in 1926.
182
not really ollapse (in this multi-dimensional treatment), so it is di ult

to see how he ould have thought that gave merely a probability
distribution over an ensemble.
The laim that the wave fun tion does not ollapse is also found,
in ee t, in se tion II of the report that Born and Heisenberg gave on
quantum me hani s. As we saw in se tion 6.1.2, in their dis ussion of
interferen e, an energy measurement is taken to indu e a randomisation
of the phases appearing in a superposition of energy states, instead of the
usual ollapse to an energy eigenstate. In their example, after an energy
measurement all of the omponents of the superposition are still present,
and the phase relations between them are randomised. In Born's example
of the loud hamber, it seems that, here too, all of the omponents of
the wave fun tion are still present at the end of a measurement. Nothing
is said, however, about the relative phases of the omponents, and we do
not know whether or not Born had in mind a similar phase randomisation
in this ase also.
6.3 Dira and Heisenberg: interferen e, state redu tion, and

delayed hoi e
Another striking feature of quantum theory, as normally understood,
is ìnterferen e between alternative histories'. Like superposition, interferen e is mathemati ally well-dened but its physi al meaning is
ambiguous, and it has long been onsidered one of the main mysteries
of quantum theory (as we saw in se tion 6.1). The question of when
interferen e an or annot take pla e is intimately bound up with the
measurement problem, in parti ular, with the question of when or how
denite out omes emerge from quantum experiments, and with the question of the boundary between the quantum and lassi al domains.
In modern times, one of the most puzzling aspe ts of interferen e was
emphasised by Wheeler (1978). In his `delayed- hoi e' experiment, it
appears that the existen e or non-existen e of interfering histories in
the past is determined by an experimental hoi e made in the present.
One version of Wheeler's experiment a `delayed- hoi e double-slit
experiment' with single photons is shown in Fig. 6.1. A single photon
is in ident on a s reen with two slits. The waves emerging from the
slits are fo ussed (by o- entred lenses) so as to ross ea h other as
shown. The insertion of a photographi plate in the interferen e region
would seem (from the interferen e pattern) to imply the past existen e of
interfering traje tories passing through both slits. On the other hand, if
6.3 Dira and Heisenberg
183
Fig. 6.1. Delayed- hoi e double-slit experiment.
no su h plate is inserted, then a dete tion at P or P seems to imply that

the parti le passed through the bottom or top slit respe tively.a Sin e
the plate ould have been inserted long after the photon ompleted most
of its journey (or journeys), it appears that an experimental hoi e now
an ae t whether or not there was a denite photon path in the past.
A similar point arose in 1927 in the general dis ussion. Dira expounded his view that quantum out omes o ur when nature makes a hoi e.
Heisenberg replied that this ould not be, be ause of the possibility of
observing interferen e later on by hoosing an appropriate experimental
arrangement, leading Heisenberg to on lude that out omes o ur when
a hoi e is made (or brought about) not by nature but by the observer.
Heisenberg's view here bears some resemblan e to Wheeler's. Dira , in
ontrast, seems to say on the one hand that sto hasti ollapse of the
wave pa ket o urs for mi ros opi systems, while on the other hand
a This inferen e is ommonly made, usually without expli it justi ation. Some
authors appeal to onservation of momentum for a free parti le, but it is not lear
how su h an argument ould be made pre ise after all there are no parti le
traje tories in standard quantum theory. In de Broglie-Bohm theory, the inferen e
is a tually wrong: parti les dete ted at P or P ome from the top or bottom slits
respe tively (Bell 1980; Bell 1987, hap. 14).
184
that if the experiment is hosen so as to allow interferen e then su h

ollapse is postponed.
Here is how Dira expresses it (p. 493):
A ording to quantum me hani s the state of the world at any time is des ribable by a wave fun tion
whi h normally varies a ording to a ausal law,
so that its initial value determines its value at any later time. It may however
happen that at a ertain time
t1 ,
=
an be expanded in the form
cn n ,
where the
n 's
are wave fun tions of su h a nature that they annot interfere
with one another at any time subsequent to
t1 .
If su h is the ase, then the
world at times later than t1 will be des ribed not by

The parti ular
but by one of the
n 's.
that it shall be must be regarded as hosen by nature.
Note rst of all that Dira regards as des ribing the state òf the
world' presumably the whole world. Then, in ir umstan es where
may be expanded in terms of non-interfering states n , the world
is subsequently des ribed by one of the n (the hoi e being made by
nature, the probability for n being |cn |2 ). Dira does not elaborate on
pre isely when or why a de omposition into non-interfering states should
exist, nor does he address the question of whether su h a de omposition
is likely to be unique. Su h questions, of ourse, go to the heart of the
measurement problem, and are lively topi s of urrent resear h.
It is interesting that, in Dira 's view (apparently), there are ir umstan es in whi h interferen e is ompletely and irreversibly destroyed.
For him, the parti ular n results from (p. 494):
an irrevo able hoi e of nature, whi h must ae t the whole of the future
ourse of events.
It seems that a ording to Dira , on e nature makes a hoi e of one

bran h, interferen e with the other bran hes is impossible for the whole
of the future. A denite ollapse has o urred, after whi h interferen e
between the alternative out omes is no longer possible, even in prin iple.
This view learly violates the S hrdinger equation as applied to the
whole world: as Dira states, the wave fun tion of the world `normally'
evolves a ording to a ausal law, but not always.
But Dira goes further, and re ognises that there are ir umstan es
where the hoi e made by nature annot have o urred at the point
where it might have been expe ted. Dira onsiders the spe i example
of the s attering of an ele tron. He rst notes that, after the s attering,
one must take the wave fun tion to be not the whole s attered wave
185
but a pa ket moving in a spe i dire tion (that is, one of the n ). He
laims (p. 494) that one ould infer that nature had hosen this spe i
dire tion:
From the results of an experiment, by tra ing ba k a hain of ausally onne ted events one ould determine in whi h dire tion the ele tron was s attered
and one would thus infer that nature had hosen this dire tion.
This is illustrated in Fig. 6.2(a). If the ele tron is dete ted at P, for
example, one may arguably infer a orresponding hoi e of dire tion at
the time of s attering. (Note that Figs. 6.2(a)( ) are ours.)
On the other hand, Dira goes on to make the following observation
(p. 494):
If, now, one arranged a mirror to ree t the ele tron wave s attered in one
dire tion
d1
so as to make it interfere with the ele tron wave s attered in
another dire tion
d2 ,
one would not be able to distinguish between the ase
when the ele tron is s attered in the dire tion

the dire tion
d1
and ree ted ba k into
d2 .
d2
and when it is s attered in
One would then not be able to
tra e ba k the hain of ausal events so far, and one would not be able to say
that nature had hosen a dire tion as soon as the ollision o urred, but only
[that at a later time nature hose where the ele tron should appear.
Dira 's modied s enario is sket hed in Fig. 6.2(b). The presen e of
the mirror leads to interferen e, at Q, between parts of the ele tron wave
s attered in dierent dire tions d1 , d2 . And Dira 's interpretation is that
this interferen e is intimately related to the fa t that an experimenter
observing the outgoing ele tron in a dire tion d2 `would not be able
to distinguish between the ase when the ele tron is s attered in the
dire tion d2 and when it is s attered in the dire tion d1 and ree ted
ba k into d2 '. The experimenter would not be able to `tra e ba k the
hain of ausal events' to the point where he ould say that `nature had
hosen a dire tion as soon as the ollision o urred'. For Dira , in this
ase, nature did not make a hoi e at the time of the ollision, and only
later nature ` hose where the ele tron should appear'.
What Dira des ribes here is pre isely the viewpoint popularised by
Feynman in his famous le tures (Feynman 1965), a ording to whi h if a
pro ess o urs by dierent routes that are subsequently indistinguishable
(in the sense that afterwards an experimenter is in prin iple unable to tell
whi h route was taken) then the probability amplitudes for the dierent
routes are to be added; whereas if the dierent routes are subsequently
distinguishable in prin iple, then the probabilities are to be added.a
a Note that Feynman's path-integral formulation of quantum theory developed in
186
Fig. 6.2. Re onstru tion of s attering s enarios dis ussed by Dira (Figs. (a),
(b)) and Heisenberg (Fig. ( )).
Dira 's presentation of the s attering experiment with the mirror ends
with the statement (p. 494):
The interferen e between the
n 's
ompels nature to postpone her hoi e.
his PhD thesis and elsewhere (Feynman 1942, 1948) was anti ipated by Dira
(1933). Feynman's thesis and Dira 's paper are reprinted in Brown (2005).
187
In his manus ript, a an elled version of the senten e begins with `Thus a
possibility of interferen e between .... ', while another an elled version
begins as `Thus the existen e of .... ' (itali s added). Possibly, Dira
hesitated here be ause he saw that the mirror ould be added by the
experimenter after the s attering had taken pla e, leading to di ulties
with his view that without the mirror nature makes a hoi e at the time
of s attering. For if, in the absen e of the mirror, nature indeed makes
a hoi e at the time of s attering, how ould this hoi e be undone by
subsequent addition of the mirror? Whether Dira really foresaw this
di ulty is hard to say. In any ase, pre isely this point was made by
Heisenberg, and Dira 's hesitation here ertainly ree ts a deep di ulty
that lies at the heart of the measurement problem.
Heisenberg makes his point with disarming simpli ity (p. 496):
I do not agree with Mr Dira when he says that, in the des ribed experiment,
nature makes a hoi e. Even if you pla e yourself very far away from your
s attering material, and if you measure after a very long time, you are able
to obtain interferen e by taking two mirrors. If nature had made a hoi e, it
would be di ult to imagine how the interferen e is produ ed.
What Heisenberg had in mind seems to have been something like the
set-up shown in Fig. 6.2( ), where a pair of mirrors is pla ed far away
from the s attering region, ausing dierent parts of the s attered wave
to re-overlap and interfere at R. A ording to Dira 's a ount, in the
absen e of any mirrors (Fig. 6.2(a)), upon dete tion of the parti le one
might say that nature hose a spe i dire tion at the time of s attering.
Heisenberg points out that, by pla ing mirrors far away (and removing
the dete tors at P and P), interferen e may be observed a long time
after the s attering took pla e.
While Heisenberg does not mention it expli itly, in this example the
hoi e between `whi h-way information' on the one hand, or interfering
paths on the other, may be made long after the parti le has ompleted
most of its journey (or journeys), just as in Wheeler's delayed- hoi e
experiment. Dira 's set-up with no mirrors at all provides whi h-way
information, sin e dete tion of the parti le at a point far away may be
interpreted as providing information on the dire tion hosen at the time
of s attering (Fig. 6.2(a)). Heisenberg's modi ation, with the two mirrors, demonstrates interferen e between alternative paths starting from
the s attering region (Fig. 6.2( )). Unlike Wheeler, however, Heisenberg
does not expli itly emphasise that the hoi e of whether or not to add
the mirrors ould be made at the last moment, long after the s attering
188
takes pla e. On the other hand, Heisenberg does emphasise that the
measurement with the mirrors ould be done `very far away' and àfter
a very long time', and notes the ontradi tion with nature having made
a hoi e at the time of s attering. Thus, Heisenberg's remarks arguably
ontain the essen e of Wheeler's delayed- hoi e experiment.
Heisenberg goes on to say (p. 496) that, instead of nature making a
hoi e,
I should rather say, as I did in my last paper, that the
observer himself
makes
the hoi e, be ause it is only at the moment where the observation is made
that the ` hoi e' has be ome a physi al reality and that the phase relationship
in the waves, the power of interferen e, is destroyed.
From the hronology of Heisenberg's publi ations, here he must be referring to his un ertainty paper (published in May 1927), in whi h he
writes that àll per eiving is a hoi e from a plenitude of possibilities'
(Heisenberg 1927, p. 197). Heisenberg's statement above that the observer `makes' the hoi e seems to be meant in the sense of the observer
`bringing about' the hoi e. Thus it would seem that, for Heisenberg,
a denite out ome o urs and there is no longer any possibility of
interferen e only when an experimenter makes an observation. Similar
views have been expressed by Wheeler (1986).
One may, however, obje t to this viewpoint, on the grounds that there
is no reason in prin iple why a more advan ed being ould not observe
interferen e between the alternative states of the dete tor registering
interferen e, or, between the alternative states of the human observer
wat hing the dete tor. (Cf. the dis ussion of Wigner's paradox in se tion 5.2.2.) After all, the dete tor is ertainly just another physi al system,
built out of atoms. And as far as we an tell, human observers an
likewise be treated as physi al systems built out of atoms. To say that
`the power of interferen e' is `destroyed' when and only when a human
observer intervenes is to make a remarkable assertion to the ee t that
human beings, unlike any other physi al systems, have spe ial properties
by virtue of whi h they annot be treated by ordinary physi al laws but
generate deviations from those laws. As we have already mentioned,
there is no eviden e that human beings are able to violate, for example,
the laws of gravity or of thermodynami s, and it would be remarkable
if they were indeed able to violate the laws of quantum physi s.
It is interesting to note that, while for Heisenberg the human observer
seems to play a ru ial role at the end of a quantum experiment, for
Dira the human observer and his `freewill' seems to play a ru ial
6.3 Further remarks on Born and Heisenberg
189
role at the beginning, in the preparation stage. For as Dira puts it

(p. 493, Dira 's itali s): `The disturban es that an experimenter applies
to a system to observe it are dire tly under his ontrol, and are a ts of
freewill by him. It is only the numbers that des ribe these a ts of freewill
that an be taken as initial numbers for a al ulation in the quantum
theory'. (Cf. the dis ussion about determinism in se tion 8.2.)
Returning to Dira 's view of quantum out omes, Heisenberg's obje tion ertainly auses a di ulty. If a hoi e or ollapse to a parti ular
n really does o ur around the time of s attering, then a `delayed
interferen e experiment' of the form des ribed by Heisenberg should
show no interferen e, and Dira 's view would amount to a violation
of the quantum formalism along the lines of dynami al models of wave
fun tion ollapse (Pearle 1976, 1979, 1989; Ghirardi, Rimini and Weber
1986). And Dira 's aveats on erning the possibility of tra ing ba k
a hain of ausal events do not lead to a really satisfa tory position
either. As in Feynman's view that interferen e o urs only for paths
that are subsequently indistinguishable, the question is begged as to
the pre ise denition of subsequently distinguishable or subsequently
indistinguishable paths: for in a delayed- hoi e set-up, it appears to be
at the later whim of the experimenter to de ide whether ertain paths
taken in the past are subsequently distinguishable or not. This pro edure
orre tly predi ts the experimental results (or statisti s thereof), but it
has the pe uliar onsequen e that whether or not there is a matter of fa t
about the past depends on what the experimenter does in the present.
Finally, as dis ussed in se tion 6.1.1, interferen e was onsidered from
a pilot-wave perspe tive by de Broglie in his report and by Brillouin
in the dis ussion that followed. De Broglie did not omment, however,
on the ex hange between Dira and Heisenberg. From a modern point
of view it is lear that, in his theory, the parti le traje tory does take
one parti ular route after a s attering pro ess, while at the same time
there are portions of the s attered wave travelling along the alternative
routes. An èmpty' part of the wave an be subsequently ree ted by a
mirror, and if the ree ted wave later reoverlaps with the part of the
wave arrying the parti le, then in the interferen e zone the parti le is
indeed ae ted by both omponents. Similarly, de Broglie's theory provides a straightforward a ount of Wheeler's delayed- hoi e experiment,
without present a tions inuen ing the past in any way (Bell 1980; Bell
1987, hap. 14; Bohm, Dewdney and Hiley 1985).
190
6.4 Further remarks on Born and Heisenberg's quantum

me hani s
As we saw in se tion 6.1.2, Born and Heisenberg's report ontains some
remarkable omments about the nature of interferen e (in se tion II,
`Physi al interpretation'). These omments are perhaps related to a
on eptual transition that seems to o ur at around this point in their
presentation. In the earlier part of their se tion II, Born and Heisenberg
des ribe a theory in whi h probabilisti transitions o ur between possessed values of energy; while later in the same se tion, in their dis ussion
of arbitrary observables, they emphasise probabilisti transitions from
one measurement to the next (still noting the presen e of interferen e).
Earlier in that se tion they expli itly assert that a system always o upies a denite energy state at any one time, while in the later treatment
of arbitrary observables nothing is said about whether a system always
possesses denite values or not. This is perhaps not surprising, given
that the dis ussion of interferen e (for the ase of energy measurements)
2
2
made it lear that taking the quantities |cn (t)| = |hn |(t)i| to be
population probabilities for energies En led to a di ulty in the presen e
of interferen e. As we saw in se tion 6.1.2, Born and Heisenberg resolved
the di ulty by asserting that unmeasured population probabilities are
somehow not appli able or meaningful. This does not seem onsistent
with the view they expressed earlier, that atoms always have denite
energy states even when these are not measured. How ould an ensemble
of atoms have denite energy states, without the energy distribution
being meaningful?
Consideration of interferen e, then, was likely to for e a shift away
from the view that atoms are always in denite stationary states. Later, in his book of 1930, Heisenberg did in fa t expli itly deny that
atoms are always in su h states. Considering again the example from
his un ertainty paper, of atoms passing through two su essive regions
of inhomogeneous eld (see se tion 6.1.2), Heisenberg notes that if the
energies are not a tually measured in the intermediate region, then,
be ause of the resulting ìnterferen e of probabilities',
it is not reasonable to speak of the atom as having been in a stationary state
between
F1
and
F2
[that is, in the intermediate region. (Heisenberg 1930b,
p. 61)
As we have already noted, again in se tion 6.1.2, Born and Heisenberg's dis ussion of interferen e seems to dispense with the standard ollapse postulate for quantum states. Upon performing an energy measure-
191
ment, instead of the usual ollapse to a single energy eigenstate |ni, Born
P
and Heisenberg in ee t repla e the superposition n |cn (0)| ein |ni by
a similar expression with randomised phases n . And the justi ation
given for this appears to be some form of un ertainty relation or omplementarity between population probabilities and phases: if the former
have been measured, then the latter are ill-dened, and vi e versa.
On this view, the denite phase relationships asso iated with interferen e pre lude the possibility of speaking of a well-dened population
probability, so that the usual formulas of probability al ulus annot
be properly applied; on the other hand, if the population probability
has been measured experimentally, then the phases are ill-dened, and
averaging over the random phases destroys interferen e.
One ru ial point is not entirely lear, however. Was the phase randomisation thought to o ur only upon measurement of energy, or upon
measurement of any arbitrary observable?
The phase randomisation expli itly appealed to by Born and Heisenberg takes the following form: for a quantum state
X
X
|(t)i =
(6)
|Ei hE |(t)i =
|Ei hE |(0)i eiEt
E
an energy measurement indu es a random hange in ea h phase fa tor

eiEt eiEt ei(E) , where ea h (E) is random on the unit ir le.
This pro edure might be generalised to, for example, measurements of
position, as follows: for a state
!
X
X X
|i
(7)
|xi hx |i
|pi eipx hx |i
x
(writing as if x and p were dis rete, for simpli ity) one might suppose that a position measurement indu es a random hange eipx
eipx ei(x) , resulting in a state
X
(8)
|xi hx |i ei(x)
x
with random relative phases. Averaging over the random phases (x)
would then destroy interferen e between dierent positions, just as in
the ase of energy measurements.
However, we have found no lear eviden e that Born or Heisenberg
onsidered any su h generalisation. It then seems possible that the phase
randomisation argument for the suppression of interferen e was to be
applied to the ase of energy measurements only. On the other hand,
192
there is a suggestive remark by Heisenberg in the general dis ussion

(p. 496), quoted in the last se tion. When expressing his view that denite out omes o ur only when an experimenter makes an observation,
Heisenberg refers to an example where position measurements are made
at the end of a s attering pro ess (see Fig. 6.2( )), and he states that
it is only when the observation is made that `the phase relationship
in the waves, the power of interferen e, is destroyed'. This might be
read as suggesting that the waves ontinue to exist, but may or may
not have the apa ity to interfere depending on whether or not the
phase relations have been randomised by the position measurement.
If Heisenberg did take su h a view, his use of wave pa ket redu tion
for position measurements in the un ertainty paper would have to be
interpreted as some sort of ee tive des ription.
Even in the ase of energy measurements, the status of the phase
randomisation argument is not lear. After all, Born and Heisenberg
assert that the time-dependent S hrdinger equation itself (whi h they
use to dis uss interferen e) is only phenomenologi al, and appli able to
subsystems only. Fundamentally, they have a time-independent theory
for a losed system. Presumably, the phase randomisation for energy
measurements was also seen as phenomenologi al only, with the measured system being treated as a subsystem.
As we saw in se tion 6.3, in the general dis ussion Dira des ribes
what is re ognisably the pro ess of wave pa ket redu tion. Born and
Heisenberg, in ontrast, seem to speak only of the ordinary redu tion
(or onditionalisation) of `probability fun tions' as it appears in standard
probability al ulus. As they put it, near the end of se tion III of their
report (p. 431):
the result of ea h measurement an be expressed by the hoi e of appropriate
initial values for probability fun tions .... . Ea h new experiment repla es the
probability fun tions valid until now with new ones, whi h orrespond to the
result of the observation .... .
On the other hand, Born and Heisenberg's ontributions at the Solvay

onferen e do not seem su iently lear or omplete to warrant denite
on lusions as to what they believed on erning the pre ise relationship
between probabilities and the wave fun tion; sometimes it is un lear
whether or not they mean to draw a distin tion between probability
distributions on the one hand and wave fun tions on the other.
It may be that, in O tober 1927, Born and Heisenberg had in some
respe ts not yet rea hed a denitive point of view, perhaps partly be au-
193
se of the dierent perspe tives that Born and Heisenberg ea h brought

to the subje t. Born's re ent thinking (in 1926) had been inuen ed by
Einstein's idea of a guiding eld, while Heisenberg's re ent thinking (in
his un ertainty paper) had been inuen ed by the operational approa h
to physi s.
Con erning the question of wave pa ket ollapse, it should also be
remembered that Pauli seems to have played an important role in Born
and Heisenberg's thinking at the time. In parti ular, as we saw in Born's
dis ussion of the loud hamber (se tion 6.2.1), Pauli had been riti al of
Heisenberg's use of the redu tion of the wave pa ket in the un ertainty
paper (in a dis ussion of lassi al ele tron orbits), and Born who by
his own a ount was following Pauli's suggestion tried to show that
su h redu tion was unne essary.
7
7.1 Einstein's 1927 argument for in ompleteness

A huge literature arose out of the famous ÈPR' paper by Einstein, Podolsky and Rosen (1935), entitled `Can quantum-me hani al des ription
of physi al reality be onsidered omplete?'. The EPR paper argued, on
the basis of (among other things) the absen e of a tion at a distan e,
that quantum theory must be in omplete.a It is less well-known that a
mu h simpler argument, leading to the same on lusion, was presented
by Einstein eight years earlier in the general dis ussion at the fth Solvay
onferen e (pp. 485 .).
Einstein ompares and ontrasts two views about the nature of the
wave fun tion , for the spe i ase of a single ele tron. A ording to
view I, represents an ensemble (or ` loud') of ele trons; while a ording
to view II, is a omplete des ription of an individual ele tron. Einstein
argues that view II is in ompatible with lo ality, and that to avoid
this, in addition to there should exist a lo alised parti le (along the
lines of de Broglie's theory). Thus, a ording to this reasoning, if one
assumes lo ality, then quantum theory (as normally understood today)
is in omplete.
The on lusion of Einstein's argument in 1927 is the same as that
of EPR in 1935, even if the form of the argument is rather dierent.
Einstein onsiders ele trons striking a s reen with a small hole that
dira ts the ele tron wave, whi h on the far side of the s reen spreads
out uniformly in all dire tions and strikes a photographi lm in the
shape of a hemisphere with large radius (see Einstein's gure). Einstein's
argument against view II is then as follows:
a Note that, as pointed out by Fine (1986) and dis ussed further by Howard (1990),
the logi al stru ture of the EPR paper (whi h was a tually written by Podolsky)
is more ompli ated and less dire t than Einstein had intended.
194

If
||2
195
were simply regarded as the probability that at a ertain point a given
the same elementary

in two or several pla es on the s reen. But the
2
to whi h || expresses the probability that this
parti le is found at a given time, it ould happen that

pro ess produ es an a tion
interpretation, a ording
parti le is found at a given point, assumes an entirely pe uliar me hanism

of a tion at a distan e, whi h prevents the wave ontinuously distributed in
spa e from produ ing an a tion in
two
pla es on the s reen.
The key point here is that, if there is no a tion at a distan e, and

if the extended eld is indeed a omplete des ription of the physi al
situation, then if the ele tron is dete ted at a point P on the lm, it
ould happen that the ele tron is also dete ted at another point Q, or
indeed at any point where ||2 is non-zero. Upon dete tion at P , it
appears that a `me hanism of a tion at a distan e' prevents dete tion
elsewhere.
Einstein's argument is so on ise that its point is easily missed, and
one might well dismiss it as arising from an elementary onfusion about
the nature of probability. (Indeed, Bohr omments that he does not
ùnderstand what pre isely is the point' Einstein is making.) For example, it might be thought that, sin e we are talking about a probability
distribution for just one parti le, it is a matter of pure logi that only one
dete tion an o ur.a But this would be to beg the question on erning
the nature of . Einstein's wording above attempts to onvey a distin tion between probability for a `given' parti le (leading to the possibility of
multiple dete tions) and probability for `this' parti le (leading to single
dete tion only). The wording is not su h as to onvey the distin tion
very learly, perhaps indi ating an inadequate translation of Einstein's
German into Fren h.b But from the ontext, the words `probability that
this parti le is found' are learly being used to express the assumption
that in this ase indeed expresses the probability for just one parti le
dete tion.
As shown by Hardy (1995), Einstein's argument may be readily put
into the same rigorous form as the later EPR argument. (See Norsen
(2005) for a areful and extensive dis ussion.) Hardy simplies Einstein's
example, and onsiders a single parti le in ident on a beam splitter
(Fig. 7.1), so that there are only two points P1 , P2 at whi h the parti le might be dete ted. One may then adopt the following su ient
ondition, given by EPR, for the existen e of an element of reality:
a Cf. Shimony (2005).
b Unfortunately, the full German text of Einstein's ontribution to the general
dis ussion seems to have been lost; the Einstein ar hives ontain only a fragment,
onsisting of just the rst four paragraphs (AEA 16-617.00).
196
If, without in any way disturbing a system, we an predi t with ertainty (i.e.,
with probability equal to unity) the value of a physi al quantity, then there
exists an element of physi al reality orresponding to this physi al quantity.
(Einstein, Podolsky and Rosen 1935, p. 777)
Now, if a dete tor is pla ed at P1 , either it will re or it will not. In either
ase, from the state of the dete tor at P1 one ould dedu e with ertainty
whether or not a dete tor pla ed at P2 would re. Su h dedu tions ould
be made for any individual run of the experiment. Even though the
out ome at P1 annot be predi ted in advan e, in ea h ase the out ome
allows us to infer the existen e of a denite element of reality at P2 .
If lo ality holds, an element of reality at P2 annot be ae ted by the
presen e or absen e of a dete tor at P1 . Therefore, even if no dete tor is
pla ed at P1 , there must still be an element of reality at P2 orresponding
to dete tion or no dete tion at P2 . Sin e is a superposition of dete tion
and no dete tion at P2 , ontains nothing orresponding to the dedu ed
element of reality at P2 . Therefore, is not a omplete des ription of a
single parti le.a
Thus, `the essential points in the EPR argument had already been
made by Einstein some eight years earlier at the fth Solvay onferen e'
(Hardy 1995, p. 600).
Einstein on ludes that:
In my opinion, one an remove this obje tion only in the following way, that
one does not des ribe the pro ess solely by the S hrdinger wave, but that at
the same time one lo alises the parti le during the propagation. I think that
Mr de Broglie is right to sear h in this dire tion. If one works solely with the
2
S hrdinger waves, interpretation II of || implies to my mind a ontradi tion
with the postulate of relativity.
In other words, for Einstein, a tion at a distan e an be avoided only by

admitting that the wave fun tion is in omplete.
A ording to Einstein's argument, quantum theory is either nonlo al
or in omplete. For the rest of his life, Einstein ontinued to believe
that lo ality was a fundamental prin iple of physi s, and so he adhered
to the view that quantum theory must be in omplete. However, further reasoning by Bell (1964) showed that any ompletion of quantum
theory would still require nonlo ality, in order to reprodu e the details
of quantum orrelations for entangled states (assuming the absen e of
a Note that in this argument the in ompleteness of quantum theory is inferred from
the assumption of lo ality. Cf. Bell (1987, p. 143).
197
Fig. 7.1. Hardy's simplied version of Einstein's argument.
ba kwards ausation or of many worldsa). It then appears that, whether omplete or in omplete, quantum theory is ne essarily nonlo al, a
on lusion that would surely have been deeply disturbing to Einstein.
It is ironi that Einstein's (and EPR's) argument started out by
holding steadfast to lo ality and dedu ing that quantum theory is in omplete. But then the argument, as arried further by Bell, led to
a ontradi tion between lo ality and quantum orrelations, so that in
the end one fails to establish in ompleteness and instead establishes
nonlo ality (with ompleteness or in ompleteness remaining an open
question).
a Bell's argument assumes that there is no ommon ause between the hidden
variables (dened at the time of preparation) and the settings of the measuring
apparatus. It also assumes that there is no ba kwards ausation, so that the hidden
variables are not ae ted by the future out omes or apparatus settings. Further,
the derivation of the Bell inequalities assumes that a quantum measurement has
only one out ome, and therefore does not apply in the many-worlds interpretation.
198
7.2 A pre ursor: Einstein at Salzburg in 1909

In September 1909, at a meeting in Salzburg, Einstein gave a le ture
entitled Òn the development of our views on erning the nature and
onstitution of radiation' (Einstein 1909). Einstein summarised what he
saw as eviden e for the dual nature of radiation: he held that light had
both parti le and wave aspe ts, and argued that lassi al ele tromagneti
theory would have to be abandoned. It seems to have gone unnoti ed
that one of Einstein's arguments at Salzburg was essentially the same
as the argument he presented at the 1927 Solvay onferen e (though
applied to light quanta instead of to ele trons).
Einstein began his le ture by noting that the phenomena of interferen e and dira tion make it plain that, at least in some respe ts,
light behaves like a wave. He then went on to des ribe how, in other
respe ts, light behaves as if it onsisted of parti les. In experiments
involving the photoele tri ee t, it had been found that the velo ity of
the photoele trons was independent of the radiation intensity. A ording
to Einstein, this was more onsistent with `Newton's emission theory
of light' than with the wave theory. Einstein also dis ussed pressure
u tuations in bla kbody radiation, and showed that these ontained
two terms, whi h ould be naturally identied as ontributions from
parti le-like and wave-like aspe ts of the radiation.
Of spe ial interest here is another argument Einstein gave for the existen e of lo alised light quanta. Einstein onsidered a beam of ele trons
(`primary athode rays') in ident upon a metal plate P1 and produ ing
X-rays (see Fig. 7.2). These X-rays, in turn, strike a se ond metal plate
P2 leading to the produ tion of ele trons (`se ondary athode rays') from
P2 . Experimentally, it had been found that the velo ity of the se ondary
ele trons had the same order of magnitude as the velo ity of the primary
ele trons. Further, the available eviden e suggested that the velo ity of
the se ondary ele trons did not depend at all on the distan e between the
plates P1 and P2 , or on the intensity of the primary ele tron beam, but
only on the velo ity of the primary ele trons. Assuming this to be stri tly
true, Einstein then asked what would happen if the primary intensity
were so small, or the area of the plate P1 so small, that one ould onsider
just one ele tron striking the plate, as in Fig. 7.2. A ording to Einstein,
we will have to assume that on
ele tron on
P1 )
P2
(as a result of the impinging of the above
either nothing is being produ ed or that a se ondary emission
of an ele tron o urs on it with a velo ity of the same order of magnitude as of
the ele tron impinging on
P1 . In other words, the elementary radiation pro ess
seems to pro eed su h that it does not, as the wave theory would require,
199
Fig. 7.2. Figure based on Einstein's 1909 argument for the existen e of
lo alised light quanta. Assuming the prin iple of lo al a tion, the delo alised
X-ray wave an produ e an ele tron (of energy omparable to that of the
primary ele tron) in a small region of the se ond plate only if, in addition to
the wave, there is a lo alised energy fragment propagating in spa e from
to
P1
P2 .
distribute and s atter the energy of the primary ele tron in a spheri al wave
propagating in all dire tions. Rather, it seems that at least a large part of this
energy is available at some lo ation of
P2
or somewhere else. (Einstein 1909,
English translation, p. 388)
Einstein's argument, then, is that a ording to the wave theory the

point of emission of the X-ray from the rst plate must be the sour e of
waves spreading out in spa e, waves whose amplitude will spread over
the region o upied by the se ond plate. And yet, in the se ond plate, all
the energy of the X-ray be omes on entrated in the vi inity of a single
point, leading to the produ tion of an ele tron with velo ity omparable
to that of the primary ele tron.a Einstein on luded from this that, in
addition to the wave spreading from the point of emission, there seems
also to be a lo alised energy fragment propagating from the point of
a Cf. Compton's report, p. 339: Ìt is learly impossible that all the energy of an
X-ray pulse whi h has spread out in a spheri al wave should spend itself on this
[small region'.
200
emission of the X-ray wave to the point of produ tion of the se ondary
ele tron. As Einstein put it: `the elementary pro ess of radiation seems
to be dire ted '.
Now, Einstein's argument of 1909 impli itly assumes a prin iple of
lo al a tion, similar to that expli itly assumed in his published ritique of
quantum theory at the 1927 Solvay onferen e. Be ause the distan e between the plates P1 and P2 an be arbitrarily large, the wave impinging
on P2 an be spread over an arbitrarily large area. The produ tion of an
ele tron in a highly lo alised region of P2 an then be a ounted for only
if, in addition to the delo alised wave, there is a lo alised energy fragment
propagating through spa e for otherwise, there would have to be some
me hanism by means of whi h energy spread out over arbitrarily large
regions of spa e suddenly be omes on entrated in the neighbourhood
of a single point.
It should be quite lear, then, that Einstein's 1927 argument for the
existen e of lo alised ele trons (a ompanying de Broglie-S hrdinger
waves) was identi al in form to one of his 1909 arguments for the existen e of lo alised light quanta (a ompanying ele tromagneti waves).
In his 1927 argument, the small hole in the s reen (see his gure) a ts
as a sour e for an ele tron wave, whi h spreads over the area of the
photographi lm just as, in the 1909 argument, the point where
the primary ele tron strikes the rst plate a ts as a sour e for an X-ray
wave, whi h spreads over the area of the se ond plate. Both arguments
depend ru ially on the assumption (impli it in 1909, expli it in 1927)
that there is no a tion at a distan e.
As we shall dis uss further in hapter 9, by 1927 Einstein had already
spent over twenty years trying to re on ile lo alised energy quanta
whi h he had postulated in 1905 with the wave aspe t of radiation.
And for mu h of that time, he had been more or less alone in his belief
in the existen e of su h quanta. It is then perhaps not so surprising that
at the 1927 Solvay meeting Einstein was able to raise su h a penetrating
ritique of the view that the wave fun tion is a omplete des ription of a
single ele tron: from his long and largely solitary experien e pondering
the wave-parti le duality of light, Einstein ould immediately see that, in
the analogous ase of ele tron waves, the prin iple of lo al a tion entailed
the existen e of lo alised parti les moving through spa e, in addition to
the wave fun tion.
The meeting in Salzburg took pla e four years before Bohr published
his model of the atom. After 1913, one might have simply rephrased
Einstein's argument in terms of atomi transitions. Consider an atom
201
A that makes a transition from an initial stationary state with energy

Ei to a nal stationary state with lower energy Ef . At a later time,
the energy Ei Ef lost by atom A may be wholly absorbed by an
arbitrarily distant atom B, if there exists an appropriate transition from
the initial state of B to a nal state orresponding to an energy in rease
Ei Ef . This pro ess may seem unmysterious, if one imagines atom
A emitting a photon, or `lo alised energy quantum', whi h somehow
propagates through spa e from A to B. However, if one tries to make
do without the photon on ept, and represents the ele tromagneti eld
in terms of ( lassi al) waves only whi h spread out in all dire tions
from A then it is hard to understand how the energy lost by A
may be wholly transferred to B: instead, one would expe t the energy to
spread out in spa e like the waves themselves, so that the energy density
be omes diluted.
We have laboured this point be ause the power of Einstein's simple
argument seems to have been generally missed, not only in 1909, but also
in 1927, and for de ades afterwards. Indeed, it appears that Einstein's
point did not start to be ome widely appre iated until the late twentieth
entury (see, again, Norsen (2005)).
In retrospe t, it seems quite puzzling that Einstein's simple argument
should have taken so long to be understood. A perhaps related puzzle,
emphasised by Pais (1982, pp. 3826), is why Einstein's light-quantum
hypothesis itself should have been largely ignored by so many physi ists
until the advent of the Compton ee t in 1923. Even after Millikan's
experimental onrmation of Einstein's photoele tri equation in 1916,
àlmost no one but Einstein himself would have anything to do with
light-quanta' (Pais 1982, p. 386).a
We do not wish to suggest, of ourse, that Einstein's lo ality argument should today be regarded as establishing the existen e of lo alised
photons: for the impli it premise of Einstein's argument the prin iple
of lo ality today seems to be ruled out by Bell's theorem. Our point,
rather, is that prior to Bell (and ertainly in 1909) Einstein's arguments
were indeed ompelling and should have been taken more seriously.
a Though a ording to Brillouin's re olle tions of 1962, the situation was rather
dierent in Fran e, where Einstein's light quantum was a epted (by Langevin,
Perrin and Marie Curie) mu h earlier than it was elsewhere (Mehra and
Re henberg 1982a, p. 580).
202
7.3 More on nonlo ality and relativity

At the end of his long ontribution to the general dis ussion (in whi h
he argued for the in ompleteness of quantum theory), Einstein obje ted
to the multi-dimensional representation in onguration spa e, on the
grounds that (p. 487)
.... the feature of for es of a ting only at small
spatial
distan es nds a less
natural expression in onguration spa e than in the spa e of three or four

dimensions.
As Einstein himself stated, he was here adding another argument against

what he alled view II (the view that is a omplete des ription of
an individual system), a view that he laimed is èssentially tied to a
multi-dimensional representation ( onguration spa e)'.
Einstein's point seems to be that, if physi s is fundamentally grounded
in onguration spa e, there will be no reason to expe t physi s to
be hara terised by lo al a tion. This obje tion should be seen in the
ontext of Einstein's on erns, in the period 192627, over the nonseparability of S hrdinger's wave me hani s for many-body systems
(Howard 1990, pp. 8391; f. se tion 12.2).
A ertain form of lassi al lo ality survives, of ourse, in modern
quantum theory and quantum eld theory, in the stru ture of the Hamiltonian or Lagrangian, a stru ture that ensures the absen e of ontrollable
nonlo al signals at the statisti al level. But even so, we understand today
that, in fa t, quantum physi s is hara terised by nonlo ality. And the
nonlo ality may indeed be tra ed to the fa t that, unlike lassi al theory,
quantum theory is not grounded in ordinary three-dimensional spa e.
The setting for standard quantum theory is Hilbert spa e, whose
tensor-produ t stru ture allows for entanglement and asso iated nonlo al ee ts. In the pilot-wave formulation of quantum theory, the setting is
onguration spa e (in whi h the pilot wave propagates), and in general
the motions of spatially-separated parti les are nonlo ally onne ted.
Further, Bell's theorem shows that, if we leave aside ba kwards ausation
or many worlds, then quantum theory is in some sense nonlo al under
any interpretation or formulation. As Ballentine on e pointed out, while
dis ussing the signi an e of Bell's theorem:
Perhaps what is needed is not an explanation of nonlo ality, but an explanation
of lo ality. Why, if lo ality is not true, does it work so well in so many dierent
ontexts? (Ballentine 1987, pp. 7867)
Einstein's fear, that there would be di ulties with lo ality in quantum
physi s, has ertainly been borne out by subsequent developments. In
7.3 More on nonlo ality and relativity
203
standard quantum theory, there appears to be a pea eful but uneasy

` oexisten e' with relativity. While from a pilot-wave (or more generally,
from a deterministi hidden-variables) point of view, statisti al lo ality
appears as an a idental feature of the `quantum equilibrium' state
(Valentini 1991b, 2002a).
In his main ontribution to the general dis ussion Dira (p. 491) also
notes that `the general theory of the wave fun tion in many-dimensional
spa e ne essarily involves the abandonment of relativity', but he suggests
that this problem might be solved by `quantising 3-dimensional waves'
(that is, by what we would now all quantum eld theory). And de
Broglie in his report, when onsidering the pilot-wave dynami s of manybody systems, notes that unlike in the ase of a single parti le ìt does not
appear easy to nd a wave that would dene the motion of the system
taking Relativity into a ount' (p. 386), a di ulty that has persisted in
pilot-wave theory right up to the present day (see, for example, Berndl
et al. (1996)).
In 1927, then, there was a fairly broad re ognition that the fundamental use of onguration spa e did not bode well for onsisten y with
relativity.
8
Time, determinism, and the spa etime
framework
8.1 Time in quantum theory

By 1920, the spe ta ular onrmation of general relativity, during the
solar e lipse of 1919, had made Einstein a household name. Not only
did relativity theory (both spe ial and general) upset the long-re eived
Newtonian ideas of spa e and time, it also stimulated a widespread
òperationalist' attitude to physi al theories. Physi al quantities ame to
be seen as inextri ably interwoven with our means of measuring them,
in the sense that any limits on our means of measurement were taken
to imply limits on the denability, or `meaningfulness', of the physi al
quantities themselves. In parti ular, Einstein's relativity paper of 1905
with its operational analysis of simultaneity ame to be widely
regarded as a model for the new operationalist approa h to physi s.
Not surprisingly, then, as the puzzles ontinued to emerge from atomi experiments, in the 1920s a number of workers suggested that the
on epts of spa e and time would require still further revision in the
atomi domain. Thus, Campbell (1921, 1926) suggested that the puzzles
in atomi physi s ould be removed if the on ept of time was given a
purely statisti al signi an e: `time, like temperature, is a purely statisti al on eption, having no meaning ex ept as applied to statisti al
aggregates' (quoted in Beller 1999, p. 97). In an operationalist vein,
Campbell onsidered ` lo ks' based on (random) radioa tive de ays. He
suggested that it might be possible to onstru t a theory that did not
involve time at all, and in whi h àll the experiments on whi h the
prevailing temporal on eptions are based an be des ribed in terms of
statisti s' (quoted in Beller 1999, p. 98). On the other hand, Senftleben
(1923) asserted that Plan k's onstant h set limits to the denability of
the on epts of spa e and time, and on luded that spa etime must be
204
205
dis ontinuous. A ording to Beller (1999, pp. 96101), both Campbell

and Senftleben had a signi ant inuen e on Heisenberg in his formulation of the un ertainty prin iple.a Certainly, in a letter to Pauli of 28
O tober 1926, Heisenberg expresses views very similar to Campbell's
(Pauli 1979, p. 350):
I have for all that a hope in a later solution of more or less the following
kind (but one should not say something like this aloud): that spa e and time
are really only statisti al on epts, su h as, say, temperature, pressure et .
in a gas. I mean that spatial and temporal on epts are meaningless for
one
orpus le and that they make more and more sense the more parti les are
present. I often try to get further in this dire tion, but until now it will not
work.
Be that as it may, at the 1927 Solvay onferen e, in their report on

quantum me hani s, Born and Heisenberg seem to express the remarkable view that temporal hanges do not o ur at all for losed systems,
and that the time-dependent S hrdinger equation emerges only as an
ee tive and approximate des ription for subsystems. How these views
related to Campbell's, or indeed if they did at all, is not lear.
In their se tion II, `Physi al interpretation', Born and Heisenberg
begin with the following statement (p. 419):
The most noti eable defe t of the original matrix me hani s onsists in the
fa t that at rst it appears to give information not about a tual phenomena,
but rather only about possible states and pro esses. .... it says nothing about
when a given state is present .... matrix me hani s deals only with losed
periodi systems, and in these there are indeed no hanges. In order to have
true pro esses .... one must dire t one's attention to a
part
of the system .... .
From a modern point of view, the original matrix me hani s did not
ontain the notion of a general state |i for a system (not even a stati ,
Heisenberg-pi ture state). The only states that appeared in the theory
were the stationary states |Ei i ( f. hapter 3). Even as regards stationary
states, there seems to have been no notion of initial state (ìt says
nothing about when a given state is present'). Instead, the matri es
provided a olle tive representation of all the energy eigenstates of a
losed system with Hamiltonian H . In modern notation, the matri es
onsisted of matrix elements of (Heisenberg-pi ture) observables (t) =
e(i/~)Ht (0)e(i/~)Ht in the energy basis:
hEi | (t) |Ej i = hEi | (0) |Ej i e(i/~)(Ei Ej )t .
(1)
a For Campbell's inuen e on Bohr's formulation of omplementarity, see Beller

(1999, pp. 1357) and also Mehra and Re henberg (2000, pp. 18990).
206
Time, determinism, and spa etime framework
As noted in hapter 3, the formal mathemati s of matrix me hani s

then seems to represent an atomi system somewhat in the manner of
the Bohr-Kramers-Slater (BKS) theory ( f. hapter 9), with ea h matrix
element orresponding to a virtual os illator of frequen y ij = (Ei
Ej )/h.
The matrix formalism, without a notion of initial state, amounts to
a stati des ription.a However, Born and Heisenberg add an intuitive
physi al pi ture to the formalism, to the ee t that a subsystem of a
larger ( losed) system is in fa t in one stationary state at any one time
and performs random, indeterministi `quantum jumps' between su h
states ( f. hapter 3).
Born and Heisenberg then go on to say that `[the lumsiness of the
matrix theory in the des ription of pro esses developing in time an
be avoided' (p. 421) by introdu ing what we would now all the timedependent S hrdinger equation. Here, it might appear that their view is
that the mentioned `defe t of the original matrix me hani s' is removed
by generalisation to a time-dependent theory. However, they add (p. 421)
that:
Essentially, the introdu tion of time as a numeri al variable redu es to thinking
of the system under onsideration as oupled to another one and negle ting
the rea tion on the latter. But this formalism is very onvenient .... .
These words give, instead, the impression that the time-dependent theory is regarded as only emergent in some approximation; the time-dependent
S hrdinger equation seems to have no fundamental status.
Even so, this ` onvenient' formalism `leads to a further development
of the statisti al view'. They in lude a time-dependent external perturbation in the (time-dependent) S hrdinger equation, and show how
to al ulate the time development of any initial wave fun tion. Born
and Heisenberg then argue that, following Bohr's original (1913) theory
of stationary states, a system an be in only one energy eigenstate at
any one time, leading to the interpretation of a superposition |i =
P
2
n cn (t) |Ei i as a statisti al mixture, with state probabilities |cn | . The
time evolution of the wave fun tion then des ribes transition probabilities from initial to nal stationary states. This might seem lear enough,
but a di ulty is then raised on erning the interpretation of a ase
where the initial wave fun tion is already a superposition, resulting in
ìnterferen e of probabilities' at later times (see se tion 6.1.2).
a Even if one added a notion of initial state, be ause the only allowed states are
energy eigenstates the des ription of a losed system would still be stati .
207
Let us now onsider what S hrdinger had to say, in his report on wave
me hani s, on erning time in quantum theory. (De Broglie's report
does not ontain any spe ial remarks on this subje t.) In his report
S hrdinger rst presents (or derives from a variational prin iple) what
we would now all the time-independent S hrdinger equation for a
nonrelativisti many-body system with oordinates q1 , q2 , ..., qn . After
noting that the eigenfun tions k , with eigenvalues Ek , may be identied
with Bohr's stationary states, S hrdinger addresses the question of
time. He rst points out (p. 450) that the time-independent theory might
be regarded as su ient, providing as it does a des ription of stationary
states, together with expressions for jump probabilities between them:
One an take the view that one should be ontent in prin iple with what has
been said so far .... . The single stationary states of Bohr's theory would then
k , whi h do not ontain time at

all. One .... an form from them .... quantities that an be aptly taken to be
jump probabilities between the single stationary states.
in a way be des ribed by the eigenfun tions
Here, the jump probabilities are to be obtained from matrix elements

su h as (in modern notation)
Z
hk| Qi |k i = dq qi k k
(2)
whi h an all be al ulated from the eigenfun tions k .

S hrdinger suggests further that intera ting systems ould be treated
in the same way, by regarding them as one single system.
S hrdinger then goes on to dis uss this point of view and its relation
with the ideas of Campbell (p. 450):
On this view the
time variable
would play absolutely no role in an isolated
system a possibility to whi h N. Campbell .... has re ently pointed. Limiting

our attention to an isolated system, we would not per eive the passage of time
in it any more than we an noti e its possible progress in spa e .... . What we
would noti e would be merely a sequen e of dis ontinuous transitions, so to
speak a inemati image, but without the possibility of omparing the time
intervals between the transitions.
A ording to these ideas, then, time does not exist at the level of isolated
atomi systems, and our usual (ma ros opi ally-dened) time emerges
only from the statisti s of large numbers of transitions between stationary states. As S hrdinger puts it:
Only se ondarily, and in fa t with in reasing pre ision the more extended
the system, would a
statisti al
denition of time result from
ounting
the
transitions taking pla e (Campbell's `radioa tive lo k'). Of ourse then one
208
annot understand the jump probability in the usual way as the probability of a
single jump probability is

two possibilities for jumps, the probability
before the other is equal to its jump probability
transition al ulated relative to unit time. Rather, a

then utterly meaningless; only with
that the one may happen
divided by the sum of the two.
S hrdinger laims that this is the only onsistent view in a theory

with quantum jumps, asserting that `[either all hanges in nature are
dis ontinuous or not a single one'.
Having sket hed a timeless view of isolated systems with dis rete
quantum jumps, S hrdinger states that su h a dis rete viewpoint `still
poses great di ulties', and he goes on to develop his own theory of
time-dependent quantum states, in whi h ( ontinuous) time evolution
does play a fundamental role even at the level of a single atomi system.
Here, a general time-dependent wave fun tion (q, t) a solution of the
time-dependent S hrdinger equation, with arbitrary initial onditions
is regarded as the des ription of the ontinuous time development of
a single isolated system.
From a ontemporary perspe tive, it is lear that quantum theory as
we know it today is rather less radi al than some expe ted it to be in the
1920s, espe ially on erning the on epts of spa e and time. Both nonrelativisti quantum me hani s, and relativisti quantum eld theory,
take pla e on a lassi al spa etime ba kground; time and spa e are ontinuous and well-dened, even for losed systems. The evolution operator
U (t, t0 ) provides a ontinuous time evolution |(t)i = U (t, t0 ) |(t0 )i
for any initial quantum state, with respe t to an èxternal' time parameter t (even in quantum eld theory, in a given inertial frame). While
S hrdinger's and de Broglie's interpretations of the wave fun tion did
not gain widespread a eptan e, their view of time in quantum theory
oin ides with the one generally a epted today.a
In ontrast, the views on time expressed by Born and Heisenberg are
somewhat reminis ent of views put forward by some later workers in the
ontext of anoni al quantum gravity. There, the wave fun tional [(3) G]
on the spa e of 3-geometries (3) G ontains no expli it time parameter,
and obeys a `timeless' S hrdinger equation H = 0 (the WheelerDeWitt equation, with Hamiltonian density operator H). It is laimed
a We are of ourse referring here to the standard theories as presented in textbooks.
The literature ontains a number of proposals, along operational lines, alling
for a `quantum spa etime' that in orporates quantum-theoreti al limits on the
onstru tion of rods and lo ks. A statisti al approa h to ausal stru ture,
somewhat reminis ent of Campbell's statisti al view of time, has re ently been
proposed by Hardy (2005).
8.2 Determinism and probability
209
that `time' emerges only phenomenologi ally, through the analysis of

intera tion with quantum lo ks, or by the extra tion of an ee tive
time variable from the 3-metri (the radius of an expanding universe
being a popular hoi e) (DeWitt 1967). However, loser analysis reveals
a series of di ulties with su h proposals: for example, it is di ult to
ensure the emergen e of a well-behaved time parameter t su h that only
one physi al state is asso iated with ea h value of t. (See, for example,
Unruh and Wald (1989).) Despite some 50 years of eort, in luding the
te hni al progress made in re ent years using `loop' variables to solve
the equations (Rovelli 2004), the `problem of time' in anoni al quantum
gravity remains unresolved.a

In the published text, the rst se tion of the general dis ussion bears the
title `Causality, determinism, probability'. Lorentz's opening remarks
are mainly on erned with the importan e of having a lear and denite
pi ture of physi al pro esses (see se tion 8.3). He ends by addressing the
question of determinism (p. 478):
.... I think that this notion of probability should be pla ed at the end, and
as a on lusion, of theoreti al onsiderations, and not as an
a priori
axiom,
though I may well admit that this indetermina y orresponds to experimental

possibilities. I would always be able to keep my deterministi faith for the
fundamental phenomena .... .
Lorentz seems to demand that the fundamental phenomena be deterministi , and that indeterminism should be merely emergent or ee tive.
Probabilities should not be axiomati , and some theoreti al explanation
is needed for the experimental limitations en ountered in pra ti e. This
view would nowadays be usually asso iated with deterministi hiddenvariables theories, su h as de Broglie's pilot-wave dynami s (though it
might also be asso iated with the many-worlds interpretation of Everett).
De Broglie's basi equations the guidan e equation and S hrdinger equation are ertainly deterministi . De Broglie in his report,
and Brillouin in the subsequent dis ussion, give examples of how these
equations determine the traje tories (during interferen e and dira tion,
atomi transitions, and elasti s attering). As regards probabilities, de
a Barbour (1994a,b) has proposed a timeless formulation of lassi al and quantum
physi s. As applied to quantum gravity, the viability of Barbour's s heme seems
to depend on unproven properties of the Wheeler-DeWitt equation.
210
Broglie pointed out that if an ensemble of systems with initial wave

2
fun tion 0 begins with a Born-rule distribution P0 = |0 | , then as
evolves, the system dynami s will maintain the distribution P =
2
|| at later times. However, nothing was said about how the initial
distribution might arise in the rst pla e. Subsequent work has shown
that, in de Broglie's theory, the Born-rule distribution an arise from the
omplex evolution generated by the dynami s itself, mu h as thermal
distributions arise in lassi al dynami s, thereby providing an example
of the kind of theoreti al explanation that Lorentz wished for (Bohm
1953; Valentini 1991a, 1992, 2001; Valentini and Westman 2005).a On
this view, the initial ensemble onsidered by de Broglie orresponds to a
spe ial èquilibrium' state analogous to thermal equilibrium in lassi al
physi s.
Lorentz goes on to say (p. 478):
Could a deeper mind not be aware of the motions of these ele trons? Could one
not keep determinism by making it an obje t of belief ? Must one ne essarily
elevate indeterminism to a prin iple?
Here, again, we now know that de Broglie's theory provides an example of what Lorentz seems to have had in mind. For in prin iple, the
2
theory allows the existen e of `nonequilibrium' distributions P 6= ||
(Valentini 1991b, 1992), just as lassi al physi s allows the existen e
of non-thermal distributions (not uniformly distributed on the energy
surfa e in phase spa e). Su h distributions violate many of the standard
quantum onstraints; in parti ular, an experimenter possessing parti les
with a distribution P mu h narrower than ||2 would be able to use those
parti les to perform measurements on ordinary systems more a urate
than normally allowed by the un ertainty prin iple; an experimenter
possessing su h `nonequilibrium parti les' would in fa t be able to use
them to observe the (normally invisible) details of the traje tories of
ordinary parti les (Valentini 2002b; Pearle and Valentini 2006). From
this point of view, there is indeed no need to èlevate indeterminism to
a prin iple': for the urrent experimental limitations (embodied in the
a Bohm (1953) onsidered the parti ular ase of an ensemble of two-level mole ules
and argued that external perturbations would drive it to quantum equilibrium.
A general argument for relaxation was not given, however, and soon afterwards
Bohm and Vigier (1954) modied the dynami s by adding random u tuations
that drive any system to equilibrium. This move to a sto hasti theory seems
unne essary: a general H -theorem argument, analogous to the lassi al oarsegraining H -theorem, has been given (Valentini 1991a, 1992, 2001), and numeri al
simulations show a very e ient relaxation with an exponential de ay of the
oarse-grained H -fun tion on the basis of the purely deterministi de BroglieBohm theory (Valentini and Westman 2005).
211
un ertainty prin iple) are not built into the laws of physi s; rather, they
2
are merely ontingent features of the equilibrium state P = || .
In the general dis ussion, as we saw in se tion 6.3, Dira expressed
the view that quantum out omes o ur when nature makes a hoi e, a
view ountered by Heisenberg who laimed that the ` hoi e' is in some
sense really made by the observer. As Lorentz noted at the end of this
ex hange, the view that nature makes a hoi e amounts to a fundamental indeterminism, while at the same time, Dira and Heisenberg had
radi ally dierent views about the meaning of this indeterminism.
Dira also gave an argument for why quantum theory had to be
indeterministi . In his view, the indeterminism was ne essary be ause
of the inevitable disturban e involved in setting up an initial quantum
state (pp. 492 f.):
I should now like to express my views on determinism and the nature of
the numbers appearing in the al ulations of the quantum theory .... . In
the lassi al theory one starts from ertain numbers des ribing ompletely the
initial state of the system, and dedu es other numbers that des ribe ompletely
the nal state. This deterministi theory applies only to an isolated system.
But, as Professor Bohr has pointed out, an isolated system is by denition
unobservable. One an observe the system only by disturbing it and observing
its rea tion to the disturban e. Now sin e physi s is on erned only with
observable quantities the deterministi lassi al theory is untenable.
Dira 's argument seems unsatisfa tory. First of all, as a general philosophi al point, the laim that `physi s is on erned only with observable
quantities' is not realisti . As is well known to philosophers of s ien e
as well as to experimentalists, observation is `theory-laden': in order
to arry out observations (or measurements) some body of theory is
required in order to know how to arry out a orre t observation (for
example, some knowledge is required of how the system being measured
intera ts with the apparatus, in order to design a orre tly fun tioning
apparatus). Thus, some body of theory is ne essarily on eptually prior
to observation, and it is logi ally impossible to base physi al theory on
òbservables' only. More spe i ally, Dira laims that lassi al determinism is untenable be ause of the disturban e involved in observing a system: but there are many ases in lassi al physi s where experimenters
an use their knowledge of the intera tions involved to ompensate for
the disturban e aused by the measurement. Disturban e per se annot
be a reason for indeterminism.
It may well be that Dira had in mind the kind of ìrredu ible' or
ùn ontrollable' disturban e that textbooks ommonly asso iate with
212
the un ertainty prin iple. However, it is interesting that, in fa t, Dira

goes on to say that the disturban es applied by an experimenter are
under his ontrol (p. 493, Dira 's itali s):
In the quantum theory one also begins with ertain numbers and dedu es
others from them. .... The disturban es that an experimenter applies to a
system to observe it are dire tly under his ontrol, and are a ts of freewill by
him. It is only the numbers that des ribe these a ts of freewill that an be taken
as initial numbers for a al ulation in the quantum theory. Other numbers
des ribing the initial state of the system are inherently unobservable, and do
not appear in the quantum theoreti al treatment.
The `disturban es' refer to the experimental operations that the experimenter hooses to apply to the system, and indeed these are normally
regarded as freely ontrolled by the experimenter (at least in some
ee tive sense). The sense in whi h the word `disturban e' is being
used here is quite dierent from the textbook sense of un ontrollable
disturban e asso iated with quantum un ertainty. Dira seems to regard
quantum numbers, or eigenvalues, as representing the extent to whi h an
experimenter an ontrollably manipulate a system. Ma ros opi operations are under our ontrol, and through these we an prepare an initial
state spe ied by parti ular quantum numbers. For Dira , these initial
numbers represent à ts of freewill' in the form of laboratory operations.
(The nal remark about òther numbers' that are unobservable, and
that do not appear in quantum theory, is intriguing, and might be taken
as suggesting that there are other degrees of freedom that annot be
ontrolled by us.)
A view quite dierent from Dira 's is expressed by Born at the very
end of the general dis ussion. A ording to Born, the onstraints on
the preparation of an initial quantum state are not what distinguishes
quantum from lassi al me hani s, for in lassi al physi s too (p. 519)
.... the pre ision with whi h the future lo ation of a parti le an be predi ted
depends on the a ura y of the measurement of the initial lo ation.
For Born, the dieren e rather lies in the law of propagation of probability pa kets:
It is then not in this that the manner of des ription of quantum me hani s,
by wave pa kets, is dierent from lassi al me hani s. It is dierent be ause
the laws of propagation of pa kets are slightly dierent in the two ases.
8.3 Visualisability and the spa etime framework
213

With hindsight, from a ontemporary perspe tive, perhaps the most
hara teristi feature of quantum physi s is the apparent absen e of
visualisable pro esses taking pla e within a spa etime framework. From
Bell's theorem, it appears that any attempt to provide a omplete des ription of quantum systems (within a single world) will require some
form of nonlo ality, leading to di ulties with relativisti spa etime. At
the time of writing, we possess only one hidden-variables theory of broad
s ope the pilot-wave theory of de Broglie and Bohm and in this
theory there is a eld on onguration spa e (not 3-spa e) that ae ts
the motion of quantum systems.a Arguably, then, pilot-wave theory does
not really t into a spa etime framework: the physi s is grounded in
onguration spa e, and the intera tions en oded in the pilot wave take
pla e outside of 3-spa e. Instead of a nonlo al hidden-variables theory,
one might prefer to have a omplete a ount of quantum behaviour in
terms of many worlds: there too, one leaves behind ordinary spa etime as
a basi framework for physi s, sin e the totality of what is real annot be
mapped onto a single spa etime geometry. Generally speaking, whatever
one's view of quantum theory today, the usual spa etime framework
seems too restri tive, and unable to a omodate (at least in a natural
way) the phenomena asso iated with quantum superposition and entanglement.b
We saw in se tion 8.1 that the quantum or matrix me hani s of
Born and Heisenberg ertainly did not provide an a ount of physi al
systems in a spa etime framework. In ontrast, the initial pra ti al
su ess of S hrdinger's wave me hani s in 1926 had led some workers
to think that an understanding in terms of (wave pro esses in) spa e
and time might be possible after all. The resulting tension between
S hrdinger on the one hand, and Bohr, Heisenberg and Pauli on the
other (in the year pre eding the Solvay meeting) has been des ribed at
length in se tion 4.6, where we saw that for S hrdinger the notion
of Àns hauli hkeit' in the sense of visualisability in a spa etime
framework played a key role.
The lash between quantum physi s and the spa etime framework
was a entral theme of the fth Solvay onferen e. There was a notable
a As already mentioned in se tion 5.2.1, attempts to onstru t hidden-variables
theories without an ontologi al wave fun tion, along the lines pioneered by Fnyes
(1952) and Nelson (1966), seem to fail (Wallstrom 1994; Pearle and Valentini
2006).
b A possible ex eption here is some version of quantum theory with dynami al wave
fun tion ollapse.
214
tension between those parti ipants who still hoped for a spa etime-based
theory and those who insisted that no su h theory was possible. These
dieren es are espe ially apparent in the general dis ussion where, as we
have already dis ussed in se tion 7.3, di ulties were raised on erning
lo ality and relativity.
Lorentz, in his opening remarks at the rst session of the general
dis ussion, seems to set the tone for one side of the debate, by speaking
in favour of spa e and time as a basi framework for physi s (p. 476):
We wish to make a representation of the phenomena, to form an image of
them in our minds. Until now, we have always wanted to form these images
by means of the ordinary notions of time and spa e. These notions are perhaps
innate; in any ase, they have developed from our personal experien e, by our
daily observations. For me, these notions are lear and I onfess that I should
be unable to imagine physi s without these notions. The image that I wish to
form of phenomena must be absolutely sharp and denite, and it seems to me
that we an form su h an image only in the framework of spa e and time.
Lorentz's ommittment to pro esses taking pla e in spa e and time

was shared by de Broglie and S hrdinger, even though both men had
found themselves unable to avoid working in terms of onguration
spa e. As we saw in se tion 2.4, in his report de Broglie presented
his pilot-wave dynami s, with a guiding eld in onguration spa e,
only as a makeshift: he hoped that his pilot-wave dynami s would turn
out to be an ee tive theory only, and that underlying it would be a
theory of wave elds in 3-spa e with singularities representing parti le
motion (the double-solution theory). S hrdinger, too, despite working
with a many-body wave equation in onguration spa e, hoped that the
physi al ontent of his theory ould be ultimately interpreted in terms
of pro esses taking pla e in 3-spa e (see hapter 4).
In ontrast, some of the other parti ipants, in parti ular Bohr and
Pauli, wel omed indeed insisted upon the break with the spa etime
framework. As Bohr put it in the general dis ussion, after Einstein's
remarks on lo ality and ompleteness (p. 488):
The whole foundation for [a ausal spa etime des ription is taken away by
quantum theory, for it is based on [the assumption of observations without
interferen e.
In support of Bohr's ontention, Pauli (pp. 489 f.) provided an intriguing

argument to the ee t that intera tions between parti les annot be
understood in a spa etime framework. Spe i ally, Pauli based his argument on the quantum-theoreti al a ount of the long-range intera tions
known as van der Waals for es.
215
Pauli begins his argument by saying (in agreement with Bohr) that
the use of multi-dimensional onguration spa e is
only a te hni al means of formulating mathemati ally the laws of mutual a tion between several parti les, a tions whi h ertainly do not allow themselves
to be des ribed simply, in the ordinary way, in spa e and time.
Here, on the one hand, onguration spa e has only mathemati al signi an e. But on the other hand, as Einstein feared (see se tion 7.3),
there is a ording to Pauli no expli it a ount of lo al a tion in spa e
and time. Pauli adds that the multi-dimensional method might one day
be repla ed with what we would now all quantum eld theory. He then
goes on to say that, in any ase, in a ordan e with Bohr's point of
view, no matter what `te hni al means' are used to des ribe `the mutual
a tions of several parti les', su h a tions ` annot be des ribed in the
ordinary manner in spa e and time' (p. 489).
Pauli then illustrates his point with an example. He onsiders two
widely-separated hydrogen atoms, ea h in their ground state, and asks
what their ènergy of mutual a tion' might be. A ording to the usual
des ription in spa e and time, says Pauli, for large separations there
should be no mutual a tion at all. And yet (p. 490),
when one treats the same question by the multi-dimensional method, the result
is quite dierent, and in a ordan e with experiment.
What Pauli is referring to here is the problem of a ounting for van der
Waals for es between atoms and mole ules. Classi ally, mole ules with
dipole moments tend to align, resulting in a mean intera tion energy
1/R6 (where R is the distan e between two mole ules). However, many
mole ules exhibiting van der Waals for es have zero dipole moment; and
while the lassi al orientation ee t be omes negligible at high temperatures, van der Waals for es do not. The problem of explaining van
der Waals for es was nally solved by quantum theory, beginning with
the work of Wang in 1927.a (Note that the ee t here is quite distin t
from that of ex hange for es resulting from the Pauli ex lusion prin iple.
The latter for es are important only when the relevant ele troni wave
fun tions have signi ant overlap, whereas van der Waals for es o ur
between neutral atoms even when they are so far apart that their harge
louds have negligible overlap.)
A standard textbook al ulation of the van der Waals for e, between
two hydrogen atoms (1 and 2) separated by a displa ement R, pro eeds
a For a detailed review, see Margenau (1939).
216
as follows.a The unperturbed energy eigenstate is the produ t = 01 02

of the two ground states. If R = |R| is mu h larger than the Bohr radius,
the lassi al ele trostati potential between the atoms is

e2
(r1 R)(r2 R)
V = 3 r1 r2 3
(3)
,
R
R2
where r1 , r2 are respe tively the positions of ele trons 1, 2 relative to
their nu lei. In the state = 01 02 the mean values of r1 , r2 both vanish,
so that the expe tation value of V vanishes. However, the presen e of
V perturbs the ground state of the system, to a new and entangled
state , satisfying (H0 + V ) = (2E0 + E) , where H0 is the
unperturbed Hamiltonian, E0 is the ground-state energy of hydrogen,
and E is the energy perturbation from whi h one dedu es the van
der Waals for e. Standard perturbation methods show that E has a
leading term proportional to 1/R6 , a ounting for the (attra tive) van
der Waals for e.
Pauli regarded this result whi h had been obtained by Wang from
the S hrdinger equation in onguration spa e as eviden e that in
quantum theory there are intera tions that annot be des ribed in terms
of spa e and time. The result may be roughly understood lassi ally, as
Pauli points out, by imagining that in ea h atom there is an os illating
dipole moment that an indu e (and so intera t with) a dipole moment
in the other atom. But su h understanding is only heuristi . In a proper
treatment, using `multi-dimensional wave me hani s', the orre t result
is obtained by methods that, a ording to Pauli, annot be understood
in a spa etime framework.a
Pauli's point seems to have been that, be ause the perturbed wave
fun tion is entangled, multi-dimensional onguration spa e plays a
ru ial role in bringing about the orre t result, whi h therefore annot
be properly understood in terms of 3-spa e alone. If this argument seems
unwarranted, it ought to be remembered that, before Wang's derivation
in 1927, there had been a long history of failed attempts to explain van
der Waals for es lassi ally (Margenau 1939).
Disagreements over both the usefulness and the tenability of the spa etime framework are also apparent elsewhere in the general dis ussion. For
example, Kramers (p. 497) asks: `What advantage do you see in giving a
a See, for example, S hi (1955, pp. 17680).
a A ording to Ehrenfest, Bohr also gave an argument (in a onversation with
Einstein) against a spa etime des ription when treating many-parti le problems.
See se tion 1.5.
217
pre ise value to the velo ity v of the photons?' To this de Broglie replies
(in a spirit similar to that of Lorentz above):
This allows one to imagine the traje tory followed by the photons and to
spe ify the meaning of these entities; one an thus onsider the photon as a
material point having a position and a velo ity.
Kramers was un onvin ed:

I do not very well see, for my part, the advantage that there is, for the
des ription of experiments, in making a pi ture where the photons travel along
well-dened traje tories.
In this ex hange, Kramers had suggested that de Broglie's theory ould

not explain radiation pressure from a single photon, thereby questioning
the tenability (and not just the usefulness) of the spa etime framework
for the des ription of elementary intera tions. (Cf. se tion 10.4.)
Later in the general dis ussion, another argument against de Broglie's
theory is provided by Pauli. As we shall dis uss at length in se tion 10.2,
Pauli laims on the basis of an example that pilot-wave theory annot a ount for the dis rete energy ex hange taking pla e in inelasti ollisions.
It is interesting to note that, a ording to Pauli, the root of the di ulty
lies in the attempt to onstru t a deterministi parti le dynami s in a
spa etime framework (p. 511):
.... this di ulty .... is due dire tly to the ondition assumed by Mr de Broglie,
that in the individual ollision pro ess the behaviour of the parti les should
be ompletely determined and may at the same time be des ribed ompletely
by ordinary kinemati s in spa etime.
9
In this hapter we address proposals (by Einstein, and by Bohr, Kramers and Slater) a ording to whi h quantum events are inuen ed by
`guiding elds' in 3-spa e. These ideas led to a predi ted violation of
energy-momentum onservation for single events, in ontradi tion with
experiment. The ontradi tion was resolved only by the introdu tion of
guiding elds in onguration spa e. All this took pla e before the fth
Solvay onferen e, but nevertheless forms an important ba kground to
some of the dis ussions that took pla e there.
9.1 Einstein's early attempts to formulate a dynami al theory

of light quanta
Sin e the publi ation of his light-quantum hypothesis in 1905, Einstein
had been engaged in a solitary struggle to onstru t a detailed theory
of light quanta, and to understand the relationship between the quanta
on the one hand and the ele tromagneti eld on the other.a Einstein's
eorts in this dire tion were never published. We know of them indire tly: they are mentioned in letters, and they are alluded to in Einstein's
1909 le ture in Salzburg. Einstein's published papers on light quanta
ontinued for the most part in the same vein as his 1905 paper: using the
theory of u tuations to make dedu tions about the nature of radiation,
without giving details of a substantial theory. Einstein was essentially
alone in his dualisti view of light, in whi h lo alised energy fragments
a In 1900 Plan k had, of ourse, ee tively introdu ed a quantisation in the
intera tion between radiation and matter; but it was Einstein in 1905 who rst
proposed that radiation itself (even in free spa e) onsisted, at least in part, of
spatially lo alised energy quanta. In 1918 Einstein wrote to his friend Besso: Ì
do not doubt anymore the reality of radiation quanta, although I still stand quite
alone in this onvi tion' (original itali s, as quoted in Pais (1982, p. 411)).
218
9.1 Einstein's early attempts
219
oexisted with extended waves, until the work of de Broglie in 1923

whi h extended the dualism to all parti les, and made onsiderable
progress towards a real theory (see hapter 2).
A glimpse of Einstein's attempts to formulate a dynami al theory of
light quanta may be obtained from a lose reading of his 1909 Salzburg
le ture. There, as we saw in se tion 7.2, Einstein marshalled eviden e
that light waves ontain lo alised energy fragments. He suggested that
the ele tromagneti eld is asso iated with singular points at whi h the
energy is lo alised, and he oered the following remarkable (if heuristi )
pi ture:
I more or less imagine ea h su h singular point as being surrounded by a eld of
for e whi h has essentially the hara ter of a plane wave and whose amplitude
de reases with the distan e from the singular point. If many su h singularities
are present at separations that are small ompared with the dimensions of the
eld of for e of a singular point, then su h elds of for e will superpose, and
their totality will yield an undulatory eld of for e that may dier only slightly
from an undulatory eld as dened by the urrent ele tromagneti theory of
light. (Einstein 1909, English translation, p. 394)
Here, ea h light quantum is supposed to have an extended eld asso iated with it, and large numbers of quanta with their asso iated elds are
supposed to yield (to a good approximation) the ele tromagneti eld of
Maxwell's theory. In other words, the ele tromagneti eld as we know
it emerges from the olle tive behaviour of large numbers of underlying
elds asso iated with individual quanta.
A similar view is expressed in a letter from Einstein to Lorentz written
a few months earlier, in May 1909:
I on eive of the light quantum as a point that is surrounded by a greatly
extended ve tor eld, that somehow diminishes with distan e. Whether or not
when several light quanta are present with mutually overlapping elds one
must imagine a simple superposition of the ve tor elds, that I annot say. In
any ase, for the determination of events, one must have equations of motion
for the singular points in addition to the dierential equations for the ve tor
eld. (Quoted in Howard 1990, p. 75)
From the last senten e, it is lear that Einstein's on eption was supposed to be deterministi .
Einstein's view in 1909, then, is remarkably reminis ent of de Broglie's
pilot-wave theory as well as of his theory of the `double solution' (from
whi h de Broglie hoped pilot-wave theory would emerge, see hapter 2).
Einstein seems to have thought of ea h individual light quantum as being
220
a ompanied by some kind of eld in 3-spa e that ae ts the motion of

the quantum.
Fas inating rea tions to Einstein's ideas appear in the re orded dis ussion that took pla e after Einstein's le ture. Stark pointed out a
phenomenon that seemed to speak in favour of lo alised energy quanta in
free spa e: èven at great distan es, up to 10 m, ele tromagneti radiation
that has left an X-ray tube for the surrounding spa e an still a hieve
on entrated a tion on a single ele tron' (Einstein 1909, English translation, p. 397). Stark's point here is, again, Einstein's lo ality argument:
as we noted at the end of se tion 7.2, in retrospe t it seems puzzling
that this simple and ompelling argument was not widely understood
mu h earlier, but here Stark learly appre iates it. However and this
is probably why the argument did not gain urren y doubts were
raised as to how su h a theory ould explain interferen e. Plan k spoke
as follows:
Stark brought up something in favor of the quantum theory, and I wish
to bring up something against it; I have in mind the interferen es at the
enormous phase dieren es of hundreds of thousands of wavelengths. When a
quantum interferes with itself, it would have to have an extension of hundreds
of thousands of wavelengths. This is also a ertain di ulty. (Einstein 1909,
English translation, p. 397)
To this obje tion, Einstein gives a most interesting reply:

I pi ture a quantum as a singularity surrounded by a large ve tor eld. By
using a large number of quanta one an onstru t a ve tor eld that does
not dier mu h from the kind of ve tor eld we assume to be involved in
radiations. I an well imagine that when rays impinge upon a boundary
surfa e, a separation of the quanta takes pla e, due to intera tion at the
boundary surfa e, possibly a ording to the phase of the resulting eld at
whi h the quanta rea h the interfa e. .... I do not see any fundamental di ulty
in the interferen e phenomena. (Einstein 1909, English translation, p. 398)
Again, the similarity to de Broglie's later ideas is striking. Einstein seems

to think that the asso iated waves an ae t the motions of the quanta,
in su h a way as to a ount for interferen e.
It should be noted, though, that in this ex hange it is somewhat
un lear whether the subje t is interferen e for single photons or for
many photons. Einstein talks about interferen e in terms of the olle tive behaviour of many quanta, rather than in terms of one quantum
at a time. While single-photon interferen e with very feeble light was
observed by Taylor (1909) in the same year, Stark at least seems not
to know of Taylor's results, for at this point (just before Einstein's
9.2 The failure of energy-momentum onservation
221
reply) he interje ts that `the experiments to whi h Mr Plan k alluded

involve very dense radiation .... . With radiation of very low density,
the interferen e phenomena would most likely be dierent' (Einstein
1909, English translation, p. 397). On the other hand, Plan k may well
have thought of the light-quantum hypothesis as implying that light
quanta would move independently, like the mole ules of an ideal gas (in
whi h ase even for intense radiation one ould onsider the motion of
ea h quantum independently). Einstein ountered pre isely su h a view
at the beginning of his reply to Plan k, where he states that ìt must
not be assumed that radiations onsist of non-intera ting quanta; this
would make it impossible to explain the phenomena of interferen e'. This
might be read as implying that intera tions among dierent quanta are
essential, and that interferen e would not o ur with one photon at a
time. However, Einstein may simply have meant that the light quanta
annot be thought of as free parti les: they must be a ompanied by a
wave as well, in order to explain interferen e. (This last reading ts with
Einstein's dis ussion, in the le ture, of thermal u tuations in radiation:
these annot be obtained from a gas of free and independent parti les
alone; a wave-like omponent is also needed.)
While there is some un ertainty over the details of Einstein's proposal, in retrospe t, given our present understanding of how de Broglie's pilot-wave theory provides a straightforward explanation of parti le
interferen e (see se tion 6.1.1), Einstein's reply to Plan k seems very
reasonable. However, it appears that yet another of Einstein's arguments
was not appre iated by his ontemporaries. Seven years later, after having veried Einstein's photoele tri equation experimentally, Millikan
nevertheless ompletely reje ted Einstein's light-quantum hypothesis,
whi h he alled `re kless .... be ause it ies in the fa e of the thoroughly
established fa ts of interferen e' (Millikan 1916, p. 355).

It appears that Einstein was still thinking along similar lines in the
1920s, though again without publishing any detailed theory. As we saw
at the end of se tion 6.2.1, in his ollision papers of 1926 Born notes
the analogy between his own work and Einstein's ideas: Ì start from a
remark by Einstein .... ; he said .... that the waves are there only to show
the orpus ular light quanta the way, and in this sense he talked of a
ghost eld ' (Born 1926b, pp. 8034). Born gives no referen e to any
published paper of Einstein's, however.
222
How Einstein's thinking at this time ompared with that in 1909 is

hard to say. Certainly, in 1909 he thought of the ele tromagneti eld as
being built up from the olle tive behaviour of large numbers of ve tor
elds asso iated with individual quanta. In su h a s enario, it seems
plausible that in the right ir umstan es the intensity of the emergent
ele tromagneti eld ould a t as, in ee t, a probability eld. (Unlike
Born, Einstein would have regarded a purely probabilisti des ription as
a makeshift only.)
Just one year before Born's ollision papers, in 1925, Einstein gave
a olloquium in Berlin where he indeed dis ussed the idea that every
parti le (in luding ele trons, following de Broglie) was a ompanied by
a `Fhrungsfeld' or guiding eld (Pais 1982, p. 441; Howard 1990, p. 72).
A ording to Wigner, who was present at the olloquium:
Yet Einstein, though in a way he was fond of it, never published it. He realized
that it is in oni t with the onservation prin iples: at a ollision of a light
quantum and an ele tron for instan e, both would follow a guiding eld.
But these guiding elds give only the probabilities of the dire tions in whi h
the two omponents, the light quantum and the ele tron, will pro eed. Sin e
they follow their dire tions independently, .... the momentum and the energy
onservation laws would be obeyed only statisti ally .... . This Einstein ould
not a ept and hen e never took his idea of the guiding eld quite seriously.
(Wigner 1980, p. 463)
In the early 1920s, then, Einstein was still thinking along lines that
are reminis ent of de Broglie's work, but he never published these ideas
be ause they oni ted with the onservation laws for individual events.
The di ulty Einstein fa ed was over ome only by the introdu tion
(through the work of de Broglie, S hrdinger, and also Born) of a guiding
eld in onguration spa e a single (and generally entangled) eld
that determined probabilities for all parti les olle tively. In Einstein's
approa h, where ea h parti le had its own guiding eld, the possibility
of entanglement was pre luded, and the orrelations were not strong
enough to guarantee energy-momentum onservation for single events.
As a simplied model of what Einstein seems to have had in mind,
onsider two parti les 1 and 2 moving towards ea h other in one dimension, with equal and opposite momenta p and p respe tively. S hemati ally, let us represent this with an initial wave fun tion (x1 , x2 , 0) =
1+ (x1 , 0)2 (x2 , 0), where 1+ and 2 are broad pa kets (approximating
plane waves) moving along +x and x respe tively. Let the pa kets meet
in a region entred around the origin, where we imagine that an elasti
ollision takes pla e with probability 12 . At large times, we assume that
223
(x1 , x2 , t) takes the s hemati and entangled form

1
(x1 , x2 , t) = 1+ (x1 , t)2 (x2 , t) + 1 (x1 , t)2+ (x2 , t) ,
2
(1)
with the rst bran h orresponding to the parti les having moved freely
past ea h other and the se ond orresponding to an elasti ollision reversing their motions. The initial state has zero total momentum. In the
nal state, both possible out omes orrespond to zero total momentum.
Thus, in quantum theory, whatever the out ome of an individual run of
the experiment, momentum is always onserved. Now imagine if, instead
of using a single wave fun tion (x1 , x2 , t) in onguration spa e, we
made use of two 3-spa e waves 1 (x, t), 2 (x, t) (one for ea h parti le).
It would then seem natural that, during the ollision, the initial 3-spa e
wave 1 (x, 0) = 1+ (x, 0) for parti le 1 would evolve into
1 (x, t) = 1+ (x, t) + 1 (x, t) ,
(2)
while the initial 3-spa e wave 2 (x, 0) = 2 (x, 0) for parti le 2 would
evolve into
2 (x, t) = 2 (x, t) + 2+ (x, t) .
(3)
If the amplitude of ea h 3-spa e wave determined the probabilities for

the respe tive parti les, there would then be four (equiprobable) possible
out omes for the s attering experiment, the two stated above, together
with one in whi h both parti les move to the right and one in whi h both
parti les move to the left. These possibilities would orrespond, in ee t,
to a nal ( onguration-spa e) wave fun tion of the separable form

1+ + 1 2 + 2+ = 1+ 2 + 1+ 2+ + 1 2 + 1 2+ .
(4)
The nal total momenta for the four (equiprobable) possible out omes
are 0, +2p, 2p, 0. The total momentum would not be generally onserved for individual out omes, but onservation would hold on average.
Note the fundamental dieren e between (1) and (4): the former is
entangled, the latter is not.
Einstein onsidered su h a failure of the onservation laws reason
enough to reje t the idea. A theory bearing some resemblan e to this
s heme was, however, proposed and published by Bohr, Kramers and
Slater (1924a,b).
In the Bohr-Kramers-Slater (BKS) theory, there are no photons. Asso iated with an atom in a stationary state is a `virtual radiation eld'
ontaining all those frequen ies orresponding to transitions to and from
224
other stationary states. The virtual eld determines the transition probabilities for the atom itself, and also ontributes to the transition
probabilities for other, distant atoms. However, the resulting orrelations
between widely-separated atoms are not strong enough to yield energymomentum onservation for individual events ( f. Compton's a ount of
the BKS theory in his report, p. 331).a
To illustrate this, onsider again (in modern language) an atom A
emitting a photon of energy Ei Ef that is subsequently absorbed by
a distant atom B. In the BKS theory, the transition at B is not dire tly
aused by the transition at A (as it would be in a simple `semi lassi al'
pi ture of the emission, propagation, and subsequent absorption of a
lo alised light quantum). Rather, the virtual radiation eld of A ontains
terms orresponding to transitions of A, and this eld ontributes to the
transition probabilities at B. Yet, if B a tually undergoes a transition
orresponding to an energy in rease Ei Ef , atom A is not onstrained
to make a transition orresponding to an energy de rease Ei Ef .
The onne tion between the atoms is merely statisti al, and energy
onservation holds only on average, not for individual pro esses.a
Einstein obje ted to the BKS theory in a olloquium, and in private
letters and onversations (Mehra and Re henberg 1982a, pp. 5534; Pais
1982, p. 420; Howard 1990, pp. 714) partly for the same reason he
had not published his proposals: that, as he believed, energy-momentum
onservation would hold even in the ase of elementary intera tions
between widely-separated systems.b Einstein's expe tation was subsequently onrmed by the Bothe-Geiger and Compton-Simon experiments (Bothe and Geiger 1925b, Compton and Simon 1925).
In the Bothe-Geiger experiment, by means of ounter oin iden es,
Compton s attering (that is, relativisti ele tron-photon s attering) was
studied for individual events, to see if the outgoing s attered photon
and the outgoing re oil ele tron were produ ed simultaneously. Stri t
temporal oin iden es were expe ted on the basis of the light-quantum
hypothesis, but not on the basis of the BKS theory. Su h oin iden es
a For a detailed a ount of the BKS theory, see Darrigol (1992, hap. 9).
a Note that Slater's original theory did ontain photons, whose motions were guided
(statisti ally) by the ele tromagneti eld. The photons were removed at the
instigation of Bohr and Kramers (Mehra and Re henberg 1982a, pp. 5437).
b In a letter to Ehrenfest dated 31 May 1924, Einstein wrote: `This idea [the BKS
theory is an old a quaintan e of mine, but one whom I do not regard as a
respe table fellow' (Howard 1990, p. 72). Einstein listed ve riti isms, in luding
the violation of the onservation laws, and a di ulty with thermodynami s
(Mehra and Re henberg 1982a, p. 553).
225
were in fa t observed by Bothe and Geiger.a In the Compton-Simon

experiment, Compton s attering was again studied, with the aim of
verifying the onservation laws for individual events. Su h onservation
was in fa t observed by Compton and Simon.b As a result of these
experiments, it was widely on luded that the BKS theory was wrong.c
It appears that, even at the time of the fth Solvay onferen e, Einstein was still thinking to some extent in terms of guiding elds in
3-spa e. Eviden e for this omes from an otherwise in omprehensible
remark Einstein made during his long ontribution to the general dis ussion. There, Einstein ompared two interpretations of the wave fun tion
for a single ele tron. On Einstein's view I, represents an ensemble
or ` loud' of ele trons, while on his view II is a omplete des ription
of an individual ele tron. As we have dis ussed at length in se tion 7.1,
Einstein argued that interpretation II is in onsistent with lo ality. Now,
Einstein also made the following remark on erning the onservation of
energy and momentum a ording to interpretations I and II (p. 487):
The se ond on eption goes further than the rst .... . It is only by virtue of II
that the theory ontains the onsequen e that the onservation laws are valid
for the elementary pro ess; it is only from II that the theory an derive the
result of the experiment of Geiger and Bothe .... .
As urrently understood, of ourse, a purely `statisti al' interpretation

of the wave fun tion ( f. se tion 5.2.4) would yield orre t predi tions, in
agreement with the onservation laws for elementary pro esses (su h as
s attering). Why, then, did Einstein assert that interpretation I would
oni t with su h elementary onservation? It must surely be that, for
whatever reason, he was thinking of interpretation I as tied spe i ally
to wave fun tions in 3-spa e, resulting in a failure of the onservation
laws for single events as dis ussed above. In ontrast, spe i ally regarding interpretation II, Einstein expli itly asserted (p. 487) that it
is èssentially tied to a multi-dimensional representation ( onguration
spa e)'.
The oni t between Einstein's ideas about guiding elds in 3-spa e
and energy-momentum onservation was, as we have mentioned, resolved
only by the introdu tion of a guiding eld in onguration spa e
with the asso iated entanglement and nonseparability that Einstein was
a See Compton's a ount of `Bothe and Geiger's oin iden e experiments', pp. 350 f.,
and also Mehra and Re henberg (1982a, pp. 60912).
b See Compton's a ount of `Dire tional emission of s attered X-rays', pp. 350 f.,
and also Mehra and Re henberg (1982a, p. 612).
Cf. Bohr's remarks in the dis ussion of Compton's report, p. 359.
226
to nd so obje tionable. Einstein's early worries about nonseparability

(long before the EPR paper) have been extensively do umented by Howard (1990). Dening `separability' as the idea that `spatio-temporally
separated systems possess well-dened real states, su h that the joint
state of the omposite system is wholly determined by these two separate
states' (Howard 1990, p. 64), Howard highlights the fundamental di ulty Einstein fa ed in having to hoose, in ee t, between separability
and energy-momentum onservation:
But as long as the `guiding' or `virtual elds' [determining the probabilities for
parti le motions, or for atomi transitions are assigned separately, one to ea h
parti le or atom, one annot arrange
of individual systems
and
both
for the merely probabilisti behavior
for orrelations between intera ting systems su-
ient to se ure stri t energy-momentum onservation in all individual events.

As it turned out, it was only S hrdinger's relo ation of the wave elds from
physi al spa e to onguration spa e that made possible the assignment of
joint wave elds that ould give the strong orrelations needed to se ure stri t
onservation. (Howard 1990, p. 73)
Here, then, is a remarkable histori al and physi al onne tion between nonseparability or entanglement on the one hand, and energymomentum onservation on the other. The introdu tion of probability
waves in onguration spa e, with generi entangled states, nally made
it possible to se ure energy-momentum onservation for individual emission and s attering events, at the pri e of introdu ing a fundamental
nonseparability into physi s.
10
S attering and measurement in de Broglie's
pilot-wave theory
At the fth Solvay onferen e, some questions that are losely related
to the quantum measurement problem (as we would now all it) were
addressed in the ontext of pilot-wave theory, in both the dis ussion
following de Broglie's report and in the general dis ussion. Most of these
questions on erned the treatment of s attering (elasti and inelasti );
they were raised by Born and Pauli, and replies were given by Brillouin and de Broglie. Of spe ial interest is the famous and widely
misunderstood obje tion by Pauli on erning inelasti s attering.
Another question losely related to the measurement problem was raised
by Kramers, on erning the re oil of a single photon on a mirror.
In this hapter, we shall rst outline the pilot-wave theory of s attering, as urrently understood, and examine the extensive dis ussions of
s attering in the ontext of de Broglie's theory that took pla e at
the onferen e.
We shall see that de Broglie and Brillouin orre tly answered the query
raised by Born on erning elasti s attering. Further, we shall see that
Pauli's obje tion on erning the inelasti ase was both more subtle and
more onfused than is generally thought; in parti ular, Pauli presented
his example in terms of a misleading opti al analogy (that was originally
given by Fermi in a more restri ted ontext). Contrary to a widespread
view, de Broglie's reply to Pauli did ontain the essential points required
for a proper treatment of inelasti s attering; at the same time, Fermi's
misleading analogy onfused matters, and neither de Broglie in 1927 nor
Bohm in 1952 saw what the true fault with Pauli's example was. (As we
shall also see, a proper pilot-wave treatment of Pauli's example was not
given until 1956, by de Broglie.)
We shall also outline the pilot-wave theory of quantum measurement,
again as urrently understood, and we shall use this as a ontext in
227
228
S attering and measurement in de Broglie's theory
whi h to examine the question raised by Kramers, of the re oil of a

single photon on a mirror, whi h de Broglie was unable to answer.

Let us rst onsider elasti s attering by a xed potential asso iated
with some s attering entre or region. An in ident parti le may (for a
pure quantum state) be represented by a freely-evolving wave pa ket
inc (x, t) that is spatially nite (that is, limited both longitudinally and
laterally), and that has mean momentum k. During the s attering, the
wave fun tion evolves into = inc + sc , where sc is the s attered
wave. At large distan es from the s attering region, and o the axis
(through the s attering entre) parallel to the in ident wave ve tor k,
only the s attered wave sc ontributes to the parti le urrent density j
where j is often used in textbook derivations of the s attering ross
se tion.
As is well known, the mathemati s of a fully time-dependent al ulation of the s attering of a nite pa ket may be simplied by resorting to
a time-independent treatment in whi h inc is taken to be an innitelyextended plane wave eikx . At large distan es from the s attering region
the wave fun tion = eikx + sc (a time-independent eigenfun tion of
the total Hamiltonian) has the asymptoti form
= eikz + f (, )
eikr
r
(1)
(taking the z -axis parallel to k, using spheri al polar oordinates entred

on the s attering region, and ignoring overall normalisation). The s attering amplitude f gives the dierential ross se tion d/d = |f (, )|2 .
In the standard textbook derivation of d/d, the urrent density j
is used to al ulate the rate of probability ow into an element of solid
angle d, where j is taken to be the urrent asso iated with sc only, even
though sc overlaps with inc = eikx . This is justied be ause the plane
wave eikx is, of ourse, merely an abstra tion used for mathemati al
onvenien e: a real in ident wave will be spatially limited, and will not
overlap with the s attered wave at the lo ation of the parti le dete tor
(whi h is assumed to be lo ated o the axis of in iden e, so as not to be
bathed in the in ident beam).
The above standard dis ussion of s attering may readily be re ast in
229
pilot-wave terms, where

v=
j
2
||
S
Im
=
m
(2)
(with = || e(i/)S ) is interpreted as the a tual velo ity eld of an

ensemble of parti les with positions distributed a ording to ||2 . The
dierential ross se tion d/d measures the fra tion of in ident parti les whose a tual traje tories end (asymptoti ally) in the element of
solid angle d.
That a real in ident pa ket is always spatially nite is, of ourse, an
elementary point known to every student of wave opti s. This (often
impli it) assumption is essential to introdu tory textbook treatments of
the s attering of light, whether by a Hertzian dipole or by a dira tion
grating. If the in ident wave were a literally innite plane wave, then
the s attered wave would of ourse overlap with the in ident wave everywhere, and no matter where a dete tor was pla ed it would be ae ted
by the in ident wave as well as by the s attered wave.
Certainly, the parti ipants at the fth Solvay onferen e many of
whom had extensive laboratory experien e were aware of this simple
point. We emphasise this be ause, as we shall see, the niteness of
in ident wave pa kets played a entral role in the dis ussions that took
pla e regarding s attering in de Broglie's theory.
Let us now onsider inelasti s attering: spe i ally, the s attering of
an ele tron by a hydrogen atom initially in the ground state. The atom
an be ome ex ited by the ollision, in whi h ase the outgoing ele tron
will have lost a orresponding amount of energy. Let the s attering
ele tron have position xs and the atomi ele tron have position xa . In a
time-dependent des ription, the total wave fun tion evolves into
(xs , xa , t) =
0 (xa )eiE0 t/ inc (xs , t) +
n (xa )eiEn t/ n (xs , t) . (3)
Here, the rst term is an initial produ t state, where 0 is the groundstate wave fun tion of hydrogen with ground-state energy E0 and inc
is a (nite) in ident pa ket. The s attering terms have omponents as
shown, where the n are the nth ex ited states of hydrogen and the
n are outgoing wave pa kets. It may be shown by standard te hniques
that, asymptoti ally, the nth outgoing pa ket n is entred on a radius
rn = (kn /m)t from the s attering region, where kn is the outgoing wave
number xed by energy onservation.
230
Be ause the outgoing (asymptoti ) pa kets n expand with dierent

speeds, they eventually be ome widely separated in spa e. The a tual
s attered ele tron with position xs (t) an o upy only one of these nonoverlapping pa kets, say i , and its velo ity will then be determined by
i alone. Further, the motion of the atomi ele tron will be determined
by the orresponding i alone, and after the s attering the atom will be
(in ee t) in an energy eigenstate i .
We are using two well-known properties of pilot-wave dynami s: (a) If
a wave fun tion = || e(i/)S is a superposition of terms 1 + 2 + ....
having no overlap in onguration spa e, then the phase gradient S
at the o upied point of onguration spa e (whi h gives the velo ity of
the a tual onguration) redu es to Si , where i = |i | e(i/)Si is the
o upied pa ket. (b) If the o upied pa ket i is a produ t over ertain
onguration omponents, then the velo ities of those omponents are
determined by the asso iated fa tors in the produ t.
Applying (a) and (b) to the ase dis ussed here, on e the n have
separated the total wave fun tion be omes a sum of non-overlapping
pa kets, where only one pa ket i = i (xa )eiEi t/ i (xs , t) an ontain
the a tual onguration (xa , xs ). The velo ity of the s attered ele tron is
then given by x s = (/m) Im(i /i ), while the velo ity of the atomi
ele tron is given by x a = (/m) Im(i /i ). Thus, there takes pla e an
ee tive ` ollapse of the wave pa ket' to the state i i .
It is straightforward to show that, if the initial ensemble of xs , xa has
distribution ||2 , the probability for ending in the ith pa ket that is,
the
the atom to end in the state i will be given by
R 3probability for
2
d xs |i (xs , t)| , in a ordan e with the usual quantum result.
Further, the ee tive ` ollapse' to the state i i is for all pra ti al
purposes irreversible. As argued by Bohm (1952a, p. 178), the s attered
parti le will subsequently intera t with many other degrees of freedom
making it very unlikely that distin t states i i , j j (i 6= j ) will
interfere at later times, as this would require the asso iated bran hes
of the total wave fun tion to overlap with respe t to every degree of
freedom involved.
Again, for mathemati al onvenien e, one often onsiders the limit in
whi h the in ident wave inc is unlimited (a plane wave). This makes
the al ulation of easier. In this limit, the outgoing wave pa kets n
be ome unlimited too, and overlap with ea h other everywhere: all the
terms in (3) then overlap in every region of spa e. If one naively assumed
that this limit orresponded to a real situation, the outgoing ele tron
would never rea h a onstant velo ity be ause it would be guided by a
231
superposition of overlapping terms, rather than by a single term in (3).

Similarly, after the s attering, the atomi ele tron would not be guided
by a single eigenfun tion i , and the atom would not nish in a denite
energy state. In any real situation, of ourse, inc will be limited in spa e
and time, and at large times the outgoing pa kets n will separate: the
traje tory of the s attered ele tron will be guided by only one of the
n and (in regions outside the path of the in ident beam) will not be
ae ted by inc . (Note that longitudinal niteness of the in ident wave
inc leads to a separation of the outgoing waves n from ea h other,
while lateral niteness of inc ensures that inc does not overlap with
the n in regions o the axis of in iden e.)
As we shall see, apart from the pra ti al irreversibility of the ee tive
ollapse pro ess, the above `pilot-wave theory of s attering' seems to
have been more or less understood by de Broglie (and perhaps also by
Brillouin) in O tober 1927. A detailed treatment of s attering was given
by Bohm in his rst paper on de Broglie-Bohm theory (Bohm 1952a).
In Bohm's se ond paper, however, appendix B gives a misleading
a ount of the de Broglie-Pauli en ounter at the fth Solvay onferen e (Bohm 1952b, pp. 1912), and this seems to be the sour e of
the widespread misunderstandings on erning this en ounter. Citing the
pro eedings of the fth Solvay onferen e, Bohm wrote the following:
De Broglie's suggestions met strong obje tions on the part of Pauli, in onne tion with the problem of inelasti s attering of a parti le by a rigid rotator.
Sin e this problem is on eptually equivalent to that of inelasti s attering of
a parti le by a hydrogen atom, whi h we have already treated .... , we shall
dis uss the obje tions raised by Pauli in terms of the latter example.
Bohm then des ribes `Pauli's argument': taking the in oming parti le to
have a plane wave fun tion, all the terms in (3) overlap, so that `neither
atom nor the outgoing parti le ever seem to approa h a stationary energy', ontrary to what is observed experimentally. A ording to Bohm:
Pauli therefore on luded that the interpretation proposed by de Broglie was
untenable. De Broglie seems to have agreed with the on lusion, sin e he
subsequently gave up his suggested interpretation.
Bohm then gives what he regards as his own, original answer to Pauli's
obje tion:
.... as is well known, the use of an in ident plane wave of innite extent is an
ex essive abstra tion, not realizable in pra ti e. A tually, both the in ident
and outgoing parts of the
-eld will always take the form of bounded pa kets.

n will ultimately
Moreover, .... all pa kets orresponding to dierent values of
232
obtain lassi ally des ribable separations. The outgoing parti le must enter
one of these pa kets, .... leaving the hydrogen atom in a denite but orrelated
stationary state.
By way of on lusion, Bohm writes:

Thus, Pauli's obje tion is seen to be based on the use of the ex essively abstra t
model of an innite plane wave.
At this point, in the light of what we have said above about the use
of plane waves in elementary wave opti s and in s attering theory, it
is natural to ask how a physi ist of Pauli's abilities ould have made
the glaring mistake that Bohm laims he made. In fa t, as we shall
see in the next se tion, Pauli's obje tion was not based on a failure to
appre iate the importan e of the niteness of initial wave pa kets. On the
ontrary, Pauli's obje tion shows some understanding of the ru ial role
played by limited pa kets in pilot-wave theory. What really happened is
that Pauli's obje tion involved a pe uliar and misleading analogy with
opti s, a ording to whi h the in ident pa ket appeared to be ne essarily
unlimited, in onditions su h as to prevent the required separation into
non-overlapping omponents.
10.2 Elasti and inelasti s attering: Born and Brillouin,

Pauli and de Broglie
In the dis ussion following de Broglie's report, Born suggests (p. 398)
that de Broglie's guidan e equation will fail for an elasti ollision between an ele tron and an atom. Spe i ally, Born asks if the ele tron
speed will be the same before and after the ollision, to whi h de Broglie
simply replies that it will.
Later in the same dis ussion, Brillouin (pp. 401 .) gives an extensive
and detailed presentation, explaining how de Broglie's theory a ounts
for the elasti s attering of a photon from a mirror. Brillouin is quite expli it about the role played by the nite extension of the in ident pa ket.
In his Fig. 2, Brillouin shows an in ident photon traje tory (at angle
to the normal) guided by an in oming and laterally-limited pa ket. The
pa ket is ree ted by the mirror, produ ing an outgoing pa ket that is
again laterally-limited. Near the mirror there is an interferen e zone,
where the in oming and outgoing pa kets overlap. As Brillouin puts it:
Let us draw a diagram for the ase of a limited beam of light falling on a plane
mirror; the interferen e is produ ed in the region of overlap of two beams.
10.2 Elasti and inelasti s attering
233
Brillouin sket hes the photon traje tory, whi h urves away from the
mirror as it enters the interferen e zone, moves approximately parallel
to the mirror while in the interferen e zone, and then moves away again,
eventually settling into a re tilinear motion guided by the outgoing
pa ket (Brillouin's Fig. 2). Brillouin des ribes the traje tory thus:
.... at rst a re tilinear path in the in ident beam, then a bending at the edge
of the interferen e zone, then a re tilinear path parallel to the mirror, with the
photon travelling in a bright fringe and avoiding the dark interferen e fringes;
then, when it omes out, the photon retreats following the dire tion of the
ree ted light beam.
Here we have a lear des ription of an in ident photon, guided by a nite

pa ket and moving uniformly towards the mirror, with the pa ket then
undergoing interferen e and s attering, while the photon is eventually
arried away again with a uniform motion by a nite outgoing
pa ket. (The ingoing and outgoing motions of the photon are stri tly
uniform, of ourse, only in the limit where the guiding pa kets be ome
innitely broad.)
Despite his des ription in terms of nite pa kets, however, in order
to al ulate the pre ise motion of the photon in the interferen e zone
Brillouin uses the standard devi e of treating the in oming pa ket as an
innite plane wave. In this (abstra t) approximation, he shows that the
photon moves parallel to the mirror with a speed v = c sin . Be ause
the in oming pa ket is in reality limited, a photon motion parallel to
the mirror is (approximately) realised only in the interferen e zone, as
Brillouin sket hes in his a ompanying Fig. 2.
While Brillouin's gure shows a laterally-limited pa ket, it is lear
from subsequent dis ussion that the in ident pa ket was impli itly regarded as limited longitudinally as well (as of ourse it must be in any
realisti situation). For in the general dis ussion the question of photon
ree tion by a mirror was raised again, and Einstein asked (p. 498)
what happens in de Broglie's theory in the ase of normal in iden e
( = 0), for whi h the formula v = c sin predi ts that the photons will
have zero speed. Pi ard responded to Einstein's query, and pointed
out that, indeed, the photons are stationary in the limiting ase of
normal in iden e. The meaning of this ex hange between Einstein and
Pi ard is lear: if a longitudinally-limited pa ket were in ident normally
on the mirror, the in ident pa ket would arry the photon towards the
mirror; in the region where the in ident and ree ted pa kets overlap,
the photon would be at rest; on e the pa ket has been ree ted, the
234
photon will be arried away by the outgoing pa ket. The dis ussion of
the ase of normal in iden e impli itly assumes that the in ident pa ket
is longitudinally-limited.
Thus, Brillouin's example of photon ree tion by a mirror illustrates
the point that the use of plane waves was for al ulational onvenien e
only, and that, when it ame to the dis ussion of real physi al examples,
it was lear to all (and hardly worth mentioning expli itly) that in ident
waves were in reality limited in extent (in all dire tions). Brillouin's
mathemati al use of plane waves parallels their use in the general theory
of s attering sket hed in se tion 10.1.
Elasti s attering is also dis ussed in de Broglie's report, for the
parti ular ase of ele trons in ident on a xed, periodi potential the
potential generated by a rystal latti e. This ase is espe ially interesting
in the present ontext, be ause it involves the separation of the s attered
wave into non-overlapping pa kets the interferen e maxima of dierent orders, well-known from the theory of X-ray dira tion with the
parti le entering just one of these pa kets (a point that is relevant to a
proper understanding of the de Broglie-Pauli en ounter). Here is how de
Broglie des ribes the dira tion of an ele tron wave by a rystal latti e
(p. 391):
.... the wave
will propagate following the general equation, in whi h one
has to insert the potentials reated by the atoms of the rystal onsidered as
entres of for e. One does not know the exa t expression for these potentials
but, be ause of the regular distribution of atoms in the rystal, one easily
realises that the s attered amplitude will show maxima in the dire tions
predi ted by Mr von Laue's theory. Be ause of the role of pilot wave played
by the wave
one must then observe a sele tive s attering of the ele trons
in these dire tions.
Again, as in Brillouin's dis ussion, there is no need to mention expli itly

the obvious point that the in ident wave will be spatially limited.
Let us now turn to the inelasti ase. This was dis ussed by de Broglie
in his report, in parti ular in the nal part, whi h in ludes a review
of re ent experiments involving the inelasti s attering of ele trons by
atoms of helium. De Broglie noted that, a ording to Born's al ulations
(using his statisti al interpretation of the wave fun tion), the dierential
ross se tion should show maxima as a fun tion of the s attering angle
(p. 391):
.... Mr Born has studied .... the ollision of a narrow beam of ele trons with an
atom. A ording to him, the urve giving the number of ele trons that have
235
suered an inelasti ollision as a fun tion of the s attering angle must show
maxima and minima .... .
As de Broglie then dis ussed in detail, su h maxima had been observed

experimentally by Dymond (though the results were only in qualitative
agreement with the predi tions). Having summarised Dymond's results,
de Broglie ommented (p. 392):
The above results must very probably be interpreted with the aid of the new
Me hani s and are to be related to Mr Born's predi tions.
We are now ready for a lose examination of Pauli's obje tion in the
general dis ussion, a ording to whi h there is a di ulty with de Broglie's theory in the ase of inelasti ollisions. That there might be su h
a di ulty was in fa t already suggested by Pauli a few months earlier,
in a letter to Bohr dated 6 August 1927, already quoted in hapter 2. As
well as noting the ex eptional quality of de Broglie's `Stru ture' paper
(de Broglie 1927b), in his letter Pauli states that he is suspi ious of
de Broglie's traje tories and annot see how the theory ould a ount
for the dis rete energy ex hange seen in individual inelasti ollisions
between ele trons and atoms (Pauli 1979, pp. 4045). (Su h dis rete
ex hange had been observed, of ourse, in the Fran k-Hertz experiment.)
This is essentially the obje tion that Pauli raises less than three months
later in Brussels.
In the general dis ussion (p. 509), Pauli begins by stating his belief
that de Broglie's theory works for elasti ollisions:
It seems to me that, on erning the statisti al results of s attering experiments, the on eption of Mr de Broglie is in full agreement with Born's theory
in the ase of elasti ollisions .... .
This preliminary omment by Pauli is signi ant. For if, as Bohm asserted in 1952, Pauli's obje tion was based on `the use of the ex essively
abstra t model of an innite plane wave' (Bohm 1952b, p. 192), then
Pauli would have regarded de Broglie's theory as problemati even in
the elasti ase. For in an elasti ollision, if the in ident wave inc
is innitely extended (a plane wave), then any part of the outgoing
region will be bathed in the in ident wave: the s attered parti le will
inevitably be ae ted by both parts of the superposition inc + sc , and
will never settle down to a onstant speed. Sin e Pauli agreed that the
outgoing speed would be onstant in the elasti ase as de Broglie
had asserted (in reply to Born) in the dis ussion following his report
236
Pauli presumably understood that the nite in ident pa ket would not
ae t the s attered parti le.
Pauli goes on to laim that de Broglie's theory will not work for inelasti ollisions, in parti ular for the example of s attering by a rotator.
To understand Pauli's point, it is important to distinguish between the
real physi al situation being dis ussed, and the opti al analogy used by
Pauli an analogy that had been introdu ed by Fermi as a onvenient
(and as we shall see limited) means to solve the s attering problem.
Pauli's obje tion is framed in terms of Fermi's analogy, and therein lies
the onfusion.
The real physi al set-up onsists of an ele tron moving in the (x, y)plane and olliding with a rotator. The latter is a model s attering entre
with one rotational degree of freedom represented by an angle .a Pauli
took the initial wave fun tion to be
0 (x, y, ) e(i/)(px x+py y+p ) ,
(4)
with p restri ted to p = m (m = 0, 1, 2, ...). The inelasti s attering

of an ele tron by a rotator had been treated by Fermi (1926) using an
analogy with opti s, a ording to whi h the (time-independent) s attering of an ele tron in two spatial dimensions by a rotator is mathemati ally equivalent to the (time-independent) s attering of a (s alar)
light wave (x, y, ) in three spatial dimensions by an innite dira tion
grating, with interpreted as a third spatial oordinate ranging over the
whole real line (, +). The innite `grating' onstitutes a periodi
potential, arising mathemati ally from the periodi ity asso iated with
the original variable . Similarly, the fun tion (x, y, ) is ne essarily
unlimited along the -axis. By onstru tion, then, both the in ident
wave and the grating are unlimited along the -axis. Fermi's analogy is
useful, be ause the dierent spe tral orders for dira ted beams emerging from the grating orrespond to the possible nal (post-s attering)
energy states of the rotator. However, as we shall see, Fermi's analogy
has only a very limited validity.
Pauli, then, presents his obje tion in terms of Fermi's opti al analogy.
He says (p. 511):
It is, however, an essential point that, in the ase where the rotator is in
a stationary state before the ollision, the in ident wave is unlimited in the
a Classi ally, a rotator might onsist of a rigid body free to rotate about a xed
axis. The quantum rotator was known to have a dis rete spe trum of quantised
energy levels ( orresponding to quantised states of angular momentum), and was
sometimes onsidered as a useful and simple model of a quantum system.
237
Fig. 10.1. S attering of a laterally-limited wave by a nite dira tion grating,

showing the separation of the rst-order beams from the zeroth-order beam.
dire tion of the [-axis. For this reason, the dierent spe tral orders of the
grating will always be superposed at ea h point of onguration spa e. If we
then al ulate, a ording to the pre epts of Mr de Broglie, the angular velo ity
of the rotator after the ollision, we must nd that this velo ity is not onstant.
In Fermi's three-dimensional analogy, for an in ident beam unlimited

along (as well as along x and y ) the s attered waves will indeed
overlap everywhere. The nal onguration will then be guided by a
superposition of all the nal energy states, and the nal velo ity of
the onguration will not be onstant. It then appears (a ording to
Pauli) that the nal angular velo ity of the rotator will not be onstant,
ontrary to what is expe ted for a stationary state, and that there will
be no denite out ome for the s attering experiment (that is, no denite
nal energy state for the rotator).
In the usual dis ussion of dira tion gratings in opti s, it is of ourse
assumed that both the grating and the in ident beam are laterally
limited, so that the emerging beams separate as shown in Fig. 10.1
(where only the zeroth-order and rst-order beams are drawn). But in
Fermi's analogy, there an be no su h lateral limitation and no su h
238
Fig. 10.2. S attering of a laterally-unlimited wave by an innite dira tion

grating.
separation. It might be thought that, in the ase of no lateral limitation,

separation of an opti al beam would nevertheless take pla e if the in ident wave were longitudinally limited. However, as shown in Fig. 10.2,
symmetry di tates that there will be no beams beyond the zeroth order.a
Thus, if one a epts Fermi's analogy with the s attering of light by an
innite grating, even if one takes a longitudinally-limited in ident light
wave, there will still be no separation and the di ulty remains . The
only way to obtain a separation of the s attered beams is through a
lateral lo alisation along and a ording to Pauli's argument this is
impossible.
A ording to Fermi's analogy, then, it is ines apable that (for an initial
stationary state) the in ident wave (x, y, ) is unlimited on the -axis,
and one annot avoid the on lusion that after the s attering the rotator
a Geometri ally, the presen e of diverging higher-order beams (as in Fig. 10.1) would
dene a preferred entral point on the grating. Roughly speaking, if one onsiders
Fig. 10.1 in the limit of an innite grating (and of a laterally unlimited in ident
beam), the higher-order emerging beams are `pushed o to innity', resulting in
Fig. 10.2.
239
need not be in a denite energy eigenstate. There seems to be no way

out of Pauli's di ulty, and de Broglie's pilot-wave theory appears to
be untenable.
Today, one might answer Pauli's obje tion by onsidering the measuring apparatus used to dete t the outgoing s attered parti le (or, used to
measure the energy of the atom). By in luding the degrees of freedom of
the apparatus in the quantum des ription, one ould obtain a separation
of the total wave fun tion into non-overlapping bran hes, resulting in
a denite quantum out ome ( f. se tion 10.3). However, in 1927 the
measuring apparatus was not normally onsidered to be part of the
quantum system; and even today, in the pilot-wave theory of s attering
(sket hed in se tion 10.1), a denite out ome is generally guaranteed
by the separation of pa kets for the s attered parti le. As we shall now
show, the usual pilot-wave theory of s attering in fa t su es in Pauli's
example too.
To see how Pauli's obje tion would normally be met today, note rst
that in Pauli's example the real physi al situation onsists of an ele tron
moving in two spatial dimensions x, y and olliding with a rotator whose
angular oordinate ranges over the unit ir le (from 0 to 2 ). In de
Broglie's dynami s, the wave fun tion (x, y, , t) yields velo ities x,
y
for the ele tron and for the rotator, where these velo ities are given by
the quantum urrent divided by ||2 . As we saw in the ase of inelasti
s attering by an atom, the nal wave fun tion will take the form
(x, y, , t) =
0 ()eiE0 t/ inc (x, y, t) +
n ()eiEn t/ n (x, y, t) , (5)
where now the n are stationary states for the rotator. On e again,
for nite (lo alised) inc , at large times the outgoing wave pa kets n
will be entred on a radius rn = (kn /m)t from the s attering region, where again kn is the outgoing wave number xed by energy
onservation. As before, be ause the n expand with dierent speeds
they eventually be ome widely separated in spa e. The total (ele tronplus-rotator) onguration (x, y, ) will then o upy only one bran h
i ()eiEi t/ i (x, y, t) of the outgoing wave fun tion. The s attered
ele tron will then be guided by i only, and the speed of the ele tron will
be onstant. Further, the motion of the rotator will be determined by
the stationary state i , and will also be uniform. As long as the in ident
wave inc is lo alised in x and y , the nal wave fun tion separates in
240
onguration spa e and the s attering pro ess has a denite out ome.
Pauli's obje tion therefore has a straightforward answer.
On Fermi's analogy, however, Pauli's example seems to be equivalent
to the s attering of a laterally-unlimited light wave in three-dimensional
spa e by an innite grating, for whi h no separation of beams an take
pla e. Sin e, for the original system, we have seen that the nal state
does separate, it is lear that Fermi's analogy must be mistaken in some
way.
To see what is wrong with Fermi's analogy, one must examine Fermi's
original paper. There, Fermi introdu es the oordinates
= mx ,
= my ,
= J ,
(6)
and writes the time-independent S hrdinger equation for the ombined rotator-plus-ele tron system of total energy E in the form
2
2 2
2
+
+
+ 2 (E V ) = 0 ,
2
2
2
(7)
where
the potential energy V is a periodi fun tion of with period
2 J . Fermi then onsiders an opti al analogue of the wave equation
(7). As Fermi puts it (Fermi 1926, p. 400):
In order to see the solution of [(7), we onsider the opti al analogy for the wave
equation [(7). In the regions far from the
-axis, where V
vanishes, [(7) is the
wave equation in an opti ally homogeneous medium; in the neighbourhood of
-axis, the medium has an anomaly in the refra tive index, whi h depends
. Opti ally this is nothing more than a linear grating of period
2 J .
the
periodi ally on
Fermi then goes on to onsider a plane wave striking the grating, and
relates the outgoing beams of dierent spe tral orders to dierent types
of ollisions between the ele tron and the rotator.
Now, if we in lude the time dependen e e(i/)Et , we may write
1 2
2
E
=
2
c2 t2
(8)
p
with c E/2, and (7) indeed oin ides with the wave equation of
s alar opti s for the ase of a given frequen y. However and here is
where Fermi's analogy breaks down be ause the `speed of light' c
depends on the energy E (or frequen y E/h), the time evolution of the
system in regions far from the -axis is in general not equivalent to wave
propagation in àn opti ally homogeneous medium'. The analogy holds
only for one energy E at a time (whi h is all Fermi needed to onsider).
241
In a realisti ase where the (nite) in ident ele tron wave is a sum over
dierent momenta and as we have said, it was understood that in
any realisti ase the in ident wave would indeed be nite and therefore
equal to su h a sum the problem of s attering by the rotator annot
be made equivalent to the s attering of light by a grating in an otherwise
opti ally homogeneous medium: on the ontrary, the `speed of light' c
would have to vary as the square root of the frequen y.
Pauli's presentation does not mention that Fermi's opti al analogy is
valid only for a single frequen y. Sin e, as we have seen, the niteness
of realisti in ident pa kets was impli it in all these dis ussions, the
impression was probably given that Fermi's analogy holds generally. A
nite (in x and y ) ele tron wave in ident on the rotator would then be
expe ted to translate into a longitudinally nite but laterally innite
(along ) light wave in ident on an innite grating, with a resulting
la k of separation in the nal state. Sin e, however, the quantity c is
frequen y-dependent, in general the analogy with light is invalid and
the on lusion unwarranted.
What really happens, then, in the two-dimensional s attering of an
ele tron by a rotator, if one interprets the angular oordinate as a
third spatial axis la Fermi? The answer is found in a detailed dis ussion
of Pauli's obje tion given by de Broglie, in a book published in 1956,
whose hapter 14 bears the title `Mr Pauli's obje tion to the pilot-wave
theory' (de Broglie 1956). There, de Broglie gives what is in fa t the
rst proper analysis of Pauli's example in terms of pilot-wave theory
(with one spatial dimension suppressed for simpli ity). The result of de
Broglie's analysis an be easily seen by re onsidering the expression (5),
and allowing to range over the real line, with (x, y, , t) regarded as
a periodi fun tion of with period 2 . Be ause separates (as we have
seen) into pa kets that are non-overlapping with respe t to x and y , the
time evolution of will be as sket hed in Fig. 10.3 (where we suppress
y ) a gure that we have adapted from de Broglie's book (p. 176).
The gure shows a wave moving along the x-axis towards the rotator at
x = 0. The wave is longitudinally limited (nite along x) and laterally
unlimited (innite along ), and separates as shown into similar pa kets
moving at dierent speeds after the s attering. (Only two of the nal
pa kets are drawn.)
The dierent speeds of the pa kets after the ollision orrespond, of
ourse, to the dierent possible kineti energies of the ele tron after the
ollision. Note that the ru ial separation into pa kets moving at dierent speeds would be ruled out if one were to mistakenly a ept Fermi's
242
Fig. 10.3. True evolution of the ele tron-rotator wave fun tion in Pauli's
example. Adapted from de Broglie (1956, p. 176).
analogy with light s attering ìn an opti ally homogeneous medium'

for the outgoing pa kets would then all have to move with the same
speed (that of light).
A omplete reply to Pauli's original obje tion should then make two
points:
(1) Be ause of the frequen y-dependent `speed of light' c
p
E/2, Fermi's analogy with opti s is of very limited validity and annot
be applied to a real ase with a nite in ident wave. (2) For a nite
in ident ele tron wave, a separation of pa kets does in fa t take pla e
with respe t to the spatial oordinates of the ele tron.
Let us now examine how de Broglie replied to Pauli in O tober 1927,
in the general dis ussion. As will be ome lear, de Broglie understood
the general separation me hanism required to yield a denite out ome,
but he was misled by the (generally false) opti al analogy and phrased
his answer in terms of it.
De Broglie replies by rst pointing out the importan e of having a
laterally-limited in ident wave, to avoid overlap among the dira ted beams, and to avoid overlap between these and the in ident wave (p. 511):
243
The di ulty pointed out by Mr Pauli has an analogue in lassi al opti s.
One an speak of the beam dira ted by a grating in a given dire tion only
if the grating and the in ident wave are laterally limited, be ause otherwise
all the dira ted beams will overlap and be bathed in the in ident wave. In
Fermi's problem, one must also assume the wave
to be limited laterally in
onguration spa e.
De Broglie then notes that, if one an assume to be limited laterally,

then the system velo ity will be ome onstant on e the dira ted waves
have separated (from ea h other and from the in ident beam):
.... the velo ity of the representative point of the system will have a onstant
value, and will orrespond to a stationary state of the rotator, as soon as the
waves dira ted by the
-axis
will have separated from the in ident beam.
What de Broglie is des ribing here is pre isely the separation into nonoverlapping pa kets in onguration spa e, whi h the pilot wave must
undergo in order for the s attering experiment to have a denite out ome
just as we have dis ussed in se tion 10.1 above.
However, in Pauli's example, it appeared that the in ident ould
not be limited laterally. On this point, de Broglie laims that Fermi's
onguration spa e is arti ial, having been formed `by rolling out along
a line the y li variable '. But as de Broglie's treatment of 1956
shows, extending the range of from the unit ir le to the real line
does not really make any dieren e: it simply distributes opies of the
onguration spa e along the -axis (see Fig. 10.3).
Pauli's
presentation did not mention the frequen y-dependent speed
p
c E/2, and in his reply de Broglie did not mention it either. Indeed,
it seems that neither de Broglie in 1927 nor Bohm in 1952 noti ed this
misleading aspe t of Pauli's obje tion.
Note that niteness of the in ident wave with respe t to x and y
had not been questioned by anyone. As we have seen, the nite spatial
extension of realisti wave pa kets was impli itly assumed by all. Pauli
had raised the impossibility of niteness spe i ally with respe t to the
-axis, and de Broglie responded on this spe i point alone. While de
Broglie was misled by Fermi's analogy, even so his remarks ontain the
key point, later developed in detail by Bohm that niteness of the
initial pa ket will ensure a separation into non-overlapping nal pa kets
in onguration spa e. As de Broglie himself put it in his book of 1956,
in referen e to his 1927 reply to Pauli: `Thus I had indeed realised ....
that the answer to Mr Pauli's obje tion had to rest on the fa t that the
244
wave trains are always limited, an idea that has been taken up again by
Mr Bohm in his re ent papers' (de Broglie 1956, p. 176).
Finally, we point out that leaving aside the misleading nature of
Fermi's opti al analogy de Broglie's audien e may well have understood his des ription of wave pa kets separating in onguration spa e,
for Born had already des ribed something very similar for the Wilson
loud hamber, earlier in the general dis ussion (pp. 482 .). As we have
seen in se tion 6.2, Born dis ussed the formation of a tra k in a loud
hamber in terms of a bran hing of the total wave fun tion in a multidimensional onguration spa e (formed by all the parti les involved).
In parti ular, for the simple ase of an -parti le intera ting with two
atoms in one dimension, Born des ribed an initially lo alised pa ket
that separates into two non-overlapping bran hes in (three-dimensional)
onguration spa e as sket hed in Born's gure. By the time de
Broglie ame to reply to Pauli, then, the audien e was already familiar
with the idea of a wave fun tion evolving in onguration spa e and
developing non-overlapping bran hes. Thus, it seems more likely than
not that this aspe t of de Broglie's reply to Pauli would have been
understood.a

The general theory of quantum measurement in pilot-wave theory was
rst developed by Bohm in his se ond paper on the theory (Bohm
1952b). Bohm understood that the degrees of freedom asso iated with
the measurement apparatus were simply extra oordinates that should
be in luded in the total system (as in, for example, Born's 1927 dis ussion of the loud hamber). Thus, denoting by x the oordinates of
the `system' and y the oordinates of the àpparatus', the dynami s
generates a traje tory (x(t), y(t)) for the total onguration, guided
y)
is
by a wave fun tion (x, y, t) (where again the velo ity eld (x,
2
given by the quantum urrent of divided by || ). Bohm showed
how, in the ir umstan es orresponding to a quantum measurement,
a It may also be worth noting that, in his report, when presenting the pilot-wave
theory of many-body systems, de Broglie expli itly asserts (p. 386) that his
probability formula in onguration spa e `fully a ords .... with the results
obtained by Mr Born for the ollision of an ele tron and an atom, and by Mr
Fermi for the ollision of an ele tron and a rotator'. Be ause the printed version of
de Broglie's le ture was already omplete by the time the onferen e took pla e,
this remark ould not have been added after de Broglie's lash with Pauli. (De
Broglie wrote to Lorentz on 11 O tober 1927 AHQP-LTZ-11, in Fren h that
he had re eived the proofs of his report from Gauthier-Villars and orre ted them.)
245
separates into non-overlapping bran hes and the system oordinate x(t)
is eventually guided by an ee tively `redu ed' pa ket for the system.
As a simple example, let the system have an initial wave fun tion
X
0 (x) =
(9)
cn n (x) ,
n
where the n are eigenfun tions of some Hermitian operator Q with

eigenvalues qn . Suppose an experiment is performed that in quantum
theory would be alled à measurement of the observable Q'. This might
be done by oupling the system to a `pointer' with oordinate y and initial wave fun tion g0 (y), where g0 is narrowly peaked around y = 0. An
appropriate oupling might be des ribed by an intera tion Hamiltonian
H = aQPy , where a is a oupling onstant and Py is the momentum
operator onjugate to y . If we negle t the rest of the Hamiltonian, for
an initial produ t wave fun tion 0 (x, y) = 0 (x)g0 (y) the S hrdinger
equation

= aQ i
i
(10)
t
y
has the solution
(x, y, t) =
X
n
cn n (x)g0 (y aqn t) .
(11)
Be ause g0 is lo alised, evolves into a superposition of terms that

separate with respe t to y (in the sense of having negligible overlap with
respe t to y ). As we saw in the ase of s attering, the nal onguration
(x, y) an be in only one `bran h' of the superposition, say i (x)g0 (y
aqi t), whi h will guide (x(t), y(t)) thereafter. And be ause the a tive
bran h is a produ t in x and y , the velo ity of x will be determined
by i (x) alone. In ee t, at the end of the quantum measurement, the
system is guided by a `redu ed' wave fun tion i (x). Thus, the pointer
plays the same role in the pilot-wave theory of measurement as the
s attered parti le does in the pilot-wave theory of s attering.
It is also readily shown that, if the initial ensemble of x, y has a
2
distribution |0 (x, y)| , then the probability of x, y ending in the bran h
2
i (x)g0 (y aqi t) is given by |ci | , in agreement with the standard Born
rule.
Finally, again as we saw in the ase of s attering, the ee tive ` ollapse'
to the state i (x)g0 (y aqi t) is for all pra ti al purposes irreversible.
As noted by Bohm (1952b, p. 182), the apparatus oordinate y will
subsequently intera t with many other degrees of freedom, making it
246
very unlikely that distin t states i (x)g0 (y aqi t), j (x)g0 (y aqj t)
(i 6= j ) will interfere later on, be ause the asso iated bran hes of the
total wave fun tion would have to overlap with respe t to every degree
of freedom involved.
10.4 Re oil of a single photon: Kramers and de Broglie

We have sket hed the pilot-wave theory of quantum measurement to
emphasise that, when one leaves the limited domain of parti le s attering
(by atoms, or by xed obsta les su h as mirrors or dira ting s reens),
it an be ome essential to des ribe the apparatus itself in terms of
pilot-wave dynami s. In some situations, the oordinates of the apparatus must be in luded in the total wave fun tion (for the `supersystem'
onsisting of system plus apparatus), to ensure that the total wave
fun tion separates into non-overlapping bran hes in onguration spa e.
As we shall see in hapter 11, it seems that this point was not properly
appre iated by de Broglie in 1927. Indeed, most theoreti ians at the
time simply applied quantum theory (in whatever form they preferred)
to mi ros opi systems only. Ma ros opi apparatus was usually treated
as a given lassi al ba kground. However, Born's treatment of the loud
hamber in the general dis ussion shows that the key insight was already known: the apparatus is made of atoms, and should ultimately be
in luded in the wave fun tion, whi h will develop a bran hing stru ture
as the measurement pro eeds. All that was needed, within pilot-wave
theory, was to arry through the details properly for a general quantum
measurement something that de Broglie did not see in 1927 and that
Bohm did see in 1952.
Now, the need to in lude ma ros opi equipment in the wave fun tion
is relevant to a problem raised by Kramers in the general dis ussion,
on erning the re oil of a mirror due to the ree tion of a single photon.
The dis ussion had returned to the question of how de Broglie's theory
a ounts for radiation pressure on a mirror (a subje t that had been
dis ussed at the end of de Broglie's le ture). As Kramers put it (p. 497):
But how is radiation pressure exerted in the ase where it is so weak that
there is only one photon in the interferen e zone? .... And if there is only one
photon, how an one a ount for the sudden hange of momentum suered by
the ree ting obje t?
Neither de Broglie nor Brillouin were able to give an answer. De Broglie
10.4 Re oil of a single photon
247
laimed that pilot-wave theory in its urrent form was able to give only
the mean pressure exerted by an ensemble (or ` loud') of photons.
What was missing from de Broglie's understanding was that, in su h
a ase, the position of the mirror would have to be treated by pilot-wave
dynami s and in luded in the wave fun tion. S hemati ally, let xp be
the position of the photon (on an axis normal to the surfa e of the
mirror) and let xm be the position of the ree ting surfa e. Let us treat
the mirror as a very massive but free body, with initial wave fun tion
(xm , 0) lo alised around xm = 0 (at t = 0). The in ident photon
initially has a lo alised wave fun tion inc (xp , 0) dire ted towards the
mirror, with mean momentum k . Roughly, if the photon pa ket strikes
the mirror at time t0 , then the initial total wave fun tion (xp , xm , 0) =
inc (xp , 0)(xm , 0) will evolve into a wave fun tion of the s hemati form
(xp , xm , t) ref (xp , t)(xm (p/M )(t t0 ), 0)e(i/)(p)xm , (12)
where ref is a ree ted pa ket dire ted away from the mirror, p = 2k
is the momentum transferred to the mirror, M is the mass of the mirror,
and (xm (p/M )(t t0 ), 0) is a pa ket (whose spreading we ignore)
moving to the right with speed p/M . The a tual oordinates xp , xm
will be guided by in a ordan e with de Broglie's equation, and the
position xm (t) of the mirror will follow the moving pa ket. The re oil of
the mirror an therefore be a ounted for.
Thus, while de Broglie had the omplete pilot-wave dynami s of a
many-body system, he seems not to have understood that it is sometimes
ne essary to in lude the oordinates of ma ros opi equipment in the
pilot-wave des ription. Otherwise, he might have been able to answer
Kramers.
On the other hand, in ordinary quantum theory too, a proper explanation for the re oil of the mirror would also have to treat the mirror
as part of the quantum system. If the mirror is regarded as a lassi al
obje t, then quantum theory would stri tly speaking be as powerless
as pilot-wave theory. Perhaps this is why Brillouin made the following
remark (p. 497):
No theory urrently gives the answer to Mr Kramers' question.
11
As we dis ussed in se tion 2.4, in his Solvay le ture of 1927 de Broglie presented the pilot-wave dynami s of a nonrelativisti many-body
system, and outlined some simple appli ations of his `new dynami s of
quanta' (to interferen e, dira tion, and atomi transitions). Further, as
we saw in se tion 10.2, ontrary to a widespread misunderstanding, in
the general dis ussion de Broglie's reply to Pauli's obje tion ontained
the essential points needed to treat inelasti s attering (even if Fermi's
misleading opti al analogy onfused matters): in parti ular, de Broglie
orre tly indi ated how denite quantum out omes in s attering pro esses arise from a separation of wave pa kets in onguration spa e. We
also saw in se tion 10.4 that de Broglie was unable to reply to a query
from Kramers on erning the re oil of a single photon on a mirror: to do
so, he would have had to introdu e a joint wave fun tion for the photon
and the mirror.
De Broglie's theory was revived by Bohm 25 years later (Bohm 1952a,b)
(though with the dynami s written in terms of a law for a eleration
instead of a law for velo ity). Bohm's truly new and very important
ontribution was a pilot-wave a ount of the general quantum theory
of measurement, with ma ros opi equipment (pointers, et .) treated as
part of the quantum system. In ee t, in 1952 Bohm provided a detailed
derivation of quantum phenomenology from de Broglie's dynami s of
1927 (albeit with the dynami al equations written dierently).
Despite this su ess, until about the late 1990s most physi ists still
believed that hidden-variables theories su h as de Broglie's ould not
possibly reprodu e the predi tions of quantum theory (even for simple
ases su h as the two-slit experiment, f. se tion 6.1). Or, they believed
that su h theories had been disproved by experiments testing EPR-type
orrelations and demonstrating violations of Bell's inequality. As just one
248
249
striking example of the latter belief, in the early 1990s James T. Cushing, then a professor of physi s and of philosophy at the University of
Notre Dame, submitted a resear h proposal to the US National S ien e
Foundation `for theoreti al work to be done, within the framework of
Bohm's version of quantum theory, on some foundational questions in
quantum me hani s' (Cushing 1996, p. 6), and re eived the following
evaluation:
The subje t under onsideration, the rival Copenhagen and ausal (Bohm)
interpretations of the quantum theory, has been dis ussed for many years
and in the opinion of several members of the Physi s Division of the NSF,
the situation has been settled. The ausal interpretation is in onsistent with
experiments whi h test Bell's inequalities. Consequently .... funding .... a
resear h program in this area would be unwise. (Cushing 1996, p. 6)
This is ironi be ause, as even a super ial reading of Bell's original

papers shows, the nonlo al theory of de Broglie and Bohm was a primary
motivation for Bell's work on his famous inequalities (Bell 1964, 1966).
Bell knew that pilot-wave theory was empiri ally equivalent to quantum
theory, and wanted to nd out if the nonlo ality was a pe uliarity of this
parti ular model, or if it was a general feature of all hidden-variables
theories. Bell's on lusion was that lo al theories have to satisfy his
inequality, whi h is in onsistent with EPR-type orrelations, and that
therefore any viable theory must be nonlo al like pilot-wave theory.
There was never any question, in experimental tests of Bell's inequality,
of testing the nonlo al theory of de Broglie and Bohm; rather, it was the
lass of lo al theories that was being tested.a
Despite this and other misunderstandings, in re ent years the pilotwave theory of de Broglie and Bohm, with parti le traje tories guided
by a physi ally-real wave fun tion, has gained wide a eptan e as an
alternative (though little used) formulation of quantum theory. While it
is still o asionally asserted that any theory with traje tories (or other
hidden variables) must disagree with experiment, su h erroneous laims
have be ome mu h less frequent.b This hange in attitude seems to have
a See hapter 9 of Cushing (1994) for an extensive dis ussion of the generally hostile
rea tions to and misrepresentations of Bohm's 1952 papers.
b For example, in his book The Elegant Universe, Greene (2000) asserts that not
only are quantum parti le traje tories unmeasurable (be ause of the un ertainty
prin iple), their very existen e is ruled out by experiments testing the Bell
inequalities: `.... theoreti al progress spearheaded by the late Irish physi ist John
Bell and the experimental results of Alain Aspe t and his ollaborators have
shown onvin ingly that .... [ele trons and everything else for that matter
annot be des ribed as simultaneously being at su h-and-su h lo ation and having
su h-and-su h speed' (p. 114). This is orre ted a few years later in The Fabri of
250
been largely a result of the publi ation in 1987 of Bell's inuential book
on the foundations of quantum theory (Bell 1987), several hapters of
whi h onsisted of pedagogi al explanations of pilot-wave theory as an
obje tive and deterministi a ount of quantum phenomena. Leaving
aside the question of whether or not pilot-wave theory (or de BroglieBohm theory) is loser to the truth about the quantum world than other
formulations, it is a remarkable fa t that it took about three-quarters
of a entury for the theory to be ome widely a epted as an internally
onsistent alternative.
Even today, however, there are widespread mis on eptions not only
about the physi s of pilot-wave theory, but also about its history. It is
generally re ognised that de Broglie worked along pilot-wave lines in the
1920s, that Bohm developed and extended the theory in 1952, and that
Bell publi ised the theory in his book in 1987. But the full extent of de
Broglie's ontributions in the 1920s is usually not re ognised.
A areful examination of the pro eedings of the fth Solvay onferen e
hanges our per eption of pilot-wave theory, both as a physi al theory
and as a part of the history of quantum physi s.
11.1 Histori al mis on eptions

Many of the widespread mis on eptions about the history of pilot-wave
theory are onveniently summarised in the following extra t from the
book by Bohm and Hiley (1993, pp. 389):
The idea of a `pilot wave' that guides the movement of the ele tron was rst
suggested by de Broglie in 1927, but only in onne tion with the one-body
system. De Broglie presented this idea at the 1927 Solvay Congress where it
was strongly riti ised by Pauli. His most important riti ism was that, in
a two-body s attering pro ess, the model ould not be applied oherently. In
onsequen e de Broglie abandoned his suggestion. The idea of a pilot wave was
proposed again in 1952 by Bohm in whi h an interpretation for the many-body
system was given. This latter made it possible to answer Pauli's riti ism ....
a
.
As we have by now repeatedly emphasised, in his 1927 Solvay report

(pp. 384 f.) de Broglie did in fa t present pilot-wave theory in onguration spa e for a many-body system, not just the one-body theory
the Cosmos, where, ommenting on `Bohm's approa h' Greene (2005) writes that
it `does not fall afoul of Bell's results be ause .... possessing denite properties
forbidden by quantum un ertainty is not ruled out; only lo ality is ruled out .... '
(p. 206, original itali s).
a Similar histori al mis on eptions appear in Cushing (1994, pp. 11821, 149).
251
in 3-spa e; and further, as we saw in se tion 10.2, de Broglie's reply

to Pauli's riti ism ontained the essential ideas needed for a proper
rebuttal.a As we shall see below, the laim that de Broglie abandoned
his theory be ause of Pauli's riti ism is also not true.
Contrary to widespread belief, then, the many-body theory with a
guiding wave in onguration spa e is originally due to de Broglie and
not Bohm; and in 1927, de Broglie did understand the essentials of the
pilot-wave theory of s attering. Thus the main ontent of Bohm's rst
paper of 1952 (Bohm 1952a) whi h presents the dynami s (though
in terms of a eleration), with appli ations to s attering was already
known to de Broglie in 1927.
Note that the theory was regarded as only provisional, by both de
Broglie in 1927 and by Bohm in 1952. In parti ular, as we saw in
se tions 2.3.2 and 2.4, de Broglie regarded the introdu tion of a pilot
wave in onguration spa e as a provisional measure. Bohm, on the other
hand, suggested that the basi prin iples of the theory would break down
at nu lear distan es of order 1013 m (Bohm 1952a, pp. 1789).
As we have said, in ontrast with de Broglie's presentation of 1927,
Bohm's dynami s of 1952 was based on a eleration, not velo ity. For de
Broglie, the basi law of motion for parti les with masses mi and wave
fun tion = || e(i/~)S was the guidan e equation
mi
dxi
= i S ,
dt
(1)
whereas for Bohm, the basi law of motion was the Newtonian equation
d2 xi
dpi
= mi 2 = i (V + Q) ,
dt
dt
(2)
where
Q
X 2 2 ||
i
2m
||
i
i
(3)
is the `quantum potential'.

Taking the time derivative of (1) and using the S hrdinger equation
yields pre isely (2). For Bohm, however, (1) was not a law of motion
but rather a onstraint pi = i S to be imposed on the initial momenta
(Bohm 1952a, p. 170). This initial onstraint happens to be preserved
in time by (2), whi h Bohm regarded as the true law of motion. Indeed,
a To our knowledge, Bonk (1994) is the only other author to have noti ed that de
Broglie's reply to Pauli was indeed along the right lines.
252
Bohm suggested that the initial onstraint pi = i S ould be relaxed,

leading to orre tions to quantum theory:
.... this restri tion is not inherent in the on eptual stru ture .... it is quite
onsistent in our interpretation to ontemplate modi ations in the theory,
whi h permit an arbitrary relation between
pp. 17071)
and
S(x).
(Bohm 1952a,
For de Broglie, in ontrast, there was never any question of relaxing (1):
he regarded (1) as the basi law of motion for a new form of parti le
dynami s, and indeed for him (1) embodied as we saw in hapter 2
the uni ation of the prin iples of Maupertuis and Fermat, a uni ation
that he regarded as the guiding prin iple of his new dynami s. (De
Broglie did mention in passing, however, the alternative formulation in
terms of a eleration, both in his `Stru ture' paper ( f. se tion 2.3.1)
and in his Solvay report (p. 383).)
Some authors seem to believe that the `re asting' of Bohm's se ondorder dynami s into rst-order form was due to Bell (1987). But in fa t,
Bell's (pedagogi al) presentation of the theory based on the guidan e
equation for velo ity, and ignoring the notion of quantum potential
was identi al to de Broglie's original presentation. De Broglie's rst-order
dynami s of 1927 is sometimes referred to as `Bohmian me hani s'. As
already noted in hapter 2, this is a misnomer: rstly be ause of de
Broglie's priority, and se ondly be ause Bohm's me hani s of 1952 was
a tually se ond-order in time.
Another ommon histori al mis on eption on erns the re eption of de
Broglie's theory at the Solvay onferen e. It is usually said that de Broglie's ideas attra ted hardly any attention. It is di ult to understand
how su h an impression originated, for even a ursory perusal of the
pro eedings reveals that de Broglie's theory was extensively dis ussed,
both after de Broglie's le ture and during the general dis ussion. Nevertheless, in his lassi a ount of the histori al development of quantum
theory, Jammer asserts that when de Broglie presented his theory:
It was immediately lear that nobody a epted his ideas .... . In fa t, with the
ex eption of some remarks by Pauli .... de Broglie's ausal interpretation was
not even further dis ussed at the meeting. Only Einstein on e referred to it
en passant. (Jammer 1966, p. 357)a
a Jammer adds footnoted referen es to pp. 280 and 256 respe tively of the original
pro eedings, where Pauli's obje tion (involving Fermi's treatment of the rotator)
appears, and where, in his main ontribution to the general dis ussion, Einstein
omments that in his opinion de Broglie ìs right to sear h in this dire tion' (p. 487).
253
It appears to have es aped Jammer's attention that the general dis ussion ontains extensive and varied omments on many aspe ts of
de Broglie's theory (in luding a query by Einstein about the speed of
photons), as does the dis ussion after de Broglie's le ture, and that
support for de Broglie's ideas was expressed by Brillouin and by Einstein.
In a later histori al study, again, Jammer (1974, pp. 11011) writes
in referen e to the fth Solvay onferen e that de Broglie's theory `was
hardly dis ussed at all', and that `the only serious rea tion ame from
Pauli'. In the same study, Jammer quotes extensively from de Broglie's
report and from the general dis ussion, apparently without noti ing the
extensive dis ussions of de Broglie's theory that appear both after de
Broglie's report and in the general dis ussion.
But Jammer is by no means the only historian to have given short
shrift to de Broglie's major presen e at the 1927 Solvay onferen e.
In his book The Solvay Conferen es on Physi s, Mehra (1975, p. xvi)
quotes de Broglie himself as saying, with referen e to his presentation of
pilot-wave theory in 1927, that ìt re eived hardly any attention'. But
these words were written by de Broglie some 46 years later (de Broglie
1974), and de Broglie's re olle tion (or misre olle tion) after nearly half
a entury is belied by the ontent of the published pro eedings.a
In volume 6 of their monumental The Histori al Development of Quantum Theory, Mehra and Re henberg (2000, pp. 24650) devote several
pages to the published pro eedings of the 1927 Solvay onferen e, fo ussing on the general dis ussion mainly on the omments by Einstein,
Dira and Heisenberg as well as on the unpublished omments by
Bohr. The rest of the general dis ussion is summarised in a single senten e (p. 250):
After the Einstein-Pauli-Dira -Heisenberg ex hange, the general dis ussion
turned to more te hni al problems onne ted with the des ription of photons
and ele trons in quantum me hani s, as well as with the details of de Broglie's
re ent ideas.
It is added that `though these points possess some intrinsi interest, they
do not throw mu h light on the interpretation debate'.
As we shall see in se tion 12.1, it would appear that, for Mehra and
Re henberg, as indeed for most ommentators, the ìnterpretation debate' entred mainly around private or semi-private dis ussions between
Bohr and Einstein, and that the many pages of published dis ussions
a In fa t, several ommentators have drawn erroneous on lusions about de Broglie,
by relying on mistaken `re olle tions' written by de Broglie himself de ades later.
254
were of omparatively little interest. Thus the remarkable downplaying

of the dis ussion of de Broglie's theory, as well as of other ideas, is
oupled with a strong tenden y on the part of many authors to
portray (in orre tly) the 1927 onferen e as fo ussed primarily on the
onfrontation between Bohr and Einstein.
We have already noted the widespread histori al mis on eptions on erning the de Broglie-Pauli en ounter in the general dis ussion. A related mis on eption on erns de Broglie's thinking in the immediate
aftermath of the Solvay onferen e. It is often asserted that, soon after
the onferen e, de Broglie abandoned his theory primarily be ause of
Pauli's riti ism. This is not orre t. In his book An Introdu tion to
the Study of Wave Me hani s (de Broglie 1930), whi h was published
just three years after the Solvay meeting, de Broglie gives three main
reasons for why he onsiders his pilot-wave theory to be unsatisfa tory.
First, de Broglie onsiders (p. 120) a parti le in ident on an imperfe tly ree ting mirror, and notes that if the parti le is found in the
transmitted beam then the ree ted part of the wave must disappear
(this being à ne essary onsequen e of the interferen e prin iple'). De
Broglie on ludes that `the wave is not a physi al phenomenon in the
old sense of the word. It is of the nature of a symboli representation of
a probability .... '. Here, de Broglie did not understand how pilot-wave
theory a ounts for the ee tive (and pra ti ally irreversible) ollapse of
the wave pa ket, by means of a separation into non-overlapping bran hes
involving many degrees of freedom ( f. se tions s at-in-pwt and 10.3).
Se ond, de Broglie notes (pp. 121, 133) that a parti le in free spa e
guided by a superposition of plane waves would have a rapidly-varying
velo ity and energy, and he annot see how this ould be onsistent with
the out omes of quantum energy measurements, whi h would oin ide
stri tly with the energy eigenvalues present in the superposition. To
solve this se ond problem, de Broglie would have had to apply pilot-wave
dynami s to the pro ess of quantum measurement itself in luding the
apparatus in the wave fun tion if ne essary as done mu h later by
Bohm (see se tion10.3 ). This question of energy measurement bears
some similarity to that raised by Pauli, and perhaps Pauli's query set
de Broglie thinking about this problem. But even so, de Broglie in
ee t gave the essen e of a orre t reply to Pauli's query, and the
problem of energy measurement was posed by de Broglie himself. Third,
in applying pilot-wave theory to photons, de Broglie nds (p. 132) that
in some ir umstan es (spe i ally, in the interferen e zone lose to an
imperfe tly ree ting mirror) the photon traje tories have superluminal
255
speeds, whi h he onsiders una eptable. De Broglie's book does not

mention Pauli's riti ism.
It is also often laimed that, when de Broglie abandoned his pilot-wave
theory (soon after the Solvay onferen e), he qui kly adopted the views
of Bohr and Heisenberg. Thus, for example, Cushing (1994, p. 121)
writes: `By early 1928 he [de Broglie had de ided to adopt the views of
Bohr and Heisenberg'.a But de Broglie's book of 1930, in whi h the above
di ulties with pilot-wave theory are des ribed, ontains a `General
introdu tion' that is `the reprodu tion of a ommuni ation made by the
author at the meeting of the British Asso iation for the Advan ement
of S ien e held in Glasgow in September, 1928' (de Broglie 1930, p. 1).
This introdu tion therefore gives an overview of de Broglie's thinking
in late 1928, almost a year after the fth Solvay onferen e. While de
Broglie makes it lear (p. 7) that in his view it is `not possible to regard
the theory of the pilot-wave as satisfa tory', and states that the `point
of view developed by Heisenberg and Bohr .... appears to ontain a large
body of truth', his on luding paragraph shows that he was still not
satised:
To sum up, the physi al interpretation of the new me hani s remains an
extremely di ult question .... the dualism of waves and parti les must be
admitted .... . Unfortunately the profound nature of the two members in this
duality and the pre ise relation existing between them still remain a mystery.
(de Broglie 1930, p. 10)
At that time, de Broglie seems to have a epted the formalism of quantum

theory, and the statisti al interpretation of the wave fun tion, but he
still thought that an adequate physi al understanding of wave-parti le
duality had yet to be rea hed. Yet another year later, in De ember 1929,
doubts about the orre t interpretation of quantum theory ould still be
dis erned in de Broglie's Nobel Le ture (de Broglie 1999):
Is it even still possible to assume that at ea h moment the orpus le o upies a
well-dened position in the wave and that the wave in its propagation arries
the orpus le along in the same way as a wave would arry along a ork?
These are di ult questions and to dis uss them would take us too far and
even to the onnes of philosophy. All that I shall say about them here is
that nowadays the tenden y in general is to assume that it is not onstantly
possible to assign to the orpus le a well-dened position in the wave.
a As eviden e for this, Cushing ites later re olle tions by de Broglie in his book
Physi s and Mi rophysi s (de Broglie 1955), whi h was originally published in
Fren h in 1947. Again, de Broglie's re olle tions de ades later do not seem a reliable
guide to what a tually happened ir a 1927.
256
This brings us to another histori al mis on eption on erning de Broglie's work. Nowadays, de Broglie-Bohm theory is often presented as
a ` ompletion' of quantum theory.a Criti s sometimes view this as an
arbitrary addition to or amendment of the quantum formalism, the
traje tories being viewed as an additional `baggage' being appended
to an already given formalism. Regardless of the truth or otherwise of
pilot-wave theory as a physi al theory, su h a view ertainly does not do
justi e to the histori al fa ts. For the elements of pilot-wave theory
waves guiding parti les via de Broglie's velo ity formula were already
in pla e in de Broglie's thesis of 1924, before either matrix or wave
me hani s existed. And it was by following the lead of de Broglie's thesis
that S hrdinger developed the wave equation for de Broglie's matter
waves. While S hrdinger dropped the traje tories and onsidered only
the waves, nevertheless, histori ally speaking the wave fun tion and
the S hrdinger equation both grew out of de Broglie's phase waves.b
The pilot-wave theory of 1927 was the ulmination of de Broglie's
independent work from 1923, with a major input from S hrdinger in
1926. There is no sense in whi h de Broglie's traje tories were ever àdded
to' some pre-existing theory. And when Bohm revived the theory in
1952, while it may have seemed to Bohm's ontemporaries (and indeed
to Bohm himself) that he was adding something to quantum theory,
from a histori al point of view Bohm was simply reinstating what had
been there from the beginning.
The failure to a knowledge the priority of de Broglie's thinking from
1923 is visible even in the dis ussions of 1927. In the dis ussion following
de Broglie's le ture, Pauli (p. 399) presents what he laims is the entral
idea of de Broglie's theory:
I should like to make a small remark on what seems to me to be the mathemati al basis of Mr de Broglie's viewpoint on erning parti les in motion on
denite traje tories. His on eption is based on the prin iple of onservation
of harge .... if in a eld theory there exists a onservation prin iple .... it
is always formally possible to introdu e a velo ity ve tor ... and to imagine
furthermore orpus les that move following the urrent lines of this ve tor.
Pauli's assertion that de Broglie's theory is based on the onservation

of harge makes sense only for one parti le: for a many-body system,
de Broglie's velo ity eld is asso iated with onservation of probability
in onguration spa e, whereas onservation of harge is always tied
a An often- ited motivation for introdu ing the traje tories is, of ourse, to solve
the measurement problem.
b See also the dis ussion in se tion 4.5.
257
to 3-spa e. Still, the point remains that in any theory with a lo ally
onserved probability urrent, it is indeed possible to introdu e parti le traje tories following the ow lines of that urrent. This way of
presenting de Broglie's theory then makes the traje tories look like an
addendum to a pre-existing stru ture: given the S hrdinger equation
with its lo ally onserved urrent, one an add traje tories if one wishes.
But to present the theory in this way is a major distortion of the
histori al fa ts and priorities. The essen e of de Broglie's dynami s ame
before S hrdinger's work, not after. Further, de Broglie obtained his
velo ity law not from the S hrdinger urrent (whi h was unknown in
1923 or 1924) but from his postulated relation between the prin iples
of Maupertuis and Fermat. And nally, while de Broglie's traje tory
equation did not in fa t owe anything to the S hrdinger equation, again,
the latter equation arose out of onsiderations (of the opti al-me hani al
analogy) that had been initiated by de Broglie. It seems rather lear
that the histori al priority of de Broglie's work was being downplayed
by Pauli's remarks, as it has been more or less ever sin e.
A related histori al mis on eption on erns the status of the S hrdinger equation in pilot-wave theory. It is sometimes argued that this
equation has no natural pla e in the theory, and that therefore the theory
is arti ial. For example, ommenting on what he alls `Bohm's theory',
Polkinghorne (2002, pp. 55, 89) writes:
There is an air of ontrivan e about it that makes it unappealing. For example,
the hidden wave has to satisfy a wave equation. Where does this equation
ome from? The frank answer is out of the air or, more a urately, out of
the mind of S hrdinger. To get the right results, Bohm's wave equation
must be the S hrdinger equation, but this does not follow from any internal
logi of the theory and it is simply an
ad ho
strategy designed to produ e
empiri ally a eptable answers. .... It is on these grounds that most physi ists
nd the greatest di ulty with Bohmian ideas .... the
ad ho
but ne essary
appropriation of the S hrdinger equation as the equation for the Bohmian

wave has an unattra tively opportunist air to it.
Polkinghorne's omments are a fair riti ism of Bohm's 1952 reformulation of de Broglie's theory. For as we have seen, Bohm based his presentation on the Newtonian equation of motion (2) for a eleration, with
a `quantum potential' Q determined by the wave fun tion through
(3): a ording to Bohm, generates a `quantum for e' i Q, whi h
a ounts for quantum ee ts. From this Newtonian standpoint, the wave
equation for does indeed have nothing to do with the internal logi
of the theory: it is then fair to say that, in Bohm's formulation, the
258
S hrdinger equation is àppropriated' for a purpose quite foreign to

the origins of that equation. However, Polkinghorne's ritique does not
apply to pilot-wave theory in its original de Broglian formulation, as
a new form of dynami s in whi h parti le velo ities are determined by
guiding waves (rather than parti le a elerations being determined by
Newtonian for es). For as a matter of histori al fa t, the S hrdinger
equation did follow from the internal logi of de Broglie's theory. After
all, S hrdinger set out in the rst pla e to nd the general wave equation
for de Broglie's waves; and his derivation of that equation owed mu h
to the opti al-me hani al analogy, whi h was a key omponent of de
Broglie's approa h to dynami s.a It annot be said that de Broglie
àppropriated' the S hrdinger equation for a purpose foreign to its
origins, when the original purpose of the S hrdinger equation was in
fa t to des ribe de Broglie's waves.
11.2 Why was de Broglie's theory reje ted?

One might ask why de Broglie's theory did not gain widespread support
soon after 1927. This question has been onsidered by Bonk (1994),
who applies Bayesian reasoning to some of the dis ussions at the fth
Solvay onferen e, in an attempt to understand the rapid a eptan e
of the `Copenhagen' interpretation. It has also been suggested that
were it not for ertain histori al a idents, de Broglie's theory might
have triumphed in 1927 and emerged as the dominant interpretation of
quantum theory (Cushing 1994, hapter 10). It is di ult to evaluate
how realisti Cushing's àlternative histori al s enario' might have been.
Here, we shall simply highlight two points that are usually overlooked,
and whi h are relevant to any evaluation of why de Broglie's theory did
not arry the day.
Our rst point is that, be ause de Broglie did in fa t give a reply to
Pauli's riti ism that ontained the essen e of a orre t rebuttal and
be ause in ontrast de Broglie ompletely failed to reply to the di ulty
raised by Kramers the question of whether or not de Broglie made
a onvin ing ase at the fth Solvay onferen e (and if not, why not)
should be re onsidered.
Our se ond point is that there was a good te hni al reason for why
a Nowadays, it is ommon in textbooks to motivate the free-parti le S hrdinger
equation as the simplest equation satised by a plane de Broglie wave ei(kxt)
with the nonrelativisti dispersion relation = (k)2 /2m. This `derivation' is
just as natural in pilot-wave theory as it is in standard quantum theory.
11.2 Einstein's alternative pilot-wave theory
259
`standard' quantum theory had an advantage over pilot-wave theory in

1927. By the use of a simple ` ollapse postulate' for mi rosystems, it was
generally possible to a ount for quantum measurement out omes without having to treat the measurement pro ess (in luding the apparatus)
quantum-me hani ally. In ontrast, it was easy to nd examples where
in pilot-wave theory it was essential to use the theory itself to analyse the
measurement pro ess: as we saw in the last se tion, in his book of 1930
de Broglie ould not see how, for a parti le guided by a superposition
of energy eigenfun tions, an energy measurement would give one of the
results expe ted from quantum theory. Agreement with quantum theory
requires an analysis of the measurement pro ess in terms of pilot-wave
dynami s, with the apparatus in luded as part of the system, as shown
by Bohm (1952b). In ontrast, in ordinary quantum theory it usually
su es in pra ti e simply to apply a ollapse rule to the mi ros opi
system alone. Su h a ollapse rule is of ourse merely pragmati , and
dees pre ise formulation (there being no sharp boundary between `mi ros opi ' and `ma ros opi ', f. se tion 5.2); yet, in ordinary laboratory
situations, it yields predi tions that may be ompared with experiment.
These two points should be taken into a ount in any full evaluation
of why de Broglie's theory was reje ted in 1927 and shortly thereafter.
Further relevant material, that seems to have never been onsidered
before, onsists of omments by Heisenberg on the possibility of a deterministi pilot-wave interpretation. These omments do not appear in
the pro eedings of the fth Solvay onferen e, nor are they dire ted at
de Broglie's theory. Rather, they appear in a letter Heisenberg wrote
to Einstein a few months earlier, on 10 June 1927, and they on ern
Einstein's version of pilot-wave theory. Heisenberg's remarks ould just
as well have been dire ted at de Broglie's theory, however. Both Einstein's theory and Heisenberg's omments thereon are dis ussed in the
next se tion.
11.3 Einstein's alternative pilot-wave theory (May 1927)

As we saw in se tion 2.3.1, de Broglie rst arrived at pilot-wave theory
in a paper published in Journal de Physique in May 1927 (de Broglie
1927b). In the same month, Einstein proposed what in retrospe t appears to be an alternative version of pilot-wave theory, with parti le traje tories determined by the many-body wave fun tion but in a manner
dierent from that of de Broglie's theory. This new theory was des ribed
in a paper entitled `Does S hrdinger's wave me hani s determine the
260
motion of a system ompletely or only in the sense of statisti s?', whi h

was presented on 5 May 1927 at a meeting of the Prussian A ademy of
S ien es. On the same day Einstein wrote to Ehrenfest that `.... in a ompletely unambiguous way, one an asso iate denite movements with the
solutions [of the S hrdinger equation' (Howard 1990, p. 89). However,
on 21 May, before the paper appeared in print, Einstein withdrew it from
publi ation (Kirsten and Treder 1979, p. 135; Pais 1982, p. 444). The
paper remained unpublished, but its ontents are nevertheless known
from a manus ript version in the Einstein ar hive see Howard (1990,
pp. 8990) and Belousek (1996).a
Einstein's unpublished version of pilot-wave theory has some relevan e
to his argument for the in ompleteness of quantum theory, given in the
general dis ussion. As we saw in se tion 7.1, a ording to Einstein's
argument, lo ality requires `that one does not des ribe the [dira tion
pro ess solely by the S hrdinger wave, but that at the same time one
lo alises the parti le during the propagation' (p. 487). Einstein added: Ì
think that Mr de Broglie is right to sear h in this dire tion' without
mentioning that he himself had re ently made an attempt in the same
dire tion.
It quite possible that, before abandoning his version of pilot-wave
theory, Einstein had onsidered presenting it at the fth Solvay onferen e. Indeed, had Einstein been happy with his new theory, there is
every reason to think he would have presented it a few months later in
Brussels. As dis ussed in se tion 1.3, Lorentz had asked Einstein to give
a report on parti le statisti s, and while Einstein had agreed to do so, he
was relu tant. Less than a month after withdrawing his pilot-wave paper,
Einstein withdrew his ommittment to speak at the Solvay onferen e,
writing to Lorentz on 17 June: `.... I kept hoping to be able to ontribute
something of value in Brussels; I have now given up that hope. .... I did
not take this lightly but tried with all my strength .... ' (quoted from
Pais 1982, p. 432). It then seems indeed probable that, instead of (or
in addition to) speaking about parti le statisti s, Einstein had hoped to
present something like his version of pilot-wave theory.
Let us now des ribe what Einstein's proposal was. (For more detailed
presentations, see Belousek (1996) and Holland (2005).)
Einstein's starting point is the time-independent S hrdinger equati-
a Ar hive referen e: AEA 2-100.00 (in German); urrently available on-line at

http://www.alberteinstein.info/db/ViewDetails.do?Do umentID=34338 .
261
ona
2 2
+ V = E
2
(4)
for a many-body system with potential energy V , total energy E , and

wave fun tion on an n-dimensional onguration spa e. Einstein onsiders (4) to dene a kineti energy
EV =
2 2
,
2
(5)
whi h may also be written as

EV =
1
1
(ds/dt)2 = g q q ,
2
2
(6)
where ds2 = g dq dq is the line element in onguration spa e and

q is the system velo ity. The theory is expressed in terms of arbitrary oordinates q with metri g . The Lapla ian is then given by
2 = g , where is the ovariant derivative. Einstein
writes as (whi h he alls the `tensor of - urvature'), and
seeks unit ve tors A that extremise A A , leading to the eigenvalue
problem
( g )A = 0 ,
(7)
with n real and distin t solutions () . At ea h point, the A() dene

a lo al orthogonal oordinate system (with Eu lidean metri at that
point). In this oordinate system, both and g are diagonal, with
= () and g = . The two expressions (5) and
omponents
(6) for the kineti energy then be ome
EV =
and
2 1 X
EV =
1X 2
q
2
(8)
(9)
a The parti le masses make no expli it appearan e, be ause they have been absorbed
into the onguration-spa e metri g . Einstein is following S hrdinger's usage.
In his se ond paper on wave me hani s, S hrdinger (1926 ) introdu ed a
non-Eu lidean metri determined by the kineti energy (see also S hrdinger's
report, p. 449). The Lapla ian 2 is then understood in the Riemannian sense,
and the S hrdinger equation indeed takes the form (4) with no masses appearing
expli itly. Note that also in analyti al me hani s it is sometimes onvenient to
write the kineti energy 12 M q q as 12 (ds/dt)2 where ds2 = M dq dq is a
line element with metri M (Goldstein 1980, pp. 36970).
262
respe tively. Einstein then introdu es the hypothesis that these two
expressions mat h term by term, so that
2
q =
2
.
(10)
= () , and transforming ba k to the original oordinate

Using
P
system, where q = q A() , Einstein obtains the nal result
q =
()
A() ,
(11)
whi h expresses q in terms of () and A() , where at ea h point in

onguration spa e () and A() are determined by the lo al values
of the wave fun tion and its derivatives. (The ambiguity in sign is,
a ording to Einstein, to be expe ted for quasiperiodi motions.)
Thus, a ording to (11), the system velo ity q is lo ally determined
(up to signs) by and its derivatives. This is Einstein's proposed velo ity
law, to be ompared and ontrasted with de Broglie's.
The manus ript ontains an additional note àdded in proof', in whi h
Einstein asserts that the theory he has just outlined is physi ally una eptable, be ause it predi ts that for a system omposed of two independent subsystems, with an additive Hamiltonian and a produ t wave
fun tion, the velo ities for one subsystem generally depend on the instantaneous oordinates of the other subsystem. However, Einstein adds
that Grommer has pointed out that this problem ould be avoided by
repla ing with ln in the onstru tion of the velo ity eld. A ording
to Einstein: `The elaboration of this idea should o asion no di ulty
....' (Howard 1990, p. 90). Despite Einstein's apparent optimism that
Grommer's modi ation would work, it has been assumed (Howard 1990,
p. 90; Cushing 1994, p. 128) that Einstein withdrew the paper be ause he
soon realised that Grommer's suggestion did not work. But it has been
shown by Holland (2005) that the repla ement ln does in fa t
remove the di ulty raised by Einstein. Holland shows further, however,
that there are other di ulties with Einstein's theory with or without
Grommer's modi ation. It is not known if Einstein re ognised these
other di ulties, but if he did, they would ertainly have been onvin ing
grounds on whi h to abandon the s heme altogether.
The di ulties raised by Holland are as follows. First, the theory
applies only to a limited range of quantum states: real stationary states
with E V . Se ond, the system velo ity q is dened only in a limited
263
domain of onguration spa e (for example, without Grommer's modi ation, () / must be negative for q to be real). Third, even where
2
q is dened, the ontinuity equation (|| q ) = 0 is generally not
2
satised (where || is time-independent for a stationary state), so that
the velo ity eld q does not map an initial Born-rule distribution into a
nal Born-rule distribution. This last feature removes any realisti hope
that the theory ould reprodu e the predi tions of standard quantum
theory (Holland 2005).a
At the end of his manus ript (before the note added in proof), Einstein
writes: `.... the assignment of ompletely determined motions to solutions
of S hrdinger's dierential equation is .... just as possible as is the
assignment of determined motions to solutions of the Hamilton-Ja obi
dierential equation in lassi al me hani s' (Belousek 1996, pp. 441
2). Given the lose relationship between the Hamilton-Ja obi fun tion
and the phase S of (the latter redu ing to the former in the shortwavelength limit), it is natural to ask why Einstein did not onsider
de Broglie's velo ity eld proportional to the phase gradient S
whi h is a straightforward generalisation of the Hamilton-Ja obi velo ity
formula. As well as being mu h simpler than Einstein's, de Broglie's
velo ity eld immediately satises Einstein's desired separability of parti le motions for produ t states.a Why did Einstein instead propose
what seems a mu h more ompli ated and unwieldy s heme to generate
parti le velo ities from the wave fun tion?
Perhaps Einstein did not adopt de Broglie's velo ity eld simply be ause, for the real wave fun tions Einstein onsidered, the phase gradient
S = Im(/) vanishes; though it is not lear why Einstein thought
one ould restri t attention to su h wave fun tions. As for the seemingly
pe uliar onstru tion that Einstein did adopt, it should be noted that
Einstein was (as he himself states) following S hrdinger in his use of
a non-Eu lidean metri determined by the kineti energy (S hrdinger
1926 ). Further details of Einstein's onstru tion may have been related
to another idea Einstein was pursuing: that quantisation onditions
ould arise from a generally- ovariant and òverdetermined' eld theory
a As Holland (2005) also points out, from a modern point of view the mutual
dependen e of parti le motions for produ t states, whi h Einstein found so
una eptable, need not be a real di ulty. We now know that a hidden-variables
theory must be nonlo al, and there is no reason why in some theories the
underlying nonlo ality ould not exist for fa torisable quantum states as well as
for entangled ones. Indeed, Holland gives an example of just su h a theory.
a It is not known whether or not Einstein noti ed the nonlo ality of de Broglie's
theory for entangled wave fun tions.
264
that onstrains initial states, with parti les represented by singularities

(Pais 1982, pp. 4648), an idea that, in retrospe t, seems somewhat
reminis ent of de Broglie's double-solution theory (se tion 2.3.1).
It would of ourse have been most interesting to see how physi ists
would have rea ted had Einstein in fa t published his paper or presented
it at the fth Solvay onferen e. It so happens that Heisenberg had heard
about Einstein's theory through Born and Jordan, and on 19 May,
just two days before Einstein withdrew the paper wrote to Einstein
asking about it. On 10 June 1927, Heisenberg wrote to Einstein again,
this time with detailed omments and arguments against what Einstein
was (or had been) proposing.a
At the beginning of this letter, after thanking Einstein for his `friendly
letter', Heisenberg says he would like to explain why he believes indeterminism is ne essary and not just possible. He hara terises Einstein
as thinking that, while all experiments will agree with the statisti al
quantum theory, nevertheless it will be possible to talk about denite
parti le traje tories. Heisenberg then outlines an obje tion. He onsiders
free ele trons with a onstant and very low velo ity hen e large de
Broglie wavelength striking a grating with spa ing omparable to
. He remarks that, in Einstein's theory, the ele trons will be s attered
in dis rete spatial dire tions, and that if the initial position of a parti le
were known one ould al ulate where the parti le will hit the grating.
Heisenberg then asserts that one ould set up an obsta le at that point,
so as to dee t the parti le in an arbitrary dire tion, independently
of the rest of the grating. Heisenberg says that this ould be done,
if the for es between the parti le and the obsta le a t only at short
range (over distan es mu h smaller than the spa ing of the grating).
Heisenberg then adds that, in a tual fa t, the ele tron will be s attered
in the usual dis rete dire tions regardless of the obsta le. Heisenberg
goes on to say that one ould es ape this on lusion if one `sets the
motion of the parti le again in dire t relation to the behaviour of the
waves'. But this means, says Heisenberg, that the size of the parti le
or the range of its intera tion depends on its velo ity. Heisenberg
asserts that making su h assumptions a tually amounts to giving up
the word `parti le' and leads to a loss in understanding of why the
simple potential energy e2 /r appears in the S hrdinger equation or
in the matrix Hamiltonian fun tion. On the other hand, Heisenberg
agrees that: Ìf you use the word parti le so liberally, I onsider it
a Heisenberg to Einstein, 19 May and 10 June 1927, AEA 12-173.00 and 12-174.00
(both in German).
11.4 Obje tions: in 1927 and today
265
as very well possible that one an again also dene parti le traje tories'.
But then, adds Heisenberg, one loses the simpli ity of quantum theory,
a ording to whi h the parti le motion takes pla e lassi ally (to the
extent that one an speak about motion in quantum theory). Heisenberg
notes that Einstein seems willing to sa ri e this simpli ity for the sake of
maintaining ausality. He remarks further that, in Einstein's on eption,
many experiments are still determined only statisti ally, and `we ould
only onsole ourselves with the fa t that, while for us the prin iple of
ausality is meaningless, be ause of the un ertainty relation p1 q1 h,
however the dear God knows in addition the position of the parti le and
thereby keeps the validity of the ausal law'. Heisenberg adds that he
nds it unattra tive to try to des ribe more than just the ` onne tion
between experiments' [Zusammenhang der Experimente.
It is likely that Heisenberg had similar views of de Broglie's theory,
and that if he had ommented on de Broglie's theory at the fth Solvay
onferen e he would have said things similar to the above.
Heisenberg's obje tion on erning the ele tron and the grating seems
to be based on inappropriate reasoning taken from lassi al physi s (a
ommon feature of obje tions to pilot-wave theory even today), and
Heisenberg agrees that if the motion of the parti le is stri tly tied to
that of the waves, then it should be possible to obtain onsisten y with
observation (as we now know is indeed the ase for de Broglie's theory).
Even so, Heisenberg seems to think that su h a highly non lassi al
parti le dynami s would la k the simpli ity and intelligibility of ertain
lassi al ideas that quantum theory preserves. Finally, Heisenberg is unhappy with a theory ontaining unobservable ausal onne tions. Similar
obje tions to pilot-wave theory are onsidered in the next se tion.

It is interesting to observe that many of the obje tions to pilot-wave
theory that are ommonly heard today were already voi ed in 1927.
Regarding the existen e or non-existen e of de Broglie's traje tories,
Brillouin in the dis ussion following de Broglie's le ture (p. 399)
ould just as well have been replying to a present-day riti of de BroglieBohm theory when he said that:
Mr Born an doubt the real existen e of the traje tories al ulated by L. de
Broglie, and assert that one will never be able to observe them, but he annot
prove to us that these traje tories do not exist. There is no ontradi tion
between the point of view of L. de Broglie and that of other authors .... .
266
Here we have a oni t between those who believe in hidden entities

be ause of the explanatory role they play, and those who think that
what annot be observed in detail should play no theoreti al role. Su h
oni ts are not un ommon in the history of s ien e: for example, a
similar polarisation of views o urred in the late nineteenth entury
regarding the reality of atoms (whi h the ènergeti ists' regarded as
metaphysi al tions).
The debate over the reality of the traje tories postulated by de Broglie
has been sharpened in re ent years by the re ognition that, from the
perspe tive of pilot-wave theory itself, our inability to observe those
traje tories is not a fundamental onstraint built into the theory, but
rather an a ident of initial onditions with a Born-rule probability
2
distribution P = || (for an ensemble of systems with wave fun tion ).
The statisti al noise asso iated with su h `quantum equilibrium' distributions sets limits to what an be measured, but for more general `none2
quilibrium' distributions P 6= || , the un ertainty prin iple is violated
and observation of the traje tories be omes possible (Valentini 1991b,
1992, 2002b; Pearle and Valentini 2006). Su h nonequilibrium entails, of
ourse, a departure from the statisti al predi tions of quantum theory,
whi h are obtained in pilot-wave theory only as a spe ial èquilibrium'
ase. Arguably, the above disagreement between Born and Brillouin
might nowadays turn on the question of whether one is willing to believe
that quantum physi s is merely the physi s of a spe ial statisti al state.
Another obje tion sometimes heard today is that a velo ity law different from that assumed by de Broglie is equally possible. In the dis ussion following de Broglie's le ture, S hrdinger (p. 400) raised the
possibility of an alternative parti le velo ity dened by the momentum
density of a eld. Pauli pointed out that, for a relativisti eld, if the
parti le velo ity were obtained by dividing the momentum density by
the energy density, then the resulting traje tories would dier from those
obtained by de Broglie, who assumed a velo ity dened (in the ase of a
single parti le) by the ratio of urrent density to harge density. Possibly,
then as now, the existen e of alternative velo ities may have been interpreted as asting doubt on the reality of the velo ities a tually assumed
by de Broglie (though the true velo ities would be ome measurable in
the presen e of quantum nonequilibrium).
In addition to the riti ism that the traje tories annot be observed,
today it is also often obje ted that the traje tories are rather strange
from a lassi al perspe tive. The pe uliar nature of de Broglie's traje tories was addressed in the dis ussion following de Broglie's le ture
267
(pp. 401 .), and again in the general dis ussion (pp. 498, 507 f.). It was,
for example, pointed out that the speed of an ele tron ould be zero in
a stationary state, and that for general atomi states the orbits would
be very ompli ated. At the end of his dis ussion of photon ree tion
by a mirror, Brillouin (p. 403) argued that de Broglie's non-re tilinear
photon paths (in free spa e) were ne essary in order to avoid a paradox
posed by Lewis, in whi h in the presen e of interferen e it appeared that
photons would ollide with only one end of a mirror, ausing it to rotate,
even though from lassi al ele trodynami s the mean radiation pressure
on the mirror is expe ted to be uniform (see Brillouin's Fig. 3).
This last example of Brillouin's re alls present-day debates involving
ertain kinds of quantum measurements, in whi h the traje tories predi ted by pilot-wave theory have ounter-intuitive features that some
authors have labelled `surreal' (Englert et al. 1992, Aharonov and Vaidman 1996), while other authors regard these features as perfe tly understandable from within pilot-wave theory itself (Valentini 1992, p. 24,
Dewdney, Hardy and Squires 1993, Drr et al. 1993). A key question here
is whether it is reasonable to expe t a theory of subquantum dynami s
to onform to lassi al intuitions about measurement (given that it is the
underlying dynami s that should be used to analyse the measurement
pro ess).
Pilot-wave theory is sometimes seen as a return to lassi al physi s (wel omed by some, riti ised by others). But in fa t, de Broglie's
velo ity-based dynami s is a new form of dynami s that is simply quite
distin t from lassi al theory; therefore, it is to be expe ted that the
behaviour of the traje tories will not onform to lassi al expe tations.
As we saw in detail in hapter 2, de Broglie did indeed originally regard
his theory as a radi al departure from the prin iples of lassi al dynami s. It was Bohm's later revival of de Broglie's theory, in an unnatural
pseudo-Newtonian form, that led to the widespread and mistaken per eption that de Broglie-Bohm theory onstituted a return to lassi al
physi s. In more re ent years, de Broglie's original pilot-wave dynami s
has again be ome re ognised as a new form of dynami s in its own right
(Drr, Goldstein and Zangh 1992, Valentini 1992).
12
The fth Solvay onferen e is usually remembered for the lash that took
pla e between Bohr and Einstein, supposedly on erning in parti ular
the possibility of breaking the un ertainty relations. It might be assumed
that this lash took the form of an o ial debate that was the entrepie e
of the onferen e. However, no re ord of any su h debate appears in
the published pro eedings, where both Bohr and Einstein are in fa t
relatively silent.
The available eviden e shows that in 1927 the famous ex hanges between Bohr and Einstein a tually onsisted of informal dis ussions, whi h
took pla e semi-privately (mainly over breakfast and dinner), and whi h
were overheard by just a few of the parti ipants, in parti ular Heisenberg
and Ehrenfest. The histori al sour es for this onsist, in fa t, entirely
of a ounts given by Bohr, Heisenberg and Ehrenfest. These a ounts
essentially ignore the extensive formal dis ussions appearing in the published pro eedings.
As a result of relying on these sour es, the per eption of the onferen e
by posterity has been skewed on two ounts. First, at the fth Solvay onferen e there o urred mu h more that was memorable and important
besides the Bohr-Einstein lash. Se ond, as shown in detail by Howard
(1990), the real nature of Einstein's obje tions was in fa t misunderstood
by Bohr, Heisenberg and Ehrenfest: for Einstein's main target was not
the un ertainty relations, but what he saw as the nonseparability of
quantum theory.
Below we shall indi ate how these misunderstandings arose, summarise what now appear to have been Einstein's true on erns, and end
by urging physi ists, philosophers and historians to re onsider what
a tually took pla e in Brussels in O tober 1927, bearing in mind the deep
questions that we still fa e on erning the nature of quantum physi s.
268
269

A ording to Heisenberg, the dis ussions that took pla e at the fth
Solvay onferen e ` ontributed extraordinarily to the lari ation of the
physi al foundations of the quantum theory' and indeed led to `the
outward ompletion of the quantum theory, whi h now an be applied
without worries as a theory losed in itself' (Heisenberg 1929, p. 495).
For Heisenberg, and perhaps for others in the Copenhagen-Gttingen
amp, the 1927 onferen e seems to have played a key role in nalising
the interpretation of the theory.
However, the per eption that the interpretation had been nalised
proved to be mistaken. As we have shown at length in hapter 5, the
interpretation of quantum theory is today still an open question, and
deep on erns as to its meaning have stubbornly persisted. Further, as we
have seen throughout part II of this book, many of today's fundamental
on erns were voi ed (often at onsiderable length) at the fth Solvay
onferen e. Given that these on erns are still very mu h alive, it has
evidently been a mistake to allow re olle tions of private dis ussions
between Bohr and Einstein to overshadow our histori al memory of the
rest of the onferen e.
Note that, as we saw in hapter 1 (p. 16), while Mehra (1975, p. 152)
and also Mehra and Re henberg (2000, p. 246) state that the general
dis ussion was a dis ussion following `Bohr's report', in fa t Bohr did
not present a report at the onferen e, nor was he invited to give one.
This misunderstanding seems to have arisen be ause, at Bohr's request,
a translation of his Como le ture appears in the published pro eedings,
to repla e his remarks in the general dis ussion; and this has no doubt
ontributed to the ommon view that Bohr played a entral role at
the onferen e, when in fa t it is lear from the pro eedings that at
the o ial meetings both he and Einstein played a rather marginal
role.a Further, it seems rather lear that the standard (and unbalan ed)
version of events was propagated in parti ular by Bohr and Heisenberg,
espe ially through their writings de ades later.
There is very little independent eviden e from the time as to what
was said between Bohr and Einstein. Thus, for example, Mehra and
Re henberg (2000, p. 251) note the `little eviden e of the Bohr-Einstein
debate in the o ial onferen e do uments', and rely on eyewitness
reports by Ehrenfest, Heisenberg and Bohr to yield à fairly onsistent
a Sin e Bohr had not been invited to give a report, one might also question the
propriety of his request that his remarks in the general dis ussion be repla ed by
a translation of a rather lengthy paper he was already publishing elsewhere.
270
histori al pi ture of the great epistemologi al debate between Bohr and

Einstein' (p. 256).
One frequently- ited pie e of ontemporary eviden e is a des ription
of the onferen e written by Ehrenfest a few days later, in a letter to
his students and asso iates in Leiden. This letter is ited at length by
Mehra and Re henberg (2000, pp. 2513). An extra t reads:
Bohr towering ompletely over everybody. .... step by step defeating everybody.
.... It was delightful for me to be present during the onversations between Bohr
and Einstein. .... Einstein all the time with new examples. In a ertain sense
a sort of Perpetuum Mobile of the se ond kind to break the UNCERTAINTY
RELATION. Bohr .... onstantly sear hing for the tools to rush one example
after the other. Einstein .... jumping out fresh every morning. .... I am almost
without reservation pro Bohr and ontra Einstein.
This letter has often been taken as representative of the onferen e.

However, there is a marked ontrast with the published pro eedings (in
whi h Bohr and Einstein are mostly silent), a ontrast whi h has not
been taken into a ount.a
After examining the published pro eedings, Mehra and Re henberg
(2000, pp. 25056) go on to onsider at greater length re olle tions
by Heisenberg and Bohr written de ades after the onferen e
on erning the dis ussions between Bohr and Einstein. With hindsight,
again given that the interpretation of quantum theory is today an open
question, it would be desirable to have a more balan ed view of the
onferen e, fo ussing more on the ontent of the published pro eedings,
and rather less on these later re olle tions by just two of the parti ipants.
Here is an extra t from Heisenberg's re olle tion, written some 40
years later (Heisenberg 1967, p. 107):
The dis ussions were soon fo ussed upon a duel between Einstein and Bohr .... .
We generally met already at breakfast in the hotel, and Einstein began to
des ribe an ideal experiment in whi h he thought the inner ontradi tions of
the Copenhagen interpretation were espe ially learly visible. Einstein, Bohr
and I walked together from the hotel to the onferen e building, and I listened
to the lively dis ussion between those two people whose philosophi al attitudes
were so dierent, .... at lun h time the dis ussions ontinued between Bohr and
the others from Copenhagen. Bohr had usually nished the omplete analysis
of the ideal experiment by late afternoon and would show it to Einstein at the
supper table. Einstein had no good obje tion to this analysis, but in his heart
he was not onvin ed.
a We have attempted to ompare Ehrenfest's ontemporary a ount with that in

letters written by other parti ipants soon after the onferen e, but have found
nothing signi ant.
271
From this a ount, the Bohr-Einstein lash appears indeed to have been
a private dis ussion, with a few of Bohr's lose asso iates in attendan e.
And yet, there has been a marked tenden y to portray this dis ussion as
the entrepie e of the whole onferen e. Thus, for example, in the prefa e
to Mehra's book The Solvay Conferen es on Physi s, Heisenberg wrote
the following about the 1927 onferen e (Mehra 1975, pp. vvi):
Therefore the dis ussions at the 1927 Solvay Conferen e, from the very beginning, entred around the paradoxa of quantum theory. .... Einstein therefore
suggested spe ial experimental arrangements for whi h, in his opinion, the
un ertainty relations ould be evaded. But the analysis arried out by Bohr
and others during the Conferen e revealed errors in Einstein's arguments. In
this situation, by means of intensive dis ussions, the Conferen e ontributed
dire tly to the lari ation of the quantum-theoreti al paradoxa.
Heisenberg says nothing at all about the alternative theories of de Broglie

and S hrdinger, or about the views of Lorentz or Dira (for example),
or about the other extensive dis ussions re orded in the pro eedings.
As for the text of Mehra's book on the Solvay onferen es, the hapter devoted to the fth Solvay onferen e ontains a summary of the
general dis ussion, whi h says nothing about the published dis ussions
beyond providing a list of the parti ipants. Mehra's summary states
that a debate took pla e between Bohr and Einstein, and that the
famous Bohr-Einstein dialogue began here in 1927. Mehra then adds an
appendix, reprodu ing Bohr's famous essay `Dis ussion with Einstein on
epistemologi al problems in atomi physi s' (Bohr 1949), written more
than 20 years after the onferen e took pla e. On e again, the published
dis ussions are made to appear rather insigni ant ompared to (Bohr's
re olle tion of) the informal dis ussions between Bohr and Einstein.
The above essay by Bohr is in fa t the prin ipal and most detailed
histori al sour e for the Bohr-Einstein debate. This essay was Bohr's
ontribution to the 1949 fests hrift for Einstein's seventieth birthday.
It is reprinted as the very rst paper in Wheeler and Zurek's (1983)
inuential olle tion Quantum Theory and Measurement, as well as elsewhere. It gives a detailed a ount of Bohr's dis ussions with Einstein at
the fth Solvay onferen e (as well as at the sixth Solvay onferen e of
1930). A ording to Bohr, the dis ussions in 1927 entred around, among
other things, a version of the double-slit experiment, in whi h a ording
to Einstein it was possible to observe interferen e while at the same time
dedu ing whi h path the parti le had taken, a laim on lusively refuted
by Bohr.
Were Bohr's re olle tions a urate? Jammer ertainly thought so:
272
Bohr's masterly report of his dis ussions with Einstein on this issue, though
written more than 20 years after they had taken pla e, is undoubtedly a reliable
sour e for the history of this episode. (Jammer 1974, p. 120)
Though as Jammer himself adds (p. 120): Ìt is, however, most deplorable
that additional do umentary material on the Bohr-Einstein debate is
extremely s anty'.
We now know that, as we shall now dis uss, Bohr's re olle tion of
his dis ussions with Einstein did not properly apture Einstein's true
intentions, essentially be ause, at the time, no one understood what
Einstein's prin ipal on ern was: the nonseparability of quantum theory.

Separability the requirement that the joint state of a omposite of
spatially separated systems should be determined by the states of the
omponent parts was a ondition basi to Einstein's eld-theoreti
view of physi s (as indeed was the absen e of a tion at a distan e). As
already mentioned in se tion 9.2, Howard (1990) has shown in great detail how Einstein's on erns about the failure of separability in quantum
theory date ba k to long before the famous EPR paper of 1935.a There
is no doubt that, by 1909, Einstein understood that if light quanta
were treated like the spatially independent mole ules of an ideal gas,
then the resulting u tuations were in onsistent with Plan k's formula
for bla kbody radiation; and ertainly, in 1925, Einstein was on erned
that Bose-Einstein statisti s entailed a mysterious interdependen e of
photons. Also in 1925, as we dis ussed in se tion 9.2, Einstein's theory
of guiding elds in 3-spa e in whi h spatially separated systems ea h
had their own guiding wave, in a ordan e with Einstein's separability
riterion oni ted with energy-momentum onservation for single
events. Howard (pp. 8391) argues further that, in spring 1927, Einstein
must have realised that S hrdinger's wave me hani s in onguration
spa e violated separability, be ause a general solution to the S hrdinger
equation for a omposite system ould not be written as a produ t
over the omponents.b In other words, Einstein obje ted to what we
would now all entanglement, and on luded that the wave fun tion in
onguration spa e ould not represent anything physi al.
a See also Fine (1986, hapter 3).
b Howard's argument here is somewhat ir umstantial, appealing in part to the
di ulty with separability that Einstein had with his own hidden-variables
amendment of wave me hani s ( f. se tion 11.3).
273
Einstein's on erns about separability ontinued up to and beyond

the fth Solvay onferen e. While it seems that Einstein did have some
early doubts about the validity of the un ertainty relations, Howard's
re onstru tion shows that Einstein's main on ern lay elsewhere. The
primary aim of the famous thought experiments that Einstein dis ussed
with Bohr, in 1927 and subsequently, was not to defeat the un ertainty
relations but to highlight the (for him disturbing) feature of quantum
theory, that spatially separated systems annot be treated independently. As Howard (p. 94) puts it, regarding the 1927 Solvay onferen e:
But if the un ertainty relations
stein,
really
were the main sti king point
for Ein-
why did Einstein not say so in the published version of his remarks, or
anywhere else for that matter in orresponden e or in print in the weeks and
months following the Solvay meeting?
We have indeed seen in hapter 7 that Einstein's riti ism in the

general dis ussion on erned lo ality and ompleteness (just like the
later EPR argument), not the un ertainty relations. Bohr, in his reply,
states that he does not ùnderstand what pre isely is the point' Einstein
is making. It seems rather lear that, indeed, in 1927 Bohr did not
understand Einstein's point, and it is remarkable that what is most
often re alled about the fth Solvay onferen e was in fa t largely a
misunderstanding.
A ording to Bohr's later re olle tions (Bohr 1949), at the fth Solvay
onferen e Einstein proposed a version of the two-slit experiment in
whi h measurement of the transverse re oil of a s reen with a single slit
would enable one to dedu e the path of a parti le through a se ond s reen
with two slits, while at the same time observing interferen e on the far
side of the se ond s reen. (Consideration of this experiment must have
taken pla e informally, not in the o ial dis ussions.) This experiment
has been analysed in detail by Wootters and Zurek (1979), who show
the ru ial role played by quantum nonseparability between the parti le
and the rst s reen. While the eviden e is somewhat sket hy in this
parti ular instan e, a ording to Howard the main point that on erned
Einstein in this experiment was pre isely su h nonseparability.
That separability was indeed Einstein's entral on ern is learer in
the later `photon-box' thought experiment he dis ussed with Bohr at the
sixth Solvay onferen e of 1930, involving weighing a box from whi h a
photon es apes. Again, Bohr dis usses this experiment at length in his
re olle tions, where a ording to him it was yet another of Einstein's
attempts to ir umvent the un ertainty relations. Spe i ally, a ording
274
to Bohr, Einstein's intention was to beat the energy-time un ertainty

relation, by measuring both the energy of the emitted photon (by weighing the box before and after) and its time of emission (given by a lo k
ontrolling the shutter releasing the photon). On Bohr's a ount, this
attempt failed, ironi ally, be ause of the time dilation in a gravitational
eld implied by Einstein's own general theory of relativity.
However, it seems that in fa t, the photon-box experiment was (like
Einstein's published obje tion of 1927) really a form of the later EPR
argument for in ompleteness. This is shown by a letter Ehrenfest wrote
to Bohr on 9 July 1931, just after Ehrenfest had visited Einstein in Berlin
(Howard 1990, pp. 989). Ehrenfest reports that Einstein said he did not
invent the photon-box experiment to defeat the un ertainty relations
(whi h he had for a long time no longer doubted), but `for a totally
dierent purpose'. Ehrenfest then explains that Einstein's real intention
was to onstru t an example in whi h the hoi e of measurement at
one lo ation would enable an experimenter to predi t either one or the
other of two in ompatible quantities for a system that was far away at
the time of the measurement. In the example at hand, if the es aped
photon is ree ted ba k towards the box after having travelled a great
distan e, then the time of its return may be predi ted with ertainty if
the experimenter he ks the lo k reading (while the photon is still far
away); alternatively, the energy (or frequen y) of the returning photon
may be predi ted with ertainty if, instead, the experimenter hooses
to weigh the box (again while the photon is still far away). Be ause
the two possible operations take pla e while the photon is at a great
distan e, the assumptions of separability and lo ality imply that both
the time and energy of the returning photon are in reality determined
in advan e (even if in pra ti e an experimenter annot arry out both
predi tions simultaneously), leading to the on lusion that quantum
theory is in omplete.a
The true thrust of Einstein's argument was not appre iated at the
a The reasoning here is similar to that in Einstein's own (and simpler) version of
the EPR argument, whi h rst appears in a letter from Einstein to S hrdinger of
19 June 1935, one month after the EPR paper was published (Fine 1986, hap. 3).
The argument whi h Einstein repeated and rened between 1936 and 1949
runs essentially as follows. A omplete theory should asso iate one and only one
theoreti al state with ea h real state of a system; in an EPR-type experiment
on orrelated systems, depending on what measurement is arried out at one
wing of the experiment, quantum theory asso iates dierent wave fun tions with
what must (assuming lo ality) be the same real state at the other distant wing.
Therefore, quantum theory is in omplete. For a detailed dis ussion, see Howard
(1990).
275
time, perhaps be ause Bohr and his asso iates tended to identify the
existen e of physi al quantities with their experimental measurability: if
two quantities ould not be measured simultaneously in the same experiment, they did not exist simultaneously in the same experiment. With
this attitude in mind, it would be natural to mistake Einstein's laim of
simultaneous existen e for a laim of simultaneous measurability.
We feel that Howard's reappraisal of the Bohr-Einstein debate, as well
as being of great intrinsi interest, also provides an instru tive example
of how the history of quantum physi s should be re onsidered in the light
of our modern understanding of quantum theory and its open problems.
There was ertainly mu h more to the fth Solvay onferen e than the
Bohr-Einstein lash, and a similar reappraisal of other ru ial en ounters
at that time seems overdue.
If the history of quantum theory is written on the assumption that
Bohr, Heisenberg and Born were right, and that de Broglie, S hrdinger
and Einstein were wrong, the resulting a ount is likely to be unsatisfa tory: opposing views will tend to be misunderstood or underestimated, supporting views over-emphasised, and valid alternative approa hes
ignored.
A re onsideration of the fth Solvay onferen e ertainly entails a
re-evaluation of de Broglie's pilot-wave theory as a oherent but (until
very re ently) `forgotten' formulation of quantum theory. S hrdinger's
ideas, too, seem more plausible today, in the light of modern ollapse models. One should also re onsider what Born and Heisenberg's `quantum
me hani s' really was, in parti ular as on erns the role of time and the
ollapse of the wave fun tion.
There is no longer a denitive, widely-a epted interpretation of quantum me hani s; it is no longer lear who was right and who was wrong
in O tober 1927. Therefore, it seems parti ularly important at this time
to return to the histori al sour es and re-evaluate them. We hope that
physi ists, philosophers and historians will re onsider the signi an e
of the fth Solvay onferen e, both for the history of physi s and for
urrent resear h in the foundations of quantum theory.
Part III
The pro eedings of the 1927 Solvay
onferen e
H. A. Lorentz
Hardly a few months have gone by sin e the meeting of the fth physi s
onferen e in Brussels, and now I must, in the name of the s ienti
ommittee, re all here all that meant to the Solvay International Institute of Physi s he who was our hairman and the moving spirit of our
meetings. The illustrious tea her and physi ist, H. A. Lorentz, was taken
away in February 1928 by a sudden illness, when we had just admired,
on e again, his magni ent intelle tual gifts whi h age was unable to
diminish in the least.
Professor Lorentz, of a simple and modest demeanour, nevertheless
enjoyed an ex eptional authority, thanks to the ombination of rare
qualities in a harmonious whole. Theoreti ian with profound views
eminent tea her in the highest forms of instru tion and tirelessly devoted
to this task fervent advo ate of all international s ienti ollaboration he found, wherever he went, a grateful ir le of pupils, dis iples
and those who arried on his work. Ernest Solvay had an unfailing
appre iation of this moral and intelle tual for e, and it was on this
that he relied to arry through a plan that was dear to him, that of
serving S ien e by organising onferen es omposed of a limited number
of physi ists, gathered together to dis uss subje ts where the need for
new insights is felt with parti ular intensity. Thus was born the Solvay
International Institute of Physi s, of whi h Ernest Solvay followed the
beginnings with a tou hing on ern and to whi h Lorentz devoted a loyal
and fruitful a tivity.
All those who had the honour to be his ollaborators know what he
was as hairman of these onferen es and of the preparatory meetings.
His thorough knowledge of physi s gave him an overall view of the
problems to be examined. His lear judgement, his fair and benevolent
spirit guided the s ienti ommittee in the hoi e of the assistan e it
279
280
H. A. Lorentz
was appropriate to all upon. When we then were gathered together at

a onferen e, one ould only admire without reservations the mastery
with whi h he ondu ted the hairmanship. His shining intelle t dominated the dis ussion and followed it also in the details, stimulating it
or preventing it from drifting, making sure that all opinions ould be
usefully expressed, bringing out the nal on lusion as far as possible.
His perfe t knowledge of languages allowed him to interpret, with equal
fa ility, the words uttered by ea h one. Our hairman appeared to us,
in fa t, gifted with an invin ible youth, in his passion for s ienti truth
and in the joy he had in omparing opinions, sometimes with a shrewd
smile on his fa e, and even a little mis hievousness when onfronted
with an unforeseen aspe t of the question. Respe t and ae tion went
to him spontaneously, reating a ordial and friendly atmosphere, whi h
fa ilitated the ommon work and in reased its e ien y.
True reator of the theoreti al edi e that explains opti al and ele tromagneti phenomena by the ex hange of energy between ele trons
ontained in matter and radiation viewed in a ordan e with Maxwell's
theory, Lorentz retained a devotion to this lassi al theory. All the
more remarkable is the exibility of mind with whi h he followed the
dis on erting evolution of the quantum theory and of the new me hani s.
The impetus that he gave to the Solvay institute will be a memory
and an example for the s ienti ommittee. May this volume, faithful
report of the work of the re ent physi s onferen e, be a tribute to the
memory of he who, for the fth and last time, honoured the onferen e
by his presen e and by his guidan e.
M. Curie
The fth of the physi s onferen es, provided for by arti le 10 of the statutes of the international institute of physi s founded by Ernest Solvay,
held its sessions in Brussels on the premises of the institute from 24 to
29 O tober 1927.
The following took part in the onferen e:
Lorentz , of Haarlem, Chairman.

Mrs P. Curie, of Paris; Messrs N. Bohr, of Copenhagen; M. Born,
of Gttingen; W. L. Bragg, of Man hester; L. Brillouin, of Paris;
A. H. Compton, of Chi ago; L.-V. de Broglie, of Paris; P. Debye,
of Leipzig; P. A. M. Dira , of Cambridge; P. Ehrenfest, of Leiden;
A. Einstein, of Berlin; R. H. Fowler, of Cambridge; Ch.-E. Guye,
of Geneva; W. Heisenberg, of Copenhagen; M. Knudsen, of Copenhagen; H. A. Kramers, of Utre ht; P. Langevin, of Paris; W. Pauli,
of Hamburg; M. Plan k, of Berlin; O. W. Ri hardson, of London;
[E. S hrdinger, of Zuri h; C. T. R. Wilson, of Cambridge, Members.
Mr J.-E. Vers haffelt, of Gent, fullled the duties of Se retary.
Messrs Th. De Donder, E. Henriot and Aug. Pi ard, professors
at the University of Brussels, attended the meetings of the onferen e
as guests of the s ienti ommittee, Mr Ed. Herzen, professor at the
Mr H. A.
ole des Hautes tudes de Bruxelles, as representative of the Solvay

family.
281
282
Langmuir
Professor I.
, of S hene tady (U. S. of Ameri a), visiting
Europe, attended the meetings as a guest.
Mr Edm. van Aubel, member of the S ienti Committee, and Mr H.
Deslandres, dire tor of the Meudon observatory, invited to parti ipate
in the onferen e meetings, had been ex used.
Bragg
Sir W. H.
, member of the s ienti ommittee, who had handed
in his resignation before the meetings and requested to be ex used, also
did not attend the sessions.
The administrative ommission of the institute was omposed of:
Bordet
Messrs Jules
, professor at the University of Brussels, appointed
by H. M. the King of the Belgians; Armand
, engineer, manager of Solvay and Co.; Mauri e
, professor at the University
of Brussels; mile
, professor at the University of Brussels;
Ch.
, engineer, appointed by the family of Mr Ernest Solvay,
Administrative Se retary.
Lefbure
Henriot
Bourquin
Solvay
The s ienti ommittee was omposed of:
Lorentz
Knudsen
Bragg
Curie
Einstein
Guye
Langevin
Ri hardson
Aubel
Sir W. H. Bragg, resigning member, was repla ed by Mr B. Cabrera,
, professor at the University of Leiden, ChairMessrs H. A.

man; M.
, professor at the University of Copenhagen, Se retary; W. H.
, professor at the University of London, president of the
Royal Institution; Mrs Pierre
, professor at the Fa ulty of S ien es
of Paris; Messrs A.
,a professor, in Berlin; Charles-Eug.
,
professor at the University of Geneva; P.
, professor at the
Collge de Fran e, in Paris; O. W.
, professor at the University of London; Edm. van
, professor at the University of Gent.
professor at the University of Madrid.
To repla e its late hairman, the s ienti ommittee hose Professor
P. Langevin.
a Chosen in repla ement of Mr H. Kamerlingh Onnes, de eased.
The intensity of X-ray ree tion
By Mr W. L. BRAGG
1. The lassi al treatment of x-ray diffra tion
phenomena
The earliest experiments on the dira tion of X-rays by rystals showed
that the dire tions in whi h the rays were dira ted were governed by the
lassi al laws of opti s. Laue's original paper on the dira tion of white1
radiation by a rystal, and the work whi h my father and I initiated on
the ree tion of lines2 in the X-ray spe trum, were alike based on the
laws of opti s whi h hold for the dira tion grating. The high a ura y
whi h has been developed by Siegbahn and others in the realm of X-ray
spe tros opy is the best eviden e of the truth of these laws. Advan e in
a ura y has shown the ne essity of taking into a ount the very small
refra tion of X-rays by the rystal, but this refra tion is also determined
by the lassi al laws and provides no ex eption3 to the above statement.
The rst attempts at rystal analysis showed further that the strength
of the dira ted beam was related to the stru ture of the rystal in a way
to be expe ted by the opti al analogy. This has been the basis of most
work on the analysis of rystal stru ture. When mono hromati X-rays
are ree ted from a set of rystal planes, the orders of ree tion are
strong, weak, or absent in a way whi h an be a ounted for qualitatively
a We follow Bragg's original English types ript, from the opy in the Ri hardson
olle tion, AHQP-RDN, do ument M-0059 (indexed as ùnidentied author' in
the mi rolmed atalogue). Obvious typos are orre ted mostly ta itly and some
of the spelling has been harmonised with that used in the rest of the volume.
Dis repan ies between the original English and the published Fren h are endnoted
(eds.).
283
284
W. L. Bragg
by the arrangement of atoms4 parallel to these planes. In the analysis

of many stru tures, it is not ne essary to make a stri t examination of
the strength of the dira ted beams. Slight displa ements of the atoms
ause the intensities of the higher orders to u tuate so rapidly, that
it is possible to x the atomi positions with high a ura y by using a
rough estimate of the relative intensity of the dierent orders.
When we atta k the problem of developing an a urate quantitative
theory of intensity of dira tion, many di ulties present themselves.
These di ulties are so great, and the interpretation of the experimental
results has often been so un ertain, that it has led5 to a natural distrust
of dedu tions drawn from intensity measurements. Investigators of rystal stru tures have relied on qualitative methods,6 sin e these were in
many ases quite adequate. The development of the quantitative analysis
has always interested me personally, parti ularly as a means of atta king
the more ompli ated rystalline stru tures, and it would seem that at
the present time the te hnique has rea hed a stage when we an rely on
the results. It is my purpose in this paper to attempt a riti al survey
of the present development of the subje t. It is of onsiderable interest
be ause it is our most dire t way of analysing atomi and mole ular
stru ture.
In any X-ray examination of a rystalline body, what we a tually
measure is a series of samples7 of the oherent radiation s attered in
ertain denite dire tions by the unit of the stru ture. This unit is, in
general, the element of pattern of the rystal, while in ertain simple
ases it may be a single atom.
In the examination of a small body by the mi ros ope, the obje tive
re eives the radiation s attered in dierent dire tions by the body, and
the information about its stru ture, whi h we get by viewing the nal
image, is ontained at an earlier stage8 in these s attered beams. Though
the two ases of mi ros opi and X-ray examination are so similar, there
are ertain important dieren es. The s attered beams in the mi ros ope
an be ombined again to form an image, and in the formation of
the image the phase relationship between beams s attered in dierent
dire tions plays an essential part. In the X-ray problem, sin e we an only
measure the intensity of s attering in ea h dire tion, this phase relationship annot be determined experimentally, though in many ases it an
be inferred.9 Further, the mi ros ope re eives the s attered beams over
a ontinuous range of dire tions, whereas the geometry of the rystalline
stru ture limits our examination to ertain dire tions of s attering. Thus
we annot form dire tly an image of the rystalline unit whi h is being
285
illuminated by X-rays. We an only measure experimentally the strength

of the s attered beams, and then build up an image pie e by pie e from
the information we have obtained.
It is important to note that in the ase of X-ray examination all
work is being arried out at what is very nearly the theoreti al limit
of the resolving power of our instruments. The range of wavelength
whi h it is onvenient to use lies between 0.6 and 1.5 . This range
is of su iently small wavelength for work with the details of rystal
stru ture, whi h is always on a s ale of several ngstrm units, but
the wavelengths are in onveniently great for an examination into atomi
stru ture. It is unfortunate from a pra ti al point of view that there
is no onvenient steady sour e of radiation between the K lines of the
metal palladium, and the very mu h shorter K lines of tungsten. This
di ulty will no doubt be over ome, and a te hnique of ùltraviolet'
X-ray mi ros opy will be developed, but at present all the a urate
work on intensity of ree tion has been done with wavelengths in the
neighbourhood of 0.7 .
We may onveniently10 divide the pro ess of analysis into three stages.
a) The experimental measurement of the intensities of the dira ted
beams.
b) The redu tion of these observations, with the aid of theoreti al formulae, to measurements of the amplitudes of the waves s attered by a
single unit of the stru ture, when a wave train of given amplitude falls
on it.
) The building up of the image, or dedu tion of the form of the unit,
from these measurements of s attering in dierent dire tions.
2. History of the use of quantitative methods

The fundamental prin iples of a mathemati al analysis of X-ray ree tion were given in Laue's original paper [1, but the pre ise treatment of
intensity of ree tion may be said to have been initiated by Darwin [2
with two papers in the Philosophi al Magazine early in 1914, in whi h
he laid down the basis for a omplete theory of X-ray ree tion based
on the lassi al laws of ele trodynami s.11 The very fundamental and
independent treatment of the whole problem by Ewald [3, along quite
286
W. L. Bragg
dierent lines, has onrmed Darwin's on lusions in all essentials. These

papers established the following important points.
1. Two formulae for the intensity of X-ray ree tion an be dedu ed,
depending on the assumptions whi h are made. The rst of these has
sin e ome to be known as the formula for the ìdeally imperfe t rystal'
or `mosai rystal'.a It holds for a rystal in whi h the homogeneous
blo ks are so small that the redu tion in intensity of a ray passing
through ea h blo k, and being partly ree ted by it, is wholly a ounted
for by the ordinary absorption oe ient. This ase is simple to treat
from a mathemati al point of view, and in a tual fa t many rystals
approa h this physi al ondition of a perfe t mosai .
The se ond formula applies to ree tion by an ideally perfe t rystal.
Here ordinary12 absorption plays no part in intensity of ree tion. This is
perfe t over a nite range of glan ing angles, all radiation being ree ted
within this range. The range depends on the e ien y of the atom planes
in s attering. The se ond formula is entirely dierent from the rst, and
leads to numeri al results of a dierent order of magnitude.
2. The a tual intensity of ree tion in the ase of ro ksalt is of the order
to be expe ted from the imperfe t rystal formula.
3. The observed rapid de line in intensity of the high orders is only partly
a ounted for by the formula for ree tion, and must be due in addition
to the spatial distribution of s attering matter in the atoms (ele tron
distribution).
4. When a rystal is so perfe t that it is ne essary to allow for the
intera tion of the separate planes, the transmitted beam is extinguished
more rapidly than orresponds to the true absorption of the rystal
(extin tion).
5. There exists a refra tive index for both rystalline and amorphous
substan es, slightly less than unity, whi h auses small deviations from
the law n = 2d sin .
Another important fa tor in intensity of ree tion had been already examined theoreti ally by Debye [4, this being the diminution in
a I believe we owe to Ewald the happy suggestion of the word `mosai '.
287
intensity with rising temperature due to atomi movement. Though

subsequent work has put Debye's and Darwin's formulae in modied
and more onvenient forms, the essential features were all ontained in
these early papers.
On the experimental side,the rst a urate quantitative measurements
were made by W. H. Bragg [5.13 The rystal was moved with onstant
angular velo ity through the ree ting position, and the total amount
of ree ted radiation measured. He showed that the ree tion14 from
ro ksalt for a series of fa es lay on a smooth urve when plotted against
the sine of the glan ing angle, emphasising that a denite physi al onstant was being measured. This method of measurement has sin e been
15
widely used. The quantity E
I , where E is the total energy of radiation
ree ted, the angular velo ity of rotation, and I the total radiation
falling on the rystal fa e per se ond, is independent of the experimental
arrangements, and is a onstant for a given ree tion from a mosai
rystal; it is generally termed the ìntegrated ree tion'.16 It is related
in a simple way to the energy measurements from a powdered rystal,
whi h have also been employed for a urate quantitative work. W. H.
Bragg's original measurements were omparisons17 of this quantity for
dierent fa es, not absolute measurements in whi h the strength of an
in ident beam was onsidered.
W. H. Bragg further demonstrated the existen e of the extin tion
ee t predi ted by Darwin, by passing X-rays through a diamond rystal set for ree tion and obtaining an in reased absorption. He made
measurements of the diminution in intensity of ree tion18 with rising
temperature predi ted by Debye, and observed19 by Laue, and showed
that the ee t was of the expe ted order. In the Bakerian Le ture in
1915 [6 he des ribed measurements in the intensity of a very perfe t
rystal, al ite, whi h seemed to show that the intensity was proportional
to the s attering power of the atomi planes and not to the square of the
power (this is to be expe ted from the formula for ree tion by a perfe t
rystal). In the same address he proposed the use of the Fourier method
of interpreting the measurements20 whi h has been re ently used with
su h su ess by Duane, Havighurst, and Compton, and whi h is dealt
with in the fourth se tion of this summary.21 At about the same time,
Debye and Compton independently dis ussed the inuen e of ele troni
distribution in the atom on the intensity of ree tion.
The next step was made by Compton [7 in 1917. Darwin's formula for
the mosai rystal was dedu ed by a dierent method, and was applied
to the interpretation of W. H. Bragg's results with ro ksalt. Compton
288
W. L. Bragg
on luded that the ele troni distribution in the atoms was of the type to
be expe ted from Bohr's atomi model. Compton then published the rst
measurements of the absolute intensity of ref1exion. A mono hromati
beam of X-rays was obtained by ree tion from a rystal, and this
was ree ted by a se ond rotating rystal (ro ksalt and22 al ite). The
absolute value of the integrated ree tion E
I was found to be of the right
order for ro ksalt when al ulated by the imperfe t rystal formula, but
to be very low for al ite indi ating strong extin tion or a wrong formula,
in the se ond ase.
In 1921 and 1922 I published with James and Bosanquet a series
of measurements on ro ksalt in whi h we tried to obtain a high a ura y. We made absolute measurements of intensity for the strongest
ree tions,a and ompared the weaker ree tions with them. Our main
ontributions in these papers were a more a urate set of measurements
of integrated ree tion for a large number of planes, and a method for
estimating and orre ting for the ee t of extin tion. As Darwin showed
in a paper in 1922 [9 on the theoreti al interpretation of our results,
we only su eeded in orre ting for extin tion of the kind he termed `se ondary' and not for `primary' extin tion.b Sin e then measurements by
Havighurst [10, by Harris, Bates and M Innes23 [11 and by Bearden [12
have been made on the ree ting power of powdered sodium hloride
when extin tion is absent. Their measurements have agreed with ours
very losely indeed, onrming one's faith in intensity measurements,
and showing that we were fortunate in hoosing a rystal for our examination where primary extin tion was very small. In the same papers
we tried to make a areful analysis of the results in order to nd how
mu h information about atomi stru ture ould be legitimately dedu ed
from them, and we published urves showing the ele tron distribution
in sodium and hlorine24 atoms.
In this dis ussion, I have refrained from any referen e to the question
of ree tion by `perfe t' rystals. The formula for ree tion by su h
rystals was rst obtained by Darwin, and has been arrived at independently by Ewald. The ree tion by su h rystals has been examined
amongst others by Bergen Davis25 and Stempel [13, and by Mark [14
and predi tions of the theory have been veried. It is not onsidered
a In our paper we failed to give due a knowledgement to Compton's absolute
measurements in 1917 of whi h we were not aware at the time.
b Primary extin tion is an ex essive absorption of the beam whi h is being ree ted
in ea h homogeneous blo k of rystal, se ondary extin tion a statisti al ex essive
absorption of the beam in the many small blo ks of a mosai rystal.
289
here, be ause I wish to onne the dis ussion to those ases where a
omparison of the intensity of in ident and ree ted radiation leads to
a urate quantitative estimates of the distribution of s attering matter.
This ideal an be attained with a tual rystals,26 when they are of the
imperfe t or mosai type, though allowan e for extin tion is sometimes
di ult in the ase of the stronger ree tions. On the other hand, it is
far more di ult to know what one is measuring in the ase of rystals
whi h approximate to the perfe t type. It is a fortunate ir umstan e
that mathemati al formulae an be applied most easily to the type of
imperfe t rystal more ommon in nature.
3. Results of quantitative analysis

For the sake of on iseness, only one of the many intensity formulae will
be given here, for it illustrates the essential features of them all. Let us
suppose that the integrated ree tion is being measured when X-rays
fall on the fa e of a rotating rystal of the mosai type. We then have
=
Q 1 + cos2 2
.
2
2
(a) is the ee tive absorption oe ient, whi h may be greater than the
normal oe ient, owing to the existen e of extin tion at the ree ting
angle.
2
2
is the `polarisation fa tor', whi h arises be ause
(b) The fa tor 1+cos
2
the in ident rays are assumed to be unpolarised.
( )
Q=
N e2
F
mc2
2
3
,
sin 2
where e and m are the ele troni onstants,27 c the velo ity of light,
the wavelength used, N the number of s attering units per unit volume,
and the glan ing angle.
(d) F is the quantity we are seeking to dedu e. It represents the s attering power of the rystal unit in the dire tion under onsideration,
measured in terms of the s attering power of a single28 ele tron a ording
to the lassi al formula of J. J. Thomson. It is dened by Ewald as the
Àtomfaktor'29 when it applies to a single atom.
290
W. L. Bragg
Formulae appli able to other experimental arrangements (the powder

method for instan e) are very similar, and ontain the same quantity Q.
Our measurements of ree tion thus lead to values of Q, and so of F ,
sin e all other quantities in the formulae are known. Measurements on
a given rystal yield a series of values for F , and all the information
that an be found out about this rystalline or atomi stru ture is
represented by these values. They are the same for the same rystal
whatever wavelength is employed (sin e F is a fun tion of sin ), though
of ourse with shorter wavelength we have the advantage of measuring
a mu h greater number of these oe ients (in reased resolving power).
At this stage the ee t of the thermal agitation of the atom will be
onsidered as inuen ing the value of F . If we wish to make dedu tions about atomi stru ture, the thermal agitation must be taken into
a ount. Allowan e for it is a ompli ated matter, be ause not only do
some atoms move more than others, but also they hange their relative
mean positions as the temperature alters in the more omplex rystals.
This will be dealt with more fully below.
A series of examples will now be given to show that these quantitative
formulae, when tested, lead to results whi h indi ate that the theory is
on the right lines. It is perhaps more onvin ing to study the results
obtained with very simple rystals, though I think that the su ess of
the theory in analysing highly omplex stru tures is also very strong
eviden e, be ause we have overed su h a wide range of substan es.
In the simple rystals, where the positions of the atoms are denite,
we an get the s attering power of individual atoms. The results should
both indi ate the orre t number of ele trons in the atom, and should
outline an atom of about the right size. When F is plotted against sin
its value should tend to the number, N , of ele trons in the atom for small
values of sin , and should fall away as sin in reases, at a rate whi h is
reasonably explained by the spatial extension of the atom. In Fig. 1, the
full lines give F urves obtained experimentally by various observers.
The dotted lines are F urves al ulated for the generalised atomi
model of Thomas [15, of appropriate atomi number. The Thomas
atomi model, whi h has been shown for omparison, is most useful
as it gives us the approximate ele troni distribution in an atom of any
atomi number. Thomas al ulates an ideal distribution of ele trons in
an atom of high atomi number. He assumes spheri al symmetry for
the atom, and supposes that èle trons are distributed uniformly in the
six-dimensional phase-spa e for the motion of an ele tron, at the rate
291
Fig. 1.
of two for ea h h3 of (six) volume'.30 He thus obtains an ideal ele tron
atmosphere around the nu leus, the onstants of whi h an be simply
adjusted31 so as to be suitable for any given nu lear harge. It is of ourse
to be expe ted that the lower the atomi weight, the more the a tual
distribution of s attering matter will depart from this arrangement, and
will ree t the idiosyn rasies of the parti ular atom in question. The
gure will show, however, that the a tual urves are very similar to those
al ulated for Thomas' models. In parti ular, it will be lear that they
tend to maximum values not far removed from the number of ele trons in
the atom in ea h ase. The general agreement between the observed and
al ulated F urves must mean that our measurements of F are outlining
a pi ture of the atom. The agreement holds also for other atomi models
than those of Thomas, whi h all lead to atoms with approximately the
same spatial extension and ele troni distribution, as is well known.
All these measurements of F ne essitate absolute values for the integrated ree tion. It is not ne essary to measure these dire tly in ea h
292
W. L. Bragg
ase. When any one ree tion has been measured in absolute value (by
omparison of in ident and ree ted radiation), other rystals may be
ompared with it. The standard whi h has been used in every ase, as
far as I am aware, is the ro ksalt rystal. Absolute measurements on this
have been made by Compton [7, by Bragg, James and Bosanquet [8,
and by Wasastjerna [18 whi h agree satisfa torily with ea h other.
4. Interpretation of measurements of F
In interpreting these measurements of s attering power, we may either
al ulate the s attering of a proposed atomi model and ompare it
with the observed F urve, or we may use the observations to al ulate
the distribution of s attering matter dire tly. The latter method is the
more attra tive, and in the hands of Duane, Havighurst, and Compton it has yielded highly interesting ìmages' of the atomi stru ture
seen by X-rays. There is a lose analogy between the examination of
a series of parallel planes by means of X-rays, and the examination of
a dira tion grating, by a mi ros ope, whi h is onsidered in Abbe's
theory of mi ros opi vision.a The obje tive of the mi ros ope may be
onsidered as re eiving a limited number of orders of spe tra from the
grating. These spe tra in their turn build up the image viewed by the
eyepie e, and the perfe tion of this image depends on the number of
spe tra re eived. The strength of ea h spe tral order depends on the
magnitude of the orresponding oe ient in that Fourier series whi h
represents the amplitude of the light transmitted at ea h point of the
grating. The extension of this well-known opti al prin iple to the X-ray
eld was suggested by W. H. Bragg [6 in 1915. He had formed the
on lusion32 that the amplitudes of the s attered wave from ro ksalt
were inversely proportional to the square of the order of ree tion, and
he showed that33 `the periodi fun tion whi h represents the density of
the medium must therefore be of the form34
const +
cos 4 xd
cos 2n xd
cos 2 xd
+
+
...
+
+ ... '
12
22
n2
and in this way built up a urve showing the periodi density of the
ro ksalt grating. The method was not applied, however, to the mu h
more a urate measurements whi h are now available until re ently,
when Duane and Havighurst showed how mu h ould be done with it.
a See for instan e the dis ussion of this theory and of A. B. Porter's experiments to
illustrate it in Wood's Opti s, Chapter VIII.
293
Duane independently arrived at a more general formula of the same type,

giving the density of s attering matter at any point in the whole rystal
as a triple Fourier series, whose oe ients depend on the intensity of
ree tion from planes of all possible indi es. Havighurst applied this
prin iple to our measurements of ro ksalt, and to measurements whi h he
has made on other rystals, and obtained a pi ture of the relative density
of s attering matter along ertain lines in these rystals. Compton made
the further step of putting the formulae in a form whi h gives the
absolute density of ele troni distribution (assuming the s attering to
be by ele trons obeying the lassi al laws). Compton gives a very full
dis ussion of the whole matter in his book X-rays and Ele trons.35 It
is not only an extremely attra tive way of making lear just what has
been a hieved by the X-ray analysis, but also the most dire t method of
determining the stru ture.
The formula for the distribution of s attering matter in parallel sheets,
for a rystal with a entre of symmetry, is given by Compton as follows
Pz =
2X
2nz
Z
+
.
Fn cos
a
a 1
a
Here z is measured perpendi ularly to the planes whi h are spa ed a

matter between planes
distan e a apart. Pz dz is the amount of s attering
Ra
at distan es z and z + dz , and Z (= 0 Pz dz ) is the total s attering
matter of the rystal unit. This is a simplied form of Duane's formula
for a Fourier series of whi h the general term is

2n1 x
2n2 y
2n3 z
An1 n2 n3 sin
n1 sin
n2 sin
n3 ,
a1
a2
a3
An1 n2 n3 being proportional to the amplitude of the s attered wave from
the plane (n1 n2 n3 ).
Another Fourier series, due to Compton, gives the radial distribution
of s attering matter, i.e. the values of Un where Un dr is the amount of
s attering matter between radii r and r + dr
Un =
8r X
2nr
,
nFn sin
a2 1
a
where a is hosen so that values of F o ur at onvenient intervals on

the graph for F .
If we know the values of F for a given atom over a su iently wide
range, we an build up an image of the atom either as a `sheet distribution' parallel to a plane, or as a radial distribution of s attering
294
W. L. Bragg
matter around the nu leus. In using these methods of analysis, however, it is very ne essary to remember that we are working right at the
limit of resolving power of our instruments, and in fa t are attempting
a more ambitious problem than in the orresponding opti al ase. In
A. B. Porter's experiments to test Abbe's theory, he viewed the image
of a dira tion grating and removed any desired group of dira ted rays
by utting them o with a s reen. The rst order gives blurred lines, four
or ve orders give sharper lines with a ne dark line down the entre,
eight orders give two dark lines down the entre of ea h bright line and
so forth. These imperfe t images are due to the absen es of the higher
members in building up the Fourier series. In exa tly the same way we
get false detail in our X-ray image, owing to ignoran e of the values of
the higher members in the F urve. Similarly, the ne stru ture whi h
a tually exists may be glossed over, sin e by using a wavelength of 0.7 ,
we annot hope to `resolve' details of atomi stru ture on a s ale of less
than half this value.
The ignoran e of the values of higher members of the Fourier series
matters mu h less in the urve of sheet distribution than in that for
radial distribution, sin e the latter onverges far more slowly. Examples
of the Fourier method of analysis are given in the next paragraph.
As opposed to this method of building up an image from the X-ray
results,36 we may make an atomi model and test it by al ulating an F
urve for it whi h an be ompared with that obtained experimentally.
This is the most satisfa tory method of testing models arrived at by
other lines of resear h, for nothing has to be assumed about the values
of the higher oe ients F . It is of ourse again true that our test only
applies to details of the proposed model on a s ale omparable with the
wavelength we are using. Sin e we an ree t X-rays right ba k from an
atomi plane, we may get a resolving power for a given wavelength with
the X-ray method twi e as great as the best the mi ros ope an yield.
It is perhaps worth mentioning the methods I used with James and
Bosanquet in our determination of the ele troni distribution in sodium
and hlorine in 1922. We tried to avoid extrapolations of the F urve
beyond the limit of experimental investigation. We divided the atom
arbitrarily into a set of shells, with an unknown number of ele trons
in ea h shell. These unknowns were evaluated by making the s attering
due to them t the F urve over the observed range, this being simply
done by solving simultaneous linear equations. We found we got mu h
the same type of distribution however the shells were hosen, and that a
limit to the ele troni distribution at a radius of about 1.1 in sodium
295
Fig. 2.
and 1.8 in hlorine was learly indi ated. Our distribution orresponds
in its general outline to that found by the mu h more dire t Fourier
analysis, as the examples in paragraph 7 will show.
5. Examples of analysis
We owe to Duane [20 the appre iation of the very attra tive way in
whi h the Fourier analysis represents the results of X-ray examinations.
It has the great merit of representing, in the form of a single urve,
the information yielded by all orders of ree tion from a given plane,
or from the whole rystal. It is of ourse only an alternative way of
interpreting the results, and the dedu tions we an make about atomi
or mole ular stru tures depend in the end on the extent to whi h we an
296
W. L. Bragg
Fig. 3a. Distribution of ele trons in sheets parallel to 0001.37

trust our experimental observations, and not on the method of analysis
we use. The Fourier method is so dire t however, and its signi an e so
easy to grasp, that Duane's introdu tion of it marks a great advan e in
te hnique of analysis.
I have reserved to paragraph 7 the more di ult problem of the arrangement of s attering matter in the atoms themselves, and the examples
given here are of a simpler hara ter. They illustrate the appli ation of
analysis to the general problem of the distribution of s attering matter in
the whole rystal, when we are not so near the limit of resolving power.
The urves in Fig. 2 represent the rst appli ation of the new method
of Fourier analysis to a urate data, arried out by Havighurst [21 in
1925. He used our determinations of F for sodium hloride, and Duane's
three-dimensional Fourier series, and al ulated the density of s attering
matter along a ube edge through sodium and hlorine entres, along a
ube diagonal through the same atoms, and along two fa e diagonals
hosen so as to pass through hlorine atoms alone or sodium atoms
alone in the rystal. The atoms show as peaks in the density distribution.
In the other examples, the formula for distribution in sheets has been
297
Fig. 3b. Distribution of ele trons in sheets parallel to 1010.39
applied to some results we have obtained in our work on rystal stru ture
at Man hester. I have given them be ause I feel they are onvin ing
eviden e of the power of quantitative measurements, and show that all
methods of interpretations lead to the same results.
Mr West and I [22 re ently analysed the hexagonal rystal beryl,
Be3 A12 Si6 O18 ,38 whi h has a stru ture of some omplexity, depending
on seven parameters. We obtained the atomi positions by the usual
method of analysis, using more or less known F urves for the atoms in
the rystal, and moving them about till we explained the observed F s due
to the rystal unit. Fig. 3 shows the reinterpretations of this result by the
Fourier method. Fig. 3a gives the ele tron density in sheets perpendi ular
to the prin ipal axis of the rystal, whi h is of a very simple type. The
parti ular point to be noted is the orresponden e between the position
of the line B in the gure and the hump of the Fourier analysis. The line
B marks the position of a group of oxygen atoms whi h lies between two
other groups A and C xed by symmetry, the position of B being xed
by a parameter found by familiar methods of rystal analysis. The hump
represents the same group xed by the Fourier analysis, and it will be
seen how losely they orrespond. In Figs. 3b and 3 more omplex sets
of planes are shown. The dotted urve represents the interpretation of
our results by Fourier analysis. The full urve is got by adding together
the humps due to the separate atoms shown below, the position of these
having been obtained by our X-ray analysis and their sizes by the aid
of the urve in Fig. 3a in whi h the ontribution of the atoms an be
separated out. The orresponden e between the two shows that the older
methods and the Fourier analysis agree. It is to be noted that the rystal
298
W. L. Bragg
Fig. 3 . Distribution of ele trons in sheets parallel to 1120.42
had rst to be analysed by the older methods, in order that the sizes of
the Fourier oe ients might be known.
In Fig. 4 I have given a set of urves for the alums, re ently analysed by
Professor Cork [23. The alums are ompli ated ubi rystals with su h
formulae as KAl(SO4 )2 .12 H2 O. Wy ko40 has shown that the potassium
and aluminium atoms41 o upy the same positions in the ubi ell
as the sodium atom in ro ksalt. Now we an repla e the potassium
by ammonium, rubidium, aesium, or thallium, and the aluminium by
hromium, or other trivalent metals. Though the positions of the other
atoms in the rystals are not yet known, they will presumably be mu h
the same in all these rystals. If we represent by a Fourier series the
quantitative measurements of the alums, we would expe t the density of
s attering matter to vary from rystal to rystal at the points o upied
by the metal atoms, but to remain onstant elsewhere. The urves show
this in the most interesting43 way.
299
Fig. 4.
The ee t of heat motion on the movements of the atoms has already
been mentioned. It was rst treated theoreti ally by Debye [4. Re ently Waller [24 has re al ulated Debye's formula, and has arrived at a
modied form of it. Debye found that the intensities of the interferen e
maxima in a simple rystal should be multiplied by a fa tor eM , where
M=
x=
6h2 (x) sin2

,
k x
2
characteristic temperature of crystal

=
.
T
absolute temperature
Without going into further detail, it is su ient to note that Waller's
formula diers from Debye's by making the fa tor e2M , not eM . James
and Miss Firth [25 have re ently arried out a series of measurements
300
W. L. Bragg
Fig. 5.
for ro ksalt between the temperatures 86 abs. and 900 abs. They nd
that Waller's formula is very losely followed up to 500 abs., though
at higher temperatures the de line in intensity is even more rapid, as
is perhaps to be expe ted owing to the rystal be oming more loosely
bound. I have given the results of the measurements in Figs. 5 and 6,
both as an example of the type of information whi h an be got from
X-ray measurements, and be ause these a tual gures are of interest as
a set of areful and a urate measurements of s attering power.
Fig. 5 shows the F urves for sodium and hlorine at dierent temperatures. The rapid de line in intensity for the higher orders will be
realised when it is remembered that they are proportional to F 2 . The
301
Fig. 6.
urve for absolute zero is an extrapolation from the others, following the
Debye formula as modied by Waller.
In Fig. 6 the same results are interpreted by the Fourier analysis.
The urve at room temperatures for NaCl is pra ti ally identi al with
the interpretation of our earlier gures by Compton, in his book X-rays
and Ele trons,44 though the gures on whi h it is based should be more
a urate.45 The urves show the manner in whi h the sharply dened
peaks due to Cl and Na at low temperatures be ome diuse owing to
heat motion at the higher temperatures.
Several interesting points arise in onne tion with this analysis. In
the rst pla e, James and Firth46 nd that the heat fa tor is dierent
for sodium and hlorine, the sodium atoms moving with greater average
amplitudes than the hlorine atoms. This has a very interesting bearing
on the rystal dynami s whi h is being further investigated by Waller.
To a rst approximation both atoms are ae ted equally by the elasti
waves travelling through the rystal, but in a further approximation it
an be seen that the sodium atoms are more loosely bound than the
302
W. L. Bragg
Fig. 7.
hlorine atoms. If an atom of either kind were only xed in position
by the six atoms immediately surrounding it, Waller has shown that
there would be no dieren e between the motions of a sodium atom
between six hlorine atoms, or a hlorine atom between six sodium
atoms. However, the hlorine is more rmly pinned in position be ause
it has in addition twelve large hlorine neighbours, whereas the sodium
atom is mu h less inuen ed by the twelve nearest sodium atoms. Hen e
arises the dieren e in their heat motions. It is important to nd the
orre t method for redu ing observations to absolute zero, and this
dieren e in heat motion must be satisfa torily analysed before this is
possible.
In the se ond pla e, the a ura y whi h an be attained by the experimental measurements holds out some hope that we may be able to
test dire tly whether there is zero-point energy47 or not. This is being
investigated by James and Waller. If a reliable atomi model is available,
it would seem that the measurements an tell whether there is vibration
at absolute zero or not, for the theoreti al diminution in intensity due to
the vibration is mu h larger than the experimental error in measuring
303
F . I feel onsiderable diden e in speaking of the question of zero-point

energy, and would like to have the advi e of the mathemati al physi ists
present.
We may al ulate, either from the measured heat fa tor or dire tly
from the Fourier analyses, the average amplitude of vibration for dierent temperatures. James and Firth nd by both methods, for instan e,
that at room temperature the mean amplitude of vibration for both
atoms is 0.21 , and at 900 abs. it is about 0.58 . They examined the
form whi h the Fourier urve at 0 abs. assumes when it is deformed
by supposing all the atoms to be in vibration with the same mean
amplitude.
It has been already remarked that the observed F urves for atoms
are very similar to those al ulated for the Thomas atomi model. The
same omparison may be made between the distributions of s attering
matter. In Fig. 7 the distribution in sheets for NaCl at absolute zero is
shown as a full urve. The dotted urve shows the horizontal distribution
in sheets for atoms of atomi number 17 and 11. In Thomas' model the
density rises towards an innite value very lose to the nu leus, and this
is represented by the very sharp peaks at the atomi entres in the dotted
urve. We would not expe t the observed distribution to orrespond to
the a tual Thomas distribution at these points. Throughout the rest of
the rystal the distribution is very similar. The omparison is interesting,
be ause it shows how deli ate a matter it is to get the ne detail of
atomi stru ture from the observations. Thomas' distribution is quite
ontinuous and takes no a ount of K, L and M sets of ele trons. The
slight departures of the observed urve from the smooth Thomas urve
represent the experimental eviden e for the existen e of all the individual
features of the atom.
6. The me hanism of X-ray s attering

Before going on to dis uss the appli ation of the analysis to atomi
stru ture, it is ne essary to onsider what is being measured when a
distribution of s attering matter is dedu ed from the X-ray results. The
lassi al treatment regards the atom as ontaining a number of ele trons,
ea h of whi h s atters radiation a ording to the formula of J. J. Thomson. Sin e a vast number of atoms ontribute to the ree tion by a
single rystal plane, we should obtain a pi ture of the average ele troni
distribution. The quantity F should thus tend to a maximum value, at
small angles of s attering, equal to the number of ele trons in the atom,
304
W. L. Bragg
and should fall away owing to their spatial distribution as sin in reases.
The observed48 F urves are of this hara ter, as has been seen. When
interpreted as an atomi distribution, they give atoms ontaining the
orre t number of ele trons, and this seems satisfa tory from the lassi al
viewpoint. On the other hand, the eviden e of the Compton ee t would
appear at rst sight to ast doubt on the whole of our analysis. What
we are measuring is essentially the oherent radiation dira ted by the
rystal, whereas the Compton ee t shows that a part of the radiation
whi h is s attered is of dierent wavelength. Further, this radiation of
dierent wavelength is in luded with the oherent radiation, when the
total amount of s attered radiation is measured, and found to agree
under suitable onditions with the amount predi ted by J. J. Thomson's
formula. It would therefore seem wrong to assume that we obtain a
true pi ture of ele troni distribution by the aid of measurements on the
oherent radiation alone.
Even before the advent of the new me hani s, Compton's original
treatment of the ee t whi h he dis overed suggested a way out of this
di ulty. The re oil ele tron is given an amount of energy
2
h2
m
sin
2
where and are the frequen ies of the modied and unmodied
radiations. If the ele tron is eje ted from the atom the radiation is
modied in wavelength, if not oherent waves are s attered. Sin e there
is little modied s attering at small angles, the F urve will tend to
a maximum equal to the number of ele trons in the atom, and any
interpretation of the urve will give an atom ontaining the orre t
number of ele trons. As sin in reases, more and more of the s attered
radiation will be modied, and in al ulating the F urve this must be
taken into a ount. However, if is not far from unity, the F urve
will remain a fun tion of sin , sin e whatever riterion is applied for
the s attering of modied or unmodied radiation, it will depend on the
energy imparted to the s attering ele tron, whi h is itself a fun tion of
sin
. Our X-ray analysis would thus give us an untrue pi ture of the
atom, but one whi h is onsistently the same whatever wavelength is
employed. Williams [27 and Jaun ey [28 have re al ulated F urves
from atomi models using this riterion, and found a better t to the
experimental urves when the Compton ee t was taken into a ount.
(Examples of this loser approximation will be found in the paper by
305
Fig. 8.
Williams [27 in 1926. See also a dis ussion by Kallmann and Mark
[26).49
The point at issue is illustrated by the urves in Fig. 8. Three F urves
for hlorine are plotted in the gure. The dotted line represents the
observed F urve (James and Firth). The ontinuous line is the F urve
al ulated from Hartree's [29 atomi model for hlorine. It shows a hump
at a value of sin of 0.4, whi h is not present in the observed urve. This
hump arises from the fa t that the outer ele trons in the hlorine model
give negative values for F just short of this point,50 and positive values
306
W. L. Bragg
again at the point itself. All atomi models al ulated with ele troni
orbits show similar irregularities whi h are not a tually observed. When,
however, the Compton ee t is taken into onsideration, these outer
ele trons are found to give a very small ontribution to the F urve at
the large angles where the humps51 o ur, be ause they s atter so mu h
modied radiation. The allowan e for the Compton ee t smooths out
the hump, and leads to F urves mu h more like those observed. The
third urve shows the F urve due to the ontinuous Thomas distribution
and is a lose t to the observed urve.
I have quoted from a note by Dr Ivar Waller, in the following tentative
summary of the interpretation whi h the new me hani s gives us of this
phenomenon.a In a re ent letter to Nature [30, Waller dis usses the
transition for the whole range from ordinary dispersion into Compton
ee t. His note only refers to s attering by a single ele tron, but it an
probably be extended to many-ele tron atoms. Waves of ontinually
de reasing wavelength are supposed to fall upon the atom, and the
transition is tra ed through the following stages.
a) While the wavelength of the radiation remains long ompared with
atomi dimensions, the dispersion formula for opti al frequen ies gradually transforms into the s attering for free ele trons given by the
lassi al J. J. Thomson formula. This formula holds approximately to52
wavelengths approa hing atomi dimensions.
b) At this point the s attering of oherent radiation will diminish, owing
to interferen e, and be ome more on entrated in the forward dire tion
of the in ident light. This is the phenomenon we are studying, with Xrays, and our F urves map out the distribution of the oherent radiation
where the wavelength is of atomi dimensions.
) At the same time, the s attering of in oherent radiation will be ome
appre iable, and approximate more and more losely in hange of wavelength and intensity distribution to the Compton ee t. It will have
pra ti ally merged into the Compton ee t when the momentum of a
quantum of the in ident light is large ompared with that orresponding
to ele troni motions in the atom.
d) Up to this point the Thomson formula holds for the total intensity
a Spa e forbids a referen e to the many theoreti al papers whi h have ontributed
towards this interpretation.
307
of light s attered in any dire tion, oherent and in oherent radiation

being summed together. It rst eases to hold, when the frequen y
displa ement due to the Compton ee t is no longer small ompared
with the frequen y of the in ident light.
The point of importan e for our present problem is that `the oherent
part of the radiation is to be dire tly al ulated from that ontinuous
distribution of ele tri ity whi h is dened by the S hrdinger densitydistribution in the initial state of the atom'. The lassi al treatment
supposes ea h point ele tron to s atter a ording to the J. J. Thomson
formula in all dire tions. In the new treatment, the ele tron is repla ed
by a spatial distribution of s attering matter, and so ea h ele tron has
an `F urve' of its own. It will still s atter oherent radiation in all
dire tions, but its amount will fall away from that given by the lassi al
formula owing to interferen e as sin in reases, and this de line will be
mu h more rapid for the more diuse outer ele trons than for the on entrated inner ele trons. The total amount of radiation T s attered in any
dire tion by the ele tron is given by the Thomson formula. A fra tion
f 2 T will be oherent, and will be al ulated by the laws of interferen e
from the S hrdinger distribution, and the remainder, (1 f 2 )T , will
be in oherent. Thus the total oherent radiation will be F 2 T where
F is al ulated from the S hrdinger distribution for the whole atom.
P
An amount (N f 2 )T will be s attered with hange of wavelength.
Our measurements of X-ray dira tion, if this be true, an be trusted
to measure the S hrdinger ontinuous distribution of ele tri ity in the
rystal latti e.
A very interesting point arises in the ase where hara teristi absorption frequen ies of the s attering atom are of shorter wavelength than
the radiation whi h is being s attered. In general, this has not been
so when areful intensity measurements have been made sin e atoms of
low atomi weight have alone been investigated. On the lassi al analogy,
we would expe t a reversal in phase of the s attered radiation, when an
ele tron has a hara teristi frequen y greater than that of the in ident
light. A fas inating53 experiment by Mark and Szilard [31 has shown
that something very like this takes pla e. They investigated the (111)
and (333) ree tions of RbBr, whi h are extremely weak be ause Rb and
Br oppose ea h other and are nearly equal in atomi number. They found
that these `forbidden' ree tions were indeed absent when the soft CuK
or hard BaK radiation was used, but that SrK radiation was appre iably
ree ted (SrK 0.871 ;54 absorption edges of RbK and BrK , 0.814
308
W. L. Bragg
and 0.918 ). The atoms are dierentiated be ause a reversal of phase

in s attering by the K ele trons takes pla e in the one ase and not in
the other.
7. The analysis of atomi stru ture by X-ray intensity

measurements
It has been seen that the intensity measurements assign the orre t
number of ele trons to ea h atom in a rystal, and indi ate a spatial
extension of the atoms of the right order. In attempting to make the
further step of dedu ing the arrangement of the ele trons in the atom,
the limitations of the method begin to be very apparent.
In all ases where analysis has been attempted, the atom has been
treated as spheri ally symmetri al. The analysis is used to determine
the amount of s attering matter Un dr between radii r and r + dr. All
methods of analysis give a distribution of the same general type. I have
given, for instan e, a series of analyses of sodium and hlorine in Fig. 9.
In these gures, Un is plotted as ordinate against r as abs issa. The
total area of the urve
R in ea h ase is equal to the number of ele trons
in ea h atom, sin e 0 Un dr = N . The full-line urves are our original
interpretations of the distribution in sodium and hlorine, based on our
1921 gures.55 The other urves are the interpretations of the same
or losely similar sets of gures57 by Havighurst [32 and by Compton
(X-rays and Ele trons) using the Fourier method of analysis.
In Fig. 9a are in luded our analysis of sodium in NaCl, two analyses
by Havighurst of sodium in NaCl and NaF obtained by using Duane's
triple Fourier series, and an analysis of our gures58 by Compton using
the Fourier formula for radial distribution. It will be seen that the
general distribution of s attering matter and the limits of the atom are
approximately the same in ea h ase. The same holds for the hlorine
urves in Fig. 9b.
The interesting point whi h is raised is the reality of the humps whi h
are shown by the Fourier analysis. We obtained similar humps in our
analysis by means of shells but doubted their reality be ause we found
that if we smoothed them out and re al ulated the F urve, it agreed
with the observed urve within the limits of experimental error. The
te hnique of measurement has greatly improved sin e then, and it would
even appear from later results that we over-estimated the possible errors
of our rst determinations of F . It is obvious, however, that great are
must still be taken in basing on lusions on the ner details shown by any
309
Fig. 9.56
method of analysis. The formula whi h is used in the Fourier analysis,
Un =
2nr
4r X 2nFn
sin
,
a 1
a
a
is one whi h onverges very slowly, sin e the su essive oe ients Fn
are multiplied by n. The observed F urve must be extrapolated to a
point when F is supposed to fall to zero, and the pre ise form of the
urve rea ts very sensitively to the way in whi h this extrapolation is
arried out.
310
W. L. Bragg
Fig. 10.
The urves in Fig. 10 will illustrate the extent to whi h the analysis
an be onsidered to give us information about the a tual atomi distribution. In Fig. 10a the urve shows the F values for uorine obtained by
James and Randall [17. The ir les are points obtained59 by Havighurst
from measurements on CaF, LiF, NaF;60 it will be seen that the two sets
of experimental data are in very satisfa tory agreement. In Fig. 10b I
have shown on the one hand Havighurst's interpretations of the F urve
drawn through his points, and on the other an analysis arried out by
311
Fig. 11.62
Claassen [16 of James and Randall's using the Fourier method. The
distributions are the same in their main outlines, but the peaks o ur in
quite dierent pla es.
Compton (X-rays and Ele trons, p. 167) in dis ussing his diagrams of
radial distribution has remarked that slight dieren es in the F urves
lead to wide dieren es in details of the urves, and that too mu h
onden e should not be pla ed on these details. Havighurst [32 dis usses the signi an e of the analysis very fully in his paper on ele tron
distribution in the atoms. Our data are not yet su iently a urate or
extensive. Nevertheless, we are so near to attaining an a ura y of a
312
W. L. Bragg
satisfa tory order, and the results of the analysis seem to indi ate so
learly its fundamental orre tness, that it appears to be well worth
while to pursue enquiry further. Work with shorter wavelengths, and at
low temperatures, when heat motion is small and a large range of F
values an be measured, should yield us a urate pi tures of the atomi
stru ture itself. Given a urate data,61 the Fourier method of analysis
provides a dire t way of utilising them.
The radial distribution of s attering power outlined in this way is in
general agreement with any reasonable atomi model. We have seen,
in parti ular, that the F urves, and therefore the radial distributions,
of Thomas' model63 are in approximate a ord with those a tually observed. If it is true that the s attering of oherent radiation is to be
al ulated in all ases by the S hrdinger density distribution, we should
test our model against this distribution.
An interesting attempt along these lines has been re ently made by
Pauling [33. He has used ertain simplifying assumptions to obtain an
approximate S hrdinger density-distribution for many-ele tron atoms.
I have shown in Fig. 11 four sets of urves. The radial ele tron distributions dedu ed by Havighurst and by Compton are shown as one urve
sin e they are very similar. The gure shows also our rst analysis of
ele tron distribution. Mat hed against these are plotted the generalised
distribution of the Thomas model, and the S hrdinger density distribution al ulated by Pauling.
We have obviously not yet rea hed a point when we an be satised
with the agreement between theory and experiment, yet the su ess
attained so far is a distin t en ouragement to further investigation.
8. The refra tion of X-rays

At Professor Lorentz's64 suggestion I have added a very brief note on the
refra tion of X-rays, sin e the phenomenon is so intimately onne ted
with the question of intensity of ree tion and s attering, and is another
example of the su essful appli ation of lassi al laws. The dira tion
phenomenon dealt with above (intensity of ree tion) arises from the
s attering of oherent radiation in all dire tions by the atoms of a rystal.
The refra tive index may be onsidered as being due to the s attering
in the forward dire tion of oherent radiation, whi h interferes with the
primary beam. The arrangement of the s attering matter plays no part,
so that the body may be rystalline or amorphous. The measurement of
313
the refra tive index is thus a dire t measure of the amount of oherent
radiation s attered in the forward dire tion of the in ident beam.
1. Darwin [2 appears to have been rst in pointing out that theory
assigns a refra tive index for X-rays diering from unity by about one
part in a million. He predi ted that a very slight departure from the law
of ree tion
n = 2d sin 0
would be found, the a tual angle being given by Darwin's formula

1
.
sin cos
Ewald's [34 independent treatment of X-ray ree tion leads to an equivalent result, though the problem is approa hed along quite dierent
lines.
As is well known, the rst experimental eviden e of an index of refra tion was found in a departure from the ree tion laws. Stenstrm [35
A) as
observed dieren es in the apparent wavelength of soft X-rays (3
measured in the dierent orders, whi h were explained by Ewald's laws
of X-ray ree tion. The in reased a ura y of X-ray spe tros opy has
shown that similar deviations from the simple law of ree tion exist for
harder rays, though the deviations are mu h smaller than in the ordinary
X-ray region.65 Thus the deviations have been dete ted for hard rays
by Duane and Patterson [36 and by Siegbahn and Hjalmar [37. It is
di ult to measure the refra tive index by means of these deviations in
the ordinary way, sin e they are so small, but Davis [38, 39 developed
a very ingenious way of greatly in reasing the ee t. A rystal is ground
so that the rays ree ted by the atomi planes enter or leave a fa e at a
very ne glan ing angle, and thus suer a omparatively great dee tion.
Compton [40 dis overed the total ree tion of X-rays, and measured
the index of refra tion in this way. The refra tive index is slightly less
than unity, hen e X-rays falling at a very ne glan ing angle on a plane
surfa e of a body are totally ree ted, none of the radiation passing
into the body. Compton showed that, although the refra tive index is so
nearly unity, yet the riti al glan ing angle is quite appre iable.
Finally, the dire t ee t of refra tion by a prism has been observed
by Larsson, Siegbahn and Waller [41. X-rays entered one fa e of a glass
prism at a very ne glan ing angle, and suered a measurable dee tion.
They obtained in this way a dispersion spe trum of X-rays.
0 =
314
W. L. Bragg
2. In all ases where the frequen y of the X-radiation is great ompared with any frequen y hara teristi of the atom, the refra tive index
measured by any of these methods is in lose a ord66 with the formula
ne2
,
2m 2
where n is the number of ele trons per unit volume in the body, e
and m are the ele troni onstants, and the frequen y of the in ident
radiation. The formula follows dire tly from the lassi al Drude-Lorentz
theory of dispersion, in the limiting ase where the frequen y of the
radiation is large ompared with the `free periods' of the ele trons in the
atom. It an be put in the form [4267
1=
1 = 2.71 106
Z 2
,
A
where is the wavelength in ngstrm units of the in ident radiation,

the density of the substan e, Z and A the average atomi number and
atomi weight of its onstituents (for all light atoms Z/A is very nearly
0.5).68 Expressed in this form, the order of magnitude of 1 is easily
grasped. The riti al glan ing angle for total ree tion is given by
cos = ,
when e
=
ne2
.
m 2
Expressing in minutes of ar , and in ngstrm units as before,

r
Z
.
= 8.0
A
Measurements of refra tive index have been made by Compton and by
Doan using the method of total ree tion, by Davis, Hatley and Nardro
using ree tion in a rystal, and by Larsson, Siegbahn and Waller with
a prism. A variety of substan es has been examined, and wavelengths
between 0.5 and 2 have been used. The a ura y of the experimental
determination of 1 is of the order of one to ve per ent. As long as
the riti al frequen ies of the atom have not been approa hed, the results
have agreed with the above formula within experimental error. Just as in
the measurements of intensity of ree tion the F urves approa h a limit
at small angles equal to the number of ele trons in the atom, so these
measurements of refra tive index when interpreted by lassi al theory
lead to a very a urate numbering of the ele trons in the s attering units.
315
3. A highly interesting eld is opened up by the measurements of refra tive index for wavelengths in the neighbourhood of a riti al frequen y
of the atom. It is a striking fa t that the simple dispersion formula
n
1=
e 2 X ns
2m 1 s2 2
still gives values for the refra tive index agreeing with experiment in
this region, ex ept when the riti al frequen y is very losely approa hed
indeed. Davis and von Nardro ree ted CuK and CuK X-rays69 from
iron pyrites, and found that the refra tive indi es ould be reprodu ed by
substituting onstants in the formulae orresponding to two K ele trons
in iron with the frequen y of the K absorption edge.70 R. L. Doan [44 has
re ently made a series of measurements by the total ree tion method.
His a urate data support the on lusion that the Drude-Lorentz theory
of dispersion represents the fa ts, `not only in regions remote from the
absorption edge,71 but also in some instan es in whi h the radiation
approa hes the natural frequen ies of ertain groups of ele trons'. The
existen e of two K ele trons72 is very denitely indi ated. Kallmann and
Mark [43 have gone more deeply into the form of the dispersion urve in
the neighbourhood of the riti al frequen ies. The hange in s attering
power of an atom as the frequen y of the s attered radiation passes
through a riti al value is of ourse another aspe t of this anomalous
dispersion; the experiment of Mark and Szilard whi h showed this ee t
has been des ribed above. There is ample eviden e that measurements
of refra tive index will in future prove to be a most fruitful means of
investigating the response of the atom to in ident radiation of frequen y
very near ea h of its own hara teristi frequen ies.
316
W. L. Bragg
Referen esa
[1 [W. Friedri h, P. Knipping and M. v. Laue, Bayr. Akad. d. Wiss.
Math. phys. Kl. (1912), 303.
[2 C. G. Darwin, Phil. Mag., 27 (1914), 315, 675.
[3 P. P. Ewald, Ann. d. Phys., 54 ([1917), 519.
[4 P. Debye, Ann. d. Phys., 43 (1914), 49.
[5 W. H. Bragg, Phil. Mag., 27 (1914), 881.
[6 W. H. Bragg, Phil. Trans. Roy. So . [A, 215 (1915), 253.
[7 A. H. Compton, Phys. Rev., 9 (1917), 29; 10 (1917), 95.
[8 W. L. Bragg, R. W. James and C. H. Bosanquet, Phil. Mag., 41
(1921), 309; 42 (1921), 1; 44 (1922), 433.
[9 C. G. Darwin, Phil. Mag., 43 (1922), 800.
[10 R. J. Havighurst, Phys. Rev., 28 (1926), n. 5, 869 and 882.
[11 L. Harris, S. J. Bates and D. A. Ma Innes,73 Phys. Rev., 28 (1926),
235.
[12 J. A. Bearden, Phys. Rev., 27 (1926), 796; 29 (1927), 20.
[13 Bergen Davis and W. M. Stempel, Phys. Rev., [17 (1921), 608.
[14 H. Mark, Naturwiss., 13 (1925), n. 49/50, 1042.
[15 L. H. Thomas, Pro . Camb. Phil. So ., 23 (1927), 542.
[16 A. Claassen, Pro . Phys. So . London, 38 [pt 5 (1926), 482.
[17 R. W. James and J. T. Randall, Phil. Mag. [(7), 1 (1926), 1202.
[18 J. A. Wasastjerna, Comm. Fenn., 2 (1925), 15.
[19 W. L. Bragg, C. G. Darwin and R. W. James, Phil. Mag. (7), 1
(1926), 897.
[20 W. Duane, Pro . Nat. A ad. S i., 11 (1925), 489.
[21 R. J. Havighurst, Pro . Nat. A ad. S i., 11 (1925), 502.
[22 W. L. Bragg and J. West, Roy. So . Pro . A, 111 (1926), 691.
[23 [J. M. Cork, Phil. Mag. [(7), 4 (1927), 688.
[24 I. Waller, Upsala Univ. rsskr. 1925, [11; Ann. d. Phys., 83 (1927),
153.
[25 R. W. James and E. Firth, Roy. So . Pro . [A, 117 (1927), 62.
[26 [H. Kallmann and H. Mark, Zeit. f. Phys., 26 (1926), [n. 2, [120.
[27 E. J. Williams, Phil. Mag. [(7), 2 (1926), 657.
[28 [G. E. M. Jaun ey, Phys. Rev., 29 (1927), 605.
[29 D. R. Hartree, Phil. Mag., 50 (1925), 289.
[30 I. Waller, Nature, [120 (July 1927), [155.
[31 H. Mark and L. Szilard, Zeit. f. Phys., 33 (1925), 688.
a The style of the referen es has been modernised and uniformised (eds.).
317
[32 R. J. Havighurst, Phys. Rev., 29 (1927), 1.

[33 L. Pauling, Roy. So . Pro . A, 114 (1927), 181.
[34 P. P. Ewald, Phys. Zeits h., 21 (1920), 617; Zeits hr. f. Physik, 2
(1920), 332.
[35 W. Stenstrm, Exper[imentelle Unters[u hungen d[er Rntgenspektra. Dissertation, Lund (1919).
[36 [W. Duane and [R. A. Patterson, Phys. Rev., 16 (1920), [526.74
[37 M. Siegbahn, Spektroskopie der Rntgenstrahlen [(Berlin: Springer,
1924).
[38 [C. C. Hatley and Bergen Davis, Phys. Rev., 23 (1924), 290.75
[39 B[ergen Davis and R. von Nardro, Phys. Rev., 23 (1924), 291.
[40 A. H. Compton, Phil. Mag., 45 (1923), 1121.
[41 [A. Larsson, [M. Siegbahn and [I. Waller, Naturwiss., 12 ([1924),
1212.
[42 I. Waller, [Theoretis he Studien zur Interferenz- und Dispersionstheorie der Rntgenstrahlen. [Dissertation, Upsala (1925).
[43 H. Kallmann and H. Mark, Ann. d. Physik, 82 (1927), 585.
[44 R. L. Doan, Phil. Mag. [(7), 4 n. 20 (1927), [100.
A very omplete a ount of work on intensity of ree tion is given
by Compton in X-rays and Ele trons and by Ewald in volume 24 of the
Handbu h der Physik by H. Geiger and K. S heel,76 Aufbau der festen
Materie und seine Erfors hung dur h Rntgenstrahlen, se tion 18.
318
W. L. Bragg
Dis ussion of Mr Bragg's report
Mr Debye. To what extent an you on lude that there exists an

energy at absolute zero?
Mr Bragg.
Waller and James have re ently submitted a paper

to the Royal So iety in whi h they dis uss the relation between the
inuen e of temperature on the intensity of ree tion (Debye ee t) and
the elasti onstants of a rystal. Using the experimentally determined
value of the Debye oe ient, they dedu e the s attering by an atom at
rest from the s attering by the atom at the temperature of liquid air (86
abs.). The urve dedu ed for the s attering by a perfe tly motionless
atom an of ourse take two forms, a ording to whether or not, in
interpreting the results of the experiment, one assumes the existen e of
an energy at absolute zero.
If one assumes the existen e of su h an energy, the urve dedu ed
from the experimental results agrees with that al ulated by Hartree
by applying S hrdinger's me hani s. The agreement is really very good
for sodium as well as for hlorine. On the other hand, the urve that
one obtains if one does not assume any energy at absolute zero deviates
onsiderably from the al ulated urve by an amount that ex eeds the
possible experimental error.
If these experimental resultsa are onrmed by new experiments, they
provide a dire t and onvin ing proof of the existen e of an energy at
absolute zero.
Mr Debye.
Would the ee t not be larger if one did the experiments with diamond?
Mr Bragg.
In the ase of diamond, it is di ult to interpret

the results obtained using a single rystal, be ause the stru ture is very
perfe t and the èxtin tion' is strong. One would have to work with
diamond powder. But I annot say if it would be easy to nd that there
exists an energy at absolute zero in diamond; I should onsider it further.
Mr Fowler.
Here is how Hartree al ulates the atomi elds.

Starting from Thomas' atomi eld, taken as a rst approximation, he
a Note added 5 April 1928. The results to whi h allusion is made here have just
been published in detail by Messrs James, Waller and Hartree in a paper entitled:
Àn investigation into the existen e of zero-point energy in the ro k-salt latti e by
an X-ray dira tion method' (Pro . Roy. So . A, 118 (1928), 334).
Dis ussion
319
al ulates the S hrdinger fun tions for an ele tron pla ed in this eld,
then the density of harge in the atom orresponding to the S hrdinger
fun tions, and then the orresponding atomi eld, whi h will dier
from that of Thomas. By su essive approximations one modies the
eld until the al ulations yield the eld whi h served as a starting
point. This method gives very good values for the levels orresponding
to X-rays and to visible light, and leads to the atom that Mr Bragg
onsidered for omparison with experiments.
Mr Heisenberg. How an you say that Hartree's method gives
exa t results, if it has not given any for the hydrogen atom? In the ase
of hydrogen the S hrdinger fun tions must be al ulated with the aid
of his dierential equation, in whi h one introdu es only the ele tri
potential due to the nu leus. One would not obtain orre t results if
one added to this potential the one oming from a harge distribution
by whi h one had repla ed the ele tron. One may then obtain exa t
results only by taking the harge density of all the ele trons, ex ept
the one whose motion one wishes to al ulate. Hartree's method is
ertainly very useful and I have no obje tion to it, but it is essentially
an approximation.
Mr Fowler. I may add to what I have just said that Hartree is
always areful to leave out the eld of the ele tron itself in ea h state, so
that, when he onsiders an L ele tron, for example, the entral part of
the eld of the whole atom is diminished by the eld of an L ele tron, as
far as this may be onsidered as entral. Hartree's method would then be
entirely exa t for hydrogen and in fa t he has shown that it is extremely
lose to being exa t for helium. (One nds a re ent theoreti al dis ussion
of Hartree's method, by Gaunt, in Pro . Cambr. Phil. So ., 24 (1928),
328.)
Mr Pauli.
In my opinion one must not perform the al ulations,

2
as in wave me hani s, by onsidering a density |(x, y, z)| in threedimensional spa e,77 but must onsider a density in several dimensions
|(x1 , y1 , z1 , ..., xN , yN , zN )|2 ,
whi h depends on the N parti les in the atom. For su iently short
waves the intensity of oherent s attered radiation is then proportional
320
W. L. Bragg
to78
Z
...
Z X
N
1
2i
, r
2
(n
u
d
k)
|(x1 , ..., zN )| dx1 ... dzN ,
n
where is the wavelength of the in ident radiation,
u a unit ve tor in
the dire tion of propagation, and nd the orresponding unit ve tor for
the s attered radiation; the sum must be taken over all the parti les.
The result that one obtains by assuming a three-dimensional density
annot be rigorously exa t; it an only be so to a ertain degree of
approximation.
Mr Lorentz.
How have you al ulated the s attering of radiation

by a harge distributed over a region omparable to the volume o upied
by the atom?
Mr Bragg.
To interpret the results of observation as produ ed by

an average distribution of the s attering material, we applied J. J. Thomson's lassi al formula for the amplitude of the wave s attered by a single
ele tron.
Mr Compton.
If we assume that there is always a onstant

ratio between the harge and mass of the ele tron, the result of the
lassi al al ulation of ree tion by a rystal is exa tly the same, whether
the harge and mass are assumed on entrated in parti les (ele trons)
or distributed irregularly in the atom. The intensity of ree tion is
determined by the average density of the ele tri harge in dierent
parts of the atom. That may be represented either by the probability
that a point harge o upies this region or by the volume density of an
ele tri harge distributed in a ontinuous manner through this region.
Mr Kramers. The use that one may make of the simple Thomas
model of the atom in the sear h for the laws of ree tion is extremely
interesting. It would perhaps not be superuous to investigate what
result would be obtained for the ele tron distribution if, instead of
restri ting oneself to onsidering a single entre of attra tion, one applied
Thomas' dierential equation to an innity of entres distributed as in a
rystal grating. Has anyone already tried to solve the problem of whi h
Mr Bragg has just spoken, of the al ulation of the general distribution
of the ele troni density around the nu leus of a heavy atom, in the ase
Dis ussion
321
where there are many nu lei, as in a rystal?
Mr Bragg.
No, no one has yet atta ked this problem, whi h I

only mentioned be ause it is interesting.
Mr Dira .
Do the s attering urves depend on the phase relations

between the os illations of dierent atoms?
Mr Bragg.
No, be ause the results of our experiments give only

the average s attering produ ed in ea h dire tion by a very large number
of atoms.
Mr Dira .
What would happen if you had two simple os illators

performing harmoni vibrations? Would they produ e a dierent s attering when in phase than when out of phase?
Mr Born.
The orre t answer to the question of s attering by an

atom is ontained in the remark by Mr Pauli. Stri tly speaking there is
no three-dimensional harge distribution that may des ribe exa tly how
an atom behaves; one always has to onsider the total onguration of all
the ele trons in the spa e of 3n dimensions. A model in three dimensions
only ever gives a more or less rude approximation.
Mr Kramers
asks a question on erning the inuen e of the Compton ee t on the s attering.
Mr Bragg.
I have already said something on that subje t in my

report.a Assuming a model of the atom of the old type, Jaun ey and
Williams have used the riterion that the wavelength is modied when
the re oil of the s attering ele tron is su ient to take it entirely outside
the atom. Williams was the rst to apply this riterion to s attering
urves obtained with rystals. He pointed out that while the speed of the
re oil ele tron depends on both the s attering angle and the wavelength,
any riterion one uses is a fun tion of sin , just as the interferen e
ee ts depend on sin . This implies that the existen e of the Compton
ee t modies the s attering urve su h that we an always assign the
same s attering urve to no matter what type of atom, whatever the
a Cf. Bragg's report, se tion 6 (eds ).
322
W. L. Bragg
wavelength may be.
Mr Fowler. If I have understood properly, Mr Bragg uses theore-
ti al al ulations by Waller that have not yet been published. When light
is s attered by an atom in a ordan e with the interpretation given by
Mr Waller by means of the new me hani s, the total amount of s attered
light is given exa tly by J. J. Thomson's lassi al formula (ex ept for very
hard -rays). This light is omposed of the oherent s attered radiation
and of the modied light (Compton s attering). In the theorem of the
ree tion of X-rays only the oherent s attered light must be used, and
indeed it is; and this light is given exa tly by the F urves like those
proposed by Hartree. These F urves for atomi s attering are obviously
given simply by the lassi al s attering for ea h ele tron, diminished by
interferen e.
Mr Bragg.
I should like to develop Mr Fowler's remark by re alling Waller and Wentzel's on lusions briey sket hed in my report.
The s attering by one of the ele trons in an atom partly remains the
same and partly is modied. Within ertain limits the total amount of
s attered radiation is given by J. J. Thomson's formula. A fra tion f 2
of this amount is not modied, f being a oe ient smaller than 1,
depending on the interferen e of the spatial distribution of the harge
a ording to S hrdinger and al ulated a ording to the lassi al laws
of opti s. The remaining fra tion 1 f 2 is modied.79
Mr Lorentz.
It is, without doubt, extremely noteworthy that the

total s attering, omposed of two parts of quite dierent origin, agrees
with Thomson's formula.
Mr Kramers makes two remarks:

1. As Mr Bragg has pointed out the importan e of there being interest in
having more experimental data on erning the refrangibility of X-rays in
the neighbourhood of the absorption limit, I should like to draw attention
to experiments performed re ently by Mr Prins in the laboratory of Professor Coster at Groningen. By means of his apparatus (the details of the
experiments and the results obtained are des ribed in a paper published
re ently in Zeits hrift fr Physik, 47 (1928), [479), Mr Prins nds in
a single test the angle of total ree tion orresponding to an extended
region of frequen ies. In the region of the absorption limit of the metal,
Dis ussion
323
he nds an abnormal ee t, whi h onsists mainly of a strong de rease

in the angle of total ree tion on the side of the absorption limit lo ated
towards the short wavelengths. This ee t is easily explained taking into
a ount the inuen e of absorption on the total ree tion, without it
being ne essary to enter into the question of the hange in refrangibility
of the X-rays. In fa t, the absorption may be des ribed by onsidering
the refra tive index n as a omplex number, whose imaginary part is
related in a simple manner to the absorption oe ient. Introdu ing
this omplex value for n in the well-known formulas of Fresnel for the
intensity of ree ted rays, one nds that the sharp limit of total ree tion
disappears, and that the manner in whi h the intensity of ree ted rays
depends on the angle of in iden e is su h that the experiment must
give an èe tive angle of total ree tion' that is smaller than in the
ase where there is no absorption and that de reases as the absorption
in reases.
A ording to the atomi theory one would also expe t to nd, in
the region of the absorption limit, anomalies in the real part of the
refra tive index, produ ing a similar though less noti eable de rease of
the ee tive angle of total ree tion on the side of the absorption edge
dire ted towards the large wavelengths. Mr Prins has not yet su eeded
in showing that the experiments really demonstrate this ee t.a
The theory of these anomalies in the real part of the refra tive index
onstitutes the subje t of my se ond remark.
2. Let us onsider plane and polarised ele tromagneti waves, in whi h
the ele tri for e an be represented by the real part of Ee2it , striking
an atom whi h for further simpli ity we shall assume to be isotropi .
The waves make the atom behave like an os illating dipole, giving, by
expansion in a Fourier series, a term with frequen y . Let us represent
this term by the real part of P e2it , where P is a omplex ve tor having
the same dire tion as the ve tor E to whi h it is, moreover, proportional.
If we set
P
(1)
= f + ig ,
E
where f and g are real fun tions of , the real and imaginary parts of
the refra tive index of a sample of matter are related in a simple way to
the fun tions f and g of the atoms ontained in the sample.
Extending the domain of values that may take into the negative
a Continuing his resear h Mr Prins has established (February 1928) the existen e of
this ee t, in agreement with the theory.
324
W. L. Bragg
region and dening f as an even fun tion of , g as an odd fun tion,

one easily veries that the dispersion formulas of Lorentz's lassi al
theory and also those of modern quantum me hani s are equivalent to
the formula
Z
1 + g( )
f () =
(2)
d ,

R
where the sign indi ates the `prin ipal' value of the integral.
This formula an easily be applied to atoms showing ontinuous absorption regions and is equivalent to the formulas proposed for these
ases by R. de Laer Kronig and by Mark and Kallmann. There is hardly
any doubt that this general formula may be derived from quantum
me hani s, if one duly takes into a ount the absorption of radiation,
basing oneself on Dira 's theory, for example.
From a mathemati al point of view, formula (2) gives us the means
to onstru t an analyti fun tion of a omplex variable that is holomorphi below the real axis and whose real part takes the values g( )
on this axis. If one onsiders as a real variable, the integral equation
(2) has the solution
Z
1 + f ( )
g() =
(3)
d ,

whi h shows that the imaginary part of the refra tive index depends on
the real part in nearly the same way as the real part depends on the
imaginary part. The fa t that the analyti fun tion f of the omplex
variable , dened by (2) for the lower half of the omplex plane, has
no singularity in this half-plane, means that dispersion phenomena,
when one studies them by means of waves whose amplitude grows in
an exponential manner ( omplex), an never give rise to singular
behaviour for the atoms.
Mr Compton.
The measurements of refra tive indi es of X-rays

made by Doan agree better with the Drude-Lorentz formula than with
the expression derived by Kronig based on the quantum theory of dispersion.
Mr de Broglie.
I should like to draw attention to re ent experiments arried out by Messrs J. Thibaud and A. Soltan,a whi h
tou h on the questions raised by Mr Bragg. In these experiments Messrs
a C. R. A ad. S ., 185 (1927), 642.
Dis ussion
325
Thibaud and Soltan measured, by the tangent grating method, the

wavelength of a ertain number of X-rays in the domain 20 to 70 .
Some of these wavelengths had already been determined by Mr Dauvillier
using dira tion by fatty-a id gratings. Now, omparing the results of
Dauvillier with those of Thibaud and Soltan, one noti es that there is
a systemati dis repan y between them that in reases with wavelength.
Thus for the K line of boron, Thibaud and Soltan nd 68 , while
Dauvillier had found 73.5 , that is, a dieren e of 5.5 . This systemati
dis repan y appears to be due to the in rease of the refra tive index
with wavelength. The index does not a tually play a role in the tangent
grating method, while it distorts in a systemati way the results obtained
by rystalline dira tion when one uses the Bragg formula. Starting from
the dieren e between their results and those of Mr Dauvillier, Messrs
Thibaud and Soltan have al ulated the value of the refra tive index of
fatty a ids around 70 and found
= 1 = 102
thereabouts. This agrees well with a law of the form = K2 ; sin e in the
ordinary X-ray domain the wavelengths are about 100 times smaller,
is of order 106 . One ould obje t that, a ording to the Drude-Lorentz
law, the presen e of K dis ontinuities of oxygen, nitrogen and arbon
between 30 and 45 should perturb the law in 2 . But in the X-ray
domain the validity of Drude's law is doubtful, and if one uses in its pla e
the formula proposed by Kallmann and Marka the agreement with the
experimental results is very good. Let us note nally that the existen e
of an index appre iably dierent from 1 an ontribute to explaining why
large-wavelength lines, obtained with a fatty-a id grating, are broad and
spread out.
Mr Lorentz
makes a remark on erning the refra tive index of a

rystal for Rntgen rays and the deviations from the Bragg law. It is
lear that, a ording to the lassi al theory, the index must be less than
unity, be ause the ele trons ontained in the atoms have eigenfrequen ies
smaller than the frequen y of the rays, whi h gives rise to a speed of
propagation greater than c. But in order to speak of this speed, one must
adopt the ma ros opi point of view, abstra ting away the mole ular
dis ontinuity. Now, if one wishes to explain Laue's phenomenon in all
its details, one must onsider, for example, the a tion of the vibrations
a Ann. d. Phys., 82 (1927), 585.
326
W. L. Bragg
ex ited in the parti les of a rystallographi layer on a parti le of a

neighbouring layer. This gives rise to series that one annot repla e
by integrals. It is for this reason that I found some di ulty in the
explanation of deviations from the Bragg law.80
Mr Debye. Ewald has tried to do similar al ulations.

Mr Lorentz. It is very interesting to note that with Rntgen
rays one nds, in the vi inity of an absorption edge, phenomena similar

to those that in lassi al opti s are produ ed lose to an absorption
band. There is, however, a profound dieren e between the two ases,
the absorption edge not orresponding to a frequen y that really exists
in the parti les.
Notes to pp. 283298
327

1 Here and in a few other pla es, the Fren h adds (or omits) inverted
ommas.
2 [rexion des radiations des raies
3 [ne fait prvoir au un art
4 The Fren h edition adds èn ou hes' [in layers.
5 Types ript: `have often been .... it has led'; Fren h version: à souvent
t .... elles ont onduit'.
6 [ont eu onan e dans les mthodes quantitatives
7 [portions
8 [sous une forme plus primitive
9 [il soit possible de les trouver
10 [logiquement
11 [thermodynamique
12 Word omitted in the Fren h version.
13 The Fren h edition adds `Sir'.
14 [les donnes obtenues par rexion
15 Here following the Fren h edition; the types ript reads `total radiation'.
16 Emphasis omitted in the Fren h edition.
17 [servirent omparer
18 The Fren h omits òf ree tion'.
19 [dj observe
20 [il proposa d'employer, pour l'interprtation des mesures, la mthode de
Fourier
21 [rapport
22 [ou
23 [M Innes
24 [potassium
25 The types ript has a spurious omma after `Bergen'.
26 [ l'aide de ristaux
27 [les deux onstantes le troniques
28 Here and in some other instan es, the Fren h renders `single' as `simple'.
29 [`fa teur atomique'
30 Not printed as a quotation in the Fren h edition.
31 hoisies simplement
32 The Fren h adds `de ses exprien es'.
33 The Fren h adds `dans es onditions'.
34 This is indeed a quotation from p. 272 from the le ture by W. H. Bragg.
The types ript has a omma instead of the losing quotation mark, while
the Fren h edition omits the opening quotation mark. The types ript has
a spurious denominator à' instead of `d' in the se ond and third terms
(but ta itly orre ts another typo in the original).
35 [dans son livre sur `les rayons X et les le trons'
36 The Fren h adds òn peut pro der de la faon inverse, 'est- dire'.
37 `Beryl.' omitted in Fren h edition.
38 The Fren h edition uses supers ripts throughout.
39 The Fren h edition omits the overbar in the aption.
40 [Wy kho
41 Word omitted in Fren h edition.
42 Again, the Fren h edition omits the overbar in the aption.
328
Notes to pp. 301325
43 [frappante
44 [son livre sur les rayons X et les le trons
45 [bien que les gures (si ) sur lesquelles la nouvelle ourbe se base soient
plus exa tes
46 Here and in several other pla es, the Fren h adds `Mlle'.
47 The Fren h reads ùne nergie au zro absolu (nergie de stru ture)'.
48 Word missing in the Fren h edition.
49 Bra ket printed as a footnote in the Fren h edition.
50 [tout prs de e point
51 [irrgularits
52 [pour
53 [brillante
54 [ SrK = 0.871A
55 [faites d'aprs 1921 gures
56 The Fren h omits `B. J. B. gures for NaCl'.
57 [gures
58 Again, in the Fren h, the false friend `gures'.
59 [dduits
60 [CaFl, LiFl, NaFl
61 [Une fois que nous disposerons de donnes pr ises
62 The Fren h edition omits `& Compton' and has `Modle de Pauling et
S hrdinger'.
63 The Fren h translates as if the omma were after òf Thomas' model'
rather than before.
64 [M. Lorentz
65 The types ript reads `mu h smaller in the ordinary X-ray region', but
given the ontext the text should be amended as shown (as also done in
the Fren h version).
66 [parfaitement d'a ord
67 Referen e omitted in the Fren h edition.
Z
68 [ A la valeur moyenne du rapport du nombre atomique au poids atomique
pour ses divers onstituants (pour tous les atomes lgers e rapport est
peu prs gal 0.5
69 [rayons
70 Types ript: òf the K adsorption edge'; Fren h version: `de la
dis ontinuit K'.
71 Types ript: àdsorption edge'; Fren h: `bord d'absorption'.
72 The Fren h adds `dans la pyrite'.
73 Fren h edition: `Ma Innes'.
74 Types ript and Fren h edition both have `532'.
75 Both types ript and Fren h edition give this referen e as `B. Davis and
C. C. Hatley'. The types ript has `291'.
76 Authors added in the Fren h edition.
77 Here and in the following displayed formula, the published version has
square bra kets instead of absolute bars.
78 Arrow missing on
rk
in the published volume.
79 The original text mistakenly states that both fra tions are `not modied'.
80 The mixing of rst and third person, here and in a few similar instan es
throughout the dis ussions, is as in the published text.
Disagreements between experiment and the

ele tromagneti theory of radiation
a
By Mr Arthur H. COMPTON
Introdu tion
Professor W. L. Bragg has just dis ussed a whole series of radiation
phenomena in whi h the ele tromagneti theory is onrmed. He has
even dwelt on some of the limiting ases, su h as the ree tion of X-rays
by rystals, in whi h the ele tromagneti theory of radiation gives us, at
least approximately, a orre t interpretation of the fa ts, although there
are reasons to doubt that its predi tions are truly exa t. I have been left
the task of pleading the opposing ause to that of the ele tromagneti
theory of radiation, seen from the experimental viewpoint.
I have to de lare from the outset that in playing this role of the a user
I have no intention of diminishing the importan e of the ele tromagneti
theory as applied to a great variety of problems.b It is, however, only by
a An English version of this report (Compton 1928) was published in the Journal of
the Franklin Institute. The Fren h version appears to be essentially a translation
of the English paper with some additions. Whenever there are no dis repan ies,
we reprodu e Compton's own English (we have orre ted some obvious typos
and harmonised some of the spelling). Interesting variants are footnoted. Other
dis repan ies between the two versions are reported in the endnotes (eds.).
b The opening has been translated from the Fren h edition. The English version has
the following dierent opening (eds.):
During the last few years it has be ome in reasingly evident that the lassi al
ele tromagneti theory of radiation is in apable of a ounting for ertain large
lasses of phenomena, espe ially those on erned with the intera tion between
radiation and matter. It is not that we question the wave hara ter of light
the striking su esses of this on eption in explaining polarisation and
interferen e of light an leave no doubt that radiation has the hara teristi s
329
330
A. H. Compton
a quainting ourselves with the real or apparent1 failures of this powerful

theory that we an hope to develop a more omplete theory of radiation
whi h will des ribe the fa ts as we know them.
The more serious di ulties whi h present themselves in onne tion
with the theory that radiation onsists of ele tromagneti waves, propagated through spa e in a ord with the demands of Maxwell's equations,
may be lassied onveniently under ve heads:2
(1) Is there an ether? If there are os illations, there must be a medium
in whi h these os illations are produ ed. Assuming the existen e of su h
a medium, however, one en ounters great di ulties.
(2) How are the waves produ ed? The lassi al ele trodynami s requires
as a sour e of an ele tromagneti wave an os illator of the same frequen y
as that of the waves it radiates. Our studies of spe tra,3 however, make
it appear impossible that an atom should ontain os illators of the same
frequen ies as the emitted rays.
(3) The photoele tri ee t. This phenomenon is wholly anomalous when
viewed from the standpoint of waves.
(4) The s attering of X-rays, and the re oil ele trons, phenomena in
whi h we nd gradually in reasing departures from the predi tions of
the lassi al wave theory as the frequen y in reases.
(5) Experiments on individual intera tions between quanta of radiation
and ele trons. If the results of the experiments of this type are reliable,
they seem to show denitely that individual quanta of radiation, of
energy h , pro eed in denite dire tions.
The photon hypothesis.4 In order to exhibit more learly the di ulties with the lassi al theory of radiation, it will be helpful to keep
in mind the suggestion that5 light onsists of orpus les. We need not
think of these two views as ne essarily alternative. It may well be that
the two on eptions are omplementary. Perhaps the orpus le is related
of waves; but it is equally true that ertain other properties of radiation are
not easily interpreted in terms of waves. The power of the ele tromagneti
theory as applied to a great variety of problems of radiation is too well known
to require emphasis.
Experiment and the ele tromagneti theory
331
to the wave in somewhat the same manner that the mole ule is related
to matter in bulk; or there may be a guiding wave whi h dire ts the
orpus les whi h arry the energy. In any ase, the phenomena whi h
we have just mentioned suggest the hypothesis that radiation is divisible
into units possessing energy h , and whi h pro eed in denite dire tions
with momentum h/c. This is obviously similar to Newton's old on eption of light orpus les. It was revived in its present form by Professor
Einstein,6 it was defended under the name of the `Neutron Theory' by Sir
William [H. Bragg, and has been given new life by the re ent dis overies
asso iated with the s attering of X-rays.
In referring to this unit of radiation I shall use the name `photon',
suggested re ently by G. N. Lewis.a This word avoids any impli ation
regarding the nature of the unit, as ontained for example in the name
`needle ray'. As ompared with the terms `radiation quantum' and `light
quant',7 this name has the advantages of brevity and of avoiding any
implied dependen e upon the mu h more general quantum me hani s or
quantum theory of atomi stru ture.
Virtual radiation. Another on eption of the nature of radiation whi h
it will be desirable to ompare with the experiments is Bohr, Kramers
and Slater's important theory of virtual radiation.b A ording to this
theory, an atom in an ex ited state is ontinually emitting virtual radiation, to whi h no energy hara teristi s are to be as ribed. The normal
atoms have asso iated with them virtual os illators, of the frequen ies
orresponding to jumps of the atom to all of the stationary states of
higher energy. The virtual radiation may be thought of as being absorbed
by these virtual os illators, and any atom whi h has a virtual os illator
absorbing this virtual radiation has a ertain probability of jumping
suddenly to the higher state of energy orresponding to the frequen y
of the parti ular virtual os illator. On the average, if the radiation is
ompletely absorbed, the number of su h jumps to levels of higher energy
is equal to the number of emitting atoms whi h pass from higher to
lower states. But there is no dire t onne tion between the falling of
one atom from a higher to a lower state and a orresponding rise of a
se ond atom from a lower to a higher state. Thus on this view the energy
of the emitting atoms and of the absorbing atoms is only statisti ally
onserved.
a G. N. Lewis, Nature, [118, [874 (De . 18, 1926).
b N. Bohr, H. A. Kramers and J. C. Slater, Phil. Mag., 47 (1924), 785; Zeits. f.
Phys., 24 (1924), 69.
332
A. H. Compton
The problem of the ether
The onstan y of the speed of radiation of dierent wavelengths has

long been onsidered as one of the most powerful arguments in favour
of the wave theory of light. This onstan y suggests that a perturbation
is travelling through a xed medium in spa e, the ether.
If experiments like those by Mi helson and Morley's were to show
the existen e of a relative motion with respe t to su h a medium, this
argument would be onsiderably strengthened. For then we ould imagine light as having a speed determined with referen e to a xed axis
in spa e. But, ex ept for the re ent and quite doubtful experiments by
Miller,a no-one has ever dete ted su h a relative motion. We thus nd
ourselves in the di ult position of having to imagine a medium in
whi h perturbations travel with a denite speed, not with referen e to
a xed system of axes, but with referen e to ea h individual observer,
whatever his motion. If we think of the omplex properties a medium
must have in order to transmit a perturbation in this way, we nd that
the medium diers so onsiderably from the simple ether from whi h
we started that the analogy between a wave in su h a medium and a
pertubation travelling in an elasti medium is very distant. It is true
that doubts have often been expressed as to the usefulness of retaining
the notion of the ether. Nevertheless, if light is truly a wave motion, in
the sense of Maxwell, there must be a medium in order to transmit this
motion, without whi h the notion of wave would have no meaning. This
means that, instead of being a support for the wave theory, the on ept
of the ether has be ome an un omfortable burden of whi h the wave
theory has been unable to rid itself.
If, on the other hand, we a ept the view suggested by the theory of
relativity, in whi h for the motion of matter or energy there is a limiting
speed relative to the observer, it is not surprising to nd a form of energy
that moves at this limiting speed. If we abandon the idea of an ether, it
is simpler to suppose that this energy moves in the form of orpus les
rather than waves.
a D. C. Miller, Nat. A ad. S i. Pro ., 11 (1925), 306.
The emission of radiation
333
When we tra e a sound to its origin, we nd it oming from an os illator

vibrating with the frequen y of the sound itself. The same is true of
ele tri waves, su h as radio waves, where the sour e of the radiation
is a stream of ele trons os illating ba k and forth in a wire. But when
we tra e a light ray or an X-ray ba k to its origin, we fail to nd any
os illator whi h has the same frequen y as the ray itself. The more omplete our knowledge be omes of the origin of spe trum lines, the more
learly we see that if we are to assign any frequen ies to the ele trons
within the atoms, these frequen ies are not the frequen ies of the emitted
rays, but are the frequen ies asso iated with the stationary states of the
atom. This result annot be re on iled with the ele tromagneti theory
of radiation, nor has any me hanism been suggested whereby radiation
of one frequen y an be ex ited by an os illator of another frequen y.
The wave theory of radiation is thus powerless to suggest how the waves
originate.
The origin of the radiation is onsiderably simpler when we onsider
it from the photon viewpoint. We nd that an atom hanges from a
stationary state of one energy to a state of less energy, and asso iated
with this hange radiation is emitted. What is simpler than to suppose
that the energy lost by the atom is radiated away as a single photon? It
is on this view unne essary to say anything regarding the frequen y of
the radiation. We are on erned only with the energy of the photon, its
dire tion of emission, and its state of polarisation.
The problem of the emission of radiation takes an espe ially interesting form when we onsider the produ tion of the ontinuous X-ray
spe trum.a Experiment shows that both the intensity and the average
frequen y of the X-rays emitted at angles less than 90 degrees with the
athode-ray stream are greater than at angles greater than 90 degrees.
This is just what we should expe t due to the Doppler ee t if the X-rays
are emitted by a radiator moving in the dire tion of the athode rays. In
order to a ount for the observed dissymmetry between the rays in the
forward and ba kward dire tions, the parti les emitting the radiation
must be moving with a speed of the order of 25 per ent that of light.
This means that the emitting parti les must be free ele trons, sin e it
would require an impossibly large energy to set an atom into motion
with su h a speed.
a The di ulty here dis ussed was rst emphasised by D. L. Webster, Phys. Rev.,
13 (1919), 303.
334
A. H. Compton
But it will be re alled that the ontinuous X-ray spe trum has a sharp
upper limit. Su h a sharp limit is, however, possible on the wave theory
only in ase the rays ome in trains of waves of onsiderable length, so
that the interferen e between the waves in dierent parts of the train
an be omplete at small glan ing angles of ree tion from the rystal.
This implies that the os illator whi h emits the rays must vibrate ba k
and forth with onstant frequen y a large number of times while the ray
is being emitted. Su h an os illation might be imagined for an ele tron
within an atom; but it is impossible for an ele tron moving through
an irregular assemblage of atoms with a speed omparable with that of
light.
Thus the Doppler ee t in the primary X-rays demands that the rays
shall be emitted by rapidly moving ele trons, while the sharp limit to
the ontinuous spe trum requires that the rays be emitted by an ele tron
bound within an atom.
The only possible es ape from this dilemma on the wave theory is
to suppose that the ele tron is itself apable of internal os illation of
su h a hara ter as to emit radiation. This would, however, introdu e
an undesirable omplexity into our on eption of the ele tron, and would
as ribe the ontinuous X-rays to an origin entirely dierent from that of
other known sour es of radiation.
Here again the photon theory aords a simple solution. It is a onsequen e of Ehrenfest's adiabati prin iplea that photons emitted by a
moving radiator will show the same Doppler ee t, with regard to both
frequen y and intensity, as does a beam of waves.b But if we suppose
that photons are radiated by the moving athode ele trons, the energy
of ea h photon will be the energy lost by the ele tron, and the limit of
the X-ray spe trum is ne essarily rea hed when the energy of the photon
is equal to the initial energy of the ele tron, i.e., h = eV . In this ase,
if we onsider the initial state as an ele tron approa hing an atom with
large kineti energy and the nal state as the ele tron leaving the atom
with a smaller kineti energy, we see that the emission of the ontinuous
X-ray spe trum is the same kind of event as the emission of any other
type of radiation.
a The adiabati prin iple onsists in the following. Sin e for a quantised quantity
there should be no quantum jumps indu ed by an innitely slowly varying
external for e (in this ase, one that gently a elerates a radiator), there is an
analogy between these quantities and the lassi al adiabati invariants. Ehrenfest
(1917) a ordingly formulated a prin iple identifying the lassi al quantities to be
quantised as the adiabati invariants of a system (eds.).
b Cf., e.g., A. H. Compton, Phys. Rev., 21 (1923), 483.
335
Absorption of radiation. A ording to the photon theory, absorption

o urs when a photon meets an atom and imparts its energy to the atom.
The atom is thereby raised to a stationary state of higher energy
pre isely the reverse of the emission pro ess.
On the wave theory, absorption is ne essarily a ontinuous pro ess, if
we admit the onservation of energy, sin e on no part of the wave front is
there enough energy available to hange the atom suddenly from a state
of low energy to a state of higher energy. What eviden e we have is,
however, strongly against the atom having for any onsiderable length
of time an energy intermediate between two stationary states; and if su h
intermediate states annot exist, the gradual absorption of radiation is
not possible. Thus the absorption of energy from waves9 is irre on ilable
with the on eption of stationary states.
We have seen that on the theory of virtual radiation the energy of
the emitting atoms and of the absorbing atoms is only statisti ally
onserved. There is a ording to this view therefore no di ulty with
supposing that the absorbing atom suddenly jumps to a higher level
of energy, even though it has not re eived from the radiation as mu h
energy as is ne essary to make the jump. It is thus possible through
virtual os illators and virtual radiation to re on ile the wave theory of
radiation with the sudden absorption of energy, and hen e to retain the
idea of stationary states.
The photoele tri effe t

It is well known that the photon hypothesis was introdu ed by Einstein
to a ount for the photoele tri ee t.a The assumption that light onsists of dis rete units whi h an be absorbed by atoms only as units, ea h
giving rise to a photoele tron, a ounted at on e for the fa t that the
number of photoele trons is proportional to the intensity of the light;
and the assumption that the energy of the light unit is equal to h ,
where h is Plan k's onstant, made it possible to predi t the kineti
energy with whi h the photoele trons should be eje ted, as expressed by
Einstein's well-known photoele tri equation,

1
mc2 p
1 = h wp .
(1)
1 2
a A. Einstein, Ann. d. Phys., 17 (1905), [132.10
336
A. H. Compton
Seven years elapsed before experiments by Ri hardson and Comptona

and by Hughesb showed that the energy of the emitted ele trons was
indeed proportional to the frequen y less a onstant,12 and that the
fa tor of proportionality was lose to the value of h al ulated from
Plan k's radiation formula. Millikan's more re ent pre ision photoele tri experiments with the alkali metalsc onrmed the identity of the
onstant h in the photoele tri equation with that in Plan k's radiation
formula. De Broglie's beautiful experimentsd with the magneti spe trograph showed that in the region of X-ray frequen ies the same equation
holds, if only we interpret the work fun tion wp as the work required to
remove the ele tron from the pth energy level of the atom. Thibaud has
made use of this resulte in omparing the velo ities of the photoele trons
eje ted by -rays from dierent elements, and has thus shown that the
photoele tri equation (1) holds with pre ision even for -rays of the
highest speed. Thus from light of frequen y so low that it is barely able
to eje t photoele trons from metals to -rays that eje t photoele trons
with a speed almost as great as that of light, the photon theory expresses
a urately the speed of the photoele trons.
The dire tion in whi h the photoele trons are emitted is no less instru tive than is the velo ity. Experiments using the loud expansion
method, performed13 by C. T. R. Wilsona and others,b have shown that
the most probable dire tion in whi h the photoele tron is eje ted from
an atom is nearly the dire tion of the ele tri ve tor of the in ident
wave, but with an appre iable forward omponent to its motion. There
is, however, a very onsiderable variation in the dire tion of emission.
For example, if we plot the number of photoele trons eje ted at dierent angles with the primary beam we nd, a ording to Auger, the
distribution shown in Fig. 1.
Ea h of these urves, taken at a dierent potential, represents the
distribution of about 200 photoele tron tra ks. It will be seen that as the
potential on the X-ray tube in reases, the average forward omponent
of the photoele tron's motion also in reases.
When polarised X-rays are used, there is a strong preponderan e of the
a
b

d
e
a
b
O. W. Ri hardson and K. T. Compton, Phil. Mag., 24 (1912), 575.

A. L. Hughes, Phil. Trans. A, 212 (1912), 205.11
R. A. Millikan, Phys. Rev., 7 (1916), 355.
M. de Broglie, Jour. de Phys., 2 (1921), 265.
J. Thibaud, C. R., 179 (1924), 165, 1053 and 1322.
C. T. R. Wilson, Pro . Roy. So . A, 104 (1923), 1.
A. H. Compton, Bull. Natl. Res. Coun., No. 20 (1922), 25; F. W. Bubb, Phys.
Rev., 23 (1924), 137; P. Auger, C. R., 178 (1924), 1535; D. H. Loughridge, Phys.
Rev., 26 (1925), 697; F. Kir hner, Zeits. f. Phys., 27 (1926), 385.
337
Fig. 1. Longitudinal distribution of photoele trons for X-rays of three dierent

ee tive wavelengths, a ording to Auger.
photoele trons in or near the plane in luding the ele tri ve tor of the
in ident rays. Thus Fig. 2 shows the distribution found by Bubb of the
dire tion of the photoele trons eje ted from moist air when traversed by
X-rays that have been polarised by s attering at right angles from a blo k
of paran. Be ause of multiple s attering in the paran, the s attered
rays are not ompletely polarised, and this is probably su ient to
a ount for the fa t that some photoele trons appear to start at right
angles with the ele tri ve tor. This ee t with X-rays is doubtless
similar in hara ter to the sele tive photoele tri ee t dis overed many
years ago by Pohl and Pringsheim, in whi h the number of ele trons
eje ted by light from the liquid surfa e of sodium-potassium alloy is
greater when the ele tri ve tor is in a plane perpendi ular to the surfa e
than when parallel to the surfa e.
Re ent experiments have shown that the dire tion in whi h the photoele trons are eje ted by X-rays is at least very nearly independent of
the material from whi h the ele trons ome.a
Can ele tromagneti waves produ e photoele trons? Before dis ussing
the produ tion of photoele trons from the standpoint of radiation quanta, let us see what su ess meets the attempt15 to a ount for them
on the basis of ele tromagneti waves. The fa t that they are emitted
a E. A. Owen, Pro . Phys. So ., 30 (1918), 133; Auger, Kir hner, Loughridge, lo .
it.14
338
A. H. Compton
Fig. 2. Lateral distribution of photoele trons for in ompletely polarised X-rays,

a ording to Bubb.
approximately in the dire tion of the ele tri ve tor would suggest that
the photoele trons are eje ted by the dire t a tion of the ele tri eld
of the in ident rays. If this were the ase, however, we should expe t
the speed of the eje ted ele trons to be greater for greater intensity
of radiation, whereas experiment shows that for the same wavelength
intense sunlight eje ts an ele tron no faster than does the feeble light
from a star. Furthermore, the energy available from the ele tromagneti
wave is wholly inadequate. Thus in a re ent experiment performed by
Joe and Dobronrawov,a X-rays were produ ed by the impa t on a target
of 104 to 105 ele trons per se ond. Sin e on the ele tromagneti theory
an X-ray pulse is of the order of 103 waves in length or 1016 se onds
a A. Joe and N. Dobronrawov, Zeits. f. Phys., 34, 889 (1925).
339
in duration, the X-ray pulses must have followed ea h other at widely

separated intervals. It was found, however, that photoele trons were
o asionally eje ted from a bismuth parti le whi h subtended a solid
angle not greater than 105 . It is learly impossible that all the energy
of an X-ray pulse whi h has spread out in a spheri al wave should spend
itself on this bismuth parti le. Thus on the wave theory the eje tion
of the photoele tron, whi h has almost as mu h energy as the original
athode ele tron, ould not have been a omplished by a single16 pulse.
It annot therefore be the dire t a tion of the ele tri ve tor of the wave,
taken in the usual sense,17 whi h has eje ted the ele tron.
We may assume, on the other hand, that the energy is gradually
absorbed in the bismuth parti le of Joe's experiment until an amount
h has a umulated, whi h is then spent in eje ting the photoele tron.
We have already alled attention to the fa t that this gradual absorption
hypothesis implies the existen e of stationary states in the atom having
innitesimal gradations of energy, whereas the eviden e is very strong
that atoms annot endure ex ept in ertain denitely dened stationary
states. But new di ulties also arise. Why do the photoele trons tend
to start in the dire tion of the ele tri eld of the in ident wave? If we
suppose that it is the gradual absorption of energy from a wave whi h
liberates the ele tron, why does there exist a tenden y for the ele tron
to start with a large omponent of its motion in a forward dire tion?18
The forward impulse due to the radiation pressure as19 the energy is
gradually absorbed will be transferred to the atom and not left with
[the absorbing ele tron. The a umulation hypothesis is thus di ult
to defend.
Photons and photoele trons. On the photon theory it is possible to
a ount in a simple manner for most of the properties of the photoele trons. We have seen how Einstein was able to predi t a urately the
velo ity of the photoele trons, assuming only that energy is onserved
when a photon a ts on an ele tron. In order to a ount for the dire tion of
emission we must as ribe to the photon some of the properties of an ele tromagneti pulse. Bubb introdu ed the suggestiona that we as ribe to
the photon a ve tor property similar to the ele tri ve tor of an ele tromagneti wave, so that when the photon traverses an atom the ele trons
and the nu leus re eive impulses in opposite dire tions perpendi ular
to the dire tion of propagation. Asso iated with this ele tri ve tor, we
a F. W. Bubb, Phys. Rev., 23 (1924), 137.
340
A. H. Compton
should also expe t to nd a magneti ve tor. Thus if an ele tron is set
in motion by the ele tri ve tor of the photon at right angles to the
dire tion of propagation, the magneti ve tor of the photon will a t
on the moving ele tron in the dire tion of propagation. This is stri tly
analogous to the radiation pressure exerted by an ele tromagneti wave
on an ele tron whi h it traverses, and means that the forward momentum
of the absorbed photon is transferred to the photoele tron.
In the simplest ase, where we negle t the initial momentum of the
ele tron in its orbital motion in the atom, the angle between the dire tion
of the in ident ray and the dire tion of eje tion is found from these
assumptions to be
p
= tan1 2/ ,
(2)
where = /, and = h/mc = 0.0242 . The quantity is small

ompared with unity, ex ept for very hard X-rays and -rays. Thus for
light, equation (2) predi ts the expulsion of photoele trons at nearly 90
degrees. This is in a ord with the rather un ertain data whi h have
been obtained with visible and ultra-violet light.a
The only really signi ant test of this result is in its appli ation to
X-ray photoele trons. In Fig. 1 are drawn the lines 1 , 2 and 3 for the
three urves, at the angles al ulated by Auger from equation (2). It will
be seen that they fall very satisfa torily in the dire tion of maximum
emission of the photoele trons. Similar results have been obtained by
other investigators.b This may be taken as proof that a photon imparts
not only its energy, but also its momentum to the photoele trons.c
a Cf. A. Parts h and W. Hallwa hs, Ann. d. Phys., 41 (1913), 247.
b W. Bothe, Zeits. f. Phys., 26 (1925), 59; F. Kir hner, Zeits. f. Phys., 27 (1926),
385.20
The English version in ludes here the following footnote. Cf. also the omments
by Bragg on p. 356 and the ensuing dis ussion (eds.).
Sin e this was written, experiments by [D. H. Loughridge (Phys.
Rev., 30
(1927), [488) have been published whi h show a forward omponent to the
photoele tron's motion whi h seems to be greater than that predi ted by
equation (2). Williams, in experiments as yet unpublished, nds that the
forward omponent is almost twi e as great as that predi ted by this theory.
These results indi ate that the me hanism of intera tion between the photon
and the atom must be more omplex than here postulated. The fa t that
the forward momentum of the photoele tron is found to be of the same
order of magnitude as that of the in ident photon, however, suggests that the
momentum of the photon is a quired by the photoele tron, while an additional
forward impulse is imparted by the atom. Thus these more re ent experiments
341
Honestya obliges me to point out a di ulty that arises in this explanation of the motion of the photoele trons. It is the failure of the attempts
made to a ount properly for the fa t that the photoele trons are emitted
over a wide range of angles instead of in a denite dire tion, as would
be suggested by the al ulation just outlined. The most interesting of
these attempts is that of Bubb,b who takes into a ount the momentum
of the ele tron immediately before the absorption of the photon. Bubb
nds a dispersion of the dire tions of emission of the photoele trons of
the orre t order of magnitude, but whi h is larger when the ele tron
issues from a heavy atom than when it issues from a light one. We
have seen, however, that experiment has shown this dispersion of the
dire tions of emission to be notably independent of the element from
whi h the photoele tron originates.
Whatever may be the ause of the dispersion in the dire tions of motion of the photoele trons,21 it will readily be seen that if the time during
whi h the photon exerts a for e on the ele tron is omparable with the
natural period of the ele tron22 in the atom, the impulse imparted to the
ele tron will be transferred in part to the positive nu leus about whi h
the ele tron is moving. The fa t that the photoele trons are eje ted with
a forward omponent equal, within the limits of experimental error, to
the momentum of the in ident photon23 means that no appre iable part
of the photon's momentum is spent on the remainder of the atom. This
an only be the ase if the time of a tion of the photon on the ele tron is
short ompared with the time of revolution of the ele tron in its orbit.a
also support the view that the photoele tron a quires both the energy and the
momentum of the photon.
a This paragraph is present only in the Fren h edition. The orresponding one in
the English edition reads:
If the angular momentum of the atomi system from whi h the photoele tron
is eje ted is to be onserved when a ted upon by the radiation, the ele tron
annot be eje ted exa tly in the dire tion of
, but must re eive an impulse in a
dire tion determined by the position of the ele tron in the atom at the instant
it is traversed by the photon. Thus we should probably onsider the ele tri
ve tor of the X-ray wave as dening merely the most probable dire tion in
whi h the impulse should be imparted to the ele tron. This is doubtless the
hief reason why the photoele trons are emitted over a wide range of angles
instead of in a denite dire tion, as would be suggested by the al ulation just
outlined.
With the footnote: Cf. A. H. Compton, Phys. Rev., [31 (1928), [59 (eds.).
b F. W. Bubb, Phil. Mag., 49 (1925), 824.
a The English edition in ludes the further senten e: `Su h a short duration of
342
A. H. Compton
The photoele tri ee t and virtual radiation. It is to be noted that

none of these properties of the photoele tron is in onsistent with the
virtual radiation theory of Bohr, Kramers and Slater. The di ulties
whi h applied to the lassi al wave theory do not apply here, sin e the
energy and momentum are onserved only statisti ally. There is nothing
in this theory, however, whi h would enable us to predi t anything
regarding the motion of the photoele trons. The degree of su ess that
has attended the appli ation of the photon hypothesis to the motion of
these ele trons has ome dire tly from the appli ation of the onservation
prin iples to the individual a tion of a photon on an ele tron. The power
of these prin iples as applied to this ase is surprising if the assumption
is orre t that they are only statisti ally valid.
Phenomena asso iated with the s attering of X-rays

As is now well known, there is a group of phenomena asso iated with
the s attering of X-rays for whi h the lassi al wave theory of radiation
fails to a ount. These phenomena may be onsidered under the heads
of: (1) The hange of wavelength of X-rays due to s attering, (2) the
intensity of s attered X-rays, and (3) the re oil ele trons.
The earliest experiments on se ondary X-rays and -rays24 showed
a dieren e in the penetrating power of the primary and the se ondary
rays. In the ase of X-rays, Barkla and his ollaboratorsa showed that the
se ondary rays from the heavy elements onsisted largely of uores ent
radiations hara teristi of the radiator, and that it was the presen e of
these softer rays whi h was hiey responsible for the greater absorption
of the se ondary rays. When later experimentsb showed a measurable
dieren e in penetration even for light elements su h as arbon, from
whi h no uores ent K or L radiation appears, it was natural to as ribec
this dieren e to a new type of uores ent radiation, similar to the K and
L types, but of shorter wavelength. Careful absorption measurementsd
failed, however, to reveal any riti al absorption limit for these assumed
a
b

d
intera tion is a natural onsequen e of the photon on eption of radiation, but

is quite ontrary to the onsequen es of the ele tromagneti theory' (eds.).
C. [G. Barkla and C. A. Sadler, Phil. Mag., 16, 550 (1908).25
C. A. Sadler and P. Mesham, Phil. Mag., 24 (1912), 138; J. Laub, Ann. d. Phys.,
46 (1915), 785.
[C. G. Barkla and [M. P. White, Phil. Mag., 34 (1917), 270; J. Laub, Ann. d.
Phys., 46 (1915), 785, et al.
E.g., [F. K. Ri htmyer and [K. Grant, Phys. Rev., 15 (1920), 547.
343
`J' radiations similar to those orresponding to the K and L radiations.

Moreover, dire t spe tros opi observationsa failed to reveal the existen e of any spe trum lines26 under onditions for whi h the supposed
J-rays should appear. It thus be ame evident that the softening of the
se ondary X-rays from the lighter elements was due to a dierent kind
of pro ess than the softening of the se ondary rays from heavy elements
where uores ent X-rays are present.
A series of skilfully devised absorption experiments performed by
J. A. Graya showed, on the other hand, that both in the ase of rays and in that of X-rays an in rease in wavelength a ompanies the
s attering of the rays of light elements.
It was at this stage that the rst spe tros opi investigations of the
se ondary X-rays from light elements were made.b A ording to the
usual ele tron theory of s attering it is obvious that the s attered rays
will be of the same frequen y as the for ed os illations of the ele trons
whi h emit them, and hen e will be identi al in frequen y with the
primary waves whi h set the ele trons in motion. Instead of showing
s attered rays of the same wavelength as the primary rays, however,
these spe tra revealed lines in the se ondary rays orresponding to those
in the primary beam, but with ea h line displa ed slightly toward the
longer wavelengths.
This result might have been predi ted from Gray's absorption measurements; but the spe trum measurements had the advantage of aording a quantitative measurement of the hange in wavelength, whi h
gave a basis for its theoreti al interpretation.
The spe tros opi experiments whi h have shown this hange in wavelength are too well knownc to require dis ussion. The interpretation of the wavelength hange in terms of photons being dee ted by
individual27 ele trons and imparting a part of their energy to the s attering ele trons is also very familiar. For purposes of dis ussion, however,
let us re all that when we onsider the intera tion of a single photon
with a single ele tron the prin iples of the onservation of energy and
momentum lead usd to the result that the hange in wavelength of the
a E.g., [W. Duane and [T. Shimizu, Phys. Rev., 13 (1919), [289; ibid., 14 (1919),
389.
a J. A. Gray, Phil. Mag., 26 (1913), 611; Jour. Frank. Inst., [190, 643 (Nov. 1920).
b A. H. Compton, Bull. Natl. Res. Coun., No. 20, [18 ([O tober 1922); Phys. Rev.,
22 (1923), 409.
Cf., e.g., A. H. Compton, Phys. Rev., 22 (1923), 409; P. A. Ross, Pro . Nat. A ad.,
10 (1924), 304.
d A. H. Compton, Phys. Rev., [21 (1923), 483; P. Debye, Phys. Zeits., 24 (1923),
161.
344
A. H. Compton
dee ted photon is

h
(3)
(1 cos ) ,
mc
where is the angle through whi h the photon is dee ted. The ele tron
at the same time re oils from the photon at an angle of given by,28
=
1
cot = (1 + ) tan ;
2
and the kineti energy of the re oiling ele tron is,
Ekin = h
2 cos2
.
(1 + )2 2 cos2
(4)
(5)
The experiments show in the spe trum of the s attered rays two lines
orresponding to ea h line of the primary ray. One of these lines is of
pre isely the same wavelength as the primary ray, and the se ond line,
though somewhat broadened, has its entre of gravity displa ed by the
amount predi ted by equation (3). A ording to experiments by Kallman
and Marka and by Sharp,b this agreement between the theoreti al29 and
the observed shift is pre ise within a small fra tion of 1 per ent.
The re oil ele trons. From the quantitative agreement between the
theoreti al and the observed wavelengths of the s attered rays, the re oil
ele trons predi ted by the photon theory of s attering were looked for
with some onden e.30 When this theory was proposed, there was
no dire t eviden e for the existen e of su h ele trons, though indire t
eviden e suggested that the se ondary -rays eje ted from matter by
hard -rays are mostly of this type. Within a few months of their
predi tion, however, C. T. R. Wilsonc and W. Bothed independently
announ ed their dis overy. The re oil ele trons show as short tra ks,
pointed in the dire tion of the primary X-ray beam, mixed among the
mu h longer tra ks due to the photoele trons eje ted by the X-rays.
Perhaps the most onvin ing reason for asso iating these short tra ks
with the s attered X-rays omes from a study of their number. Ea h
photoele tron in a loud photograph represents a quantum of truly
absorbed X-ray energy. If the short tra ks are due to re oil ele trons,
ea h one should represent the s attering of a photon. Thus the ratio
Nr /Np of the number of short tra ks to the number of long tra ks should
a
b

d
H. Kallman and H. Mark, Naturwiss., 13 (1925), 297.

H. M. Sharp, Phys. Rev., 26 (1925), 691.
C. T. R. Wilson, Pro . Roy. So . [A, 104 (1923), 1.
W. Bothe, Zeits. f. Phys., 16 (1923), 319.
345
be the same as the ratio / of the s attered to the truly absorbed

energy31 when the X-rays pass through air. The latter ratio is known
from absorption measurements, and the former ratio an be determined
by ounting the tra ks on the photographs. The satisfa tory agreement
between the two ratiosa for X-rays of dierent wavelengths means that
on the average there is about one quantum of energy s attered for ea h
short tra k that is produ ed.
This result is in itself ontrary to the predi tions of the lassi al wave
theory, sin e on this basis all the energy spent on a free ele tron (ex ept
the insigni ant ee t of radiation pressure) should reappear as s attered
X-rays. In these experiments, on the ontrary, 5 or 10 per ent as mu h
energy appears in the motion of the re oil ele trons as appears in the
s attered X-rays.
That these short tra ks asso iated with the s attered X-rays orrespond to the re oil ele trons predi ted by the photon theory of s attering be omes lear from a study of their energies. The energy of the
ele tron whi h produ es a tra k an be al ulated from the range of the
tra k. The ranges of tra ks whi h start in dierent dire tions have been
studieda using primary X-rays of dierent wavelengths, with the result
that equation (5)33 has been satisfa torily veried.
In view of the fa t that ele trons of this type were unknown at the time
the photon theory of s attering was presented, their existen e, and the
lose agreement with the predi tions as to their number, dire tion and
velo ity, supply strong eviden e in favour of the fundamental hypotheses
of the theory.
Interpretation of these experiments. It is impossible to a ount for
s attered rays of altered frequen y, and for the existen e of the re oil
ele trons, if we assume that X-rays onsist of ele tromagneti waves
in the usual sense. Yet some progress has been made on the basis of
semi- lassi al theories. It is an interesting fa t that the wavelength of
the s attered ray a ording to equation (3)34 varies with the angle just
as one would expe t from a Doppler ee t if the rays are s attered from
an ele tron moving in the dire tion of the primary beam. Moreover,
the velo ity that must be assigned to the ele tron in order to give the
proper magnitude to the hange of wavelength is that whi h the ele tron
would a quire by radiation pressure if it should absorb a quantum of the
a A. H. Compton and A. W. Simon, Phys. Rev., 25 (1925), 306; J. M. Nuttall and
E. J. Williams, Man hester Memoirs, 70 (1926), 1.
a Compton and Simon, lo . it.32
346
A. H. Compton
in ident rays. Several writersa have therefore assumed that an ele tron
takes from the in ident beam a whole quantum of the in ident radiation,
and then emits this energy as a spheri al wave while moving forward36
with high velo ity.
This on eption that the radiation o urs in spheri al waves, and that
the s attering ele tron an nevertheless a quire suddenly the impulses
from a whole quantum of in ident radiation is in onsistent with the prin iple of energy onservation. But there is the more serious experimental
di ulty that this theory predi ts re oil ele trons all moving in the same
dire tion and with the same velo ity. The experiments show, on the
other hand, a variety of dire tions and velo ities, with the velo ity and
dire tion orrelated as demanded by the photon hypothesis. Moreover,
the maximum range of the re oil ele trons, though in agreement with
the predi tions of the photon theory, is found to be about four times as
great as that predi ted by the semi- lassi al theory.
There is nothing in these experiments, as far as we have des ribed
them, whi h is in onsistent with the idea of virtual os illators ontinually s attering virtual radiation. In order to a ount for the hange of
wavelength on this view, Bohr, Kramers and Slater assumed that the
virtual os illators s atter as if moving in the dire tion of the primary
beam, a ounting for the hange of wavelength as a Doppler ee t. They
then supposed that o asionally an ele tron, under the stimulation of
the primary virtual rays, will suddenly move forward with a momentum
large ompared with the impulse re eived from the radiation pressure.
Though we have seen that not all of the re oil ele trons move dire tly
forward, but in a variety of dierent dire tions, the theory ould easily
be extended to in lude the type of motion that is a tually observed.
The only obje tion that one an raise against this virtual radiation
theory in onne tion with the s attering phenomena as viewed on a large
s ale, is that it is di ult to see how su h a theory ould by itself predi t
the hange of wavelength and the motion of the re oil ele trons. These
phenomena are dire tly predi table if the onservation of energy and
momentum are assumed to apply to the individual a tions of radiation
on ele trons; but this is pre isely where the virtual radiation theory
denies the validity of the onservation prin iples.
We may on lude that the photon theory predi ts quantitatively and
in detail the hange of wavelength of the s attered X-rays and the hara a C. R. Bauer, C. R., 177 (1923), 1211; C. T. R. Wilson, Pro . Roy. So . [A, 104
(1923), 1; K. Fosterling, Phys. Zeits., 25 (1924), 313; O. Halpern, Zeits. f. Phys.,
30 (1924), 153.35
347
teristi s of the re oil ele trons. The virtual radiation theory is probably
not in onsistent with these experiments, but is in apable of predi ting
the results. The lassi al theory, however, is altogether helpless to deal
with these phenomena.
The origin of the unmodied line The unmodied line is probably
due to X-rays whi h are s attered by ele trons so rmly held within
the atom that they are not eje ted by the impulse from the dee ted
photons. This view is adequate to a ount for the major hara teristi s of
the unmodied rays, though as yet no quantitatively satisfa tory theory
of their origin has been published.a It is probable that a detailed a ount
of these rays will involve denite assumptions regarding the nature and
the duration of the intera tion between a photon and an ele tron; but it
is doubtful whether su h investigations will add new eviden e as to the
existen e of the photons themselves.
A similar situation holds regarding the intensity of the s attered Xrays. Histori ally it was the fa t that the lassi al ele tromagneti theory
is unable to a ount for the low intensity of the s attered X-rays whi h
alled attention to the importan e of the problem of s attering. But
the solutions whi h have been oered by Breit,b Dira c and othersd
of this intensity problem as distinguished from that of the hange of
wavelength, seem to introdu e no new on epts regarding the nature of
radiation or of the s attering pro ess. Let us therefore turn our attention
to the experiments that have been performed on the individual pro ess
of intera tion between photons and ele trons.
Intera tions between radiation and single ele trons
39
The most signi ant of the experiments whi h show departures from
the predi tions of the lassi al wave theory are those that study the
a tion of radiation on individual atoms or on individual ele trons. Two
a Cf., however, G. E. M. Jaun ey, Phys. Rev., 25 (1925), 314 and ibid., 723;
G. Wentzel, Zeits. f. Phys., 43 (1927), 14, 779; I. Waller, Nature, [120, 155 (July
30, 1927).37 [The footnote in the English edition ontinues with the senten e: Ìt is
possible that the theories of the latter authors may be satisfa tory, but they have
not yet been stated in a form suitable for quantitative test' (eds.).
b G. Breit, Phys. Rev., 27 (1926), 242.
P. A. M. Dira , Pro . Roy. So . A, [111 (1926), [405.
d W. Gordon, Zeits. f. Phys., 40 (1926), 117; E. S hrdinger, Ann. d. Phys., 82
(1927), 257; O. Klein, Zeits. f. Phys., 41 (1927), 407; G. Wentzel, Zeits. f. Phys.,
43 (1927), 1, 779.38
348
A. H. Compton
methods have been found suitable for performing these experiments,

Geiger's point ounters, and Wilson's loud expansion photographs.
(1) Test for oin iden es with uores ent X-rays. Bothe has performed an experimenta in whi h uores ent K radiation from a thin opper
foil is ex ited by a beam of in ident X-rays. The emitted rays are so
feeble that only about ve quanta of energy are radiated per se ond.
Two point ounters are mounted, one on either side of the opper foil in
ea h of whi h an average of one photoele tron is produ ed and re orded
for about twenty quanta radiated by the foil. If we assume that the
uores ent radiation is emitted in quanta of energy, but pro eed[s in
spheri al waves in all dire tions, there should thus be about 1 han e
in 20 that the re ording of a photoele tron in one hamber should be
simultaneous with the re ording of a photoele tron in the other.
The experiments showed no oin iden es other than those whi h were
expli able by su h sour es as high-speed -parti les whi h traverse both
ounting hambers.
This result is in a ord with the photon hypothesis,b a ording to
whi h oin iden es should not o ur. It is, nevertheless, equally in a ord
with the virtual radiation hypothesis, if one assumes that the virtual
os illators in the opper ontinuously emit virtual uores ent radiation,
so that the photoele trons should be observed in the ounting hambers
at arbitrary intervals.c
a W. Bothe, Zeits. f. Phys., 37 (1926), 547.
b The English edition ontinues: `For if a photon of uores ent radiation produ es a
-ray in one ounting hamber it annot traverse the se ond hamber. Coin iden es
should therefore not o ur' (eds.).
At this point in the English version Compton is mu h more riti al of the BKS
theory (eds.):
A ording to the virtual radiation hypothesis, however, oin iden es should
have been observed. For on this view the uores ent K radiation is emitted by
virtual os illators asso iated with atoms in whi h there is a va an y in the K
shell. That is, the opper foil an emit uores ent K radiation only during the
short interval of time following the expulsion of a photoele tron from the K
shell, until the shell is again o upied by another ele tron. This time interval is
15
so short (of the order of 10
se .) as to be sensibly instantaneous on the s ale
of Bothe's experiments. Sin e on this view the virtual uores ent radiation is
emitted in spheri al waves, the ounting hambers on both sides of the foil
should be simultaneously ae ted, and oin ident pulses in the two hambers
should frequently o ur. The results of the experiment are thus ontrary to
the predi tions of the virtual radiation hypothesis.
349
But the experiment is important in the sense that it refutes the often
suggested idea that a quantum of radiation energy is suddenly emitted
in the form of a spheri al wave when an atom passes from one stationary
state to another.
(2) The omposite photoele tri ee t.40 Wilsona and Augerb have
noti ed in their loud expansion photographs that when X-rays eje t
photoele trons from heavy atoms, it often o urs that two or more
ele trons are eje ted simultaneously from the same atom. Auger has
dedu ed from studying the ranges of these ele trons that, when this
o urs, the total energy of all the emitted ele trons is no larger than
that of a quantum of the in ident radiation. When two ele trons are
emitted simultaneously it is usually the ase that the enegy of one of
them is
Ekin = h hK ,
whi h a ording to the photon theory means that this ele tron is due to
the absorption of an in ident photon a ompanied by the eje tion of an
ele tron from the K energy level. The se ond ele tron has in general the
energy
Ekin = hK hL .
This ele tron an be explained as the result of the absorption by an L

ele tron of the K-ray emitted when another L ele tron o upies the
pla e left va ant in the K orbit by the primary photoele tron. It is
established that all the ele trons that are observed in the omposite
photoele tri ee t have to be interpreted in the same way. Their interpretation a ording to the photon theory thus meets with no di ulties.
With regard to the virtual radiation theory, we an take two points
of view: rst, under the inuen e of the ex itation produ ed by the
primary virtual radiation, virtual uores ent K radiation is emitted by
virtual os illators asso iated with all the atoms traversed by the primary
beam. In this view, the probability that this virtual uores ent radiation
will ause the eje tion of a photoele tron from the same atom as the one
that has emitted the primary photoele tron is so small that su h an
event will almost never o ur; se ond, we an alternatively assume that
a virtual os illator emitting virtual K radiation is asso iated only with
an atom in whi h there is a va ant pla e in the K shell. In this ase,
a C. T. R. Wilson, Pro . Roy. So . A, 104 (1923), 1.
b P. Auger, Journ. d. Phys., 6 (1926), 183.
350
A. H. Compton
sin e the virtual radiation pro eeds from the atom that has emitted the
primary photoele tron, we ould expe t with extremely large probability
that it should ex ite a photoele tron from the L shell of its own atom,
thus a ounting for the omposite photoele tri ee t. But in this view
the virtual uores ent radiation is emitted only during a very short
interval after the eje tion of the primary photoele tron, in whi h ase
Bothe's uores en e experiment, des ribed above, should have shown
some oin iden es.
One sees thus that the virtual radiation hypothesis is irre on ilable
both with the omposite photoele tri ee t and with the absen e of
oin iden es in Bothe's uores en e experiment. The photon hypothesis,
instead, is in omplete a ord with both these experimental fa ts.
(3) Bothe and Geiger's oin iden e experiments.41 We have seen that
a ording to Bohr, Kramers and Slater's theory, virtual radiation42 is
being ontinually s attered by matter traversed by X-rays, but only
o asionally is a re oil ele tron emitted. This is in sharp ontrast with
the photon theory, a ording to whi h a re oil ele tron appears every
time a photon is s attered. A ru ial test between the two points of
view is aorded by an experiment devised and brilliantly performed43
by Bothe and Geiger.a X-rays were passed through hydrogen gas, and
the resulting re oil ele trons and s attered rays were dete ted by means
of two dierent point ounters pla ed on opposite sides of the olumn of
gas. The hamber for ounting the re oil ele trons was left open, but a
sheet of thin platinum prevented the re oil ele trons from entering the
hamber for ounting the s attered rays. Of ourse not every photon
entering the se ond ounter ould be noti ed, for its dete tion depends
upon the produ tion of a -ray. It was found that there were about ten
re oil ele trons for every s attered photon that re orded itself.
The impulses from the ounting hambers were re orded on a moving
photographi lm. In observations over a total period of over ve hours,
sixty-six su h oin iden es were observed. Bothe and Geiger al ulate
that a ording to the statisti s of the virtual radiation theory the han e
was only 1 in 400 000 that so many oin iden es should have o urred.
This result therefore is in a ord with the predi tions of the photon
theory, but is dire tly ontrary to the statisti al view of the s attering
pro ess.
(4) Dire tional emission of s attered X-rays. Additional information
a W. Bothe and H. Geiger, Zeits. f. Phys., 26 (1924), 44; 32 (1925), 639.
Fig. 3. If the X-rays ex ite a re oil ele tron at an angle

predi ts a se ondary
-parti le
at an angle
351
the photon theory
regarding the nature of s attered X-rays has been obtained by studying

the relation between the dire tion of eje tion of the re oil ele tron and
the dire tion in whi h the asso iated photon pro eeds. A ording to
the photon theory, we have a denite relation (equation (4)) between
the angle at whi h the photon is s attered and the angle at whi h the
re oil ele tron is eje ted. But a ording to any form of spreading wave
theory, in luding that of Bohr, Kramers and Slater, the s attered rays
may produ e ee ts in any dire tion whatever, and there should be no
orrelation between the dire tions in whi h the re oil ele trons pro eed
and the dire tions in whi h the se ondary -rays are eje ted by the
s attered X-rays.
A test to see whether su h a relation exists has been made,a using
Wilson's loud apparatus, in the manner shown diagrammati ally in
Fig. 3. Ea h re oil ele tron produ es a visible tra k, and o asionally a
se ondary tra k is produ ed by the s attered X-ray. When but one re oil
a A. H. Compton and A. W. Simon, Phys. Rev., 26 (1925), 289.
352
A. H. Compton
ele tron appears on the same plate with the tra k due to the s attered
rays, it is possible to tell at on e whether the angles satisfy equation (4).
If two or three re oil tra ks appear,44 the measurements on ea h tra k
an be approximately45 weighted.
Out of 850 plates taken in the nal series of readings, thirty-eight show
both re oil tra ks and se ondary -ray tra ks. On eighteen of these plates
the observed angle 46 is within 20 degrees of the angle al ulated from
the measured value of , while the other twenty tra ks are distributed
at random angles. This ratio 18:20 is about that to be expe ted for the
ratio of the rays s attered by the part of the air from whi h the re oil
tra ks ould be measured to the stray rays from various sour es. There
is only about 1 han e in 250 that so many se ondary -rays should have
appeared at the theoreti al angle.
If this experiment is reliable, it means that there is s attered X-ray
energy asso iated with ea h re oil ele tron su ient to produ e a -ray,
and pro eeding in a dire tion determined at the moment of eje tion
of the re oil ele tron. In other words, the s attered X-rays pro eed in
photons, that is47 in dire ted quanta of radiant energy.
This result, like that of Bothe and Geiger, is irre on ilable with Bohr,
Kramers and Slater's hypothesis of the statisti al produ tion of re oil
and photoele trons. On the other hand, both of these experiments are
in omplete a ord with the predi tions of the photon theory.
Reliability of experimental eviden e

While all of the experiments that we have onsidered are di ult to
re on ile with the lassi al theory that radiation onsists of ele tromagneti waves, only those dealing with the individual s attering pro ess48
aord ru ial tests between the photon theory and the statisti al theory
of virtual radiation. It be omes of espe ial importan e, therefore, to
onsider the errors to whi h these experiments are subje t.
When two point ounters are set side by side, it is very easy to obtain
oin iden es from extraneous sour es. Thus, for example, the apparatus
must be ele tri ally shielded so perfe tly that a spark on the high-tension
outt that operates the X-ray tube may not produ e oin ident impulses
in the two ounters. Then there are high-speed - and -rays, due to
radium emanation in the air and other radioa tive impurities, whi h
may pass through both hambers and produ e spurious oin iden es.
The method whi h Bothe and Geiger used to dete t the oin iden es,
of49 re ording on a photographi lm the time of ea h pulse, makes it
353
possible to estimate reliably50 the probability that the oin iden es are
due to han e. Moreover, it is possible by auxiliary tests to determine
whether spurious oin iden es are o urring for example, by operating
the outt as usual, ex ept that the X-rays are absorbed by a sheet of lead.
It is espe ially worthy of note that in the uores en e experiment the
photon theory predi ted absen e of oin iden es, while in the s attering
experiment it predi ted their presen e. It is thus di ult to see how
both of these ounter experiments an have been seriously ae ted by
systemati errors.
In the loud expansion experiment the ee t of stray radiation is to
hide the ee t sought for, rather than to introdu e a spurious ee t. It
is possible that due to radioa tive ontamination and to stray s attered
X-rays -parti les may appear in dierent parts of the hamber, but
it will be only a matter of han e if these -parti les appear in the
position predi ted from the dire tion of eje tion of the re oil ele trons.
It was in fa t only by taking great are to redu e su h stray radiations
to a minimum that the dire tional relations were learly observed in the
photographs. It would seem that the only form of onsistent error that
ould vitiate the result of this experiment would be the psy hologi al51
one of misjudging the angles at whi h the -parti les appear. It hardly
seems possible, however, that errors in the measurement of these angles
ould be large enough to a ount for the strong apparent tenden y for
the angles to t with the theoreti al formula.
It is perhaps worth mentioning further that the initial publi ations
of the two experiments on the individual s attering pro ess were made
simultaneously, whi h means that both sets of experimenters had independently rea hed a on lusion opposed to the statisti al theory of the
produ tion of the -rays.
Nevertheless,52 given the di ulty of the experiments and the importan e of the on lusions to whi h they have led, it is highly desirable that
both experiments should be repeated by physi ists from other laboratories.
Summary
The lassi al theory that radiation onsists of ele tromagneti waves
propagated in all dire tions through spa e53 is intimately onne ted to
the idea of the ether, whi h is di ult to on eive. It aords no adequate
pi ture of the manner in whi h radiation is emitted or absorbed. It is
in onsistent with the experiments on the photoele tri ee t, and is
354
A. H. Compton
entirely helpless to a ount for the hange of wavelength of s attered

radiation or the produ tion of re oil ele trons.
The theory of virtual os illators and virtual radiation whi h are asso iated statisti ally with sudden jumps of atomi energy and the emission
of photoele trons and re oil ele trons, does not seem to be in onsistent
with any of these phenomena as viewed on a ma ros opi s ale. This
theory, however,54 retains the di ulties inherent in the on eption of
the ether and seems powerless to predi t the hara teristi s of the photoele trons and the re oil ele trons. It55 is further di ult to re on ile
with the omposite photoele tri ee t and is also ontrary to Bothe's
and Bothe and Geiger's oin iden e experiments and to the ray tra k
experiments relating the dire tions of eje tion of a re oil ele tron and of
emission of the asso iated s attered X-ray.
The photon theory avoids the di ulties asso iated with the on eption of the ether.56 The produ tion and absorption of radiation is very
simply onne ted with the modern idea of stationary states. It supplies
a straightforward explanation of the major hara teristi s of the photoele tri ee t, and it a ounts in the simplest possible manner for the
hange of wavelength a ompanying s attering and the existen e of re oil
ele trons. Moreover, it predi ts a urately the results of the experiments
with individual radiation quanta, where the statisti al theory fails.
Unless the four57 experiments on the individual events58 are subje t
to improbably large experimental errors, the on lusion is, I believe,
unes apable that radiation onsists of dire ted quanta of energy, i.e.,
of photons, and that energy and momentum are onserved when these
photons intera t with ele trons or atoms.
Let me say again that this result does not mean that there is no
truth in the on ept of waves of radiation. The on lusion is rather that
energy is not transmitted by su h waves. The power of the wave on ept
in problems of interferen e, refra tion, et ., is too well known to require
emphasis. Whether the waves serve to guide the photons, or whether
there is some other relation between photons and waves is another and
a di ult question.
Dis ussion
355
Dis ussion of Mr Compton's report
Mr Lorentz. I would like to make two omments. First on the
question of the ether. Mr Compton onsiders it an advantage of the

photon theory that it allows us to do without the hypothesis of an ether
whi h leads to great di ulties. I must say that these di ulties do not
seem so great to me and that in my opinion the theory of relativity
does not ne essarily rule out the on ept of a universal medium. Indeed,
Maxwell's equations are ompatible with relativity, and one an well
imagine a medium for whi h these equations hold. One an even, as
Maxwell and other physi ists have done with some su ess, onstru t
a me hani al model of su h a medium. One would have to add only
the hypothesis of the permeability of ponderable matter by the ether to
have all that is required. Of ourse, in making these remarks, I should
not wish to return in any way to these me hani al models, from whi h
physi s has turned away for good reasons. One an be satised with the
on ept of a medium that an pass freely through matter and to whi h
Maxwell's equations an be applied.
In the se ond pla e: it is quite ertain that, in the phenomena of light,
there must yet be something other than the photons. For instqn em in a
dira tion experiment performed with very weak light, it an happen
that the number of photons present at a given instant between the
dira ting s reen and the plane on whi h one observes the distribution
of light, is very limited. The average number an even be smaller than
one, whi h means that there are instants when no photon is present in
the spa e under onsideration.
This learly shows that the dira tion phenomena annot be produ ed
by some novel a tion among the photons. There must be something that
guides them in their progress and it is natural to seek this something
in the ele tromagneti eld as determined by the lassi al theory. This
notion of ele tromagneti eld, with its waves and vibrations would bring
us ba k, in Mr Compton's view, to the notion of ether.
Mr Compton.
It seems, indeed, di ult to avoid the idea of

waves in the dis ussion of opti al phenomena. A ording to Maxwell's
theory the ele tri and magneti properties of spa e lead to the idea of
waves as dire tly as did the elasti ether imagined by Fresnel. Why the
spa e having su h magneti properties should bear the name of ether
is perhaps simply a matter of words. The fa t that these properties of
spa e immediately lead to the wave equation with velo ity c is a mu h
356
A. H. Compton
more solid basis for the hypothesis of the existen e of waves than the
old elasti ether. That something (E and H ) propagates like a wave
with velo ity c seems evident. However, experiments of the kind we have
just dis ussed show, if they are orre t, that the energy of the bundle
of X-rays propagates in the form of parti les and not in the form of
extended waves. So then, not even the ele tromagneti ether appears to
be satisfa tory.
Mr Bragg.
In his report Mr Compton has dis ussed the average

momentum omponent of the ele trons in the dire tion of motion of
the photon, and he has informed us of the on lusion, at whi h several
experimenters have arrived, that this forward average omponent is equal
to the momentum of the light quantum whose energy has been absorbed
and is found again in that of the photoele tron.
I would like to report in this onne tion some results obtained by
Mr Williams.a Mono hromati X-rays, with wavelength lying between
0.5 and 0.7 , enter a Wilson loud hamber ontaining oxygen or
nitrogen. The traje tories of the photoele trons are observed through a
stereos ope and their initial dire tions are measured. Sin e the speed of
the photoele trons is exa tly known (the ionisation energy being weak by
omparison to the quantity h ), a measurement of the initial dire tion
is equivalent to a measurement of momentum in the forward dire tion.
Williams nds that the average momentum omponent in this dire tion
h
is in all ases markedly larger than the quantity h
c or . These results
an be summarised by a omparison with the s heme proposed by Perrin
and Auger ([P. Auger and F. Perrin, Journ. d. Phys. [6th series, vol. 8
(February 1927), [93). They are in perfe t agreement with the cos2
law, provided one assumes that the magneti impulse Tm is equal to
h
1.8 h
c and not just c as these authors assume. One should not atta h
any parti ular importan e to this number 1.8, be ause the range of the
examined wavelengths is too small. I mention it only to show that it
is possible that the simple law proposed by Mr Compton might not be
exa t.
I would like to point out that this method of measuring the forward
omponent of the momentum is more pre ise than an attempt made to
establish results about the most probable dire tion of emission.
Mr Wilson says that his own observations, dis ussed in his Memoir
a Cf. the relevant footnote on p. 340 (eds.).
Dis ussion
357
of 1923a (but whi h do not pretend to be very pre ise) seem to show
that in fa t the forward momentum omponent of the photoele trons
is, on average, mu h larger than what one would derive from the idea
that the absorbed quantum yields all of its momentum to the expelled
ele tron.
Mr Ri hardson. When they are expelled by ertain X-rays, the
ele trons have a momentum in the dire tion of propagation of the rays
equal to 1.8 h
c . If I have understood Mr Bragg orre tly, this result is not
the ee t of some spe i elementary pro ess [a tion, but the average
result for a great number of observations in whi h the ele trons were
expelled in dierent dire tions. Whether or not the laws of energy and
momentum onservation apply to an elementary pro ess, it is ertain
that they apply to the average result for a great number of these pro esses. Therefore, the pro ess [pro essus we are talking about must be
governed by the equations for momentum and energy. If for simpli ity
we ignore the renements introdu ed by relativity, these equations are
h
= mv + M V
c
and
h =
1
1
2
mv 2 + M V ,
2
2
where m and M are the masses, v and V the velo ities of the ele trons
and of the positive residue; the overbars express that these are averages.
The experiments show that the average value of mv is 1.8 h
c and not
h
.
This
means
that
M
V
is
not
zero,
so
that
we
annot
ignore
this term
c
in the equation. If we onsider, for instan e, the photoele tri ee t on
a hydrogen atom, we have to take the ollision energy of the hydrogen
nu leus into a ount in the energy equation.
Mr Lorentz. The term
1
2 59
2 mv ?
2
1
2MV
will however be mu h smaller than
Mr Ri hardson.
It is approximately its 1850th part: that annot

always be onsidered negligible.
Mr Born thinks that he is speaking also for several other members

a Referen ed in footnote on p. 336 (eds.).
358
A. H. Compton
in asking Mr Compton to explain why one should expe t that the momentum imparted to the ele tron be equal to h
c .
Mr Compton.
When radiation of energy h is absorbed by an

atom whi h one surely has to assume in order to a ount for the
kineti energy of the photoele tron the momentum imparted to the
atom by this radiation is h
c . A ording to the lassi al ele tron theory,
when an atom omposed of a negative harge e of mass m and a positive
harge +e of mass M absorbs energy from an ele tromagneti wave,
the momenta imparted to the two elementary harges [le trons are
inversely proportional to their masses. This depends on the fa t that
the forward momentum is due to the magneti ve tor, whi h a ts with
a for e proportional to the velo ity and onsequently more strongly on
the harge having the smaller mass.60 Ee tively, the momentum is thus
re eived by the harge with the smaller mass.
Mr Debye.
Is the reason why you think that the rest of the atom
does not re eive any of the forward momentum purely theoreti al?
Mr Compton. The photographs of the traje tories of the photo-
ele trons show, in a ordan e with Auger's predi tion, that the forward
omponent of the momentum of the photoele tron is, on average, the
same as that of the photon. That means, learly, that on average the
rest of the atom does not re eive any momentum.
Mr Dira .
I have examined the motion of an ele tron pla ed

in an arbitrary for e eld a ording to the lassi al theory, when it is
subje t to in ident radiation, and I have shown in a ompletely general
way that at every instant the fra tion of the rate of hange [vitesse de
variation of the forward momentum of the ele tron due to the in ident
radiation is equal to 1c times the fra tion of the rate of hange of the
energy due to the in ident radiation. The nu leus and the other ele trons
of the atom produ e hanges of momentum and of energy that at ea h
instant are simply added to those produ ed by the in ident radiation.
Sin e the radiation must modify the ele tron's orbit, it must also hange
the fra tion of the rate of hange of the momentum and of the energy
that omes from the nu leus and the other ele trons, so that it would be
ne essary to integrate the motion in order to determine the total hange
produ ed by the in ident radiation in the energy and the momentum.
Dis ussion
359
Mr Born. I would like to mention here a paper by Wentzel,
whi h ontains a rigorous treatment of the s attering of light by atoms

a ording to quantum me hani s. In it, the author onsiders also the
inuen e of the magneti for e, whi h allows him to obtain the quantum
analogue of the lassi al light pressure. It is only in the limiting ase of
very short wavelengths that one nds that the momentum of light h
c is
ompletely transmitted to the ele tron; in the ase of large wavelengths
an inuen e of the binding for es appears.
Mr Ehrenfest.
One an show by a very simple example where

the surplus of forward momentum, whi h we have just dis ussed, an
have its origin. Take a box whose inner walls ree t light ompletely,
but diusedly, and assume that on the bottom there is a little hole.
Through the latter I shine a ray of light into the box whi h omes and
goes inside the box and pushes away its lid and bottom. The lid then
has a surplus of forward momentum.
Mr Bohr.
With regard to the question of waves or photons dis ussed by Mr Compton, I would like to make a few remarks, without preempting the general dis ussion. The radiation experiments have indeed
revealed features that are not easy to re on ile within a lassi al pi ture.
This di ulty arises parti ularly in the Compton ee t itself. Several
aspe ts of this phenomenon an be des ribed very simply with the aid of
photons, but we must not forget that the hange of frequen y that takes
pla e is measured using instruments whose fun tioning is interpreted
a ording to the wave theory. There seems to be a logi al ontradi tion
here, sin e the des ription of the in ident wave as well as that of the
s attered wave require that these waves be nitely extended [limites
in spa e and time, while the hange in energy and in momentum of
the ele tron is onsidered as an instantaneous phenomenon at a given
point in spa etime. It is pre isely be ause of su h di ulties that Messrs
Kramers, Slater and myself were led to think that one should ompletely
reje t the idea of the existen e of photons and assume that the laws of
onservation of energy and momentum are true only in a statisti al way.
The well-known experiments by Geiger and Bothe and by Compton
and Simon, however, have shown that this point of view is not admissible
a
a Born is presumably referring to Wentzel's se ond paper on the photoele tri ee t
(Wentzel 1927). Compare Mehra and Re henberg (1987, pp. 835 .) (eds.).
a This dis ussion ontribution by Bohr is reprinted and translated also in vol. 5 of
Bohr's Colle ted Works (Bohr 1984, pp. 20712). (eds.).
360
A. H. Compton
and that the onservation laws are valid for the individual pro esses, in
a ordan e with the on ept of photons. But the dilemma before whi h
we are pla ed regarding the nature of light is only a typi al example
of the di ulties that one en ounters when one wishes to interpret
the atomi phenomena using lassi al on epts. The logi al di ulties
with a des ription in spa e and time have sin e been removed in large
part by the fa t that it has been realised that one en ounters a similar
paradox with respe t to the nature of material parti les. A ording to
the fundamental ideas of Mr de Broglie, whi h have found su h perfe t
onrmation in the experiments of Davisson and Germer, the on ept
of waves is as indispensable in the interpretation of the properties of
material parti les as in the ase of light. We know thereby that it is
equally ne essary to attribute to the wave eld a nite extension in
spa e and in time, if one wishes to dene the energy and the momentum
of the ele tron, just as one has to assume a similar nite extension in
the ase of the light quantum in order to be able to talk about frequen y
and wavelength.
Therefore, in the ase of the s attering pro ess, in order to des ribe the
two hanges ae ting the ele tron and the light we must work with four
wave elds (two for the ele tron, before and after the phenomenon, and
two for the quantum of light, in ident and s attered), nite in extension,
whi h meet in the same region of spa etime.a In su h a representation
all possibility of in ompatibility with a des ription in spa e and time
disappears. I hope the general dis ussion will give me the opportunity
to enter more deeply into the details of this question, whi h is intimately
tied to the general problem of quantum theory.
Mr Brillouin.
I have had the opportunity to dis uss Mr Compton's report with Mr Auger,b and wish to make a few omments on this
topi . A purely orpus ular des ription of radiation is not su ient to
understand the pe uliarities of the phenomena; to assume that energy
is transported by photons h is not enough to a ount for all the ee ts
of radiation. It is essential to omplete our information by giving the
dire tion of the ele tri eld; we annot do without this eld, whose role
in the wave des ription is well known.
I shall re all in this ontext a simple argument, re ently given by Auger
a Compare also the dis ussion ontributions below, by Pauli, S hrdinger and others
(eds.).
b As noted in se tion 1.4, this and other reports had been ir ulated among the
parti ipants before the onferen e (eds.).
Dis ussion
361
Fig. 1.
and F. Perrin, and whi h illustrates learly this remark. Let us onsider
the emission of ele trons by an atom subje t to radiation, and let us
examine the distribution of the dire tions of emission. This distribution
has usually been observed in a plane ontaining the light ray and the
dire tion of the ele tri eld (the in ident radiation is assumed to be
polarised); let be the angle formed by the dire tion of emission of the
photoele trons and the ele tri eld h; as long as the in ident radiation is
not too hard, the distribution of the photoele trons is symmetri around
the ele tri eld; one an then show that the probability law ne essarily
takes the form A cos2 . Indeed, instead of observing the distribution
in the plane of in iden e (Fig. 1), let us examine it in the plane of the
wave; the same distribution law will still be valid; and it is the only one
that would allow us to obtain, through the superposition of two waves
polarised at right angles, an entirely symmetri distribution
A cos2 + A sin2 = A .
Now, from the point of view of waves, one must ne essarily obtain
this result, a beam [rayonnement of natural light having no privileged
dire tion in the plane of the wave. These symmetry onsiderations, whi h
any theory of radiation must respe t, provide a substantial di ulty
for the stru tural theories of the photon (Bubb's quantum ve tor, for
instan e).
Summing up, the dis ontinuity of the radiation manifests itself just in
the most elementary way, through the laws of onservation of energy and
momentum, but the detailed analysis of the phenomena is interpreted
more naturally from the ontinuous point of view. For the problem of
emission of the photoele trons, a omplete theory has been given by
362
A. H. Compton
Wentzel, by means of wave me hani s.a He nds the A cos2 law of

F. Perrin and Auger for radiation of low penetration; when the radiation
is harder, Wentzel obtains a more omplex law, in whi h the ele trons
tend to be emitted in larger numbers in the forward dire tion. His theory,
however, seems in omplete with regard to this point sin e, if I am not
mistaken, he has assumed the immobility of the atomi nu leus; now,
nothing tells us a priori how the momentum h
c of the photon is going
to be distributed between the nu leus and the emitted ele tron.
Mr Lorentz.
Allow me to point out that a ording to the old

ele tron theory, when one has a nu leus and an ele tron on whi h a
beam of polarised light falls, the initial angular momentum of the system
is always onserved. The angular momentum imparted to the ele tronnu leus system will be provided at the expense of the angular momentum
of the radiation eld.
Mr Compton.
The on eption of the photon diers from the

lassi al theory in that, when a photoele tron is emitted, the photon is
ompletely absorbed and no radiation eld is left. The motion of the
photoele tron must thus be su h that the nal angular momentum of
the ele tron-nu leus system will be the same as the initial momentum of
the photon-ele tron-nu leus system. This ondition restri ts the possible
traje tories of the emitted photoele tron.
Mr Kramers.
In order to interpret his experiments, Mr Compton

needs to know how the absorption is divided between a omponent ,
due to the `true' absorption, and a omponent due to the s attering.
We do not know with ertainty that, if an be written in the form Ca +
D, the onstant D truly represents the s attering for large wavelengths,
where Ca is no longer small ompared to D. In general, thus, spe i
measurements of are ne essary. Did you have su ient information
regarding the values of and in your experiments?
Mr Compton.
The most important ase in whi h it is ne essary

to distinguish between the true absorption and the absorption due to
the s attering, is that of arbon. For this ase, Hewletta has measured
a This is presumably Wentzel's treatment from his rst paper on the photoele tri
ee t (Wentzel 1926). Compare again Mehra and Re henberg (1987, pp. 835 .)
(eds.).
a [C. W. Hewlett, Phys. Rev., 17 (1921), 284.
Dis ussion
363
dire tly for the wavelength 0.71 and the total absorption over
a large range of wavelengths. The dieren e between and for the
wavelength 0.71 orresponds to for this wavelength. A ording to
Owen's formula this is proportional to 3 ; we an thus al ulate for
all wavelengths. The dieren e between this value of and the measured
value of orresponds to the value of for the wavelengths onsidered.
Sin e is relatively small in the ase of arbon, espe ially for small
wavelengths, this pro edure yields a value for that annot be very
impre ise.
Mr Bragg. When one onsults the original literature on this
subje t, one is stru k by how mu h the X-ray absorption measurements

leave to be desired, both with regard to pre ision as well as with regard
to the extent of the s ale of wavelengths for whi h they have been
performed.
Mr Pauli. How large is the broadening of the modied rays?

Mr Compton. The experiments have shown learly that the mo-
died ray is broader than the unmodied ray. In the typi al ase of
the ray 0.7 s attered by arbon, the broadening is of order 0.005
angstrm. Unfortunately, the experiments on erning this point are far
from being satisfa tory, and this number should be onsidered only as a
rough approximation.
Mr Pauli.
The broadening of the modied ray an be interpreted

theoreti ally in two ways, whi h to tell the truth redu e to the same
a ording to quantum me hani s. First, the ele tron, in a given stationary state of the atom, has a ertain velo ity distribution with regard to
magnitude and dire tion. That gives rise to a broadening of the frequen y
of the s attered rays through the Doppler ee t, a broadening whose
v
order of magnitude is
= c , where v denotes the average velo ity of
the ele tron in the atom.
In order to onvey the se ond means of explanation, I would like to
sket h briey the meaning of the Compton ee t in wave me hani s.a
This meaning is based rst of all on the wave equation
!
X 2
4 2 e2 X 2
4i e X
+ m20 c2 = 0
2
2
x
h
c
x
h
c
a For a modern dis ussion, see Bjrken and Drell (1964, Chapter 9) (eds.).
364
A. H. Compton
and further on the expression

iS =
4i e

+
,
x
x
h c
in whi h is S hrdinger's fun tion, the omplex onjugate value

and the four-potential of the ele tromagneti eld. Given S , one
al ulates the radiation from lassi al ele trodynami s. If now in the
wave equation one repla es by the potential of an in ident plane
wave, the terms that are proportional to the amplitude of this wave an
be onsidered innitely small in the rst order, and one an apply the approximation methods of perturbation theory. This now is a point where
one needs to be espe ially areful. It is all-important to know what one
will take as the unperturbed eld , whi h must orrespond to a solution
of the wave equation for the free parti le ( orresponding to = 0).61
One nds that in order to agree with the observations, it is ne essary to
take two innitely extended mono hromati wave trains as being already
present in the unperturbed solution, of whi h one orresponds to the
initial state, the other to the nal state of the Compton pro ess. In my
opinion this assumption, on whi h the theories of the Compton ee t
by S hrdinger, Gordon and Klein are based, is unsatisfa tory and this
defe t is orre ted only by Dira 's quantum ele trodynami s.a But if
one makes this assumption, the urrent distribution of the unperturbed
solution orresponds to that of an innitely extended dira tion grating
[un rseau inniment tendu that moves with a onstant speed, and the
a tion of the radiation on this grating leads to a sharp modied ray.
If one onsiders a bound ele tron in an atom, one has to repla e one
omponent of the solution in the unperturbed harge and urrent
distribution by the eigenfun tion of the atom in the stationary state
onsidered, and the other omponent by a solution orresponding to the
nal state of the Compton pro ess (belonging to the ontinuous spe trum
of the atom), whi h at great distan e from the atom behaves more or
less as a plane wave. One thus has a moving grating that rst of all
depends only on the nite extension of the atom and in the se ond pla e
has omponents no longer moving with the same speed at all. This gives
rise to a la k of sharpness of the shifted ray of the s attered radiation.
But one an show that, from the point of view of quantum me hani s,
this explanation for the la k of sharpness of the shifted ray is just another
form of the explanation given in the rst instan e and whi h relies on the
dierent dire tions of the initial velo ities of the ele trons in the atom.
a Compare below S hrdinger's ontribution and the ensuing dis ussion (eds.).
Dis ussion
365
For a ording to quantum me hani s if

= f (x, y, z)e2ivt
is the eigenfun tion orresponding to a given stationary state of the

atom, the fun tion62
Z Z Z
2i
(px , py , pz ) =
f (x, y, z)e (px x+py y+pz z) dxdydz ,
whi h one obtains by de omposing f in plane waves a ording to Fourier an be interpreted in the sense that |(p)|2 dpx dpy dpz denotes the
probability that in the given stationary state the omponents of the
momentum of the ele tron lie between px , py , pz and px + dpx , et .
Now, if through the resulting velo ity distribution of the ele trons in
the atom one al ulates the broadening of the shifted line a ording to
the rst point of view, for light of su iently short wavelength with
respe t to whi h the ele tron an be onsidered free in the atom (and
it is only under these onditions that the pro edure is legitimate), one
nds exa tly the same result as with the other method des ribed.a
Mr Compton.
Jaun ey has al ulated the broadening of the modied ray using essentially the method that Mr Pauli has just des ribed.
Jaun ey assumed, however, that the velo ities of the ele tron are the
ones given by Bohr's theory of orbital motions. The broadening thus
obtained is larger than that found experimentally.
Mrs Curie.
In his very interesting report, Professor Compton has

dwelt on emphasising the reasons that lead one to adopt the theory of
a ollision between a quantum and a free ele tron. Along the same line
of thought, I think it is useful to point out the following two views:
First, the existen e of ollision ele trons seems to play a fundamental
role in the biologi al ee ts produ ed on living tissues by very highfrequen y radiation, su h as the most penetrating -rays emitted by
radioelements. If one assumes that the biologi al ee t may be attributed
to the ionisation produ ed in the ells subje ted to radiation, this ee t
annot depend dire tly on the -rays, but is due to the emission of
se ondary -rays that a ompanies the passage of the -rays through
matter. Before the dis overy of the ollision ele trons, only a single
a Pauli was possibly the rst to introdu e the probability interpretation of the wave
fun tion in momentum spa e, in a letter to Heisenberg of 19 O tober 1926 (Pauli,
1979, pp. 3478). Cf. the footnote on p. 117 (eds.).
366
A. H. Compton
me hanism was known for the produ tion of these se ondary rays, that
onsisting in the total absorption of a quantum of radiation by the
atom, with the emission of a photoele tron. The absorption oe ient
relating to this pro ess varies with the wavelength of the primary
-radiation, as well as with the density [ of the absorbing matter
and the atomi number N of the atoms omposing it, a ording to
the well-known relation of Bragg and Peir e = N 3 3 , where is
a oe ient that has a onstant value for frequen ies higher than that
of the K dis ontinuity. If this relation valid in the domain of X-rays
an be applied to high-frequen y -rays, the resulting value of for the
light elements is so weak that the emission of photoele trons appears
unable to explain the biologi al ee ts of radiation on the living tissues
traversed.a
The issue appears altogether dierent if one takes into onsideration
the emission of ollision ele trons in these tissues, following Compton's
theory. For a ollimated primary beam of -rays, the fra tion of ele tromagneti energy onverted into kineti energy of the ele trons per unit
mass of the absorbing matter is given by the oe ient
a
0
=
,
2
(1 2)
where 0 is the s attering oe ient per unit mass valid for medium
frequen y X-rays, a ording to the theory of J. J. Thomson, and is lose
h
to 0.2, while is Compton's parameter = mc
2 (h Plan k's onstant,
primary frequen y, m rest mass of the ele tron, c speed of light). Taking
= 1.2, a value suitable for an important group of -rays (equivalent
potential 610 kilovolts), one nds a = 0.02, that is, 2 per ent of the
primary energy is onverted to energy of the ele tron per unit mass
of absorbing matter, when e a possibility of interpreting the observed
biologi al ee ts. To this dire t produ tion of ollision ele trons along
the traje tory of the primary beam is added, in an extended medium, a
supplementary produ tion, from the fa t that to ea h of these ele trons
orresponds a s attered quantum, with a smaller value than the primary
quantum, and that this s attered quantum an in turn be subje t to
the Compton ee t in the medium through whi h it propagates, with
produ tion of a new ollision ele tron and of an even smaller quantum.
This pro ess, indenitely repeatable and alled the `multiple Compton
ee t' seems in fa t to have been observed by ertain authors.a Not only
a It is true that several authors have re ently ontested the legitima y of extending
the absorption law of Bragg and Peir e to X-rays.
a [B. Rajewsky, Forts hritte auf dem Gebiet der Roentgenstrahlung, 35 (1926), 262.
Dis ussion
367
is the number of ollision ele trons thereby multiplied, but, further, the
primary quantum, redu ed by su essive ollisions takes on values for
whi h the absorption with emission of photoele trons be omes more and
more probable.
These fa ts have an important reper ussion on the te hnique of X-ray
therapy. Certain authors had, in fa t, denied the usefulness of produ ing
very high-voltage apparatus providing X-rays of very high frequen y and
very high penetrating power, whose use is otherwise onvenient owing to
the uniformity of irradiation they allow one to attain. If these rays had
been devoid of e a y, one would have had to give up on their use. Su h
is not the ase if one adopts the point of view of the Compton ee t,
and it is then legitimate to dire t the te hnique towards the use of high
voltages.
Another interesting point of view to examine is that of the emission of
-rays by radioa tive bodies. Professor Compton has pointed out that
among the -rays of se ondary origin, some ould be ollision ele trons
produ ed by the s attering of the primary -rays on the ele trons ontained in the matter they traverse.
It is in an ee t of this type that Thibaud thinks one may nd the
explanation for the appearan e of the magneti spe tra of the se ondary
-rays. These spe tra are omposed of lines that may be attributed to
groups of photoele trons of the same speed, ea h of whi h is emitted by
absorption in a thin metalli envelope of a group of homogeneous -rays
emitted by a radioelement ontained in this envelope. Ea h line of photoele tri origin is a ompanied by a band beginning at the line itself and
extending towards the region of low velo ities. Thibaud thinks that this
band ould be due to photoele trons expelled from the s reen by those
-rays that, in this same s reen, had suered the Compton ee t with
redu tion of frequen y. This interpretation appears plausible; however,
in order to prove it, it would be ne essary to study the stru ture of the
band and nd in the same spe trum the band that may be attributed
to the ollision ele trons orresponding to the s attered -rays.
An analogous problem arises regarding the emission of -rays by
radioa tive bodies with negligible thi kness, so as to eliminate, as far
as possible, the se ondary ee ts due to the supports and envelopes.
One then observes a magneti spe trum that may be attributed to the
radioelement alone and onsisting either of a ontinuous band, or of
the superposition of a ontinuous spe trum and a line spe trum. The
latter has re eived a satisfa tory interpretation in some re ent papers
(L. Meitner, Ellis, Thibaud, et .).
368
A. H. Compton
A line is due to a group of photoele trons with the same speed expelled
from the levels of the radioa tive atoms by a group of homogeneous
-rays produ ed in their nu lei. This ee t is alled ìnternal onversion',
sin e one assumes that the quantum emitted by an atomi nu leus is
reabsorbed in the ele tron loud [enveloppe le tronique of the same
atom. The great majority of observed lines nd their explanation in this
hypothesis.
The interpretation of the ontinuous spe trum appears to present
more di ulties. Some authors attribute it only to the primary -rays,
while others onsider the possibility of a se ondary origin and invoke
the Compton ee t as a possible ause of its produ tion (L. Meitner).
This would be an ìnternal' Compton ee t, su h that a -ray emitted
from the nu leus of an atom would experien e a ollision with one of
the weakly bound ele trons at the periphery of the same atom. If that
were the ase, the velo ity distribution of the emitted ollision ele trons
would not be arbitrary, but would have to onform to the predi tions of
Compton's theory.
I have losely examined this problem, whi h has a very omplex appearan e.a Ea h group of homogeneous -rays is a ompanied by s attered
-rays, so that in the dira tion spe trum of the -rays, ea h line should
experien e a broadening of 0.0485 units. The experiments on the
dira tion of -rays are di ult and not very numerous; so far the
broadening ee t has not been reported.
Ea h homogeneous group of -rays must orrespond to a group of
ollision ele trons, whose velo ity varies ontinuously from zero to an
upper limit derived from Compton's theory and whi h in the magneti
spe trum orresponds to a band bounded sharply on the side of the large
velo ities. The same group of -rays may orrespond to further groups
of photoele trons expelled from the dierent levels K, L, et . of the atom
through internal absorption of the s attered -rays. For ea h group of
photoele trons, the velo ity of emission lies between two well-dened
limits. The upper limit orresponds to the surplus energy of the primary
-rays with respe t to the extra tion work W hara teristi of the given
level; the lower limit orresponds to the surplus energy, with respe t to
the same work, of the -rays s attered in the dire tion opposite to that
of the primary rays, and having experien ed be ause of that the highest
loss of frequen y. In the magneti spe trum, ea h group of photoele trons
will be represented by a band equally well bounded on the side of the
a [M. Curie, Le Journal de Physique et le Radium, 7 (1926), 97.
Dis ussion
369
large and of the small velo ities, with the same dieren e between the
extreme energies for ea h band.
It is easy to see that in the same magneti spe trum the dierent bands
orresponding to the same group of -rays may partially overlap, making
it di ult to analyse the spe trum omparing the distribution of -rays
with that predi ted by theory. For substan es emitting several groups of
-rays, the di ulty must be ome onsiderable, unless there are large
dieren es in their relative ee tiveness in produ ing the desired ee t.
Let us also point out that the ontinuous spe trum due to the Compton
ee t may be superposed with a ontinuous spe trum independent of
this ee t (that may be attributed for instan e to the primary -rays).
Examination of the experimental data available so far does not yet
allow one to draw on lusions onvin ingly. Most of the spe tra are very
omplex, and their pre ise study with respe t to the energy distributions
of the -rays will require very detailed work. In ertain simple spe tra
su h as that of the -rays of RaD, one observes lines of photoele tri
origin that may be attributed to a single group of mono hromati rays. These lines form the upper edge of bands extending towards low
velo ities and probably arising from photoele trons produ ed by the
s attered -rays. In ertain magneti spe tra obtained from the -rays
of mesothorium 2 in the region of low velo ities, one noti es in the
ontinuous spe trum a gap that might orrespond, for the group of
primary -rays with 58 kilovolts, to the separation between the band
due to the ollision ele trons and that due to the photoele trons of the
s attered -rays.a
Mr S hrdinger
, at the invitation of Mr Ehrenfest, draws on the

bla kboard in oloured halk the system of four wave trains by whi h he
has tried to represent the Compton ee t in an ans hauli h way [d'une
faon intuitiveb (Ann. d. Phys. 4th series, vol. 82 (1927), 257).c
Mr Bohr.
The simultaneous onsideration of two systems of waves

has not the aim of giving a ausal theory in the lassi al sense, but one
an show that it leads to a symboli analogy. This has been studied
in parti ular by Klein. Furthermore, it has been possible to treat the
a D. K. Yovanovit h and A. Prola, Comptes Rendus, 183 (1926), 878.
b For dis ussions of the notion of Ans hauli hkeit, see se tions 3.4.7, 4.6 and 8.3
(eds.).
S hrdinger (1928, p. x) later remarked on a mistake pointed out to him by
Ehrenfest in the gure as published in the original paper (eds.).
370
A. H. Compton
problem in more depth through the way Dira has formulated S hrdinger's theory. We nd here an even more advan ed renun iation of
Ans hauli hkeit [intuitivit, a fa t very hara teristi of the symboli
methods in quantum theory.
Mr Lorentz.
Mr S hrdinger has shown how one an explain the

Compton ee t in wave me hani s. In this explanation one onsiders
the waves asso iated with the ele tron (e) and the photon (ph), before
(1) and after (2) the en ounter. It is natural to think that, of these
four systems of waves e1 , ph1 , e2 and ph2 , the latter two are produ ed
by the en ounter. But they are not determined by e1 and ph1 , be ause
one an for example hoose arbitrarily the dire tion of e2 . Thus, for
the problem to be well-dened, it is not su ient to know e1 and ph1 ;
another pie e of data is ne essary, just as in the ase of the ollision
of two elasti balls one must know not only their initial velo ities but
also a parameter that determines the greater or lesser e entri ity of
the ollision, for instan e the angle between the relative velo ity and
the ommon normal at the moment of the en ounter. Perhaps one ould
introdu e into the explanation given by Mr S hrdinger something that
would play the role of this a essory parameter.
Mr Born.
I think it is easy to understand why three of the four

waves have to be given in order for the pro ess to be determined; it
su es to onsider analogous ir umstan es in the lassi al theory. If
the motions of the two parti les approa hing ea h other are given, the
ee t of the ollision is not yet determined; it an be made determinate
by giving the position of losest approa h or an equivalent pie e of data.
But in wave me hani s su h mi ros opi data are not available. That
is why it is ne essary to pres ribe the motion of one of the parti les
after the ollision, if one wants the motion of the se ond parti le after
the ollision to be determined. But there is nothing surprising in this,
everything being exa tly as in lassi al me hani s. The only dieren e
is that in the old theory one introdu es mi ros opi quantities, su h as
the radii of the atoms that ollide, whi h are eliminated from subsequent
al ulations, while in the new theory one avoids the introdu tion of these
quantities.
Notes to pp. 329347
371

1 The words `rels ou apparents' are present only in the Fren h version.
2 The English version has only four headings (starting with `(1) How are
the waves produ ed?'), and a ordingly omits the next se tion, on `The
problem of the ether', and later referen es to the ether.
3 [d'aprs les rsultats de l'tude des spe tres
4 The English edition distinguishes se tions and subse tions more
systemati ally than the Fren h edition, and in this and other small
details of layout we shall mostly follow the former.
5 [rappeler qu'il existe une thorie dans laquelle
6 The words `le professeur' are present only in the Fren h edition.
7 [`lment de radiation' ou `quantum de lumire'
8 This se tion is present only in the Fren h version.
9 [nergie ondulatoire
10 The original footnote gives page `145'.
213'.
11 The English edition has `

12 [ part une onstante
13 [perfe tionne
14 The se ond part of the footnote is printed only in the English edition.
15 The Fren h edition here in ludes the lause `qui a t faite'.
16 Here and in several pla es in the following, the Fren h edition has
`simple' where the English one has `single'.
17 [l'a tion dire te du ve teur le trique de l'onde, prise dans le sens
ordinaire
18 [dans la dire tion de propagation de l'onde
19 [puisque
20 This footnote is only present in the English edition.
21 The pre eding lause is only present in the Fren h edition.
22 [la priode de l'le tron dans son mouvement orbital
23 In the English edition this reads: `The fa t that the photoele trons
re eive the momentum of the in ident photon'.
24 [sur les rayons X se ondaires et les rayons
25 This footnote appears only in the Fren h edition.

26 [ne fournirent au une preuve de l'existen e d'un spe tre de raies
27 This word is missing in the Fren h edition.
28 The English edition reads `(1
+ x)'.
29 [prdit
30 [on eut quelque onan e dans les le trons de re ul
31 The Fren h edition uses
dis ussion (where
instead of
in the text, but uses
in the
is used for matter density). The English edition uses
t.
32 Footnote mark missing in the Fren h edition.
33 The Fren h edition gives (4).
34 The Fren h edition gives (2).
35 This footnote is present only in the English edition.
36 This word is missing in the Fren h edition.
37 The Fren h edition reads `J. Waller'.
38 The page numbers for Wentzel appear only in the Fren h edition.
39 The English edition des ribes only three experiments, omitting the
372
Notes to pp. 349365
se tion on the omposite photoele tri ee t as well as referen es to it

later.
40 This se tion is present only in the Fren h edition.
41 This and the next se tion are of ourse numbered (2) and (3) in the
English edition.
42 [rayonnement de uores en e
43 [une exprien e ru iale entre les deux points de vue a t imagine et
brillamment ralise
44 The Fren h edition in ludes also èn mme temps'.
45 [d'une faon approprie
46 The Fren h edition has ` '.
47 The words `photons, 'est--dire' are present only in the Fren h edition.
48 [au phnomne de la diusion par les le trons individuels
49 [ou
50 [ave ertitude
51 [physiologique
52 This senten e is only printed in the Fren h edition.
53 The English edition omits referen e to the ether and ontinues dire tly
with àords no adequate pi ture'.
54 The English edition ontinues dire tly with: `seems powerless'.
55 The English edition ontinues: ìs also ontrary to'.
56 In the English edition this reads simply: À ording to the photon theory,
the produ tion .... '.
57 In the English edition: `three'.
58 [pro essus
59 Overbars have been added.
60 The Fren h text reads àve moins d'intensit sur la harge ayant la plus
petite masse'. This is evidently an error: the (transverse) velo ity of the
harges stems from the ele tri eld, whi h imparts the larger velo ity to
the harge with the smaller mass, whi h therefore experien es the larger
magneti for e.
61 The printed text reads ` '.
62 Bra kets in the exponent added.
The new dynami s of quanta
By Mr Louis de BROGLIE
I. Prin ipal points of view
1. First works of Mr Louis de Broglie [1. In his rst works on the

Dynami s of Quanta, the author of the present report started with the
following idea: taking the existen e of elementary orpus les of matter
and radiation as an experimental fa t, these orpus les are supposed to
be endowed with a periodi ity. In this way of seeing things, one no longer
on eives of the `material point' as a stati entity pertaining to only a
tiny region of spa e, but as the entre of a periodi phenomenon spread
all around it.
Let us onsider, then, a ompletely isolated material point and, in
a system of referen e atta hed to this point, let us attribute to the
postulated periodi phenomenon the appearan e of a stationary wave
dened by the fun tion
u(x0 , y0 , z0 , t0 ) = f (x0 , y0 , z0 ) cos 20 t0 .
In another Galilean system x, y , z , t, the material point will have a

re tilinear and uniform motion with velo ity v = c. Simple appli ation
a Our translation of the title (`La nouvelle dynamique des quanta') ree ts de
Broglie's frequent use of the word `quantum' to refer to a (pointlike) parti le,
an asso iation that would be lost if the title were translated as, for example, `The
new quantum dynami s' (eds.).
b On beginning this exposition, it seems right to underline that Mr Mar el Brillouin
was the true pre ursor of wave Me hani s, as one may realise by referring to the
following works: C. R. 168 (1919), 1318; 169 (1919), 48; 171 (1920), 1000.
Journ. Physique 3 (1922), 65.
373
374
L. de Broglie
of the Lorentz transformation shows that, as far as the phase is on erned, in the new system the periodi phenomenon has the appearan e
of a plane wave propagating in the dire tion of motion whose frequen y
and phase velo ity are
0
=p
,
1 2
V =
c2
c
= .
v
The appearan e of this phase propagation with a speed superior to c,

as an immediate onsequen e of the theory of Relativity, is quite striking.
There exists a noteworthy relation between v and V . The formulas
giving and V allow us in fa t to dene a refra tive index of the
va uum, for the waves of the material point of proper frequen y 0 ,
by the dispersion law
r
c
2
n=
= 1 02 .
V
One then easily shows that

1 (n)
1
=
,
v
c
that is, that the velo ity v of the material point is equal to the group
velo ity orresponding to the dispersion law.
With the free material point being thus dened by wave quantities,
the dynami al quantities must be related ba k to these. Now, sin e the
frequen y transforms like an energy, the obvious thing to do is to
assume the quantum relation
W = h ,
a relation that is valid in all systems, and from whi h one derives the
undulatory denition of the proper mass m0
m0 c2 = h0 .
Let us write the fun tion representing the wave in the system x, y, z, t
in the form
u(x, y, z, t) = f (x, y, z, t) cos
2
(x, y, z, t) .
h
Denoting by W and p the energy and momentum, one easily shows that
375
one hasa,b
W =
,
t
p = grad .
The fun tion is then none other than the Ja obi fun tion.a One
dedu es from this that, in the ase of uniform re tilinear motion, the
prin iples of least a tion and of Fermat are identi al.
To look for a generalisation of these results, let us now assume that
the material point moving in a eld derived from a potential fun tion
F (x, y, z, t) is represented by the fun tion
u(x, y, z, t) = f (x, y, z, t) cos
2
(x, y, z, t) ,
h
where is the Ja obi fun tion of the old Dynami s. This assimilation
of the phase into the Ja obi fun tion then leads us to assume the following two relations, whi h establish a general link between me hani al
quantities and wave quantities:
W = h =
,
t
p =
= grad .
V
One then dedu es that, for the waves of the new Me hani s, the spa e
o upied by the eld has a refra tive index
s
2
2
F
02 .
1
n=
h
Hamilton's equations show in addition that, here again, the velo ity of
the moving body is equal to the group velo ity.b
These on eptions lead to an interpretation of the stability onditions
introdu ed by quantum theory. If, indeed, one onsiders a losed traje tory, the phase must be a single-valued fun tion along this urve, and
as a result one is led to write the Plan k ondition1
I
(p dl) = k h (k integer) .
The Sommerfeld onditions for quasi-periodi motions may also be derived. The phenomena of quantum stability thus appear to be analogous
to phenomena of resonan e, and the appearan e of whole numbers here
a These are the relativisti guidan e equations of de Broglie's early pilot-wave theory
of 192324, for the spe ial ase of a free parti le (eds.).
b The ve tor `grad ' is the ve tor whose omponents are /x, /y , /z .
a Usually alled the Hamilton-Ja obi fun tion (eds.).
b In the ase of motion of a point harge in a magneti eld, spa e behaves like an
anisotropi medium (see Thesis, p. 39).
376
L. de Broglie
be omes as natural as in the theory of vibrating strings or plates. Nevertheless, as we shall see, the interpretation that has just been re alled
still onstitutes only a rst approximation.
The appli ation of the new on eptions to orpus les of light leads to
di ulties if one onsiders their proper mass to be nite. One avoids
these di ulties by assuming that the properties of the orpus les of
light are dedu ed from those of ordinary material points by letting the
proper mass tend to zero. The two speeds v and V then both tend to c,
and in the limit one obtains the two fundamental relations of the theory
of light quanta
h
=p,
h = W ,
c
with the aid of whi h one an a ount for Doppler ee ts, radiation
pressure, the photoele tri ee t and the Compton ee t.
The new wave on eption of Me hani s leads to a new statisti al
Me hani s, whi h allows us to unify the kineti theory of gases and
the theory of bla kbody radiation into a single do trine. This statisti s
oin ides with that proposed independently by Mr Bose [2; Mr Einstein
[3 has shown its s ope and laried its signi an e. Sin e then, numerous
papers [4 have developed it in various dire tions.
Let us add a few remarks. First, the author of this report has always
assumed that the material point o upies a well-dened position in
spa e. As a result, the amplitude f should ontain a singularity or at
the very least have abnormally high values in a very small region. But,
in fa t, the form of the amplitude plays no role in the results reviewed
above. Only the phase intervenes: hen e the name phase waves originally
given to the waves of the new Me hani s.
On the other hand, the author, after having redu ed the old forms
of Dynami s to geometri al Opti s, realised learly that this was only
a rst stage. The existen e of dira tion phenomena appeared to him
to require the onstru tion of a new Me hani s `whi h would be to
the old Me hani s (in luding that of Einstein) what wave Opti s is to
geometri al Opti s'.a It is Mr S hrdinger who has had the merit of
denitively onstru ting the new do trine.
2. The work of Mr E. S hrdinger [5. Mr S hrdinger's fundamental
idea seems to have been the following: the new Me hani s must begin
from wave equations, these equations being onstru ted in su h a way
a Revue Gnrale des S ien es, 30 November 1924, p. 633.
377
that in ea h ase the phase of their sinusoidal solutions should be a

solution of the Ja obi equation in the approximation of geometri al
Opti s.
Instead of onsidering waves whose amplitude ontains a singularity,
Mr S hrdinger systemati ally looks at waves of lassi al type, that is to
say, waves whose amplitude is a ontinuous fun tion. For him, the waves
of the new Me hani s are therefore represented by fun tions that one
an always write in the anoni al form
2
,
h
a being a ontinuous fun tion and being in the rst approximation a
solution of the Ja obi equation. We may understand the words ìn the
rst approximation' in two dierent ways: rst, if the onditions that
legitimate the use of geometri al Opti s are realised, the phase will
obey the equation alled the equation of geometri al Opti s, and this
equation will have to be identi al to that of Ja obi; se ond, one must
equally re over the Ja obi equation if one makes Plan k's onstant tend
to zero, be ause we know in advan e that the old Dynami s must then
be ome valid.
Let us rst onsider the ase of the motion of a single material point
in a stati eld derived from the potential fun tion F (x, y, z). In his
rst Memoir S hrdinger shows that the wave equation, at least in the
approximation of Newton's Me hani s, is in this ase
= a cos
8 2 m0
(E F ) = 0 .
h2
It is also just this equation that one arrives at beginning from the
dispersion law noted in the rst se tion.
Having obtained this equation, Mr S hrdinger used it to study the
quantisation of motion at the atomi s ale (hydrogen atom, Plan k
os illator, et .). He made the following fundamental observation: in the
problems onsidered in mi rome hani s, the approximations of geometri al Opti s are no longer valid at all. As a result, the interpretation of
the quantum onditions proposed by L. de Broglie shows only that the
Bohr-Sommerfeld formulas orrespond to the approximation of the old
Dynami s. To resolve the problem of quantisation rigorously, one must
therefore onsider the atom as the seat of stationary waves satisfying
ertain onditions. S hrdinger assumed, as is very natural, that the wave fun tions must be nite, single-valued and ontinuous over all spa e.
These onditions dene a set of fundamental fun tions (Eigenfunktionen)
+
378
L. de Broglie
for the amplitude, whi h represent the various stable states of the atomi
system being onsidered. The results obtained have proven that this new
quantisation method, to whi h Messrs Lon Brillouin, G. Wentzel and
Kramers [6 have made important ontributions, is the orre t one.
For Mr S hrdinger, one must look at ontinuous waves, that is to
say, waves whose amplitude does not have any singularities. How an
one then represent the `material point' ? Relying on the equality of
the velo ity of the moving body and the group velo ity, S hrdinger
sees the material point as a group of waves (Wellenpaket2 ) of losely
neighbouring frequen ies propagating in dire tions ontained within the
interior of a very narrow one. The material point would then not be
really pointlike; it would o upy a region of spa e that would be at
least of the order of magnitude of its wavelength. Sin e, in intra-atomi
phenomena, the domain where motion takes pla e has dimensions of the
order of the wavelengths, there the material point would no longer be
dened at all; for Mr S hrdinger, the ele tron in the atom is in some
sense `smeared out' [`fondu', and one an no longer speak of its position
or velo ity. This manner of on eiving of material points seems to us to
raise many di ulties; if, for example, the quantum of ultraviolet light
o upies a volume whose dimensions are of the order of its wavelength,
it is quite di ult to on eive that this quantum ould be absorbed by
an atom of dimensions a thousand times smaller.
Having established the wave equation for a material point in a stati eld, Mr S hrdinger then turned to the Dynami s of many-body
systems [la Dynamique des systmes. Still limiting himself to the Newtoniana approximation, and inspired by Hamilton's ideas, he arrived at
the following statement: Given an isolated system whose potential energy
is F (q1 , q2 , ..., qn ), the kineti energy is a homogeneous quadrati form
in the momenta pk and one may write
X
2T =
mkl pk pl ,
kl

the mkl being fun tions of the q . If m denotes the determinant mkl
and if E is the onstant of energy in the lassi al sense, then a ording
to S hrdinger one must begin with the wave equation

X
1
8 2
1
m+ 2
m 2 mkl l + 2 (E F ) = 0 ,
qk
q
h
kl
whi h des ribes the propagation of a wave in the onguration spa e

a That is, nonrelativisti (eds.).
379
onstru ted by means of the variables q . Setting
= a cos
2
,
h
and letting h tend to zero, in the limit one indeed re overs the Ja obi
equation
1 X kl
+F =E .
m
2
q k q l
kl
To quantise an atomi system, one will here again determine the fundamental fun tions of the orresponding wave equation.
We annot re all here the su esses obtained by this method (papers
by Messrs S hrdinger, Fues,3 Manneba k [7, et .), but we must insist
on the di ulties of a on eptual type that it raises. Indeed let us
onsider, for simpli ity, a system of N material points ea h possessing
three degrees of freedom. The onguration spa e is in an essential way
formed by means of the oordinates of the points and yet Mr S hrdinger
assumes that in atomi systems material points no longer have a learly
dened position. It seems a little paradoxi al to onstru t a onguration
spa e with the oordinates of points that do not exist. Furthermore, if
the propagation of a wave in spa e has a lear physi al meaning, it is
not the same as the propagation of a wave in the abstra t onguration
spa e, for whi h the number of dimensions is determined by the number
of degrees of freedom of the system. We shall therefore have to return
later to the exa t meaning of the S hrdinger equation for many-body
systems.
By a transformation of admirable ingenuity, Mr S hrdinger has shown
that the quantum Me hani s invented by Mr Heisenberg and developed
by Messrs Born, Jordan, Pauli, et ., an be translated into the language
of wave Me hani s. By omparison with Heisenberg's matrix elements,
he was able to derive the expression for the mean harge density of the
atom from the fun tions , an expression to whi h we shall return later.
The S hrdinger equations are not relativisti . For the ase of a single material point, various authors [8 have given a more general wave
equation that is in a ord with the prin iple of Relativity. Let e be the
ele tri harge of the point, V and A the two ele tromagneti potentials.
380
L. de Broglie
The equation that the wave , written in omplex form, must satisfy isa
"
#

4 2
e2 2
4i e V X
2
2 2
2 m0 c 2 (V A ) = 0 .
+
Ax
+
h c c t
x
h
c
xyz
As Mr O. Klein [9 and then the author [10 have shown, the theory of
the Universe with ve dimensions allows one to give the wave equation
a more elegant form in whi h the imaginary terms, whose presen e is
somewhat sho king for the physi ist, have disappeared.
We must also make a spe ial mention of the beautiful Memoirs in
whi h Mr De Donder [11 has onne ted the formulas of wave Me hani s
to his general theory of Einsteinian Gravity.
3. The ideas of Mr Born [12. Mr Born was stru k by the fa t that
the ontinuous wave fun tions do not allow us to say where the
parti le whose motion one is studying is and, reje ting the on ept of
the Wellenpaket, he onsiders the waves as giving only a statisti al
representation of the phenomena. Mr Born seems even to abandon the
idea of the determinism of individual physi al phenomena: the Quantum
Dynami s, he wrote in his letter to Nature, `would then be a singular
fusion of me hani s and statisti s .... . A knowledge of enables us
to follow the ourse of a physi al pro ess in so far as it is quantum
me hani ally determinate: not in a ausal sense, but in a statisti al one'.4
These on eptions were developed in a mathemati al form by their
author, in Memoirs of fundamental interest. Here, by way of example,
is how he treats the ollision of an ele tron and an atom. He writes the
S hrdinger equation for the ele tron-atom system, and he remarks that
before the ollision, the wave must be expressed by the produ t of
the fundamental fun tiona representing the initial state of the atom and
the plane wave fun tion orresponding to the uniform re tilinear motion
of the ele tron. During the ollision, there is an intera tion between the
ele tron and the atom, an intera tion that appears in the wave equation
as the mutual potential energy term. Starting from the initial form of
, Mr Born derives by methods of su essive approximation its nal
form after the ollision, in the ase of an elasti ollision, whi h does not
modify the internal state of the atom, as well as in the ase of an inelasti
ollision, where the atom passes from one stable state to another taking
energy from or yielding it to the ele tron. A ording to Mr Born, the
a This is the omplex, time-dependent Klein-Gordon equation in an external
ele tromagneti eld (eds.).
a That is, eigenfun tion (eds.).
381
nal form of determines the probability that the ollision may produ e
this or that result.
The ideas of Mr Born seem to us to ontain a great deal of truth, and
the onsiderations that shall now be developed show a great analogy
with them.
II. Probable meaning of the ontinuous waves [13

4. Case of a single material point in a stati eld. The body of
experimental dis overies made over forty years seems to require the idea
that matter and radiation possess an atomi stru ture. Nevertheless,
lassi al opti s has with immense su ess des ribed the propagation of
light by means of the on ept of ontinuous waves and, sin e the work of
Mr S hrdinger, also in wave Me hani s one always onsiders ontinuous
waves whi h, not showing any singularities, do not allow us to dene
the material point. If one does not wish to adopt the hypothesis of the
`Wellenpaket', whose development seems to raise di ulties, how an one
re on ile the existen e of pointlike elements of energy with the su ess
of theories that onsider the waves ? What link must one establish
between the orpus les and the waves? These are the hief questions
that arise in the present state of wave Me hani s.
To try to answer this, let us begin by onsidering the ase of a single
orpus le arrying a harge e and moving in an ele tromagneti elda
dened by the potentials V and A . Let us suppose rst that the motion
is one for whi h the old Me hani s (in relativisti form) is su ient. If
we write the wave in the anoni al form
2
= a cos ,
h
the fun tion is then, as we have seen, the Ja obi fun tion, and the velo ity of the orpus le is dened by the formula of Einsteinian Dynami s
e
grad + c A
v = c2
.
(I)
t eV
We propose to assume by indu tion that this formula is still valid
when the old Me hani s is no longer su ient, that is to say when
is no longer a solution of the Ja obi equation.b If one a epts this
hypothesis, whi h appears justied by its5 onsequen es, the formula (I)
a Here we leave aside the ase where there also exists a gravitational eld. Besides,
the onsiderations that follow extend without di ulty to that ase.
b Mr De Donder assumes equation (I) as we do, but denoting by not the phase of
382
L. de Broglie
ompletely determines the motion of the orpus le as soon as one is

given its position at an initial instant. In other words, the fun tion ,
just like the Ja obi fun tion of whi h it is the generalisation, determines
a whole lass of motions, and to know whi h of these motions is a tually
des ribed it su es to know the initial position.
Let us now onsider a whole loud of orpus les, identi al and without
intera tion, whose motions, determined by (I), orrespond to the same
fun tion but dier in the initial positions. Simple reasoning shows that
if the density of the loud at the initial moment is equal to

Ka2
eV ,
t
where K is a onstant, it will subsequently remain onstantly given by
this expression. We an state this result in another form. Let us suppose
there be only a single orpus le whose initial position we ignore; from
the pre eding, the probability for its presen e [sa probabilit de prsen e
at a given instant in a volume d of spa e will be

2
d = Ka
(II)
eV d .
t
In brief, in our hypotheses, ea h wave determines a ` lass of motions', and ea h one of these motions is governed by equation (I) when
one knows the initial position of the orpus le. If one ignores this initial
position, the formula (II) gives the probability for the presen e of the
orpus le in the element of volume d at the instant t. The wave
then appears as both a pilot wave (Fhrungsfeld of Mr Born) and a
probability wave. Sin e the motion of the orpus le seems to us to be
stri tly determined by equation (I), it does not seem to us that there
is any reason to renoun e believing in the determinism of individual
physi al phenomena,a and it is in this that our on eptions, whi h are
very similar in other respe ts to those of Mr Born, appear nevertheless
to dier from them markedly.
Let us remark that, if one limits oneself to the Newtonian approxima2
tion, in (I) and (II) one an repla e:
t eV by m0 c , and one obtains
the simplied forms

1 e
v =
(I )
grad + A ,
m0
c
the wave, but the lassi al Ja obi fun tion. As a result his theory and ours diverge
as soon as one leaves the domain where the old relativisti Me hani s is su ient.
a Here, that is, of the motion of individual orpus les.
383
= const a2 .
(II )
There is one ase where the appli ation of the pre eding ideas is done
in a remarkably lear form: when the initial motion of the orpus les is
uniform and re tilinear in a region free of all elds. In this region, the
loud of orpus les we have just imagined may be represented by the
homogeneous plane wave6
vx
2
W t 2 ;
= a cos
h
c
here a is a onstant, and this means that a orpus le has the same
probability to be at any point of the loud. The question of knowing
how this homogeneous plane wave will behave when penetrating a region
where a eld is present is analogous to that of determining the form of
an initially plane light wave that penetrates a refra ting medium. In
his Memoir `Quantenme hanik der Stossvorgnge', Mr Born has given a
general method of su essive approximation to solve this problem, and
Mr Wentzel [14 has shown that one an thus re over the Rutherford
formula for the dee tion of -rays by a harged entre.
We shall present yet another observation on the Dynami s of the
material point su h as results from equation (I): for the material point
one an always write the equations of the Dynami s of Relativity even
when the approximation of the old me hani s is not valid, on ondition
that one attributes to the body a variable proper mass M0 given by the
formula
r
h2 a
.
M0 = m20 2 2
4 c a
5. The interpretation of interferen e. The new Dynami s allows us
to interpret the phenomena of wave Opti s in exa tly the way that was
foreseen, a long time ago now, by Mr Einstein.a In the ase of light, the
wave is indeed the light wave of the lassi al theories.b,c If we onsider
the propagation of light in a region strewn with xed obsta les, the
propagation of the wave will depend on the nature and arrangement
of these obsta les, but the frequen y h1
t will not vary (no Doppler
a Cf. hapter 9 (eds.).
b We then onsider as the `light variable' without at all spe ifying the physi al
meaning of this quantity.
By ` lassi al theories' de Broglie seems to mean s alar wave opti s. In the general
dis ussion (p. 508), de Broglie states that the physi al nature of for photons is
unknown (eds.).
384
L. de Broglie
ee t). The formulas (I) and (II) will then take the form
2
v = grad ;
h
= consta2 .
The se ond of these formulas shows immediately that the bright and dark
fringes predi ted by the new theory will oin ide with those predi ted by
the old. To re ord the fringes, for example by photography, one an do an
experiment of short duration with intense irradiation, or an experiment
of long duration with feeble irradiation (Taylor's experiment); in the rst
ase one takes a mean in spa e, in the se ond ase a mean in time, but
if the light quanta do not a t on ea h other the statisti al result must
evidently be the same.
Mr Bothe [15 believed he ould dedu e, from ertain experiments on
the Compton ee t in a eld of interferen e, the inexa titude of the rst
formula written above, the one giving the velo ity of the quantum, but
in our opinion this on lusion an be ontested.
6. The energy-momentum tensor of the waves . In one of his Memoirs
[16, Mr S hrdinger gave the expression for the energy-momentum tensor in the interior of a wave .a Following the ideas expounded here, the
wave represents the motion of a loud of orpus les; examining the
expression given by S hrdinger and taking into a ount the relations
(I) and (II), one then per eives that it de omposes into one part giving
the energy and momentum of the parti les, and another that an be
interpreted as representing a state of stress existing in the wave around
the parti les. These stresses are zero in the states of motion onsistent
with the old Dynami s; they hara terise the new states predi ted by
wave Me hani s, whi h thus appear as ` onstrained states' of the material point and are intimately related to the variability of the proper
mass M0 . Mr De Donder has also drawn attention to this fa t, and he
was led to denote the amplitude of the waves that he onsidered by the
name of ìnternal stress potential'.
The existen e of these stresses allows one to explain how a mirror
ree ting a beam of light suers a radiation pressure, even though a ording to equation (I), be ause of interferen e, the orpus les of light
do not `strike' its surfa e.b
7. The dynami s of many-body systems. We must now examine how
a Cf. S hrdinger's report, se tion II (eds.).
b Cf. Brillouin's example in the dis ussion at the end of de Broglie's le ture (eds.).
385
these on eptions may serve to interpret the wave equation proposed by

S hrdinger for the Dynami s of many-body systems. We have pointed
out above the two di ulties that this equation raises. The rst, relating
to the meaning of the variables that serve to onstru t the onguration
spa e, disappears if one assumes that the material points always have
a quite denite position. The se ond di ulty remains. It appears to
us ertain that if one wants to physi ally represent the evolution of a
system of N orpus les, one must onsider the propagation of N waves
in spa e, ea h of the N propagations being determined by the a tion
of the N 1 orpus les onne ted to the other waves.a Nevertheless, if
one fo usses one's attention only on the orpus les, one an represent
their states by a point in onguration spa e, and one an try to relate
the motion of this representative point to the propagation of a titious
wave in onguration spa e. It appears to us very probable that the
waveb
2
(t, q1 , ..., qn ) ,
= a(q1 , q2 , ..., qn ) cos
h
a solution of the S hrdinger equation, is only a titious wave whi h,
in the Newtonian approximation, plays for the representative point of
the system in onguration spa e the same role of pilot wave and of
probability wave that the wave plays in ordinary spa e in the ase of
a single material point.
Let us suppose the system to be formed of N points having for re tangular oordinates
N
N
x11 , x12 , x13 , ..., xN
1 , x2 , x3 .
In the onguration spa e formed by means of these oordinates, the

representative point of the system has for [velo ity omponents along
the axis xki
vxki =
1
,
mk xki
mk being the mass of the k th orpus le. This is the relation that repla es
(I ) for many-body systems. From this, one dedu es that the probability
for the presen e of the representative point in the element of volume d
a Cf. se tion 2.4 (eds.).
b The amplitude a is time-independent be ause de Broglie is assuming the
time-independent S hrdinger equation. Later in his report, de Broglie applies
his dynami s to a non-stationary wave fun tion as well, for the ase of an atomi
transition (eds.).
386
L. de Broglie
of onguration spa e is
d = consta2 d .
This new relation repla es relation (II ) for many-body systems. It

fully a ords, it seems to us, with the results obtained by Mr Born for
the ollision of an ele tron and an atom, and by Mr Fermi [17 for the
ollision of an ele tron and a rotator.a
Contrary to what happens for a single material point, it does not
appear easy to nd a wave that would dene the motion of the system
taking Relativity into a ount.
8. The waves in mi rome hani s. Many authors think it is illusory
to wonder what the position or the velo ity of an ele tron in the atom is
at a given instant. We are, on the ontrary, in lined to believe that it is
possible to attribute to the orpus les a position and a velo ity even in
atomi systems, in a way that gives a pre ise meaning to the variables
of onguration spa e.
This leads to on lusions that deserve to be emphasised. Let us onsider a hydrogen atom in one of its stable states. A ording to S hrdinger,
in spheri al oordinatesb the orresponding fun tion n is of the form
n = F (r, )[A cos m + B sin m]
sin 2
Wn t
cos h
(m integer)
with
Wn = m0 c2
2 2 m0 e4
.
n2 h2
If we then apply our formula (I ), we on lude that the ele tron is motionless in the atom, a on lusion whi h would evidently be inadmissible
in the old Me hani s. However, the examination of various questions
and notably of the Zeeman ee t has led us to believe that, in its stable
states, the H atom must rather be represented by the fun tion

mh
2
Wn t
,
n = F (r, ) cos
h
2
whi h, being a linear ombination of expressions of the type written
a Cf. the remarks by Born and Brillouin in the dis ussion at the end of de Broglie's
le ture, and the de Broglie-Pauli en ounter in the general dis ussion at the end of
the onferen e (pp. 509 .) (eds.).
b r , radius ve tor; , latitude; , longitude.
387
above, is equally a eptable.a If this is true the ele tron will have, from
(I ), a uniform ir ular motion of speed
v=
1 mh
.
m0 r 2
It will then be motionless only in states where m = 0.

Generally speaking, the states of the atom at a given instant an
always be represented by a fun tion
X
=
cn n ,
n
the n being S hrdinger's Eigenfunktionen. In parti ular, the state of

transition i j during whi h the atom emits the frequen y ij would
be given by (this appears to be in keeping with S hrdinger's ideas)
= ci i + cj j ,
ci and cj being two fun tions of time that hange very slowly ompared
with the trigonometri fa tors of the n , the rst varying from 1 to 0
and the se ond from 0 to 1 during the transition. Writing the fun tion
in the anoni al form a cos 2

h , whi h is always possible, formula (I )
will give the velo ity of the ele tron during the transition, if one assumes
the initial position to be given. So it does not seem to be impossible to
arrive in this way at a visual representation of the transition.a
Let us now onsider an ensemble of hydrogen atoms that are all in the
same state represented by the same fun tion
X
=
cn n .
n
The position of the ele tron in ea h atom is unknown to us, but if, in our
imagination, we superpose all these atoms, we obtain a mean atom where
the probability for the presen e of one of the ele trons in an element of
volume d will be given by the formula (II),b K being determined by
the fa t that the total probability for all the possible positions must be
a In his memoir, `Les moments de rotation et le magntisme dans la m anique

ondulatoire' (Journal de Physique 8 (1927), 74), Mr Lon Brillouin has impli itly
assumed the hypothesis that we formulate in the text.
a In this example, de Broglie is applying his dynami s to a ase where the wave
fun tion has a time-dependent amplitude a (eds.).
b In this se tion on atomi physi s (`mi rome hani s') de Broglie onsiders the
non-relativisti approximation, using the limiting formula (I ) ex ept in this
paragraph where he reverts to the relativisti formulas (I) and (II), for the purpose
of omparison with the relativisti formulas for harge and urrent density obtained
by other authors (eds.).
388
L. de Broglie
equal to unity. The harge density and the urrent density J =

v
in the mean atom are then, from (I) and (II),

2
eV ,
= Kea
t

e

J = Kec2 a2 grad + A
c
and these formulas oin ide, apart from notation, with those of Messrs
Gordon, S hrdinger and O. Klein [18.
Limiting ourselves to the Newtonian approximation, and for a moment
the onjugate
denoting by the wave written in omplex form, and by
fun tion, it follows that
.
= consta2 = const
This is the formula to whi h Mr S hrdinger was led in reformulating

the matrix theory; it shows that the ele tri dipole moment of the mean
atom during the transition i j ontains a term of frequen y ij , and
thus allows us to interpret Bohr's frequen y relation.
Today it appears ertain that one an predi t the mean energy radiated by an atom by using the Maxwell-Lorentz equations, on ondition
v
that one introdu es in these equations the mean quantities and
a
whi h have just been dened. One an thus give the orresponden e
prin iple an entirely pre ise meaning, as Mr Debye [19 has in fa t shown
in the parti ular ase of motion with one degree of freedom. It seems
indeed that lassi al ele tromagnetism an from now on retain only a
statisti al value; this is an important fa t, whose meaning one will have
to try to explore more deeply.
To study the intera tion of radiation with an ensemble of atoms, it
is rather natural to onsider a `mean atom', immersed in a `mean light'
whi h one denes by a homogeneous plane wave of the ve tor potential.
The density of the mean atom is perturbed by the a tion of the
light and one dedu es from this the s attered radiation. This method,
whi h gives good mean predi tions, is related more or less dire tly to
the theories of s attering by Messrs S hrdinger and Klein [20, to the
theory of the Compton ee t by Messrs Gordon and S hrdinger [21,
and to the Memoirs of Mr Wentzel [22 on the photoele tri ee t and
the Compton ee t, et . The s ope of this report does not permit us to
dwell any further on this interesting work.
a Cf. S hrdinger's report, p. 454, and the ensuing dis ussion, and se tion 4.5 (eds.).
389
9. Con lusions and remarks. So far we have onsidered the orpus les
as èxterior' to the wave , their motion being only determined by the
propagation of the wave. This is, no doubt, only a provisional point
of view: a true theory of the atomi stru ture of matter and radiation
should, it seems to us, in orporate the orpus les in the wave phenomenon by onsidering singular solutions of the wave equations. One should
then show that there exists a orresponden e between the singular waves
and the waves , su h that the motion of the singularities is onne ted
to the propagation of the waves by the relation (I).a In the ase of no
[external eld, this orresponden e is easily established, but it is not so
in the general ase.
We have seen that the quantities and

v appearing in the MaxwellLorentz equations must be al ulated in terms of the fun tions , but
that does not su e to establish a deep link between the ele tromagneti
quantities and those of wave Me hani s. To establish this link,a one
should probably begin with singular waves, for Mr S hrdinger has very
rightly remarked that the potentials appearing in the wave equations are
those that result from the dis ontinuous stru ture of ele tri ity and not
those that ould be dedu ed from the fun tions .
Finally, we point out that Messrs Uhlenbe k and Goudsmit's hypothesis of the magneti ele tron, so ne essary to explain a great number of
phenomena, has not yet found its pla e in the s ope of wave Me hani s.
III. Experiments showing preliminary dire t eviden e for

the new Dynami s of the ele tron
10. Phenomena whose existen e is suggested by the new on eptions.
The ideas that have just been presented lead one to onsider the motion
of an ele tron as guided by the propagation of a ertain wave. In many usual ases, the old Me hani s remains entirely adequate as a rst
approximation; but our new point of view, as Elsasser7 [23 pointed out
already in 1925, ne essarily raises the following question: `Could one not
observe ele tron motions that the old Me hani s would be in apable of
predi ting, and whi h would therefore be hara teristi of wave Me hani s? In other words, for ele trons, ould one not nd the analogue of
the phenomena of dira tion and interferen e?'b
a Cf. se tion 2.3.1 (eds.).
a The few attempts made till now in this dire tion, notably by Mr Bateman (Nature
118 (1926), 839) and by the author (Ondes et mouvements, Chap. VIII, and C. R.
184 (1927), 81) an hardly be regarded as satisfa tory.
b This is not a quotation: these words do not appear in the ited 1925 paper by
390
L. de Broglie
These new phenomena, if they exist, must depend on the wavelength

of the wave asso iated with the ele tron motion. For an ele tron of speed
v , the fundamental formula
h
p=
V
gives
p
h
h 1 2
V
= =
.
=
p
m0 v
If is not too lose to 1, it su es to write
=
h
.
m0 v
Let V be the potential dieren e, expressed in volts, that is apable of

imparting the speed v to the ele tron; numeri ally, for the wavelength
in entimetres, one will havea
=
12.25
7.28
= 108 .
v
V
To do pre ise experiments, it is ne essary to use ele trons of at least

a few volts: from whi h one has an upper limit for of a few angstroms.
One then sees that, even for slow ele trons, the phenomena being sought
are analogous to those shown by X-rays and not to those of ordinary
light. As a result, it will be di ult to observe the dira tion of a
beam of ele trons by a small opening, and if one wishes to have some
han e of obtaining dira tion by a grating, one must either onsider
those natural three-dimensional gratings, the rystals, or use ordinary
gratings under a very grazing in iden e, as has been done re ently for
X-rays. On making slow ele trons pass through a rystalline powder or
an amorphous substan e, one ould also hope to noti e the appearan e of
rings analogous to those that have been obtained and interpreted in the
X-ray domain by Messrs Hull, Debye and S herrer, Debierne, Keesom
and De Smedt, et .
The exa t theoreti al predi tion of the phenomena to be observed
Elsasser. Further, it was de Broglie who rst suggested ele tron dira tion, in a
paper of 1923 (see se tion 2.2.1) (eds.).
a Here we have adopted the following values:
h = 6.55 1027 erg-se onds ,
m0 = 9 1028 gr ,
e = 4.77 1010 e.s.u. .
391
along these lines is still not very advan ed. Let us onsider the dira tion
of a beam of ele trons with the same velo ity by a rystal; the wave
will propagate following the general equation, in whi h one has to insert
the potentials reated by the atoms of the rystal onsidered as entres of
for e. One does not know the exa t expression for these potentials but,
be ause of the regular distribution of atoms in the rystal, one easily
realises that the s attered amplitude will show maxima in the dire tions
predi ted by Mr von Laue's theory. Be ause of the role of pilot wave
played by the wave , one must then observe a sele tive s attering of
the ele trons in these dire tions.
Using his methods, Mr Born has studied another problem: that of the
ollision of a narrow beam of ele trons with an atom. A ording to him,
the urve giving the number of ele trons that have suered an inelasti
ollision as a fun tion of the s attering angle must show maxima and
minima; in other words, these ele trons will display rings on a s reen
pla ed normally to the ontinuation of the in ident beam.
It would still be premature to speak of agreement between theory
and experiment; nevertheless, we shall present experiments that have
revealed phenomena showing at least broadly the predi ted hara ter.
11. Experiments by Mr E.G. Dymond [24. Without feeling obliged
to follow the hronologi al order, we shall rst present Mr Dymond's
experiments:
A ask of puried helium ontained an èle tron gun', whi h onsisted
of a brass tube ontaining an in andes ent lament of tungsten and in
whose end a slit was ut. This gun dis harged a well- ollimated beam
of ele trons into the gas, with a speed determined by the potential
dieren e (50 to 400 volts) established between the lament and the wall
of the tube. The wall of the ask had a slit through whi h the ele trons
ould enter a hamber where the pressure was kept low by pumping and
where, by urving their traje tories, a magneti eld brought them onto
a Faraday ylinder.
Mr Dymond rst kept the orientation of the gun xed and measured
the speed of the ele trons thus s attered by a given angle. He noti ed
that most of the s attered ele trons have the same energy as the primary
ele trons; they have therefore suered an elasti ollision. Quite a large
number of ele trons have a lower speed orresponding to an energy loss
from about 20 to 55 volts: this shows that they made the He atom pass
from the normal state 11 S to the ex ited state 21 S . One also observes a
lower proportion of other values for the energy of the s attered ele trons;
392
L. de Broglie
we shall not dis uss the interpretation that Mr Dymond has given them,
be ause what interests us most here is the variation of the number of
s attered parti les with the s attering angle . To determine this number,
Mr Dymond varied the orientation of the gun inside the ask, and for
dierent s attering angles olle ted the ele trons that suered an energy
loss equivalent to 20 to 55 volts; he onstru ted a series of urves of the
angular distribution of these ele trons for dierent values of the tension
V applied to the ele tron gun. The angular distribution urve shows
a very pronoun ed maximum for a low value of , and this maximum
appears to approa h = 0 for in reasing values of V .
Another, less important, maximum appears towards = 50 for a
primary energy of about a hundred volts, and then moves for in reasing
values of V towards in reasing . Finally, a very sharp maximum appears
for a primary energy of about 200 volts at = 30 , and then seems
independent of V . These fa ts are summarised in the following table
given by Dymond:8
V (volts)
48.9
72.3
97.5
195
294
400
....
....
....
....
....
....
Positions of the maxima ( )

24

8

5
50
< 2.5 30 59
< 2.5 30 69
< 2.5 30 70
The above results must very probably be interpreted with the aid
of the new Me hani s and are to be related to Mr Born's predi tions.
Nevertheless, as Mr Dymond very rightly says, `the theoreti al side of
the problem is however not yet su iently advan ed to give detailed
information on the phenomena to be expe ted, so that the results above
reported annot be said to substantiate the wave me hani s ex ept in
the most general way'.9
12. Experiments by Messrs C. Davisson and L. H. Germer. In 1923,
Messrs Davisson and Kunsman [25 published pe uliar results on the
s attering of ele trons at low speed. They dire ted a beam of ele trons,
a elerated by a potential dieren e of less than 1000 volts, onto a blo k
of platinum at an in iden e of 45 and determined the distribution
of s attered ele trons by olle ting them in a Faraday ylinder. For
potentials above 200 volts, one observed a steady de rease in s attering
for in reasing values of the deviation angle, but for smaller voltages
393
the urve of angular variation showed two maxima. By overing the

platinum with a deposit of magnesium, one obtained a single small
maximum for ele trons of less than 150 volts. Messrs Davisson and
Kunsman attributed the observed phenomena to the a tion of various
layers of intra-atomi ele trons on the in ident ele trons, but it seems
rather, a ording to Elsasser's opinion, that the interpretation of these
phenomena is a matter for the new Me hani s.
Resuming analogous experiments with Mr Germer [26, Mr Davisson
obtained very important results this year, whi h appear to onrm the
general predi tions and even the formulas of Wave Me hani s.
The two Ameri an physi ists sent homogeneous beams of ele trons
onto a rystal of ni kel, ut following one of the 111 fa es of the regular
o tahedron (ni kel is a ubi rystal). The in iden e being normal, the
phenomenon ne essarily had to show the ternary symmetry around the
dire tion of the in ident beam. In a ubi rystal ut in this manner,
the fa e of entry is ut obliquely by three series of 111 planes, three
series of 100 planes, and six series of 110 planes. If one takes as positive
orientation of the normals to these series of planes the one forming an
a ute angle with the fa e of entry, then these normals, together with the
dire tion of in iden e, determine distinguished azimuths, whi h Messrs
Davisson and Germer all azimuths (111), (100), (110), and for whi h
they studied the s attering; be ause of the ternary symmetry, it evidently
su es to explore a single azimuth of ea h type.
Let us pla e ourselves at one of the distinguished azimuths and let
us onsider only the distribution of Ni atoms on the fa e of entry of
the rystal, whi h we assume to be perfe t. These atoms form lines perpendi ular to the azimuth being onsidered and whose equidistan e d is
known from rystallographi data. The dierent dire tions of s attering
being identied in the azimuthal plane by the angle of o-latitude, the
waves s attered by the atoms in the fa e of entry must be in phase in
dire tions su h that one has

n 12.25
n
108
= arcsin
(n integer) .
= arcsin
d
d V
One must then expe t to observe maxima in these dire tions, for the
s attering of the ele trons by the rystal.
Now here is what Messrs Davisson and Germer observed. By gradually
varying the voltage V that a elerates the ele trons one observes, in the
neighbourhood of ertain values of V , very distin t s attering maxima
in dire tions whose o-latitude is a urately given by the above formula
394
L. de Broglie
(provided one sets in general n = 1, and sometimes n = 2). There is

dire t numeri al onrmation of the formulas of the new Dynami s; this
is evidently a result of the highest importan e.
However, the explanation of the phenomenon is not omplete: one
must explain why the s attering maxima are observed only in the neighbourhood of ertain parti ular values of V , and not for all values of V .
One interpretation naturally omes to mind: we assumed above that only
the fa e of entry of the rystal played a role, but one an assume that
the ele tron wave penetrates somewhat into the rystal and, further,
in reality the fa e of entry will never be perfe t and will be formed by
several parallel 111 planes forming steps. In these onditions, it is not
su ient to onsider the interferen e of the waves s attered by a single
reti ular plane at the surfa e, one must take into a ount the interferen e
of the waves s attered by several parallel reti ular planes. In order for
there to be a strong s attering in a dire tion , and V must then satisfy
not only the relation written above, but also another relation whi h is
easy to nd; the s attering must then be sele tive, that is to say, o ur
with [signi ant intensity only for ertain values of V , as experiment
shows. Of ourse, the theory that has just been outlined is a spe ial ase
of Laue's general theory.
Unfortunately, as Messrs Davisson and Germer have themselves remarked, in order to obtain an exa t predi tion of the fa ts in this way,
it is ne essary to attribute to the separation of the 111 planes next to
the fa e of entry a smaller value (of about 30%) than that provided by
Crystallography and by dire t measurements by means of X-rays. It is
moreover not unreasonable to assume that the very super ial reti ular
planes have a spa ing dierent from those of the deeper planes, and one
an even try to onne t this idea to our urrent on eptions on erning
the equilibrium of rystalline gratings.
If one a epts the pre eding hypothesis, the s attering must be produ ed by a very small number of reti ular planes in the entirely super ial
layer of the rystal; the on entration of ele trons in preferred dire tions
must then be mu h less pronoun ed than in the ase of s attering by a
whole unlimited spatial grating. Is it nevertheless su ient in order to
explain the `peaks' observed by Davisson and Germer? To this question,
Mr Patterson has re ently provided an armative answer, by showing
that the involvement of just two super ial reti ular planes already
su es to predi t exa tly the variations of the sele tive ree tion of
395
ele trons observed in the neighbourhood of

= 50 ,
V = 54 volts .
To on lude, we an do no better than quote the on lusion of Mr

Patterson [27: `The agreement of these results with al ulation seems
to indi ate that the phenomenon an be explained as a dira tion of
waves in the outermost layers of the rystal surfa e. It also appears
[.... that a omplete analysis of the results of su h experiments will give
valuable information as to the onditions prevailing in the a tual surfa e,
and that a new method has been made available for the investigation of
the stru ture of rystals in a region whi h has up to the present almost
ompletely es aped observation'.10
13. Experiments by Messrs G. P. Thomson and A. Reid [28. Very
re ently, Messrs Thomson and Reid have made the following results
known: if a narrow pen il of homogeneous athode rays passes normally
through a elluloid lm, and is then re eived on a photographi plate
pla ed parallel to the lm at 10 m behind it, one observes rings around
the entral spot. With rays of 13 000 volts, a photometri examination
has revealed the existen e of three rings. By gradually in reasing the
energy of the ele trons, one sees the rings appear around 2500 volts,
and they have been observed up to 16 500 volts. The radii of the rings
de rease when the energy in reases and, it seems, approximately in
inverse proportion to the speed, that is, to our wavelength .
These observations are very interesting, and again onrm the new
on eptions in broad outline. Is it a question here of an atomi phenomenon analogous to those observed by Dymond, or else of a phenomenon
of mutual interferen e falling into one of the ategories studied by Debye
and S herrer, Hull, Debierne, Keesom and De Smedt? We are unable to
say, and we limit ourselves to remarking that here the ele trons used are
relatively fast; this is interesting from the experimental point of view,
be ause it is mu h easier to study ele trons of a few thousand volts than
ele trons of about a hundred volts.
396
L. de Broglie
Bibliographya
[1 Louis de Broglie, C. R., 177 (1923), 507, 548 and 630; 179 (1924), 39,
676 and 1039. Do toral thesis, November 1924 (Masson, publisher), Annales de Physique (10), III [(1925), 22. J. de Phys. (6), VII [(1926), 1.
[2 S. N. Bose, Zts. f. Phys., 27 (1924), 384.
[3 A. Einstein, Berl. Ber. (1924), 261; (1925), [3.
[4 P. Jordan, Zts. f. Phys., 33 (1925), 649.
E. S hrdinger, Physik. Zts., 27 (1926), 95.
P. Dira , Pro . Roy. So . A, 112 (1926), 661.
E. Fermi, Zts. f. Phys., 36 (1926), 902.
L. S. Ornstein and H. A. Kramers, Zts. f. Phys., 42 (1927), 481.
[5 E. S hrdinger, Ann. der Phys., 79 (1926), 361, 489 and 734; 80
(1926), 437; 81 (1926), 109. Naturwissens h., 14th year [(1926), 664.
Phys. Rev., 28 (1926), 1051.
[6 L. Brillouin, C. R., 183 (1926), 24 and 270. J. de Phys. [(6), VII
(1926), 353.
G. Wentzel, Zts. f. Phys., 38 (1926), 518.
H. A. Kramers, Zts. f. Phys., 39 (1926), 828.
[7 E. Fues, Ann. der Phys., 80 (1926), 367; 81 (1926), 281.
C. Manneba k, Physik. Zts., 27 (1926), 563; 28 (1927), 72.
[8 L. de Broglie, C. R., 183 (1926), 272. J. de Phys. (6), VII (1926),
332.
O. Klein, Zts. f. Phys., 37 (1926), 895.
[V. Fo k, Zts. f. Phys., 38 (1926), 242.
W. Gordon, Zts. f. Phys., 40 (1926), 117.
L. Flamm, Physik. Zts., 27 (1926), 600.
[9 O. Klein, lo . it. in [8.
[10 L. de Broglie, J. de Phys. (6), VIII (1927), 65.
(See also L. Rosenfeld, A ad. Roy. Belg. (5), 13, n. 6.)
[11 Th. De Donder, C. R., 182 (1926), 1512; 183 (1926), 22 (with Mr
Van den Dungen); 183 (1926), 594; 184 (1927), 439 and 810. A ad. Roy.
Belg., sessions of 9 O tober 1926, of 8 January, 5 February, 5 Mar h and
2 April 1927.
[12 M. Born, Zts. f. Phys., 37 (1926), 863; 38 (1926), 803; 40 (1926),
167. Nature, 119 (1927), 354.
[13 L. de Broglie, C. R., 183 (1926), 447, and 184 (1927), 273. Nature,
a The style of the bibliography has been slightly modernised and uniformised with
that used in the other reports (eds.).
397
118 (1926), 441. J. de Phys. (6), VIII (1927), 225. C. R., 185 (1927),
380.
(See also E. Madelung, Zts. f. Phys., 40 (1926), 322.)
[14 G. Wentzel, Zts. f. Phys., 40 (1926), 590.
[15 W. Bothe, Zts. f. Phys., 41 (1927), 332.
[16 E. S hrdinger, Ann. der Phys., 82 (1927), 265.
[17 E. Fermi, Zts. f. Phys., 40 (1926), 399.
[18 W. Gordon and E. S hrdinger, lo . it. in [8 and [16.
O. Klein, Zts. f. Phys., 41 (1927), 407.
[19 P. Debye, Physik. Zts., 27 (1927), 170.
[20 E. S hrdinger and O. Klein, lo . it. in [5 and [18.
[21 W. Gordon, lo . it. in [8.
E. S hrdinger, Ann. der Phys., 82 (1927), 257.
[22 G. Wentzel, Zts. f. Phys., 43 (1927), 1; 41 (1927), 828.
(See also G. Be k, Zts. f. Phys., 41 (1927), 443; J. R. Oppenheimer, Zts.
f. Phys., 41 (1927), 268.)
[23 [W. Elsasser, Naturwissens h., 13 (1925), 711.
[24 E. G. Dymond, Nature, 118 (1926), 336. Phys. Rev., 29 (1927), 433.
[25 C. Davisson and C. H. Kunsman, Phys. Rev., 22 (1923), 242.
[26 C. Davisson and L. H. Germer, Nature, 119 (1927), 558.
[27 A. L. Patterson, Nature, 120 (1927), 46.
[28 C. P. Thomson and A. Reid, Nature, 119 (1927), 890.
398
L. de Broglie
Dis ussion of Mr de Broglie's report
Mr Lorentz. I should like to see learly how, in the rst form of
your theory, you re overed Sommerfeld's quantisation onditions. You

obtained a single ondition, appli able only to the ase where the orbit
is losed: the wave must, after travelling along the orbit, nish in phase
when it omes ba k to the initial point. But in most ases the traje tory
is not losed; this happens, for example, for the hydrogen atom when
one takes relativity into a ount; the traje tory is then a rosette, and
never omes ba k to its initial point.
How did you nd the quantisation onditions appli able to these multiperiodi problems?
Mr de Broglie.
The di ulty is resolved by onsidering pseudoperiods, as I pointed out in my Thesis ( hap. III, p. 41). When a system
is multiperiodi , with partial periods 1 , 2 , ..., n , one an prove that
one an nd quasi-periods that are nearly exa tly whole multiples of
the partial periods:
= m1 1 + 1 = m2 2 + 2 ... = mn n + n ,
the m1 , m2 , ..., mn being integers and the 1 , 2 , ..., n as small as

one likes. The traje tory then never omes ba k to its initial point, but
at the end of a quasi-period it omes ba k as losely as one likes to
the initial position. One will then be led to write that, at the end of a
quasi-period, the wave nishes in phase; now, there is an innite number
of quasi-periods, orresponding to all kinds of systems of values of the
integers m1 , m2 , ..., mn . In order that the wave nishes in phase after
any one of these quasi-periods, it is ne essary that one have11
Z
Z
Z
pn dqn = nn h ,
p2 dq2 = n2 h , ... ,
p1 dq1 = n1 h ,
1
whi h gives exa tly Sommerfeld's onditions.
Mr Born.
The denition of the traje tory of a parti le that Mr

de Broglie has given seems to me to present di ulties in the ase of
a ollision between an ele tron and an atom. In an elasti ollision, the
speed of the parti le must be the same after the ollision as before. I
a Darrigol (1993, pp. 3423, 3645) shows that this derivation is faulty: the ondition
that the wave should nish in phase after any quasi-period does not imply the n
separate onditions listed above.
Dis ussion
399
should like to ask Mr de Broglie if that follows from his formula.
Mr de Broglie. That follows from it, indeed.

Mr Brillouin. It seems to me that no serious obje tion an be
made to the point of view of L. de Broglie. Mr Born an doubt the

real existen e of the traje tories al ulated by L. de Broglie, and assert
that one will never be able to observe them, but he annot prove to us
that these traje tories do not exist. There is no ontradi tion between
the point of view of L. de Broglie and that of other authors, sin e, in
his report (8, p. 38812 ) L. de Broglie shows us that his formulasa are
in exa t agreement with those of Gordon, at present a epted by all
physi ists.
Mr Pauli. I should like to make a small remark on what seems to
me to be the mathemati al basis of Mr de Broglie's viewpoint on erning

parti les in motion on denite traje tories. His on eption is based on
the prin iple of onservation of harge:
s1
s2
s3
+
+
+
=0
t
x
y
z
or
4
X
sk
=0,
xk
(a)
k=1
whi h is a onsequen e of the wave equation, when one sets

isk =
4i e

+
k .
xk
xk
h c
Mr de Broglie introdu es, in pla e of the omplex fun tion , the two
real fun tions a and dened by
= ae
2i
h
= ae
2i
h
Substituting these expressions into the expression for sk yields:

e
4 2
+ k .
a
sk =
h
xk
c
From this follow the expressions given by Mr de Broglie for the velo ity
ve tor, dened by
s2
s3
s1
, v2 =
, v3 =
.
v1 =
(b)
Now if in a eld theory there exists a onservation prin iple of the

a That is, de Broglie's equations for the mean harge and urrent density to be used
in semi lassi al radiation theory (eds.).
400
L. de Broglie
form (a), it is always formally possible to introdu e a velo ity ve tor

(b), depending on spa e and time, and to imagine furthermore orpus les
that move following the urrent lines of this ve tor. Something similar
was already proposed in opti s by Slater; a ording to him, light quanta
should always move following the lines of the Poynting ve tor. Mr de
Broglie now introdu es an analogous representation for material parti les.
In any ase, I do not believe that this representation may be developed
in a satisfa tory manner; I intend to return to this during the general
dis ussion.a
Mr S hrdinger.
If I have properly understood Mr de Broglie,

the velo ity of the parti les must have its analogue in a ve tor eld
omposed of the three spatial omponents of the urrent in a fourdimensional spa e, after division of these by the omponent with respe t
to time (that is, the harge density). I should like simply to re all now
that there exist still other ve tor quantities of a eld, whi h an be made
to orrespond with the velo ity of the parti les, su h as the omponents
of the momentum density (see Ann. d. Phys.13 82, 265). Whi h of the
two analogies is the more onvin ing?
Mr Kramers.
The fa t that with independent parti les in motion

one annot onstru t an energy-momentum tensor having the properties
required by Maxwell's theory onstitutes nevertheless a di ulty.
Mr Pauli.
The quotient of the momentum by the energy density whi h Mr S hrdinger onsiders would in fa t lead in a relativisti
al ulation to other parti le traje tories than would the quotient of the
densities of urrent and of harge.
Mr Lorentz. In using his formulas for the velo ity of the ele tron,
has Mr de Broglie not al ulated this velo ity in parti ular ases, for
example for the hydrogen atom?
Mr de Broglie. When one applies the formula for the velo ity
to a wave fun tion representing a stable state of the hydrogen atom

a ording to Mr S hrdinger, one nds ir ular orbits. One does not
a See pp. 509 . (eds.).
Dis ussion
401
re over the ellipti al orbits of the old theory (see my report, 8).
Mr Ehrenfest. Can the speed of an ele tron in a stationary orbit
be zero?
Mr de Broglie. Yes, the speed of the ele tron an be zero.

Mr S hrdinger. Mr de Broglie says that in the ase of the
hydrogen atom his hypothesis leads to ir ular orbits. That is true for
the parti ular solutions of the wave equation that one obtains when one
separates the problem in polar oordinates in spa e; perhaps it is still
true for the solutions that one obtains by making use of paraboli or
ellipti al oordinates. But in the ase of a degenera y (as he onsiders
it here) it is, in reality, not at all the parti ular solutions whi h have
a signi an e, but only an arbitrary linear ombination, with onstant
oe ients, of all the parti ular solutions belonging to the same eigenvalue, be ause there is no means of distinguishing between them, all linear
ombinations being equally justied in prin iple. In these onditions,
mu h more ompli ated types of orbit will ertainly appear. But I do
not believe that in the atomi domain one may still speak of òrbits'.
Mr Lorentz. Does one know of su h more ompli ated orbits?

Mr S hrdinger. No, one does not know of them; but I simply
wanted to say that if one nds ir ular orbits, that is due to a fortuitous
hoi e of parti ular solutions that one onsiders, and this hoi e annot
be motivated in a way that has no arbitrariness.
Mr Brillouin.
Perhaps it is not superuous to give some examples that illustrate well the meaning of Mr L. de Broglie's formulas, and
that allow one to follow the motion of the parti les guided by the phase
wave. If the wave is plane and propagates freely, the traje tories of the
parti les are the rays normal to the wave surfa e. Let us suppose that the
wave is ree ted by a plane mirror, and let be the angle of in iden e;
the wave motion in front of the mirror is given by a superposition of the
in ident wave
1 = a1 cos 2
x sin z cos
t
402
L. de Broglie
Fig. 1.
and the ree ted wave
2 = a1 cos 2
whi h gives
x sin
t
t
x sin + z cos
2z cos
= 2a1 cos
cos 2
This wave is put in L. de Broglie's anoni al form

= a cos
with
a = 2a1 cos
2z cos
and
=h
x sin
t
Let us then apply L. de Broglie's formulas, in the simplied form given

on page 384 (5);14 and let us suppose that it is a light wave guiding the
photons; the velo ity of these is
2
v = grad .
h
We see that the proje tiles move parallel to the mirror, with a speed
vx = c sin , less than c. Their energy remains equal to h , be ause
their mass has undergone a variation, a ording to the following formula
(report by L. de Broglie, p. 383):15
r
r
h2 a
h
h
a
2
M 0 = m0 2 2
=
= 2 cos .
4 c a
2c
a
c
Dis ussion
403
Fig. 2.
The mass of the photons, whi h is zero in the ase where the wave
propagates freely, is then assumed to take a non-zero value in the whole
region where there is interferen e or deviation of the wave.
Let us draw a diagram for the ase of a limited beam of light falling on
a plane mirror; the interferen e is produ ed in the region of overlap of two
beams. The traje tory of a photon will be as follows: at rst a re tilinear
path in the in ident beam, then a bending at the edge of the interferen e
zone, then a re tilinear path parallel to the mirror, with the photon
travelling in a bright fringe and avoiding the dark interferen e fringes;
then, when it omes out, the photon retreats following the dire tion of
the ree ted light beam.
No photon a tually strikes the mirror, nevertheless the mirror suers
lassi al radiation pressure; it is in order to explain this fa t that L. de
Broglie assumes the existen e of spe ial stresses in the interferen e zone;
these stresses, when added to the tensor of momentum ux transported
by the photons, reprodu e the lassi al Maxwell tensor; there is then
no dieren e in the me hani al ee ts produ ed by the wave during its
ree tion by the mirror.
These remarks show how L. de Broglie's system of hypotheses preserves the lassi al formulas, and avoids a ertain number of awkward
paradoxes. One thus obtains, for example, the solution to a urious
problem posed by G. N. Lewis (Pro . Nat. A ad. 12 (1926), 22 and
439), whi h was the subje t of dis ussions between this author and R.
C. Tolman and S. Smith (Pro . Nat. A ad. 12 (1926), 343 and 508).
Lewis assumed that the photons always follow the path of a light
404
L. de Broglie
Fig. 3.
ray of geometri al opti s, but that they hoose, among the dierent
rays, only those that lead from the luminous sour e to a bright fringe
situated on an absorbing body. He then onsidered a sour e S whose
light is ree ted by two mirrors AA and BB; the light beams overlap,
produ ing interferen e [zones in whi h one pla es a s reen CD; the
dimensions are assumed to be su h that there is a bright fringe on one
of the edges D of the s reen and a dark fringe on the other edge C.
Following the hypothesis of Lewis, the photons would follow only the
paths SBD and SAD, whi h end at the bright fringe D; no photon will
take the path SAC or SBC. All the photons ome to strike the mirror
AA on the edge A, so one ould predi t that this mirror would suer a
torque; if one made it movable around an axis O, it would tend to turn
in the dire tion of the arrow.
This paradoxi al on lusion is entirely avoided by L. de Broglie, sin e
his system of hypotheses preserves the values of radiation pressure.
This example shows learly that there is a ontradi tion between the
hypothesis of re tilinear paths for the photons (following the light rays)
and the ne essity of nding photons only where a bright interferen e
fringe is produ ed, no photon going through the regions of dark fringes.
Mr Lorentz
draws attention to a ase where the lassi al theory

and the photon hypothesis lead to dierent results on erning the ponderomotive for es produ ed by light. Let us onsider ree tion by the
hypotenuse fa e of a glass prism, the angle of in iden e being larger than
the angle of total ree tion. Let us pla e a16 se ond prism behind the
rst, at a distan e of the order of magnitude of the wavelength, or only
Dis ussion
405
a fra tion of this length. Then, the ree tion will no longer be total.
The light waves that penetrate the layer of air rea h the se ond prism
before their intensity is too mu h weakened, and there give rise to a
beam transmitted in the dire tion of the in ident rays.
If, now, one al ulates the Maxwell stresses on a plane situated in
the layer of air and parallel to its surfa es, one nds that, if the angle
of in iden e ex eeds a ertain value (60 for example), there will be an
attra tion between the two prisms. Su h an ee t an never be produ ed
by the motion of orpus les, this motion always giving rise to a [positive
pressure as in the kineti theory of gases.
What is more, in the lassi al theory one easily sees the origin of the
`negative pressure'. One an distinguish two ases, that where the ele tri
os illations are in the plane of in iden e and that where this is so for the
magneti os illations. If the in iden e is very oblique, the os illations of
the in ident beam that I have just mentioned are only slightly in lined
with respe t to the normal to the hypotenuse fa e, and the same is true
for the orresponding os illations in the layer of air.
One then has approximately, in the rst ase an ele tri eld su h as
one nds between the ele trodes of a apa itor, and in the se ond ase
a magneti eld su h as exists between two opposite magneti poles.
The ee t would still remain if the se ond prism were repla ed by a
glass plate, but it must be very di ult to demonstrate this experimentally.
406
Notes to pp. 375404

1 The integral sign is printed as `
' in the original.

0
2 The Fren h uses `Wellenpa ket' throughout.
3 Mis-spelt as `Fuess'.
4 We follow the original English, whi h is a translation by Oppenheimer,
from Born (1927, p. 355). De Broglie translates 'me hani s' as
`dynamique' and in ludes the words `La Dynamique des Quanta' in the
quotation, where Born has ìt' (referring to `quantum me hani s').
5 The Fren h reads ` es' [these rather than `ses' [its.

6 `v ' is misprinted as `V '.
7 Consistently mis-spelt throughout the text as Èlsaesser'.

8 For larity, the presentation of the table has been slightly altered.
9 We have used the original English (Dymond 1927, p. 441).
10 We use here the original English text (Patterson 1927, p. 47). De Broglie
hanges òf these results' to `des rsultats exprimentaux', omits the
itali s and translates `valuable' by R
èxa ts'.
11 The last equality is misprinted as `

p dq = nn hn '.
n n
12 The original reads `p. 18', whi h is presumably in referen e to de
Broglie's Gauthier-Villars `preprint', in all likelihood ir ulated before the
onferen e (preprints of other le tures were ir ulated as mimeographs).
Cf. hapter 1, p. 19
13 Ànn.
de Phys.'
in the original.
14 This is `p. 117' in the original.

15 Again, `p. 117' in the original.
16 Misprinted as àu' instead of ùn'.
Quantum me hani s
by Messrs Max BORN and Werner HEISENBERG

Introdu tion
Quantum me hani s is based on the intuition that the essential dieren e
between atomi physi s and lassi al physi s is the o urren e of dis ontinuities (see in parti ular [1,4,5863).b Quantum me hani s should thus
be onsidered a dire t ontinuation of the quantum theory founded by
Plan k, Einstein and Bohr. Bohr in parti ular stressed repeatedly, already before the birth of quantum me hani s, that the dis ontinuities must
lead to the introdu tion of new kinemati al and me hani al on epts, so
that indeed lassi al me hani s and its orresponding on eptual s heme
should be abandoned [1,4. Quantum me hani s tries to introdu e the
new on epts through a pre ise analysis of what is òbservable in prin iple'. In fa t, this does not mean setting up the prin iple that a sharp division between òbservable' and ùnobservable' quantities is possible and
ne essary. As soon as a on eptual s heme is given, one an infer from
the observations to other fa ts that are a tually not observable dire tly,
and the boundary between òbservable' and ùnobservable'1 quantities
be omes altogether indeterminate. But if the on eptual s heme itself is
still unknown, it will be expedient to enquire only about the observations
themselves, without drawing on lusions from them, be ause otherwise
wrong on epts and prejudi es taken over from before will blo k the way
a Our translation follows the German types ript in AHQP-RDN, do ument M-0309.
Dis repan ies between the types ript and the published version are reported in the
endnotes. The published version is reprinted in Heisenberg (1984, ser. B, vol. 2,
pp. 5899) (eds.).
b Numbers in square bra kets refer to the bibliography at the end.
407
408
M. Born and W. Heisenberg
to re ognising the physi al relationships [Zusammenhnge. At the same

time the new on eptual s heme provides the ans hauli h ontent of the
new theory.a From a theory that is ans hauli h in this sense, one an
thus demand only that it is onsistent in itself and that it allows one to
predi t unambiguously the results for all experiments on eivable in its
domain. Quantum me hani s is meant as a theory that is in this sense
ans hauli h and omplete for the mi rome hani al pro esses [46.2
Two kinds of dis ontinuities are hara teristi of atomi physi s: the
existen e of orpus les (ele trons, light quanta) on the one hand, and
the o urren e of dis rete stationary states (dis rete3 energy values,
momentum values et .) on the other. Both kinds of dis ontinuities an
be introdu ed in the lassi al theory only through arti ial auxiliary assumptions. For quantum me hani s, the existen e of dis rete stationary
states and energy values is just as natural as the existen e of dis rete
eigenos illations in a lassi al os illation problem [4. The existen e of
orpus les will perhaps later turn out to be redu ible just as easily to
dis rete stationary states of the wave pro esses (quantisation of the
ele tromagneti waves on the one hand, and of the de Broglie waves
on the other) [4, [54.
The dis ontinuities, as the notion of `transition probabilities' already
shows, introdu e a statisti al element into atomi physi s. This statisti al element forms an essential part of the foundations of quantum
me hani s (see in parti ular [4,30,38,39,46,60,61,62);4 a ording to the
latter, for instan e, in many ases the ourse of an experiment is determinable from the initial onditions only statisti ally, at least if in
xing the initial onditions one takes into a ount only the experiments
on eivable in prin iple up to now. This onsequen e of quantum me hani s is empiri ally testable. Despite its statisti al hara ter, the theory
nevertheless a ounts for the apparently fully ausal determination of
ma ros opi 5 pro esses. In parti ular, the prin iples of onservation of
energy and momentum hold exa tly also in quantum me hani s. There
seems thus to be no empiri al argument against a epting fundamental
indeterminism for the mi ro osm.
a For the notion of Ans hauli hkeit, see the omments in se tions 3.4.7, 4.6 and 8.3
(eds.).
Quantum me hani s
I. The mathemati al methods of quantum me hani s
409
a
The phenomenon for whose study the mathemati al formalism of quantum

me hani s was rst developed is the spontaneous radiation of an ex ited
atom. After innumerable attempts to explain the stru ture of the line
spe tra with lassi al me hani al models had proved inadequate, one
returned to the dire t des ription of the phenomenon on the basis of
its simplest empiri al laws (Heisenberg [1). First among these is Ritz's
ombination prin iple, a ording to whi h the frequen y of ea h spe tral
line of an atom appears as the dieren e of two terms ik = Ti Tk ;
thus the set of all lines of the atom will be best des ribed by spe ifying
a quadrati array [S hema, and sin e ea h line possesses besides its
frequen y also an intensity and a phase, one will write in ea h position
of the array an elementary os illation fun tion with omplex amplitude:
q11 e2i11 t q12 e2i12 t . . .

q21 e2i21 t q22 e2i22 t . . . .
(1)
.........
......... ...
This array is understood as representing a oordinate q as a fun tion of
time in a similar way as the totality of terms of the Fourier series
X
q(t) =
qn e2in t , n = n0
n
in the lassi al theory; ex ept that now be ause of the two indi es the
sum no longer makes sense. The question arises of whi h expressions
orrespond to fun tions of the lassi al oordinate, for instan e to the
square q 2 . Now, su h arrays ordered by two indi es o ur as matri es in
mathemati s in the theory of quadrati forms and of linear transformations; the omposition of two linear transformations,
X
X
blj zj ,
xk =
akl yl ,
yl =
j
to form a new one,
xk =
ckj zj ,
then orresponds to the omposition or multipli ation of the matri es

X
ab = c, that is,
(2)
akl blj = ckj .
l
a Se tion 3.3 ontains additional material on the less familiar aspe ts of the
formalisms presented here (eds.).
410
This multipli ation in general is not ommutative. It is natural to apply

this re ipe to the array of the atomi os illations (Born and Jordan
[2, Dira [3); it is immediately evident that be ause of Ritz's formula
ik = Ti Tk no new frequen ies appear, just as in the lassi al theory
in the multipli ation of two Fourier series, and herein lies the rst justi ation for the pro edure. By repeated appli ation of additions and
multipli ations one an dene arbitrary matrix fun tions.
The analogy with the lassi al theory leads further to allowing as
representatives of real quantities only those matri es that are `Hermitian', that is, whose elements go over to the omplex onjugate numbers
under permutation of the indi es. The dis ontinuous nature of the atomi pro esses here is put into the theory from the start as empiri ally
established. However, this does not establish yet the onne tion with
quantum theory and its hara teristi onstant h. This is also a hieved,
by arrying over the ontent6 of the Bohr-Sommerfeld quantum onditions in a form given by Kuhn and Thomas, in whi h they are written
as relations between the Fourier oe ients of the oordinates q and
momenta p. In this way one obtains the matrix equation
h
1 ,
(3)
2i
where 1 means the unit matrix. The matrix p thus does not ` ommute'
with q . For several degrees of freedom the ommutation relation (3) holds
for every pair of onjugate quantities, while the qk ommute with ea h
other, the pk with ea h other, and also the pk with the non- orresponding
qk .
In order to onstru t the new me hani s (Born, Heisenberg and Jordan
[4), one arries over as far as possible the notions of the lassi al theory.
It is possible to dene the dierentiation of a matrix with respe t to
time and that of a matrix fun tion with respe t to an argument matrix.
One an thus arry over to the matrix theory the anoni al equations
pq qp =
H
dq
=
,
dt
p
dp
H
=
,
dt
q
where one should understand H(p, q) as the same fun tion of the matri es p, q that o urs in the lassi al theory as a fun tion of the numbers
p, q . (To be sure,7 ambiguity an o ur be ause of the non ommutativity of the multipli ation; for example, p2 q is dierent from pqp.) This
pro edure was tested in simple examples (harmoni and anharmoni
os illator). Further, one an prove the theorem of onservation of energy, whi h for non-degenerate systems (all terms Tk dierent from ea h
Quantum me hani s
411
other, or: all frequen ies ik dierent from zero) here takes the form: for
the solutions p, q of the anoni al equations the Hamiltonian fun tion
H(p, q) be omes a diagonal matrix W . It follows immediately that the
elements of this diagonal matrix represent the terms Tn of Ritz's formula
multiplied by h (Bohr's frequen y ondition). It is parti ularly important
to realise that onversely the requirement
H(p, q) = W
(diagonal matrix)
is a omplete substitute for the anoni al equations of motion, and leads

to unambiguously determined solutions even if one allows for degenera ies (equality of terms, vanishing frequen ies).
By a matrix with elements that are harmoni fun tions of time, one
an of ourse represent only quantities ( oordinates) that orrespond to
time-periodi quantities of the lassi al theory. Therefore y li oordinates (angles), whi h in rease proportionally to time, annot be treated
at present.a Nevertheless, one easily manages to subje t rotating systems
to the matrix method by representing the Cartesian omponents of the
angular momentum with matri es [4.8 One obtains thereby expressions
for the energyb that dier hara teristi ally from the orresponding
lassi al ones; for instan e the modulus9 of the p
total angular momentum
h
h
j
(j
=
0,
1,
2,
.
.
.)
, but to 2 j(j + 1), in a ordan e
is not equal to 2
with empiri al rules that Land and others had derived from the term
splitting in the Zeeman ee t.c Further, one obtains for the hanges
in the angular quantum numbers [Rotationsquantenzahlen the orre t
sele tion rules and intensity formulas, as had already been arrived at
earlier by orresponden e arguments and onrmed by the Utre ht observations.d
Pauli [6, avoiding angular variables, even managed to work out the
hydrogen atom with matrix me hani s, at least with regard to the energy
values and some aspe ts of the intensities.
Asking for the most general oordinates for whi h the quantum me hani al laws are valid leads to the generalisation of the notions of anoni al variables and anoni al transformations known from the lassi al
theory. Dira [3 has noted that the ontent of the expressions su h as
a This point is taken up again shortly after eq. (10). (eds.).
b Angular momentum is of ourse responsible for a hara teristi splitting of the
energy terms (eds.).
For Land's work on the anomalous Zeeman ee t, see Mehra and Re henberg
(1982a, se . IV.4, esp. pp. 46776 and 4825) (eds.).
d For the Ùtre ht observations' see Mehra and Re henberg (1982a, se . VI.6,
pp. 6478) and Mehra and Re henberg (1982b, se . III.4, esp. pp. 15461) (eds.).
412
ql pk ) kl , whi h appear in the ommutation relations of the

type (3) orresponds10 to that of the Poisson bra kets, whose vanishing
in lassi al me hani s hara terises a system of variables as anoni al.
Therefore also in quantum me hani s one will denote as anoni al every
system of matrix pairs p, q that satisfy the ommutation relations, and
as a anoni al transformation every transformation that leaves these
relations invariant. One an write these with the help of an arbitrary
matrix S in the forma
2i
h (pk ql
P = S 1 pS,
Q = S 1 qS ,
(4)
and in a ertain sense this is the most general anoni al transformation.

Then for an arbitrary fun tion one has
f (P, Q) = S 1 f (p, q)S .
Now one an also arry over the main idea of the Hamilton-Ja obi theory
[4. Indeed, if the Hamiltonian fun tion H is given as a fun tion of any
known anoni al matri es p0 , q0 , then the solution of the me hani al
problem dened by H redu es to nding a matrix S that satises the
equation
S 1 H(p0 , q0 )S = W .
(5)
This is an analogue of the Hamilton-Ja obi dierential equation of lassi al me hani s.

Exa tly as there, also here perturbation theory an be treated most
learly with the help of equation (5). If H is given as a power series in
some small parameter
H = H0 + H1 + 2 H2 + . . .
and the me hani al problem is solved for = 0, that is, H0 = W0 is

known as a diagonal matrix, then the solution to (5) an be obtained
easily as a power series
S = 1 + S1 + 2 S2 + . . .
by su essive approximations. Among the numerous appli ations of this

pro edure, only the derivation of Kramers' dispersion formula shall be
mentioned here, whi h results if one assumes that the light-emitting
and the s attering systems are weakly oupled and if one al ulates the
perturbation on the latter ignoring the ba krea tion [4.a
a Cf. p. 99 above (eds.).
a In other words, one onsiders just the s attering system under an external
Quantum me hani s
413
The theory of the anoni al transformations leads to a deeper on eption, whi h later be ame essential in understanding the physi al meaning
of the formalism.
To ea h matrix a = (anm ) one an asso iate a quadrati (more pre isely: Hermitian) forma
X
anm n m
nm
of a sequen e of variables 1 , 2 . . ., or also a linear transformation of

the sequen e of variables 1 , 2 . . ., into another one 1 , 2 . . .11
X
n =
anm m ,
(6)
m
where provisionally the meaning of the variables n and n shall be left

unspe ied; we shall return to this.
A transformation (6) is alled òrthogonal' if it maps the identity form
into itself
X
X
(7)
n n =
n n .
n
Now these orthogonal transformations of the auxiliary variables n immediately turn out to be essentially identi al to the anoni al transformations of the q and p matri es; the Hermitian hara ter and the
ommutation relations are preserved. Further, one an repla e the matrix
equation (5) by the equivalent requirement [4: the form
X
Hnm (q0 , p0 )n m
nm
is to be transformed orthogonally into a sum of squares

X
wn n n .
(8)
The fundamental problem of me hani s is thus none other than the

prin ipal axes problem for surfa es of se ond order in innite-dimensional
spa e, o urring everywhere in pure and applied mathemati s and variously studied. As is well known, this is equivalent to asking for the
values of the parameter W for whi h the linear equations
X
W n =
(9)
Hnm m
m
perturbation (Born, Heisenberg and Jotrdan [4, se tion 2.4, in parti ular eq. (32)).
See also Mehra and Re henberg (1982 , h. III, esp. pp. 934 and 1039) (eds.).
a denotes the omplex number onjugate to .
414
have a non-identi ally vanishing solution. The values W = W1 , W2 , . . .

are alled eigenvalues of the form H ; they are the energy values (terms)
of the me hani al system. To ea h eigenvalue Wn orresponds an eigensolution k = kn .a The set of these eigensolutions evidently again forms a
matrix and it is easy to see that this is identi al with the transformation
matrix S appearing in (5).b
The eigenvalues, as is well known, are invariant under orthogonal
transformations of the k ,12 and sin e these orrespond to the anoni al
substitutions of the p and q matri es, one re ognises immediately the
anoni al invarian e of the energy values Wn .
While the quantum theoreti al matri es do not belong to the lass of
matri es (nite and bounded innite13 ) investigated by the mathemati ians (espe ially by Hilbert and his s hool), one an nevertheless arry
over the main aspe ts of the known theory to the more general ase.
The pre ise formulation of these theorems14 has been re ently given by
J. von Neumann [42 in a paper to whi h we shall have to return.15
The most important result that is a hieved in this way is the theorem
that a form annot always be de omposed into a sum of squares (8), but
that there also o ur invariant integral omponents
Z
(10)
W (W )(W )dW ,
where the sequen e of variables 1 , 2 , . . . has to be omplemented by
the ontinuous distribution (W ).
In this way the ontinuous spe tra appear in the theory in the most
natural way. But this implies by no means that in this domain the
lassi al theory omes again into its own. Also here the hara teristi dis ontinuities of quantum theory remain; also in the ontinuous spe trum
a (spontaneous) state transition onsists of a `jump' of the system from
a point W to another one W with emission of a wave q(W, W )e2it
with the frequen y = h1 (W W ).
The main defe t of matrix me hani s onsists in its lumsiness, even
helplessness, in the treatment of non-periodi quantities, su h as angular
variables or oordinates that attain innitely large values (e.g. hyperboli
traje tories). To over ome this di ulty two essentially dierent routes
have been taken, the operator al ulus of Born and Wiener [21, and the
so- alled16 q-number theory of Dira [7.
a This is the notation used by Born and Heisenberg: the nth eigensolution is
represented by an innite ve tor with omponents labelled by k (eds.).
b This point is made more expli it after eq. (17). See also the relevant ontributions
by Dira and by Kramers in the general dis ussion, pp. 491 and 495 (eds.).
Quantum me hani s
415
The latter starts from the idea that a great part of the matrix relations an be obtained without an expli it representation of the matri es,
simply on the basis of the rules for operating with the matrix symbols.
These depart from the rules for numbers only in that the multipli ation is
generally not ommutative. Dira therefore onsiders abstra t quantities,
whi h he alls q-numbers (as opposed to the ordinary -numbers) and
with whi h he operates a ording to the rules of the non ommutative
algebra. It is therefore a kind of hyper omplex number system. The
ommutation relations are of ourse preserved. The theory a quires an
extraordinary resemblan e to the lassi al one; for instan e, one an
introdu e angle and a tion variables w, J and expand any q-number into
a Fourier series with respe t to the w; the oe ients are fun tions of the
J and turn out to be identi al to the matrix elements if one repla es the J
by integer multiples of h. By his method Dira has a hieved important
results, for instan e worked out the hydrogen atom independently of
Pauli [7 and determined the intensity of radiation in the Compton ee t
[12. A drawba k of this formalism apart from the quite tiresome
dealing with the non ommutative algebra is the ne essity to repla e
at a ertain point of the al ulation ertain q-numbers with ordinary
numbers (e.g. J = hn), in order to obtain results omparable with
experiment. Spe ial `quantum onditions' whi h had disappeared from
matrix me hani s are thus needed again.
The operator al ulus diers from the q-number method in that it does
not introdu e abstra t hyper omplex numbers, but on rete, onstru tible mathemati al obje ts that obey the same laws, namely operators
or fun tions in the spa e of innitely many variables. The method is by
E kart [22 and was then developed further by many others following on
from S hrdinger's wave me hani s, espe ially by Dira [38 and Jordan
[39 and in an impe able mathemati al form by J. von Neumann [42;
it rests roughly on the following idea.
A sequen e of variables 1 , 2 , . . . an be interpreted as a point in
P
an innite-dimensional spa e. If the sum of squares n |n |2 onverges,
then it represents a measure of distan e, a Eu lidean metri [Massbestimmung, in this spa e; this metri spa e of innitely many dimensions
is alled for short a Hilbert spa e. The anoni al transformations of
matrix me hani s orrespond thus to the rotations of the Hilbert spa e.
Now, however, one an also x a point in this spa e other than by
the spe i ation of dis rete oordinates 1 , 2 , . . . . Take for instan e
a omplete, normalised orthogonal system of fun tions f1 (q), f2 (q), . . . ,
416
that is one for whi h17

Z
1 for n = m
fn (q)fm (q)dq = nm =
0 for n 6= m ;
(11)
the variable q an range here over an arbitrary, also multi-dimensional

domain. If one then sets (Lan zos [23)18
(
P
(q)
=
n n fn (q)
(12)
P

H(q , q ) =
nm Hnm fn (q )fm (q ) ,
the linear equations (9)19 turn into the integral equation20
Z
W (q ) = H(q , q )(q )dq .
(13)
This relation established through (12) means thus nothing but a hange
of the oordinate system in the Hilbert spa e, given by the orthogonal
transformation matrix fn (q) with one dis rete and one ontinuous index.
One sees thus that the preferen e for `dis rete' oordinate systems
in the original version of the matrix theory is by no means something
essential. One an just as well use ` ontinuous matri es' su h as H(q , q ).
Indeed, the spe i representation of a point in the Hilbert spa e by
proje tion onto ertain orthogonal oordinate axes does not matter at
all; rather, one an summarise equations (9) and (13) in the more general
equation
W = H ,
(14)
where H denotes a linear operator whi h transforms the point of the

Hilbert spa e into another. The equation requires to nd those points
whi h under the operation H only suer a displa ement along the
line joining them to the origin.21 The points satisfying this ondition
determine an orthogonal system of axes, the prin ipal axes frame of
the operator H ; the number of axes is nite or innite, in the latter
ase distributed dis retely or ontinuously, and the eigenvalues W are
the lengths of the prin ipal axes. The linear operators in the Hilbert
spa e are thus the general on ept that an serve to represent a physi al
quantity mathemati ally. The al ulus with operators pro eeds obviously a ording to the same rules as the one with Dira 's q-numbers; they22
onstitute a realisation of this abstra t notion. So far we have analysed
the situation with the example of the Hamiltonian fun tion, but the
same holds for any quantum me hani al quantity. Any oordinate q
an be written, instead of as a matrix with dis rete indi es qnm , also
Quantum me hani s
417
as a fun tion of two ontinuous variables q(q , q ) by proje tion onto an

orthogonal system of fun tions, or, more generally, an also be onsidered
as a linear operator in the Hilbert spa e; then it has eigenvalues that are
invariant, and eigensolutions with respe t to ea h orthogonal oordinate
system. The same holds for a momentum p and every fun tion of q
and p, indeed for every quantum me hani al `quantity'. While in the
lassi al theory physi al quantities are represented by variables that an
take numeri al values from an arbitrary value range, a physi al quantity
in quantum theory is represented by a linear operator and the sto k of
values that it an take by the eigenvalues of the orresponding prin ipal
axes problem in the Hilbert spa e.
In this view, S hrdinger's wave me hani s [24 appears formally as a
spe ial ase. The simplest operator whose hara teristi values are all the
real numbers, is in fa t the multipli ation of a fun tion F (q) by the real
number q ; one writes it simply q . Then, however, the eigenfun tions are
ìmproper' fun tions; for a ording to (14) they must have the property
of being everywhere zero ex ept if W = q . Dira [38 has introdu ed for
the representation of su h improper fun tions the ùnit fun tion' (s),
should always be zero when s 6= 0, but for whi h nonetheless
Rwhi h
+
(s)ds
= 1 should hold. Then one an write down the (normalised)
eigenfun tions
(q, W ) = q(W q)
(15)
belonging to the operator q .

The onjugate to the operator q is the dierential operator
p=
h
;
2i q
(16)
indeed, the ommutation relation (3) holds, whi h means just the trivial
identity

h
h
d
dF
(pq qp)F (q) =
=
(qF ) q
F (q) .
2i dq
dq
2i
If one now onstru ts a Hamiltonian fun tion out of p, q (or out of

several su h onjugate pairs), then equation (14) be omes a dierential
equation for the quantity (q):
H(q,
h
)(q) = W (q) .
2i q
(17)
This is S hrdinger's wave equation, whi h appears here as a spe ial ase
of the operator theory. The most important point about this formulation
418
of the quantum laws (apart from the great advantage of onne ting
to known mathemati al methods) is the repla ement of all `quantum
onditions', su h as were still ne essary in Dira 's theory of q-numbers,
by the simple requirement that the eigenfun tion (q) = (q, W ) should
be everywhere nite in the domain of denition of the variables q ;
from this, in the event, a dis ontinuous spe trum of eigenvalues Wn
(along with ontinuous ones) arises automati ally. But S hrdinger's
eigenfun tion (q, W ) is a tually nothing but the transformation matrix
S of equation (5), whi h one an indeed also write in the form
HS = SW ,
analogous to (17).
Dira [38 has made this state of aairs even learer by writing the
operators q and p and thereby also H as integral operators, as in (13);
then one has to set
Z
qF (q ) =
q (q q )F (q )dq
= q F (q ) ,
Z
(18)
h dF
h
,
(q q )F (q )dq =
pF (q ) =
2i
2i dq
where, however, the o urren e of the derivative of the singular fun tion
has to be taken into the bargain. Then S hrdinger's equation (17)
takes the form (13).
The dire t passage to the matrix representation in the stri t sense
takes pla e by inverting the formulas (12), in whi h one identies the
orthogonal system fn (q) with the eigenfun tions (q, Wn ) belonging to
the dis rete spe trum. If T is an arbitrary operator ( onstru ted from q
h
and p = 2i
q ), dene the orresponding matrix Tnm by the oe ients
of the expansion
X
T n (q) =
(19)
Tnm m (q)
m
or
Tnm =
m (q)T n (q)dq ;
(19a)
then one easily sees that equation (17) is equivalent to (9).

The further development of the formal theory has taken pla e in lose
onne tion with its physi al interpretation, to whi h we therefore turn
rst.
Quantum me hani s
II. Physi al interpretation
419
The most noti eable defe t of the original matrix me hani s onsists in
the fa t that at rst it appears to give information not about a tual
phenomena, but rather only about possible states and pro esses. It
allows one to al ulate the possible stationary states of a system; further
it makes a statement about the nature of the harmoni os illation that
an manifest itself as a light wave in a quantum jump. But it says nothing
about when a given state is present, or when a hange is to be expe ted.
The reason for this is lear: matrix me hani s deals only with losed
periodi systems, and in these there are indeed no hanges. In order to
have true pro esses,23 as long as one remains in the domain of matrix
me hani s, one must dire t one's attention to a part of the system; this is
no longer losed and enters into intera tion with the rest of the system.
The question is what matrix me hani s an tell us about this.
Imagine, for instan e, two systems 1 and 2 weakly oupled to ea h
other (Heisenberg [35, Jordan [36).a For the total system onservation
of energy then holds; that is, H is a diagonal matrix. But for a subsystem,
for instan e 1, H (1) is not onstant, the matrix has elements o the
diagonal.24 The energy ex hange an now be interpreted in two ways: for
one, the periodi elements of the matrix of H (1) (or of H (2) ) represent
a slow beating, a ontinuous os illation of the energy to and fro; but
at the same time, one an also des ribe the pro ess with the on epts
of the dis ontinuum theory and say that system 1 performs quantum
jumps and arries over the energy that is thereby freed to system 2 as
quanta, and vi e versa. But one an now show that these two apparently
very dierent views do not ontradi t ea h other at all. This rests on a
mathemati al theorem that states the following:
(1)
(1)
Let f (Wn ) be any fun tion of the energy values Wn of the isolated
subsystem 1; if one forms the same fun tion of the matrix H (1) that
represents the energy of system 1 in the presen e of the oupling to
system 2, then f (H (1) ) is a matrix that does not onsist only of diagonal
elements f (H (1) )nn . But these represent the time-averaged value of the
quantity f (H (1) ). The ee t of the oupling is thus measured by the
dieren e25
fn = f (H (1) )nn f (Wn(1) ) .
The rst part of the said theorem now states that fn an be brought
a The form of the result as given here is similar to that in Heisenberg [35. For
further details, see the dis ussion in se tion 3.4.4 (eds.).
420
into the form26

fn =
X
m
{f (Wm ) f (Wn )}nm .
(20)
This an be interpreted thus: the time average of the hange in f due

to the oupling is the arithmeti mean, with ertain weightings nm , of
all possible jumps of f for the isolated system.
These nm will have to be alled `transition probabilities'. The se ond
part of the theorem determines the nm through the features of the
oupling. Namely, if p01 , q10 , p02 , q20 are oordinates satisfying the evolution
equations of the un oupled systems, for whi h therefore H (1) and H (2)
on their own are diagonal matri es, one an then think of the energy,
in luding the intera tion, as expressed as a fun tion of these quantities.
Then the solution of the me hani al problem a ording to (5)27 redu es
to onstru ting a matrix S that satises the equation
S 1 H(p01 , q10 , p02 , q20 )S = W .
Denoting the states of system 1 by n1 , those of system 2 by n2 , a state of

the total system is given by n1 n2 ,28 and to ea h transition n1 n2 m1 m2
orresponds an element of S , Sn1 n2 ,m1 m2 . Then the result is:29
X
|Sn1 n2 ,m1 m2 |2 .
n1 m1 =
(21)
n2 m2
The squares of the elements of the S -matrix thus determine the transition probabilities. The individual sum term |Sn1 n2 ,m1 m2 |2 in (21) obviously means that omponent of the transition probability for the jump
n1 m1 of system 1 that is indu ed by the jump n2 m2 of system 2.
By means of these results the ontradi tion between the two views
from whi h we started is removed. Indeed, for the mean values, whi h
alone may be observed, the on eption of ontinuous beating always
leads to the same result as the on eption of quantum jumps.
If one asks the question when a quantum jump o urs, the theory
provides no answer. At rst it seemed as if there were a gap here whi h
might be lled with further probing. But soon it be ame apparent that
this is not so, rather, that it is a failure of prin iple, whi h is deeply
an hored in the nature of the possibility of physi al knowledge [physikalis hes Erkenntnisvermgen.
One sees that quantum me hani s yields mean values orre tly, but
annot predi t the o urren e of an individual event. Thus determinism,
held so far to be the foundation of the exa t natural s ien es, appears
Quantum me hani s
421
here to go no longer un hallenged. Ea h further advan e in the interpretation of the formulas has shown that the system of quantum me hani al
formulas an be interpreted onsistently only from the point of view of a
fundamental indeterminism, but also, at the same time, that the totality
of empiri ally as ertainable fa ts an be reprodu ed by the system of the
theory.
In fa t, almost all observations in the eld of atomi physi s have a
statisti al hara ter; they are ountings, for instan e of atoms in a ertain
state. While the determinateness of an individual pro ess is assumed
by lassi al physi s, in fa t it plays pra ti ally no role, be ause the
mi ro oordinates that exa tly determine an atomi pro ess an never all
be given; therefore by averaging they are eliminated from the formulas,
whi h thereby be ome statisti al statements. It has be ome apparent
that quantum me hani s represents a merging of me hani s and statisti s, in whi h the unobservable mi ro oordinates are eliminated.
The lumsiness of the matrix theory in the des ription of pro esses
developing in time an be avoided by making use of the more general
formalisms30 we have des ribed above. In the general equation (14)
one an easily introdu e time expli itly by invoking the theorem of
lassi al me hani s that energy W and time t behave as anoni ally
onjugate quantities; in quantum me hani s it orresponds to having a
ommutation relation
h
W t tW =
.
2i
Thus for W one an posit the operator
h
2i t .
Equation (14) then reads
h
(22)
= H ,
2i t
and here one an onsider H as depending expli itly on time. A spe ial
ase of this is the equation

h
h
H(q,
(22a)
(q) = 0 ,
)
2i q
2i t
given by S hrdinger [24,31 whi h stands to (17) in the same relation

as (22) to (14), as well as the form:
Z
h (q )
(22b)
= H(q , q )(q )dq ,
2i t
mu h used by Dira , whi h relates to the integral formula (13). Essentially, the introdu tion of time as a numeri al variable redu es to thinking of
the system under onsideration as oupled to another one and negle ting
422
the rea tion on the latter. But this formalism is very onvenient and leads
to a further development of the statisti al view,a namely, if one onsiders
the ase where an expli itly time-dependent perturbation V (t) is added
to a time-independent energy fun tion H 0 , so that one has the equation

h 0
(23)
= H + V (t)
2i t
(Dira [37, Born [34).32 Now if 0n are the eigenfun tions of the operator
H 0 , whi h for the sake of simpli ity we assume to be dis rete, the desired
quantity an be expanded in terms of these:
X
(t) =
(24)
cn (t)0n .
n
The cn (t) are then the oordinates of in the Hilbert spa e with respe t
to the orthogonal system 0n ; they an be al ulated from the dierential
equation (23), if their initial values cn (0) are given. The result an be
expressed as:
X
cn (t) =
(25)
Snm (t)cm (0) ,
m
where Snm (t) is an orthogonal matrix depending33 on t and determined

by V (t).
The temporal pro ess is thus represented by a rotation of the Hilbert
spa e or by a anoni al transformation (4) with the time-dependent
matrix S .
Now how is one to interpret this?
From the point of view of Bohr's theory a system an always be
in only one quantum state. To ea h of these belongs an eigensolution
0n of the unperturbed system. If now one wishes to al ulate what
happens to a system that is initially in a ertain state, say the k th,
one has to hoose = 0k as the initial ondition for equation (23), i.e.
cn (0) = 0 for n 6= k , and ck (0) = 1. But then, after the perturbation is
over, cn (t) will have be ome equal to Snk (t), and the solution onsists
of a superposition of eigensolutions. A ording to Bohr's prin iples it
makes no sense to say a system is simultaneously in several states. The
only possible interpretation seems to be statisti al: the superposition
of several eigensolutions expresses that through the perturbation the
initial state an go over to any other quantum state, and it is lear that
as measure for the transition probability one has to take the quantity
nk = |Snk (t)|2 ;
a See the dis ussion in se tion 3.4 (eds.).
Quantum me hani s
423
be ause then one obtains again equation (20) for the average hange of
any state fun tion.
This interpretation is supported by the fa t that one establishes the
validity of Ehrenfest's adiabati theorem (Born [34); one an show that
under an innitely slow a tion, one has
nn 1,
nk 0
(n 6= k) ,
that is, the probability of a jump tends to zero.

But this assumption also leads immediately to an interpretation of the
cn (t) themselves: the |cn (t)|2 must be the state probabilities [Zustandswahrs heinli hkeiten.
Here, however, one runs into a di ulty of prin iple that is of great
importan e, as soon as one starts from an initial state for whi h not all
the cn (0) ex ept one vanish. Physi ally, this ase o urs if a system is
given for whi h one does not know exa tly the quantum state in whi h
it is, but knows only the probability |cn (0)|2 for ea h quantum state.
As a matter of fa t, the phases [Ar us of the omplex quantities cn (0)
still remain indenite; if one sets cn (0) = |cn (0)|ein , then the n denote
some phases whose meaning needs to be established. The probability
distribution at the end of the perturbation a ording to (25) is then
X
2

|cn (t)|2 =
(26)
Snm (t)cm (0)
m
and not
X
m
|Snm (t)|2 |cm (0)|2 ,
(27)
as one might suppose from the usual probability al ulus.

Formula (26), following Pauli, an be alled the theorem of the interferen e of probabilities; its deeper meaning has be ome lear only
through the wave me hani s of de Broglie and S hrdinger, whi h we
shall presently dis uss. Before this, however, it should be noted that
this ìnterferen e' does not represent a ontradi tion with the rules of
the probability al ulus, that is, with the assumption that the |Snk |2
are quite usual probabilities.a In fa t, the omposition rule (27) follows
from the on ept of probability for the problem treated here when and
only when the relative number, that is, the probability |cn |2 of the atoms
in the state n, has been established beforehand experimentally.34 In this
a The notation |Snm |2 would probably be learer, at least a ording to the reading
of this passage proposed in se tion 3.4.6 (eds.).
424
ase the phases n are unknown in prin iple,35 so that (26) then naturally
goes over to (27) [46.
It should be noted further that the formula (26) goes over to the
expression (27) if the perturbation fun tion pro eeds totally irregularly
as a fun tion of time. That is for instan e the ase when the perturbation
is produ ed by `white light'.a Then, on average, the surplus terms in
(26) drop out and one obtains (27). In this way it is easy to derive
the Einstein oe ient Bnm for the probability per unit radiation of the
quantum jumps indu ed by light absorption (Dira [37, Born [30). But,
in general, a ording to (26) the knowledge of the probabilities |cn (0)|2
is by no means su ient to al ulate the ourse of the perturbation,
rather one has to know also the phases n .
This ir umstan e re alls vividly the behaviour of light in interferen e
phenomena. The intensity of illumination on a s reen is by no means
always equal to the sum of the light intensities of the individual beams
of rays that impinge on the s reen, or, as one an well say, it is by no
means equal to the sum of the light quanta that move in the individual
beams; instead it depends essentially on the phases of the waves. Thus
at this point an analogy between the quantum me hani s of orpus les
and the wave theory of light be omes apparent.
As a matter of fa t this onne tion was found by de Broglie in a
quite dierent way. It is not our purpose to dis uss this. It is enough
to formulate the result of de Broglie's onsiderations, and their further
development by S hrdinger, and to put it in relation to quantum me hani s.
The dual nature of light waves, light quanta orresponds to
the analogous dual nature of material parti les; these also behave in a
ertain respe t like waves. S hrdinger has set up the laws of propagation
of these waves [24 and has arrived at equation [(17),36 here derived in a
dierent way. His view, however, that these waves exhaust the essen e of
matter and that parti les are nothing but wave pa kets, not only stands
in ontradi tion with the prin iples of Bohr's empiri ally very wellfounded theory, but also leads to impossible on lusions; here therefore
it shall be left to one side. Instead we attribute a dual nature to matter
also: its des ription requires both orpus les (dis ontinuities) and waves
( ontinuous pro esses). From the viewpoint of the statisti al approa h to
quantum me hani s it is now lear why these an be re on iled: the waves
are probability waves. Indeed, it is not the probabilities themselves,
a Compare also Born's dis ussion in Born (1926 [34) (eds.).
Quantum me hani s
425
rather ertain `probability amplitudes' that propagate ontinuously and

obey dierential or integral equations, as in lassi al ontinuum physi s;
but additionally there are dis ontinuities, orpus les whose frequen y is
governed by the square of these amplitudes.
The most denite support for this on eption is given by ollision phenomena for material parti les (Born [30). Already Einstein [16, when he
dedu ed from de Broglie's daring theory the possibility of `dira tion'
of material parti les,a ta itly assumed that it is the parti le number
that is determined by the intensity of the waves. The same o urs in
the interpretation given by Elsasser [17 of the experiments by Davisson
and Kunsman [18,19 on the ree tion of ele trons by rystals; also here
one assumes dire tly that the number of ele trons is a maximum in the
dira tion maxima. The same holds for Dymond's [20 experiments on
the dira tion of ele trons by helium atoms.
The appli ation of wave me hani s to the al ulation of ollision pro esses takes a form quite analogous to the theory of dira tion of light
by small parti les. One has to nd the solution to S hrdinger's wave
equation (17) that goes over at innity to a given in ident plane wave;
this solution behaves everywhere at innity like an outgoing37 spheri al
wave. The intensity of this spheri al wave in any dire tion ompared to
the intensity of the in oming wave determines the relative number of
parti les dee ted in this dire tion from a parallel ray. As a measure of
the intensity one has to take a ` urrent ve tor'38 whi h an be onstru ted from the solution (q, W ), and whi h is formed quite analogously
to the Poynting ve tor of the ele tromagneti theory of light, and whi h
measures the number of parti les rossing a unit surfa e in unit time.
In this way Wentzel [31 and Oppenheimer [32 have derived wave
me hani ally the famous Rutherford law for the s attering of -parti les
by heavy nu lei.b
If one wishes to al ulate the probabilities of ex itation and ionisation
of atoms [30, then one must introdu e the oordinates of the atomi
ele trons as variables on an equal footing with those of the olliding
ele tron. The waves then propagate no longer in three-dimensional spa e
but in multi-dimensional onguration spa e. From this one sees that the
quantum me hani al waves are indeed something quite dierent from the
light waves of the lassi al theory.
a Note that the rst predi tion of su h dira tion appears in fa t to have been made
by de Broglie in 1923; f. se tion 2.2.1 (eds.).
b Cf. Born (1969, Appendix XX). The urrent ve tor, as dened there, is the usual
h 1
j = 2i
( ) (eds.).
2m
426
If one onstru ts the urrent ve tor just dened for a solution of the
generalised S hrdinger equation (22), whi h des ribes time evolution,
one sees that the time derivative of the integral
Z
||2 dq ,
ranging over an arbitrary domain of the independent numeri al variables

q , an be transformed into the surfa e integral of the urrent ve tor over
the boundary of that domain. From this it emerges that
||2
has to be interpreted as parti le density or, better, as probability density.

The solution itself is alled `probability amplitude'.
The amplitude (q , W ) belonging to a stationary state thus yields via
|(q , W )|2 the probability that for given energy W the oordinate q is
in some given element dq .39 But this an be generalised immediately. In
fa t, (q , W ) is the proje tion of the prin ipal axis W of the operator
H onto the prin ipal axis q of the operator q . One an therefore say
in general (Jordan [39): if two physi al quantities are given by the
operators q and Q and if one knows the prin ipal axes of the former, for
instan e, a ording to magnitude and dire tion,40 then from the equation
Q(q , Q ) = Q (q , Q )
one an determine the prin ipal axes Q of Q41 and their proje tions
(q , Q ) on the axes of q . Then |(q , Q )|2 dq is the probability that for
given Q the value of q falls in a given interval dq .
If onversely one imagines the prin ipal axes of Q as given, then
those of q are obtained42 through the inverse rotation; from this one
easily re ognises that (Q , q ) is the orresponding amplitude,43 so that
|(Q , q )|2 dQ means the probability, given q , to nd the value of Q
h
in dQ . If for instan e one takes for Q the operator p = 2i
q , then one
has the equation
h
= p ,
2i q
thus
= Ce
2i
h q p
(28)
This is therefore the probability amplitude for a pair of onjugate quantities. For the probability density one obtains ||2 = C , that is, for given
q every value p is equally probable.
Quantum me hani s
427
This is an important result, sin e it allows one to retain the on ept

of ` onjugate quantity' even in the ase where the dierential denition
fails, namely when the quantity q has only a dis rete spe trum or even
when it is only apable of taking nitely many values. The latter for
instan e is the ase for angles with quantised dire tion [ri htungsgequantelte Winkel,a say for the magneti ele tron, or in the Stern-Gerla h
experiments. One an then, as Jordan does, all by denition a quantity
p onjugate to q , if the orresponding probability amplitude has the
expression (28).
As the amplitudes are the elements of the rotation matrix of one
orthogonal system into another, they are omposed a ording to the
matrix rule:
Z
(q , Q ) =
(q , )( , Q )d ;
(29)
in the ase of dis rete spe tra, instead of the integral one has nite
or innite sums. This is the general formulation of the theorem of the
interferen e of probabilities. As an appli ation, let us look again at
formula (24). Here cn (t) was the amplitude for the probability that
the system at time t has energy Wn ; 0n (q ) is the amplitude for the
probability that for given energy Wn the oordinate q has a given value.
Thus
X
(q , t) =
cn (t)0n (q )
n
expresses the amplitude for the probability that q at time t has a given
value.
Alongside the on ept of the relative state probability |(q , Q )|2 , there also o urs the on ept of the transition probability,44 namely, every
time one onsiders a system as depending on an external parameter,
be it time or any property of a weakly oupled external system. Then
the system of prin ipal axes of any quantity be omes dependent on this
parameter; it experien es a rotation, represented by an orthogonal transformation S(q , q ), in whi h the parameter enters (as in formula (25)).
The quantities |S(q , q )|2 are the `transition probabilities';45 in general,
however, they are not independent, instead the `transition amplitudes'
are omposed a ording to the interferen e rule.
a This was a standard term referring to the fa t that in the presen e of an external
magneti eld, the proje tion of the angular momentum in the dire tion of the
eld has to be quantised (quantum number m). Therefore, the dire tion of the
angular momentum with respe t to the magneti eld an be said to be quantised.
Cf. Born (1969, p. 121) (eds.).
428
III. Formulation of the prin iples and delimitation of

their s ope
After the general on epts of the theory have been developed through
analysis of empiri al ndings, the dual task arises, rst of giving a
system of prin iples as simple as possible and onne ted dire tly to
the observations, from whi h the entire theory an be dedu ed as from
a mathemati al system of axioms, and se ond of riti ally s rutinising
experien e to assure oneself that no observation on eivable by today's
means stands in ontradi tion to the prin iples.
Jordan [39 has formulated su h a `system of axioms', whi h takes the
following statements as fundamental:46
1) One requires for ea h pair of quantum me hani al quantities q, Q the
existen e of a probability amplitude (q , Q ), su h that ||2 gives the
probability47 that for given Q the value of q falls in a given innitesimal
interval.
2) Upon permutation of q and Q, the orresponding amplitude should
be (Q , q ).
3) The theorem (29) of the omposition of probability amplitudes.
4) To ea h quantity q there should belong a anoni ally onjugate one p,
dened by the amplitude (28). This is the only pla e where the quantum
onstant h appears.48
Finally one also takes as obvious that, if the quantities q and Q are
identi al, the amplitude (q , q ) be omes equal to the ùnit matrix'
(q q ), that is, always to zero, ex ept when q = q . This assumption
and the multipli ation theorem 3) together hara terise the amplitudes
thus dened as the oe ients of an orthogonal transformation; one
obtains the orthogonality onditions simply by stating that the omposition of the amplitude belonging to q, Q with that belonging to Q, q
must yield the identity.
One an then redu e all given quantities, in luding the operators, to
amplitudes by writing them as integral operators as in formula (13).
The non ommutative operator multipli ation is then a onsequen e of
Quantum me hani s
429
the axioms and loses all the strangeness atta hed to it in the original
matrix theory.
Dira 's method [38 is ompletely equivalent to Jordan's formulation,
ex ept in that he does not arrange the prin iples in axiomati form.a
This theory now indeed summarises all of quantum me hani s in a
system in whi h the simple on ept of the al ulable probability [bere henbare Wahrs heinli hkeit49 for a given event plays the main role.50
It also has some short omings, however. One formal short oming is the
o urren e of improper fun tions, like the Dira , whi h one needs for
the representation of the unit matrix for ontinuous ranges of variables.
More serious is the ir umstan e that the amplitudes are not dire tly
measurable quantities, rather, only the squares of their moduli; the phase
fa tors are indeed essential for how dierent phenomena are onne ted
[fr den Zusammenhang der vers hiedenen Ers heinungen wesentli h,
but are only indire tly determinable, exa tly as phases in opti s are
dedu ed indire tly by ombining measurements of intensity. It is, however, a tried and proven prin iple, parti ularly in quantum me hani s,
that one should introdu e as far as possible only dire tly observable
quantities as fundamental on epts of a theory. This defe t51 is related
mathemati ally to the fa t that the denition of probability in terms
of the amplitudes does not express the invarian e under orthogonal
transformations of the Hilbert spa e ( anoni al transformations).
These gaps in the theory have been lled by von Neumann [41,42.
There is52 an invariant denition of the eigenvalue spe trum for arbitrary
operators, and of the relative probabilities, without presupposing the
existen e of eigenfun tions or indeed using improper fun tions. Even
though this theory has not yet been elaborated in all dire tions, one
an however say with ertainty that a mathemati ally irreproa hable
grounding of quantum me hani s is possible.
Now the se ond question has to be answered: is this theory in a ord
with the totality of our experien e? In parti ular, given that the individual pro ess is only statisti ally determined, how an the usual deterministi order be preserved in the omposite ma ros opi phenomena?53
The most important step in testing the new on eptual system in
this dire tion onsists in the determination of the boundaries within
whi h the appli ation of the old ( lassi al) words and on epts is allowed,
su h as `position, velo ity, momentum, energy of a parti le (ele tron)'
(Heisenberg [46). It now turns out that all these quantities an be
a There are nevertheless some dieren es between the approa hes of Dira and
Jordan. Cf. Darrigol (1992, pp. 3434) (eds.).
430
individually exa tly measured and dened, as in the lassi al theory, but
that for simultaneous measurements of anoni ally onjugate quantities
(more generally: quantities whose operators do not ommute) one annot
get below a hara teristi limit of indetermina y [Unbestimmtheit.a To
determine this, a ording to Bohr [4754 one an start quite generally
from the empiri ally given dualism between waves and orpus les. One
has essentially the same phenomenon already in every dira tion of light
by a slit. If a wave impinges perpendi ularly on an (innitely long) slit
of width q1 , then the light distribution as a fun tion of the deviation
angle is given a ording to Kir hho by the square of the modulus of
the quantity
q

1
Z + q21
sin
sin
2i
,
a
e sin q dq = 2a
q1
q1
2
sin
and thus ranges over a domain whose order of magnitude is given by55
sin 1 = q1 and gets ever larger with de reasing slit width q1 . If one
onsiders this pro ess from the point of view of the orpus ular theory,
and if the asso iation given by de Broglie of frequen y and wavelength
with energy and momentum of the light quantum is valid,
h = W,
h
=P ,
then the momentum omponent perpendi ular to the dire tion of the
slit is
h
p = P sin = sin .
One sees thus that after the passage through the slit the light quanta
have a distribution whose amplitude is given by
e
2i
sin q
=e
2i
h pq
pre isely as quantum me hani s requires for two anoni ally onjugate
variables; further, the width of the domain of the variable p that ontains
the greatest number of light quanta is
p1 = P sin 1 =
P
h
=
.
q1
q1
a Here and in the following, the hoi e of translation ree ts the hara teristi
terminology of the original. Born and Heisenberg use the terms Ùnbestimmtheit'
(indetermina y) and Ùngenauigkeit' (impre ision), while the standard German
terms today are Ùnbestimmtheit' and Ùns hrfe' (unsharpness) (eds.).
Quantum me hani s
431
By general onsiderations of this kind one arrives at the insight that

the impre isions (average errors) of two anoni ally onjugate variables
p and q always stand in the relation
p1 q1 h .
(30)
The narrowing of the range of one variable, whi h forms the essen e of a
measurement, widens unfailingly the range of the other. The same follows
immediately from the mathemati al formalism of quantum me hani s on
the basis of formula (28). The a tual meaning of Plan k's onstant h is
thus that it is the universal measure of the indetermina y that enters
the laws of nature through the dualism of waves and orpus les.
That quantum me hani s is a mixture of stri tly me hani al and statisti al prin iples an be onsidered a onsequen e of this indetermina y.
Indeed, in the lassi al theory one may x the state of a me hani al
system by, for instan e, measuring the initial values of p and q at a
ertain instant. In quantum me hani s su h a measurement of the initial
state is possible only with the a ura y (30). Thus the values of p and q
are known also at later times only statisti ally.
The relation between the old and the new theory an therefore be
des ribed thus:
In lassi al me hani s one assumes the possibility of determining exa tly the initial state; the further development is then determined by
the laws themselves.
In quantum me hani s, be ause of the impre ision relation, the result
of ea h measurement an be expressed by the hoi e of appropriate initial
values for probability fun tions; the quantum me hani al laws determine
the hange (wave-like propagation) of these probability fun tions. The
result of future experiments however remains in general indeterminate
and only the expe tation56 of the result is statisti ally onstrained. Ea h
new experiment repla es the probability fun tions valid until now with
new ones, whi h orrespond to the result of the observation; it separates
the physi al quantities into known and unknown (more pre isely and less
pre isely known) quantities in a way hara teristi of the experiment.
That in this view ertain laws, like the prin iples of onservation of
energy and momentum, are stri tly valid, follows from the fa t that they
are relations between ommuting quantities (all quantities of the kind q
or all quantities of the kind p).a
a A similar but more expli it phrasing is used by Born (1926e, le ture 15): assuming
that H(p, q) = H(p) + H(q), the time derivatives not only of H but also of all
omponents of momentum and of angular momentum have the form f (q) + g(p)
432
The transition from mi ro- to ma rome hani s results naturally from

the impre ision relation be ause of the smallness of Plan k's onstant h.
The fa t of the propagation,57 the `melting away' of a `wave or probability pa ket' is ru ial to this. For some simple me hani al systems (free
ele tron in a magneti or ele tri eld (Kennard [50), harmoni os illator
(S hrdinger [25)), the quantum me hani al propagation of the wave
pa ket agrees with the propagation of the system traje tories that would
o ur in the lassi al theory if the initial onditions were known only with
the pre ision restri tion (30). Here the purely lassi al treatment of
and parti les, for instan e in the dis ussion of Wilson's photographs,
immediately nds its justi ation. But in general the statisti al laws of
the propagation of a `pa ket' for the lassi al and the quantum theory
are dierent; one has parti ularly extreme examples of this in the ases of
`dira tion' or ìnterferen e' of material rays, as in the already mentioned
experiments of Davisson, Kunsman and Germer [18,19 on the ree tion
of ele trons by metalli rystals.
That the totality of experien e an be tted into the system of this
theory an of ourse be established only by al ulation and dis ussion of
all the experimentally a essible ases. Individual experimental setups,58
in whi h the suspi ion of a ontradi tion with the pre ision limit (30)
might arise, have been dis ussed [46,47; every time the reason for the
impossibility of xing exa tly all determining data ould be exhibited
intuitively [ans hauli h aufgewiesen.
There remains only to survey the most important onsequen es of the
theory and their experimental veri ation.
IV. Appli ations of quantum me hani s

In this se tion we shall briey dis uss those appli ations of quantum
me hani s that stand in lose relation to questions of prin iple. Here the
Uhlenbe k-Goudsmit theory of the magneti ele tron shall be mentioned rst. Its formulation and the treatment of the anomalous Zeeman
ee ts with the matrix al ulus raise no di ulties [11; the treatment
with the method of eigenos illations su eeds only with the help of the
general Dira -Jordan theory (Pauli [45). Here, two three-dimensional
wave fun tions are asso iated with ea h ele tron. It be omes natural
thereby to look for an analogy between matter waves and polarised
with suitable fun tions f and g . Born states that sin e all q ommute with one
another and all p ommute with one another, the expressions f (q)+g(p) will vanish
under the same ir umstan es as in lassi al me hani s (eds.).
Quantum me hani s
433
light waves, whi h in fa t an be arried through to a ertain extent

(Darwin [49, Jordan [53). What is ommon to both phenomena is that
the number of terms is nite, so the representative matrix is also nite
(two arrangements [Einstellungen for the ele tron, two dire tions of
polarisation for light). Here the denition of the onjugate quantity by
means of dierentiation thus fails; one must resort to Jordan's denition
by means of the probability amplitudes (formula (28)).
From among the other appli ations, the quantum me hani s of manybody problems shall be mentioned [28,29,40. In a system that ontains
a number of similar parti les [glei her Partikel,59 there o urs between
them a kind of `resonan e' and from that results a de omposition of
the system of terms into subsystems that do not ombine (Heisenberg,
Dira [28,37). Wigner has systemati ally investigated this phenomenon
by resorting to group theoreti methods, and has set up the totality of
the non- ombining systems of terms [40; Hund has managed to derive
the majority of these results by omparatively elementary means [48. A
spe ial role is played by the `symmetri ' and àntisymmetri ' subsystems
of terms; in the former every eigenfun tion remains un hanged under
permutation of arbitrary similar parti les, in the latter it hanges sign
under permutation of any two parti les. In applying this theory to the
spe tra of atoms with several ele trons it turns out that the Pauli equivalen e rulea allows only the antisymmetri subsystem.60 On the basis
of this insight one an establish quantum me hani ally the systemati s
of the line spe tra and of the ele tron grouping throughout the whole
periodi system of elements.
If one has a large number of similar parti les, whi h are to be given a statisti al treatment (gas theory), one obtains dierent statisti s
depending on whether one hooses the orresponding wave fun tion
a ording to the one or the other subsystem. The symmetri system is
hara terised by the fa t that no new state arises under permutation of
the parti les from61 a state des ribed by a symmetri eigenfun tion; thus
all permutations that belong to the same set of quantum numbers (lie in
the same ` ell') together always have the weight 1. This orresponds to
the Bose-Einstein statisti s [56,16. In the antisymmetri term system
two quantum numbers may never be ome equal, be ause otherwise the
eigenfun tion vanishes; a set of quantum numbers orresponds therefore
either to no proper fun tion at all or at most to one, thus the weight of
a state is 0 or 1. This is the Fermi-Dira statisti s [57,37.
a That is, the Pauli prin iple applied to ele trons that are èquivalent' in the sense
of having the same quantum numbers n and l; f. Born (1969, p. 178). (eds.).
434
Bose-Einstein statisti s holds for light quanta, as emerges from the

validity of Plan k's radiation formula. Fermi-Dira statisti s ertainly
holds for (negative) ele trons, as emerges from the above-mentioned
systemati s of the spe tra on the basis of Pauli's equivalen e rule, and
with great likelihood also for the positive elementary parti les (protons);
one an infer this from observations of band spe tra [28,43 and in
parti ular from the spe i heat of hydrogen at low temperatures [55.62
The assumption of Fermi-Dira statisti s for the positive and negative
elementary parti les of matter has the onsequen e that Bose-Einstein
statisti s holds for all neutral stru tures, e.g. mole ules (symmetry of
the eigenfun tions under permutation of an even number63 of parti les
of matter). Within quantum me hani s, in whi h a many-body problem
is treated in onguration spa e, the new statisti s of Bose-Einstein and
Fermi-Dira has a perfe tly legitimate pla e, unlike in the lassi al theory, where an arbitrary modi ation of the usual statisti s is impossible;
nevertheless the restri tions made on the form of the eigenfun tions
appear as an arbitrary additional assumption. In parti ular, the example
of light quanta indi ates that the new statisti s is related in an essential
way to the wave-like properties of matter and light. If one de omposes
the ele tromagneti os illations of a avity into spatial harmoni omponents, ea h of these behaves like a harmoni os illator as regards time
evolution; it now turns out that under quantisation of this system of
os illators a solution results that behaves exa tly like a system of light
quanta obeying Bose-Einstein statisti s [4. Dira has used this fa t for
a onsistent treatment of ele trodynami al problems [51,52, to whi h
we shall return briey.
The orpus ular stru ture of light thus appears here as quantisation
of light waves, su h as vi e versa the wave nature of matter manifests
itself in the `quantisation' of the orpus ular motion. Jordan has shown
[54 that one an pro eed analogously with ele trons; one has then to
de ompose the S hrdinger fun tion of a avity into fundamental and
harmoni s and to quantise ea h of these as a harmoni os illator, in
su h a way in fa t that Fermi-Dira statisti s is obtained. The new
quantum numbers, whi h express the `weights' in the usual many-body
theory, have thus only the values 0 and 1. Therefore one has again here a
ase of nite matri es, whi h an be treated only with Jordan's general
theory. The existen e of ele trons thus plays the same role in the formal
elaboration of the theory as that of light quanta; both are dis ontinuities
no dierent in kind from the stationary states of a quantised system.
Quantum me hani s
435
However, if the material parti les stand in intera tion with ea h other,
the development of this idea might run into di ulties of a deep nature.
The results of Dira 's investigations [51,52 of quantum ele trodynami s onsist above all in a rigorous derivation of Einstein's transition
probabilities for spontaneous emission.a Here the ele tromagneti eld
(resolved into quantised harmoni os illations) and the atom are onsidered as a oupled system and quantum me hani s is applied in the form
of the integral equation (13). The intera tion energy appearing therein
is obtained by arrying over lassi al formulas. In this onne tion, the
nature of absorption and s attering of light by atoms is laried. Finally,
Dira [52 has managed to derive a dispersion formula with damping
term; this in ludes also the quantum me hani al interpretation of Wien's
experiments on the de ay in lumines en e of anal rays.b His method
onsists in onsidering the pro ess of the s attering of light by atoms
as a ollision of light quanta. However, sin e one an indeed attribute
energy and momentum to the light quantum but not easily a spatial
position, there is a failure of the wave me hani al ollision theory (Born
[30), in whi h one presupposes knowledge of the intera tion between
the ollision partners as a fun tion of the relative position. It is thus
ne essary to use the momenta as independent variables, and an operator
equation of matrix hara ter instead of S hrdinger's wave equation.
Here one has a ase where the use of the general points of view whi h
we have emphasised in this report annot be avoided. At the same time,
the theory of Dira reveals anew the deep analogy between ele trons and
light quanta.
Con lusion
By way of summary, we wish to emphasise that while we onsider the
last-mentioned enquiries, whi h relate to a quantum me hani al treatment of the ele tromagneti eld, as not yet ompleted [unabges hlossen, we onsider quantum me hani s to be a losed theory [ges hlossene
Theorie, whose fundamental physi al and mathemati al assumptions
are no longer sus eptible of any modi ation. Assumptions about the
physi al meaning of quantum me hani al quantities that ontradi t Jordan's or equivalent postulates will in our opinion also ontradi t experien e. (Su h ontradi tions an arise for example if the square of
a As opposed to the indu ed emission dis ussed on p. 424 (eds.).
b See above, p. 143 (eds.).
436
the modulus of the eigenfun tion is interpreted as harge density.a )

On the question of the `validity of the law of ausality' we have this
opinion: as long as one takes into a ount only experiments that lie in
the domain of our urrently a quired physi al and quantum me hani al
experien e, the assumption of indeterminism in prin iple, here taken as
fundamental, agrees with experien e. The further development of the
theory of radiation will hange nothing in this state of aairs, be ause
the dualism between orpus les and waves, whi h in quantum me hani s
appears as part of a ontradi tion-free, losed theory [abges hlossene
Theorie, holds in quite a similar way for radiation. The relation between
light quanta and ele tromagneti waves must be just as statisti al as that
between de Broglie waves and ele trons. The di ulties still standing at
present in the way of a omplete theory of radiation thus do not lie
in the dualism between light quanta and waves whi h is entirely
intelligible instead they appear only when one attempts to arrive
at a relativisti ally invariant, losed formulation of the ele tromagneti
laws; all questions for whi h su h a formulation is unne essary an be
treated by Dira 's method [51,52. However, the rst steps also towards
over oming these relativisti di ulties have already been made.
a See S hrdinger's report, espe ially his se tion I, and se tion 4.4 above (eds.).
Quantum me hani s
437
Bibliographya
[1 W. Heisenberg, ber quanten[theoretis he Umdeutung kinematis her und me hanis her Beziehungen, Z. f. Phys., 33 (1925), 879.
[2 M. Born and P. Jordan, Zur Quantenme hanik, I, Z. f. Phys., 34
(1925), 858.
[3 P. Dira , The fundamental equations of quantum me hani s, Pro .
Roy. So . A, 109 (1925), 642.
[4 M. Born, W. Heisenberg and P. Jordan, Zur Quantenme hanik, II,
Z. f. Phys., 35 (1926), 557.
[5 N. Bohr, Atomtheorie und Me hanik, Naturwiss., 14 (1926), 1.
[6 W. Pauli, ber das Wasserstospektrum vom Standpunkt der neuen
Quantenme hanik, Z. f. Phys., 36 (1926), 336.
[7 P. Dira , Quantum me hani s and a preliminary investigation of the
hydrogen atom, Pro . Roy. So . A, 110 (1926), 561.
[8 G. E. Uhlenbe k and S. Goudsmit, Ersetzung der Hypothese vom
unme hanis hen Zwang dur h eine Forderung bezgli h des inneren Verhaltens jedes einzelnen Elektrons, Naturw., 13 (1925), 953.64
[9 L. H. Thomas, The motion of the spinning ele tron, Nature, 117
(1926), 514.
[10 J. Frenkel, Die Elektrodynamik des rotierenden Elektrons, Z. f.
Phys., 37 (1926), 243.
[11 W. Heisenberg and P. Jordan, Anwendung der Quantenme hanik
auf das Problem der anomalen Zeemaneekte, Z. f. Phys., 37 (1926),
263.
[12 P. Dira , Relativity quantum me hani s with an appli ation to
Compton s attering, Pro . Roy. So . [A, 111 (1926), 405.
[13 P. Dira , The elimination of the nodes in quantum me hani s, Pro .
Roy. So . [A, 111 (1926), 281.
[14 P. Dira , On quantum algebra, Pro . Cambridge Phil. So ., 23
(1926), 412.
[15 P. Dira , The Compton ee t in wave me hani s, Pro . Cambridge
Phil. So ., 23 (1926), 500.
a The style of the bibliography has been both modernised and uniformised.
Amendments and fuller details (when missing) are given in square bra kets,
mostly without ommentary. Amendments in the Fren h edition of mistakes in
the types ript (wrong initials, spelling of names et .) are taken over also mostly
without ommentary. Mistakes o urring only in the Fren h edition are endnoted
(eds.).
438
[16 A. Einstein, Quantentheorie des einatomigen idealen Gases, II, Berl.

Ber. (1925), [3.65
[17 W. Elsasser, Bemerkungen zur Quanten[me hanik freier Elektronen, Naturw., 13 (1925), 711.
[18 C. Davisson and [C. Kunsman, [The s attering of low speed
ele trons by [Platinum and [Magnesium, Phys. Rev., 22 (1923), [242.66
[19 C. Davisson and L. Germer, The s attering of ele trons by a single
rystal of Ni kel, Nature, 119 (1927), 558.
[20 E. G. Dymond, On ele tron s attering in Helium, Phys. Rev., 29
(1927), 433.
[21 M. Born and N. Wiener, Eine neue Formulierung der Quantengesetze fr periodis he und [ni htperiodis he Vorgnge, Z. f. Phys., 36
(1926), 174.
[22 C. E kart, Operator al ulus and the solution of the equation[s of
quantum dynami s, Phys. Rev., 28 (1926), 711.
[23 K. Lan zos, ber eine feldmssige Darstellung der neuen Quantenme hanik, Z. f. Phys., 35 (1926), 812.
[24 E. S hrdinger, Quantisierung als Eigenwertproblem, I to IV, Ann.
d. Phys., 79 (1926), 361; ibid., 489; ibid., 80 (1926), 437; ibid., 81 (1926),
109.
[25 E. S hrdinger, Der stetige bergang von der Mikro- zur Makrome hanik, Naturw., 14 (1926), 664.
[26 E. S hrdinger, ber das Verhltnis der Heisenberg-Born-Jordans hen Quantenme hanik zu der meinen, Ann. d. Phys., 79 (1926), 734.
[27 P. Jordan, Bemerkung ber einen Zusammenhang zwis hen Duanes
Quantentheorie der Interferenz und den de Broglies hen Wellen, Z. f.
Phys., 37 (1926), 376.
[28 W. Heisenberg, Mehrkrperproblem und Resonanz in der Quantenme hanik, I and II, Z. f. Phys., 38 (1926), 411; Ibid. 41 (1927), 239.
[29 W. Heisenberg, ber die Spektren von Atomsystemen mit zwei
Elektronen, Z. f. Phys., 39 (1926), 499.67
[30 M. Born, Zur Quantenme hanik der Stossvorgnge, Z. f. Phys., 37
(1926), 863; [Quantenme hanik der Stossvorgnge, ibid., 38 (1926), 803.
[31 G. Wentzel, Zwei Bemerkungen ber die Zerstreuung korpuskularer
Strahlen als Beugungsers heinung, Z. f. Phys., 40 (1926), 590.
[32 J. R. Oppenheimer, Bemerkung zur Zerstreuung der -Teil hen, Z.
f. Phys., 43 (1927), 413.
[33 W. Elsasser, Diss. Gttingen, [Zur Theorie der Stossprozesse bei
Wassersto, Z. f. Phys., [45 (1927), 522.68
Quantum me hani s
439
[34 M. Born, Das Adiabatenprinzip in der Quantenme hanik, Z. f.

Phys., 40 ([1926), 167.69
[35 W. Heisenberg, S hwankungsers heinungen und Quantenme hanik,
Z. f. Phys., 40 (1926), 501.
[36 P. Jordan, ber quantenme hanis he Darstellung von Quantensprngen, Z. f. Phys., 40 ([1927), 661.
[37 P. Dira , On the theory of quantum me hani s, Pro . Roy. So . A,
112 (1926), 661.
[38 P. Dira , The physi al interpretation of quantum dynami s, Pro .
Roy. So . A, 113 ([1927), 621.
[39 P. Jordan, ber eine neue Begrndung der Quantenme hanik, Z.
f. Phys., 40 ([1927), [809; Se ond Part, ibid., 44 (1927), 1.
[40 E. Wigner, ber ni htkombinierende Terme in der neueren Quantentheorie, First Part, Z. f. Phys., 40 (1926), 492; Se ond Part, ibid., 40
(1927), 883.
[41 D. Hilbert, [J.. v. Neumann and L. Nordheim, [ber die Grundlagen der Quantenme hanik, Mathem. Ann., 98 [1928, 1.
[42 [J. v. Neumann, [Mathematis he Begrndung der Quantenme hanik, Gtt. Na hr., 20 May 1927, [1.
[43 F. Hund, Zur Deutung der Molekelspektr[en, Z. f. Phys., 40 ([1927),
742; ibid., 42 (1927), 93; ibid., 43 (1927), 805.
[44 W. Pauli, ber Gasentartung und Paramagnetismus, Z. f. Phys.,
41 (1927), 81.
[45 W. Pauli, Zur Quantenme hanik des magnetis hen Elektrons, Z. f.
Phys., 43 (1927), 601.
[46 W. Heisenberg, ber den ans hauli hen Inhalt der quantentheoretis hen Kinematik und Me hanik, Z. f. Phys., 43 (1927), 172.70
[47 N. Bohr, ber den begrii hen Aufbau der Quantentheorie, forth oming [im Ers heinen.
[48 F. Hund, Symmetrie haraktere von Termen bei Systemen mit glei hen Partikeln in der Quantenme hanik, Z. f. Phys., 43 (1927), 788.
[49 [C. G. Darwin, The ele tron as a ve tor wave, Nature, 119 (1927),
282.
[50 E. Kennard, Zur Quantenme hanik einfa her Bewegungstypen, Z.
f. Phys., 44 (1927), 326.
[51 P. Dira , The quantum theory of emission and absorption of radiation, Pro . Roy. So . A, 114 (1927), 243.
[52 P. Dira , The quantum theory of dispersion, Pro . Roy. So . A,
114 (1927), 710.
440
[53 P. Jordan, ber die Polarisation der Li htquanten, Z. f. Phys., 44

(1927), 292.
[54 P. Jordan, Zur Quantenme hanik der Gasentartung, Z. f. Phys.,
forth oming [im Ers heinen [44 (1927), 473.
[55 D. Dennison, A note on the spe i heat of [the Hydrogen [mole ule, Pro . Roy. So . A, [115 (1927), 483.
On statisti s also:
[56 N. S. Bose, Plan ks Gesetz und Li htquantenhypothese, Z. f. Phys.,
26 (1924), 178.
[57 E. Fermi, [Sulla quantizzazione del gas perfetto monatomi o, Lin ei
Rend., 3 (1926), 145.
General surveys:
[58 M. Born (Theorie des Atombaus?) [Probleme der Atomdynamik ,
Le tures given at the Massa husetts Institute of Te hnology (Springer,
[1926).71
[59 W. Heisenberg, [ber quantentheoretis he Kinematik und Me hanik, Mathem. Ann., 95 (1926), 683.
[60 W. Heisenberg, Quantenme hanik, Naturw., 14 (1926), [989.
[61 M. Born, Quantenme hanik und Statistik, Naturw., 15 (1927),
238.72
[62 P. Jordan, Kausalitt und Statistik in der modernen Physik, Naturw., 15 (1927), 105.
[63 P. Jordan, Die Entwi klung der neuen Quantenme hanik, Naturw.,
15 (1927), 614, 636.73
Dis ussion
441
Dis ussion of Messrs Born and Heisenberg's reporta
Mr Dira . I should like to point out the exa t nature of the orre-
sponden e between the matrix me hani s and the lassi al me hani s. In

lassi al me hani s one an work out a problem by two methods: (1) by
taking all the variables to be numbers and working out the motion, e.g.
by Newton's laws, whi h means one is al ulating the motion resulting
from one parti ular set of numeri al values for the initial oordinates
and momenta, and (2) by onsidering the variables to be fun tions of
the J 's (a tion variables)74 and using the general transformation theory
of dynami s and thus determining simultaneously the motion resulting
from all possible initial onditions.75 The matrix theory orresponds to
this se ond lassi al method. It gives information about all the states
of the system simultaneously. A dieren e between the matrix method
and the se ond lassi al method arises sin e in the latter one requires to
treat simultaneously only states having nearly the same J 's (one uses,
for instan e, the operators J

), while in the matrix theory one must
treat simultaneously states whose J 's dier by nite amounts.
To get results omparable with experiment when one uses the se ond
lassi al method,76 one must substitute numeri al values for the J 's in
the fun tions of the J 's obtained from the general treatment. One has to
do the same in the matrix theory. This gives rise to a di ulty sin e the
results of the general treatment are now matrix elements, ea h referring
in general to two dierent sets of J 's. It is only the diagonal elements,
for whi h these two sets of J 's oin ide, that have a dire t physi al
interpretation.
Mr Lorentz.
I was very surprised to see that the matri es satisfy

equations of motion. In theory that is very beautiful, but to me it
is a great mystery, whi h, I hope, will be laried. I am told that
by all these onsiderations one has ome to onstru t matri es that
represent what one an observe in the atom, for instan e the frequen ies
of the emitted radiation. Nevertheless, the fa t that the oordinates, the
potential energy, and so on, are now represented by matri es indi ates
that the quantities have lost their original meaning and that one has
made a huge step in the dire tion of abstra tion.
Allow me to draw attention to another point that has stru k me. Let
a The two dis ussion ontributions by Dira follow his manus ript in AHQP,
mi rolm 36, se tion 10. Deviations in the Fren h edition (whi h may or may
not be due to Dira ) are reported in endnote, as well as interesting variants or
an ellations in the manus ript, and pun tuation has been slightly altered (eds.).
442
us onsider the elements of the matri es representing the oordinates of

a parti le in an atom, a hydrogen atom for instan e, and satisfying the
equations of motion. One an then hange the phase of ea h element of
the matri es without these easing to satisfy the equations of motion;
one an, for instan e, hange the time. But one an go even further and
hange the phases, not arbitrarily, but by multiplying ea h element by
a fa tor of the form ei(m n ) , and this is quite dierent from a hange
of time origin.a
Now these matrix elements ought to represent emitted radiation. If
the emitted radiation were what is at the basis of all this, one ould
expe t to be able to hange all phases in an arbitrary way. The abovementioned fa t then leads us very naturally to the idea that it is not the
radiation that is the fundamental thing: it leads us to think that behind
the emitted os illations are hidden some true os illations, of whi h the
emitted os illations are dieren e os illations.
In this way then, in the end there would be os illations of whi h the
emitted os illations are dieren es, as in S hrdinger's theory,a and it
seems to me that this is ontained in the matri es. This ir umstan e
indi ates the existen e of a simpler wave substrate.
Mr Born.
Mr Lorentz is surprised that the matri es satisfy the

equations of motion; with regard to this I would like to note the analogy
with omplex numbers. Also here we have a ase where in an extension
of the number system the formal laws are preserved almost ompletely.
Matri es are some kind of hyper omplex numbers, whi h are distinguished from the ordinary numbers by the fa t that the law of ommutativity
no longer holds.
Mr Dira .
The arbitrary phases o urring in the matrix method

orrespond exa tly77 to the arbitrary phases in the se ond lassi al
method, where the variables are fun tions of the J 's and w's (a tion
and angle variables). There are arbitrary78 phases in the w's, whi h may
have dierent values for ea h dierent set of values for the J 's. This is
ompletely analogous79 to the matrix theory, in whi h ea h arbitrary
phase is asso iated with a row and olumn, and therefore with a set of
a This orresponds of ourse to the hoi e of a phase fa tor ein for ea h stationary
state. This point (among others) had been raised in the orresponden e between
Lorentz and Ehrenfest in the months pre eding the onferen e. See Lorentz to
Ehrenfest, 4 July 1927, AHQP-EHR-23 (in Dut h) (eds.).
a See se tion 4.4 (eds.).
Dis ussion
443
values for the J 's.
Mr Born. The phases
whi h Mr Lorentz has just mentioned

are asso iated with the dierent energy levels, quite like in lassi al
me hani s. I do not think there is anything mysterious hiding behind
this.
n
Mr Bohr.
The issue of the meaning of the arbitrary phases, raised

by Mr Lorentz, is of very great importan e, I think, in the dis ussion
of the onsisten y of the methods of quantum theory. Although the
on ept of phase is indispensable in the al ulations, it hardly enters
the interpretation of the observations.
444
Notes to pp. 407424

1 Here and in a number of pla es in the following, the Fren h edition omits
quotation marks present in the German types ript. They are ta itly
restored in this edition.
2 The Fren h edition gives `[47'.
3 [diskrete [dtermines
4 The Fren h edition omits `[60'.
5 [makroskopis he [mi ros opiques
6 [dur h sinngemsse bertragung [par une extension logique
7 In the Fren h edition, the parentheti al remark is given as a footnote.
8 The types ript does not give the referen e number, only the bra kets.
The Fren h edition omits the referen e entirely. The mentioned results
are to be found in se tion 4.1 of Born, Heisenberg and Jordan (1926 [4).
9 Word omitted in the Fren h edition.
10 [sind sinngemsse bertragungen [sont des extensions logiques
11 Misprint in the Fren h edition: summation index `n' in the equation.
12 [orthogonale Transformationen der
[transformations orthogonales
k
13 [bes hrnkte unendli he [partiellement innies
14 [Stze [prin ipes
15 [no h zur kzukommen haben [n'avons pas revenir
16 Word omitted in the Fren h edition.
17 The overbar is missing in the original types ript (only here), but is
in luded in the Fren h edition.
18 The types ript reads: `Lan zos [ ', the referen e number is added in the
Fren h edition.
19 The types ript onsistently gives this referen e as `(q)', the Fren h
edition as `(9)'.
20 Equation number missing in the Fren h edition.
21 [eine Vers hiebung lngs ihrer Verbindungslinie mit dem Nullpunkt
[un dpla ement de leur droite de jon tion ave l'origine
22 [sie [ es rgles
23 [Vorgnge [phnomnes
24 Both the manus ript and the Fren h edition read `H1 ' and `H2 ' in this
(1)
paragraph and two paragraphs later, and `H
' in the intervening
paragraph. We have uniformised the notation.
25 The Fren h edition onsistently reads `fn '.
m {f (Wn ) f (Wm )}nm '

in both the types ript and the Fren h edition, but it should be as shown
26 The right-hand side of this equation reads `

(see above, p. 115).
27 The Fren h edition gives `(2)'.
28 Both the types ript and the Fren h edition read (only here) `n1 , n2 '.
29 Both the types ript and the Fren h edition read `nm '.
30 Singular in the Fren h edition.
31 The types ript in ludes the square bra kets but no referen e number.
The Fren h edition omits the referen e entirely.
32 Only bra kets in the types ript, referen es omitted in the Fren h edition.
33 [abhngige [indpendante
2
2
34 The Fren h edition reads `(cnk ) ' instead of `|cn | ' and `nk ' instead of `n'.
35 The Fren h edition reads `pnk ' instead of `n '.
Notes to pp. 425433
445
36 Both the types ript and the Fren h edition give `(11)', but this should
evidently be either `(17)' or `(22a)'.
37 The adje tive is omitted in the Fren h edition.
38 [`Strahlvektor' [`ve teur radiant'
39 The absolute square is missing in the German types ript, but is added in
the Fren h edition.
40 [des einen, etwa, na h Grsse und Ri htung [de l'un, par exemple en
grandeur et en dire tion
41 [von
[et
42 The Fren h edition has a prime on `q '.

43 The overbar is missing in the German types ript, but is added in the
Fren h edition.
44 Throughout this paragraph, the Fren h edition translates `bergang' as
`transformation' instead of `transition'.
45 `S ' missing in the Fren h edition.
46 [das folgende Stze zugrunde legt [qui est la base des thormes
suivants. Note that `Satz' an indeed mean both `statement' and
`theorem'.
47 The Fren h edition omits absolute bars.
48 The `h' is present in the Fren h edition but not in the types ript.
49 In the types ript, this is typed over an (illegible) previous alternative.
Jordan in his habilitation le ture (1927f [62) uses the term àngebbare
Wahrs heinli hkeit' (àssignable probability' in Oppenheimer's
translation (Jordan 1927g)).
50 [in dem der einfa he Begri der bere henbaren Wahrs heinli hkeit fr ein
bestimmtes Ereignis die Hauptrolle spielt [dans lequel la simple
notion de la probabilit al ulable joue le rle prin ipale pour un
vnement dtermin
51 [berstand [dfaut. The word `berstand' may be hara terising the
phases as some kind of surplus stru ture, but it is quite likely a mistyping
of `belstand', whi h an indeed be translated as `defe t', as in the
Fren h version.
52 [Es gibt [Cet auteur donne
53 [Wie kann insbesondere bei der nur statistis hen Bestimmtheit des
Einzelvorgangs in den zusammengesetzten makroskopis hen
Ers heinungen die gewohnte deterministis he Ordnung aufre ht erhalten
werden? [En parti ulier omment, vu la dtermination uniquement
statistique des pro essus individuels dans les phnomnes ma ros opiques
ompliqus, l'ordre dterministe auquel nous sommes a outums peut-il
tre onserv?
54 This referen e is to a supposedly forth oming `ber den begrii hen
Aufbau der Quantentheorie'. Yet, no su h published or unpublished work
by Bohr is extant. Some pages titled `Zur Frage des begrii hen Aufbaus
der Quantentheorie' are ontained in the folder `Como le ture II' in the
Niels Bohr ar hive, mi rolmed in AHQP-BMSS-11, se tion 4. See also
Bohr (1985, p. 478). We wish to thank Feli ity Pors, of the Niels Bohr
ar hive, for orresponden e on this point.
55 The Fren h edition in orre tly reads `sin 1
56 [Erwartung [attente
q1
'.
57 [Ausbreitung [extension
58 [einzelne Versu hsanordnungen [des essais isols
446
Notes to pp. 433442
59 Again, the terminology has hanged both in German and in English. The
term `similar parti les' for ìdenti al parti les' is used for instan e by
Dira (1927a [37).
60 [nur das antisymmetris he Teilsystem zulsst [ne permet pas le
systme antisymtrique
61 [aus [dans
62 The Fren h edition gives `[56'.
63 Both the German version followed here and the Fren h version (à whole
number of parti les') seem rather infeli itous.
64 Both the types ript and the Fren h edition add `(Magnetelektron)'. The
Fren h edition reads `Nature'.
65 Both the types ript and the Fren h edition read `p. 5'.
66 Types ript and published volume read `Pt' and `Mg', as well as `243'.
67 In the Fren h edition: `409'.
68 This is indeed the (abridged) published version of Elsasser's Gttingen
dissertation.
69 In the Fren h edition: `Das Adiabatenprinzip in den Quanten', as well as
`1927' (the latter as in the types ript).
70 In both the types ript and the Fren h edition the title of the paper is
given as `ber den ans hauli hen Inhalt der Quantenme hanik'.
71 Date given as `1927' in types ript and volume.
72 In the Fren h edition: `288'.
73 The Fren h edition omits `614'.
74 The manus ript in ludes also ànd
w's'
and ànd angle', both an elled.
75 The Fren h edition breaks up and rearranges this senten e.

76 The Fren h edition omits the temporal lause.
77 The Fren h edition reads `trouvent une analogie'.
78 In the manus ript this repla es the an elled word ùnknown'.
79 In the manus ript this repla es ` orresponds exa tly'.
Wave me hani s
By Mr E. SCHRDINGER
Introdu tion
Under this name at present two theories are being arried on, whi h
are indeed losely related but not identi al. The rst, whi h follows on
dire tly from the famous do toral thesis by L. de Broglie, on erns waves
in three-dimensional spa e. Be ause of the stri tly relativisti treatment
that is adopted in this version from the outset, we shall refer to it as the
four-dimensional wave me hani s. The other theory is more remote from
Mr de Broglie's original ideas, insofar as it is based on a wave-like pro ess
in the spa e of position oordinates (q -spa e) of an arbitrary me hani al
system.1 We shall therefore all it the multi-dimensional wave me hani s.
Of ourse this use of the q -spa e is to be seen only as a mathemati al tool,
as it is often applied also in the old me hani s; ultimately, in this version
also, the pro ess to be des ribed is one in spa e and time. In truth,
however, a omplete uni ation of the two on eptions has not yet been
a hieved. Anything over and above the motion of a single ele tron ould
be treated so far only in the multi-dimensional version; also, this is the
one that provides the mathemati al solution to the problems posed by
the Heisenberg-Born matrix me hani s. For these reasons I shall pla e
a Our translation follows S hrdinger's German types ript in AHQP-RDN,
do ument M-1354. Dis repan ies between the types ript and the Fren h edition
are endnoted. Interspersed in the German text, S hrdinger provided his own
summary of the paper (in Fren h). We translate this in the footnotes. The Fren h
version of this report is also reprinted in S hrdinger (1984, vol. 3, pp. 30223)
(eds.).
447
448
E. S hrdinger
it rst, hoping in this way also to illustrate better the hara teristi
di ulties of the as su h more beautiful four-dimensional version.a
I. Multi-dimensional theory
Given a system whose onguration is des ribed by the generalised

position oordinates q1 , q2 , . . . , qn , lassi al me hani s onsiders its task
as being that of determining the qk as fun tions of time, that is, of
exhibiting all systems of fun tions q1 (t), q2 (t), . . . , qn (t) that orrespond
to a dynami ally possible motion of the system. Instead, a ording to
wave me hani s the solution to the problem of motion is not given by a
system of n fun tions of the single variable t, but by a single fun tion
of the n variables q1 , q2 , . . . , qn and perhaps of time (see below).
This is determined by a partial dierential equation with q1 , q2 , . . . , qn
(and perhaps t) as independent variables. This hange of role of the
qk , whi h from dependent be ome independent variables, appears to be
the ru ial point. More later on the meaning of the fun tion , whi h
is still ontroversial. We rst des ribe how it is determined, thus what
orresponds to the equations of motion of the old me hani s.
First let the system be a onservative one. We start from its Hamiltonian fun tion
H =T +V ,
that is, from the total energy expressed as a fun tion of the qk and
the anoni ally- onjugate momenta pk . We take H to be a homogeneous
quadrati fun tion of the qk and of unity and repla e in it ea h pk by
h
2 qk and unity by . We all the fun tion of the qk , qk and thus
obtained L (be ause in wave me hani s it plays the role of a Lagrange
fun tion). Thus

h
+ V 2 .
L = T qk ,
(1)
2 qk
Now we determine (q1 , q2 , . . . , qn ) by the requirement that under variation of ,
Z
Z
Ld = 0 with
(2)
2 d = 1 .
a Summary of the introdu tion: Currently there are in fa t two [theories of

wave me hani s, very losely related to ea h other but not identi al, that is,
the relativisti or four-dimensional theory, whi h on erns waves in ordinary
spa e, and the multi-dimensional theory, whi h originally on erns waves in the
onguration spa e of an arbitrary system. The former, until now, is able to deal
only with the ase of a single ele tron, while the latter, whi h provides the solution
to the matrix problems of Heisenberg-Born, omes up against the di ulty of being
put in relativisti form. We start with the latter.
Wave me hani s
449
The integration is to be performed over the whole of q -spa e (on whose

perhaps innitely distant boundary, must disappear). However, d
is not simply the produ t of the dqk , rather the `rationally measured'
volume element in q -spa e:

12
2T

d = dq1 dq2 . . . dqn
(3)

p1 . . . pk
(it is the volume element of a Riemannian q -spa e, whose metri , as
for instan e also in Hertz's me hani s,a is determined by the kineti
energy). Performing the variation, taking the normalisation onstraint
with the multiplier [Fa tor E , yields the Euler equation
8 2
(4)
(E V ) = 0
h2
( stands for the analogue of the Lapla e operator in the generalised
Riemannian sense). As is well known,
Z
Ld = E
+
for a fun tion that satises the Euler equation (4) and the onstraint in
(2).
Now, it turns out that equation (4) in general does not have, for
every E -value, a solution that is single-valued and always nite and
ontinuous together with its rst and se ond derivatives; instead, in all
spe ial ases examined so far, this is the ase pre isely for the E -values
that Bohr's theory would des ribe as stationary energy levels of the
system (in the ase of dis repan ies, the re al ulated values explain the
fa ts of experien e better than the old ones). The word `stationary' used
by Bohr is thus given a very pregnant meaning by the variation problem
(4).
We shall refer to these values as eigenvalues, Ek , and to the orresponding solutions k as eigenfun tions.b We shall number the eigenvalues always in in reasing order and shall number repeatedly those with
multiple eigensolutions. The k form a normalised omplete orthogonal
system in the q -spa e, with respe t to whi h every well-behaved fun tion
of the qk an be expanded in a series. Of ourse this does not mean that
every well-behaved fun tion solves the homogeneous equation (4) and
a For S hrdinger's interest in Hertz's work on me hani s, see Mehra and Re henberg
(1987, pp. 52232) (eds.).
b As a rule, in ertain domains of the energy axis2 the eigenvalue spe trum is
ontinuous, so that the index k is repla ed by a ontinuous parameter. In the
notation we shall generally not take this into a ount.
450
E. S hrdinger
thus the variation problem, be ause (4) is indeed an equation system,

ea h single eigensolution k satisfying a dierent element of the system,
namely the one with E = Ek .a
One an take the view that one should be ontent in prin iple with
what has been said so far and its very diverse spe ial appli ations.
The single stationary states of Bohr's theory would then in a way be
des ribed by the eigenfun tions k , whi h do not ontain time at all .a
One would nd that one an derive mu h more from them that is worth
knowing, in parti ular, one an form from them, by xed general rules,
quantities that an be aptly taken to be jump probabilities between the
single stationary states. Indeed, it an be shown for instan e that the
integral
Z
(5)
qi k k d ,
extended to the whole of q -spa e, yields pre isely the matrix element
bearing the indi es k and k of the `matrix q ' in the Heisenberg-Born
theory; similarly, the elements of all matri es o urring there an be
al ulated from the wave me hani al eigenfun tions.
The theory as it stands, restri ted to onservative systems, ould treat
already even the intera tion between two or more systems, by onsidering these as one single system, with the addition of a suitable term
in the potential energy depending on the oordinates of all subsystems.
Even the intera tion of a material system with the radiation eld is
not out of rea h, if one imagines the system together with ertain ether
os illators (eigenos illations of a avity) as a single onservative system,
positing suitable intera tion terms.
On this view the time variable would play absolutely no role in an
isolated system a possibility to whi h N. Campbell (Phil. Mag., [1
(1926), [1106) has re ently pointed. Limiting our attention to an isolated
system, we would not per eive the passage of time in it any more than
a Summary of the above: Wave me hani s demands that events in a me hani al
system that is in motion be des ribed not by giving n generalised oordinates
q1 , q2 . . . qn as fun tions of the time t, but by giving a single fun tion [ of the n
variables q1 , q2 . . . qn and maybe of the time t. The system of equations of motion
of lassi al me hani s orresponds in wave me hani s to a single partial dierential
equation, eq. (4), whi h an be obtained by a ertain variational pro edure. E is
a Lagrange multiplier, V is the potential energy, a fun tion of the oordinates;
h is Plan k's onstant, denotes the Lapla ian in q -spa e, generalised in the
sense of Riemann. One nds in spe i ases that nite and ontinuous solutions,
èigenfun tions' k of eq. (4), exist only for ertain èigenvalues' Ek of E . The set
of these fun tions forms a omplete orthogonal system in the oordinate spa e.
The eigenvalues are pre isely the `stationary energy levels' of Bohr's theory.
a Cf. se tion 8.1 (eds.).
Wave me hani s
451
we an noti e its possible progress in spa e, an assimilation of time

to the spatial oordinates that is very mu h in the spirit of relativity.
What we would noti e would be merely a sequen e of dis ontinuous
transitions, so to speak a inemati image, but without the possibility of
omparing the time intervals between the transitions. Only se ondarily,
and in fa t with in reasing pre ision the more extended the system,
would a statisti al denition of time result from ounting the transitions
taking pla e (Campbell's `radioa tive lo k'). Of ourse then one annot
understand the jump probability in the usual way as the probability
of a transition al ulated relative to unit time. Rather, a single jump
probability is then utterly meaningless; only with two possibilities for
jumps, the probability that the one may happen before the other is equal
to its jump probability divided by the sum of the two.
I onsider this view the only one that would make it possible to hold
on to `quantum jumps' in a oherent way. Either all hanges in nature
are dis ontinuous or not a single one. The rst view may have many
attra tions; for the time being however, it still poses great di ulties.
If one does not wish to be so radi al and give up in prin iple the use
of the time variable also for the single atomisti system, then it is very
natural to assume that it is ontained hidden also in equation (4). One
will onje ture that equation system (4) is the amplitude equation of an
os illation equation, from whi h time has been eliminated by settinga
(6)
e2it .
E must then be proportional to a power of , and it is natural to set

E = h . Then the following is the os illation equation that leads to (4)
with the ansatz (6):b

4i
8 2
V
=0.
2
h
h t
(7)
Now this is satised not just by a single3

k e2ik t
(k =
Ek
),
h
a S hrdinger introdu es the time-dependent equation in his fourth paper on

quatisation (1926g). There (p. 112), S hrdinger leaves the sign of time
undetermined, settling on the same onvention as in (6) the opposite of today's
onvention on pp. 114-15. As late as S hrdinger (1926h, p. 1065), one reads
that `the most general solution of the wave-problem will be (the real part of)
[eq. (27) of that paper'. Instead the wave fun tion is hara terised as èssentially
omplex' in S hrdinger (1927 , fn. 3 on p. 957) (eds.).
b Re all that S hrdinger does not in fa t set m = 1, but absorbs the mass in the
denition of (eds.).
452
E. S hrdinger
but by an arbitrary linear ombination

=
ck k e2ik t
(8)
k=1
with arbitrary (even omplex) onstants ck . If one onsiders this as

the des ription4 of a ertain sequen e of phenomena in the system, then
this is now given by a ( omplex) fun tion of the q1 , q2 , . . . , qn and of
time, a fun tion whi h an even be given arbitrarily at t = 0 (be ause of
the ompleteness5 and orthogonality of the k ); the os illation equation
(7), or its solution (8) with suitably hosen ck , then governs the temporal
development. Bohr's stationary states orrespond to the eigenos illations
of the stru ture (one ck = 1, all others = 0).
There now seems to be no obsta le to assuming that equation (7) is
valid immediately also for non- onservative systems (that is, V may
ontain time expli itly). Then, however,6 the solution no longer has
the simple form (8). A parti ularly interesting appli ation hereof is the
perturbation of an atomi system by an ele tri alternating eld. This
leads to a theory of dispersion, but we must forgo here a more detailed
des ription of the same. From (7) there always follows
Z
d
(9)
d = 0 .
dt
(An asterisk shall always denote the omplex onjugate.7 ) Instead of the
earlier normalisation ondition (2), one an thus require
Z
d = 1 ,
(10)
whi h in the onservative ase, equation (8), means
ck ck = 1 .a
(11)
k=1
a Summary of the above: Even limiting oneself to what has been said up to now, it
would be possible to derive mu h of interest from these results, for instan e the
transition probabilities, formula (5) yielding pre isely the matrix element qi (k, k )
for the same me hani al problem formulated a ording to the Heisenberg-Born
theory. Although we have restri ted ourselves so far to onservative systems, it
would be possible to treat in this way also the mutual a tion between several
systems and even between a material system and the radiation eld; one would
only have to add all relevant systems to the system under onsideration. Time
does not appear at all in our onsiderations and one ould imagine that the only
events that o ur are sudden transitions from one quantum state of the total
system to another quantum state, as Mr N. Campbell has re ently thought. Time
would be dened only statisti ally by ounting the quantum jumps (Mr Campbell's
Wave me hani s
453
What does the -fun tion mean now, that is, how does the system
des ribed by it really look like in three dimensions? Many physi ists
today are of the opinion that it does not des ribe8 the o urren es in an
individual system,9 but only the pro esses in an ensemble of very many
like onstituted systems that do not sensibly inuen e one another10
and are all under the very same onditions. I shall skip this point of
view, sin e others are presenting it.a I myself have so far found useful
the following perhaps somewhat naive but quite on rete idea [dafr
re ht greifbare Vorstellung. The lassi al system of material points does
not really exist, instead there exists something that ontinuously lls
the entire spa e and of whi h one would obtain a `snapshot' if one
dragged the lassi al system, with the amera shutter open, through
all its ongurations, the representative point in q -spa e spending in
ea h volume element d a time that is proportional to the instantaneous
value of . (The value of for only one value of the argument t is
thus in question.) Otherwise stated: the real system is a superposition of
the lassi al one in all its possible states, using as `weight fun tion'.
The systems to whi h the theory is applied onsist lassi ally of several12
harged point masses. In the interpretation just dis ussed13 the harge
of every single one of these is distributed ontinuously a ross spa e,
the individual point mass with harge e yielding to the harge in the
three-dimensional volume element dx dy dz the ontribution14
e
d .
(12)
The prime on the integration sign means: one has to integrate only over
the part of the q -spa e orresponding to a position of the distinguished
point mass within dxdydz . Sin e in general depends on time,
these harges u tuate; only in the spe ial ase of a onservative system
`radioa tive lo k'). Another, less radi al, point of view is to assume that time
is hidden already in the family of equations (4) parametrised by E , this family
being the amplitude equation of an os illation equation, from whi h time has been
eliminated by the ansatz (6). Assuming h = E one arrives at eq. (7), whi h,
be ause it no longer ontains the frequen y , is solved by the series (8), where
the ck are arbitrary, generally omplex, onstants. Now is a fun tion of the
q1 , q2 . . . qn as well as of time t and, by a suitable hoi e of the ck , it an be adjusted
to an arbitrary initial state. Nothing prevents us now from making the time appear
also in the fun tion V this is the theory of non- onservative systems, one of
whose most important appli ations is the theory of dispersion. The important
relation (9), whi h follows from eq. (7), allows one in all ases to normalise
a ording to eq. (10).
a See the report by Messrs Born and Heisenberg.11
454
E. S hrdinger
os illating with a single eigenos illation are they distributed permanently, so to speak stati ally.
It must now be emphasised that by the laim that there are15 these
harge densities (and the urrent densities arising from their u tuation),
we an mean at best half of what lassi al ele trodynami s would mean
by that. Classi ally, harge and urrent densities are (1) appli ation
points, (2) sour e points of the ele tromagneti eld. As appli ation
points they are ompletely out of the question here; the assumption
that these harges and urrents a t, say, a ording to Coulomb's or
Biot-Savart's law dire tly on one another, or are dire tly ae ted in
su h a way by external elds, this assumption is either superuous or
wrong (N.B. de fa to wrong), be ause the hanges in the fun tion
and thereby in the harges are indeed to be determined through the
os illation equation (7) thus we must not think of them as determined
also in another way, by for es a ting on them. An external ele tri eld
is to be taken a ount of in (7) in the potential fun tion V , an external
magneti eld in a similar way to be dis ussed below, this is the way
their appli ation to the harge distribution is expressed in the present
theory.
Instead, our spatially distributed harges prove themselves ex ellently
as sour e points of the eld, at least for the external a tion of the
system, in parti ular with respe t to its radiation. Considered as sour e
points in the sense of the usual ele trodynami s, they yield largely16
orre t information about its frequen y, intensity and polarisation.a In
most ases, the harge is in pra ti e onned to a region that is small
ompared to the wavelengths of the emitted light. The radiation is then
determined by the resulting dipole moment [elektris hes Moment of the
harge distribution. A ording to the prin iples determined above, this is
al ulated from the lassi al dipole moment of an arbitrary onguration
by performing an average using
Z
(13)
Mqu = M l d .
A glan e at (8) shows that in Mqu the dieren es of the k will appear as
emission frequen ies; sin e the k are the spe tros opi term values, our
pi ture provides an understanding17 of Bohr's frequen y ondition. The
integrals that appear as amplitudes of the dierent partial os illations
of the dipole moment represent a ording to the remarks on (5) the
elements of Born and Heisenberg's `dipole moment matrix'. By evaluaa See the dis ussion after the report, as well as se tion 4.4 (eds.).
Wave me hani s
455
ting these integrals one obtained the orre t polarisations and intensities
of the emitted light in many spe ial ases, in parti ular intensity zero
in all ases where a line allowed by the frequen y ondition is missing
a ording to experien e (understanding 18 of the sele tion prin iple).
Even though all these results, if one so wishes, an be deta hed from the
pi ture of the u tuating harges and be represented in a more abstra t
form, yet they put quite beyond doubt that the pi ture is tremendously
useful for one who has the need for Ans hauli hkeit!19,a
In no way should one laim that the provisional attempt of a lassi alele trodynami oupling of the eld to the harges generating the eld
is already the last word on this issue. There are internal20 reasons for
doubting this. First, there is a serious di ulty in the question of the
rea tion of the emitted radiation on the emitting system, whi h is not yet
expressed by the wave equation (7), a ording to whi h also su h wave
forms of the system that ontinuously emit radiation ould and would in
fa t always persist unabated. Further, one should onsider the following.
We always observe the radiation emitted by an atom only through its
a tion on another atom or mole ule. Now, from the wave me hani al
standpoint we an onsider two harged point masses that belong to the
same atom, neither as a ting dire tly on ea h other in their pointlike
form (standpoint of lassi al me hani s), nor are we allowed to think this
of their `smeared out' wave me hani al harge distributions (the wrong
move taunted above). Rather, we have to take a ount of their lassi al
potential energy, onsidered as a fun tion in q -spa e, in the oe ient V
of the wave equation (7). But then, when we have two dierent atoms,
it will surely not be orre t in prin iple to insert the elds generated by
the spread-out harges of the rst at the position of the se ond in the
a Summary of the above: The physi al meaning of the fun tion appears to be that
the system of harged point parti les imagined by lassi al me hani s does not
in fa t exist, but that there is a ontinuous distribution of ele tri harge, whose
density an be al ulated at ea h point of ordinary spa e using or rather ,
the square of the absolute value of . A ording to this idea, the quantum (or: real)
system is a superposition of all the possible ongurations of the lassi al system,
the real fun tion in q -spa e o urring as `weighting fun tion'. Sin e in
general ontains time, u tuations of harge must o ur. What we mean by the
existen e of these ontinuous and u tuating harges is not at all that they should
a t on ea h other a ording to Coulomb's or Biot-Savart's law the motion of
these harges is already ompletely governed by eq. (7). But what we mean is
that they are the sour es of the ele tri elds and magneti elds pro eeding from
the atom, above all the sour es of the observed radiation. In many a ase one
has obtained wonderful agreement with experiment by al ulating the radiation
of these u tuating harges using lassi al ele trodynami s. In parti ular, they
yield a omplete and general explanation of Bohr's `frequen y ondition' and of
the spe tral `polarisation and sele tion rules'.
456
E. S hrdinger
wave equation for the latter. And yet we do this when we al ulate the
radiation of an atom in the way des ribed above and now treat wave
me hani ally the behaviour of another atom in this radiation eld. I say
this way of al ulating the intera tion between the harges of dierent
atoms an be at most approximate, but not orre t in prin iple. For
within one system it is ertainly wrong. But if we bring the two atoms
loser together, then the distin tion between the harges of one and
those of the other gradually disappears, it is a tually never a distin tion
of prin iple.21 The oherent wave me hani al route would surely be
to ombine both the emitting and the re eiving system into a single one
and to des ribe them through a single wave equation with appropriate
oupling terms, however large the distan e between emitter and re eiver
may be. Then one ould be ompletely silent about the pro esses in
the radiation eld. But what would be the orre t oupling terms? Of
ourse not the usual Coulomb potentials, as soon as the distan e is equal
to several wavelengths!22 (One realises from here that without important
amendments the entire theory in reality an only be applied to very small
systems.) Perhaps one should use the retarded potentials. But these are
not fun tions in the ( ommon) q -spa e, instead they are something mu h
more ompli ated. Evidently we en ounter here the provisional limits of
the theory and must be happy to possess in the pro edure depi ted above
an approximate treatment that appears to be very useful.a
II. Four-dimensional theory

If one applies the multi-dimensional version of wave me hani s to a single
ele tron of mass m and harge e moving in a spa e with the ele trostati
potential and to be des ribed by the three re tangular oordinates
a Summary of the above: However, there are reasons to believe that our u tuating
and purely lassi ally radiating harges do not provide the last word on this
question. Sin e we observe the radiation of an atom only by its ee t on another
atom or mole ule (whi h we shall thus also treat quite naturally by the methods of
wave me hani s), our pro edure redu es to substituting into the wave equation of
one system the potentials that would be produ ed a ording to the lassi al laws
by the extended harges of another system. This way of a ounting for the mutual
a tion of the harges belonging to two dierent systems annot be absolutely
orre t, sin e for the harges belonging to the same system it is not. The orre t
method of al ulating the inuen e of a radiating atom on another atom would
be perhaps to treat them as one total system a ording to the methods of wave
me hani s. But that does not seem at all possible, sin e the retarded potentials,
whi h should no doubt o ur, are not simply fun tions of the onguration of the
systems, but something mu h more ompli ated. Evidently, at present these are
the limits of the method!
Wave me hani s
x, y, z , then the wave equation (7) be omes

1 2 2 2
8 2
4i
2 e
+
+
=0.
2
2
2
m x
y
z
h
h t
457
(14)
1
(N.B. The fa tor m
derives from the fa t that, given the way of determi
ning the metri of the q -spa e through the kineti energy, x m, y m,
z m should be used as oordinates rather than x, y, z .23 ) It now turns

out that the present equation is nothing else but the ordinary threedimensional wave equation for de Broglie's `phase waves' of the ele tron,
ex ept that the equation in the above form is shortened or trun ated in
a way that one an all `negle ting the inuen e of relativity'.
In fa t, in the ele trostati eld de Broglie gives the following expressiona for the wave velo ity u of his phase waves, depending on the
potential (i.e. on position) and on the frequen y :24

mc2
h
=
(15)
.
u = cp
0
h
(h e)2 h2 02
If one inserts this into the ordinary three-dimensional wave equation

2
1 2
2 2
+
+
=0,
x2
y 2
z 2
u2 t2
and uses (6) to eliminate the frequen y from the equation, one has25
2
2
2
( = x
2 + y 2 + z 2 )

1 2 4ie 4 2 e2 2
2
+
2 2 +
(16)
0 = 0 .
c t
hc2 t
c2
h2
Now if one onsiders that in the ase of `slow ele tron motion' (a) the
o urring frequen ies are always very nearly equal to the rest frequen y
0 , so that in order of magnitude the derivative with respe t to time in
26
(16) is equal to a multipli ation by 2i0 , and that (b) e
h in this ase
is always small with respe t to 0 ; and if one then sets in equation (16)
= e2i0 t ,
(17)
and disregarding squares of small quantities, one obtains for exa tly
equation (14) derived from the multi-dimensional version of wave me hani s. As laimed, this is thus indeed the ` lassi al approximation' of the
wave equation holding for de Broglie's phase waves.b The transformation
(17) here shows us that, onsidered from de Broglie's point of view, the
a Cf. the formula for the refra tive index on p. 375 of de Broglie's report (eds.).
b That is, the nonrelativisti approximation (eds.).
458
E. S hrdinger
multi-dimensional theory is ommitted to a so to speak trun ated view

of the frequen y, in that it subtra ts on e and for all from all frequen ies
the rest frequen y 0 (N.B. In al ulating the harge density from ,27
the additional fa tor is of ourse irrelevant sin e it has modulus28 1.29 )a
Let us now keep to the form (16) of the wave equation. It still requires
an important generalisation. In order to be truly relativisti it must
be invariant with respe t to Lorentz transformations. But if we perform
su h a transformation on our ele tri eld, hitherto assumed to be stati ,
then it loses this feature and a magneti eld appears by itself next to
it. In this way one derives almost unavoidably the form of the wave
equation in an arbitrary ele tromagneti eld. The result an be put in
the following transparent form, whi h makes the omplete equivalen e
[Glei hbere htigung of time and the three spatial oordinates fully expli it :
"
2
2
2ie
2ie
+
ax +
+
ay
x
hc
y
hc
#
(18)
2
2

2ie
2ie
1
4 2 02
=
0
.
+
az +
+
i
+
z
hc
ic t
hc
c2
(N.B. a is the ve tor potential.30 In evaluating the squares one has
to take a ount of the order of the fa tors, sin e one is dealing with
operators, and further of Maxwell's relation:
ay
az
1 (i)
ax
+
+
+
= 0 .)
x
y
z
ic t
(19)
This wave equation is of very manifold interest. First, as shown by

Gordon,a it an be derived in a way very similar to what we have
seen above for the amplitude equation of onservative systems, from a
variational prin iple, whi h now obtains in four dimensions, and where
time plays a perfe tly symmetri al role with respe t to the three spatial
oordinates. Further: if one adds to the Lagrange fun tion of Gordon's
variational prin iple the well-known Lagrange fun tion of the Maxwell
a Summary of the above: The three-dimensional wave equation, eq. (14), obtained
by applying the multi-dimensional theory to a single ele tron in an ele trostati
potential eld , is none other than the nonrelativisti approximation of the
wave equation that results from Mr L. [de Broglie's ideas for his `phase waves'.
The latter, eq. (16), is obtained by substituting into the ordinary wave equation
expression (15), whi h Mr [de Broglie has derived for the phase velo ity u as a
fun tion of the frequen y and of the potential (that is, of the oordinates
x, y, z , on whi h will depend) and by eliminating from the resulting formula the
frequen y by means of (6).
a W. Gordon, Zeits hr. f. Phys., 40 (1926), 117.
Wave me hani s
459
eld in va uo (that is, the half-dieren e of the squares of the magneti

and the ele tri eld strenghts) and varies in the spa etime integral of the
new Lagrange fun tion thus obtained not only , but also the potential
omponents , ax , ay , az , one obtains as the ve Euler equations along
with the wave equation (18) also the four retarded potential equations
for , ax , ay , az .a (One ould also say: Maxwell's se ond quadruple of
equations, while the rst, as is well-known, holds identi ally in the
potentials.32 ) It ontains as harge and urrent density quadrati forms
in and its rst derivatives33 that agree ompletely with the rule whi h
we had given in the multi-dimensional theory for al ulating the true
harge distribution from the -fun tion. Se ond, one an further deneb
a stress-energy-momentum tensor of the harges, whose ten omponents
are also quadrati forms of and its rst derivatives, and whi h together
with the well-known Maxwell tensor obeys the laws of onservation of
energy and of momentum (that is, the sum of the two tensors has a
vanishing divergen e).c
But I shall not bother you here with the rather omplex mathemati al development of these issues, sin e the view still ontains a serious
in onsisten y. Indeed, a ording to it, it would be the same potential
omponents , ax , ay , az whi h on the one hand a t to modify the wave
equation (18) (one ould say: they a t on the harges as movers 35 )
and whi h on the other hand are determined in turn, via the retarded
potential equations, by these same harges, whi h o ur as sour es in
the latter equations. (That is: the wave equation (18) determines the
fun tion, from the latter one derives the harge and urrent densities,
whi h as sour es determine the potential omponents.) In reality,
however, one operates otherwise in the appli ation of the wave equation
(18) to the hydrogen ele tron, and one must operate otherwise to obtain
the orre t result: one substitutes in the wave equation (18) the already
a E. S hrdinger, Ann. d. Phys., 82 (1927), [265.31
b E. S hrdinger, lo . it.34
Summary of the above: In order to generalise equation (16) so that it may apply
to an arbitrary ele tromagneti eld, one subje ts it to a Lorentz transformation,
whi h automati ally makes a magneti eld appear. One arrives at eq. (18), in
whi h time enters in a perfe tly symmetri al way with the spatial oordinates.
Gordon has shown that this equation derives from a four-dimensional variational
prin iple. By adding to Gordon's Lagrangian the well-known Lagrangian of the
free eld and by varying along with also the four omponents of the potential,
one derives from a single variational prin iple besides eq. (18) also the laws of
ele tromagnetism with ertain homogeneous quadrati fun tions of and its rst
derivatives as harge and urrent densities. These agree well with what was said
in the previous hapter regarding the al ulation of the u tuating harges using
the fun tion. One nds a denition of the stress-energy-momentum tensor,
whi h, added to Maxwell's tensor, satises the onservation laws.
460
E. S hrdinger
given potentials of the nu leus and of possible external elds (Stark and
Zeeman ee t). From the solution for thus obtained one derives the
u tuating harge densities dis ussed above, whi h one in fa t36 has to
use for the determination from sour es of the emitted radiation; but one
must not add a posteriori to the eld of the nu leus and the possible
external elds also the elds produ ed by these harges at the position
of the atom itself in equation (18) 37 something totally wrong would
result.
Clearly this is a painful la una. The pure eld theory is not enough,
it has to be supplemented by performing a kind of individualisation
of the harge densities oming from the single point harges of the
lassi al model, where however ea h single ìndividual' may be spread
over the whole of spa e, so that they overlap. In the wave equation
for the single individual one would have to take into a ount only the
elds produ ed by the other individuals but not its self-eld. These
remarks, however, are only meant to hara terise the general nature
of the required supplement, not to onstitute a programme to be taken
ompletely literally.a
We wish to present also the remarkable spe ial result yielded by the
relativisti form (18) of the wave equation for the hydrogen atom. One
would at rst expe t and hope to nd the well-known Sommerfeld formula for the ne stru ture of terms. Indeed one does obtain a ne
stru ture and one does obtain Sommerfeld's formula, however the result
ontradi ts experien e, be ause it is exa tly what one would nd in
the Bohr-Sommerfeld theory, if one were to posit the radial as well as
the azimuthal quantum number as half-integers [halbzahlig, that is,
half of an odd integer. Today this result is not as disquieting as
when it was rst en ountered.b In fa t, it is well-known that the extension of Bohr's theory through the Uhlenbe k-Goudsmit ele tron spin
[Elektronendrall, required by many other fa ts of experien e, has to be
supplemented in turn by the move to se ondary quantum `half'-numbers
[`halbe' Nebenquantenzahlen in order to obtain good results. How the
a Summary of the above: However, these last developments run into a great di ulty.
From their dire t appli ation would follow the logi al ne essity of taking into
a ount in the wave equation, for instan e in the ase of the hydrogen atom,
not only the potential arising from the nu leus, but also the potentials arising
from the u tuating harges; whi h, apart from the enormous mathemati al
ompli ations that would arise, would give ompletely wrong results. The eld
theory (`Feldtheorie') appears thus inadequate; it should be supplemented by a
kind of individualisation of the ele trons, despite these being extended over the
whole of spa e.
b E. S hrdinger, Ann. d. Phys., 79 (1926), [361, p. 372.
Wave me hani s
461
spin is represented in wave me hani s is still un ertain. Very promising

suggestionsa point in the dire tion that instead of the s alar a ve tor
should be introdu ed. We annot dis uss here this latest turn in the
theory.b
III. The many-ele tron problem

The attemptsa to derive numeri al results by means of approximation
methods for the atom with several ele trons, whose amplitude equation (4) or wave equation (7) annot be solved dire tly, have led to
the remarkable result that a tually, despite the multi-dimensionality of
the original equation, in this pro edure one always needs to al ulate
only with the well-known three-dimensional eigenfun tions of hydrogen;
indeed one has to al ulate ertain three-dimensional harge distributions
that result from the hydrogen eigenfun tions a ording to the prin iples
presented above, and one has to al ulate a ording the prin iples of
lassi al ele trostati s the self-potentials and intera tion potentials of
these harge distributions; these onstants then enter as oe ients in
a system of equations that in a simple way determines in prin iple the
behaviour of the many-ele tron atom. Herein, I think, lies a hint that
with the furthering of our understanding ìn the end everything will
indeed be ome intelligible in three dimensions again'.38 For this reason
I want to elaborate a little on what has just been said.
Let
k (x, y, z) and Ek ;
(k = 1, 2, 3, . . .)
be the normalised eigenfun tions (for simpli ity assumed as real) and
orresponding eigenvalues of the one-ele tron atom with Z -fold positive
nu leus, whi h for brevity we shall all the hydrogen problem. They saa C. G. Darwin, Nature, 119 (1927), 282, Pro . Roy. So . A, 116 (1927), 227.
b Summary of the above: For the hydrogen atom the relativisti equation (18) yields
a result that, although disagreeing with experien e, is rather remarkable, that
is: one obtains the same ne stru ture as the one that would result from the
Bohr-Sommerfeld theory by assuming the radial and azimuthal quantum numbers
to be ìntegral and a half', that is, half an odd integer. The theory has evidently
to be ompleted by taking into a ount what in Bohr's theory is alled the spin
of the ele tron. In wave me hani s this is perhaps expressed (C. G. Darwin) by a
polarisation of the waves, this quantity having to be modied from a s alar to
a ve tor.
a See in parti ular A. Unsold, Ann. d. Phys., 82 (1927), 355.
462
E. S hrdinger
tisfy the three-dimensional amplitude equation ( ompare equation (4)):

8 2
Ze2
2 2
1 2
= 0
m x2 + y 2 + z 2 + h2 E + r
(20)
(r = x2 + y 2 + z 2 ) .
If only one eigenos illation is present, one has the stati harge distribution39
kk = ek2 .
(21)
If one imagines two being ex ited with maximal strength, one adds to
kk + ll a harge distribution os illating with frequen y |Ek El |/h,
whose amplitude distribution is given by
2lk = 2ek l .
(22)
The spatial integral of kl vanishes when k 6= l (be ause of the orthogonality of the k ) and it is e for k = l. The harge distribution resulting
from the presen e of two eigenos illations together has thus at every
instant the sum zero. One an now form the ele trostati potential
energies
Z
Z
kl (x, y, z)k l (x , y , z )
, (23)
pk,l;k ,l = . . . dx dy dz dx dy dz
r
p
where r = (x x )2 + (y y )2 + (z z )2 and the indi es k, l, k , l
may exhibit arbitrary degenera ies (to be sure, in the ase k = k , l = l ,
p is twi e the potential self-energy of the harge distribution kl ; but
that is of no importan e). It is the onstants p that ontrol also the
many-ele tron atom.
Let us sket h this. Let the lassi al model now onsist of n ele trons
and a Z -fold positively harged nu leus at the origin. We shall use the
wave equation in the form (7). It be omes 3n-dimensional,40 say thus
4i
8 2
1
(1 + 2 + . . . + n ) 2 (Vn + Ve )
=0.
m
h
h t
(24)
Here
=
2
2
2
+
+
;
x2
y2
z2
= 1, 2, 3, . . . , n .
(25)
We have onsidered the potential energy fun tion as de omposed in two

parts, Vn + Ve ; Vn should orrespond to the intera tion of all n ele trons
Wave me hani s
463
with the nu leus, Ve to their intera tion with one another, thereforea
n
X
1
,
Vn = Ze
r
=1
(26)
Ve = +e2
X 1
r
(27)
(, )
h
i
p
p
r = x2 + y2 + z2 , r = (x x )2 + (y y )2 + (z z )2 .
As the starting point for an approximation pro edure we hoose now the
eigensolutions of equation (24) with Ve = 0, that is with the intera tion
between the ele trons disregarded. The eigenfun tions are then produ ts
of hydrogen eigenfun tions, and the eigenvalues are sums of the orresponding eigenvalues of hydrogen. As a matter of fa t, one easily shows
that41
k1 ...kn = k1 (x1 y1 z1 ) . . . kn (xn yn zn )e
2it
h (Ek1 +...+Ekn )
(28)
always satises equation (24) (with Ve = 0). And if one takes all possible
sequen es of numbers [Zahlenkombinationen for the k1 , k2 , . . . , kn , then
these produ ts of k form a omplete orthogonal system in the 3ndimensional q -spa e one has thus integrated the approximate equation
ompletely.
One now aims to solve the full42 equation (24) (with Ve 6= 0) by
expansion with respe t to this omplete orthogonal system, that is one
makes this ansatz:
=
k1 =1
...
ak1 ...kn k1 ...kn .
(29)
kn =1
But of ourse the oe ients a annot be onstants, otherwise the above
sum would again be only a solution of the trun ated equation with Ve =
0. It turns out, however, that it is enough to onsider the a as fun tions
of time alone (`method of the variation of onstants').a Substituting (29)
into (24) one nds that the following onditions on the time dependen e
a Analogously to eq. (12), the prime on the summation sign should be interpreted
as meaning that the sum is to be taken over all pairs with 6= (eds.).
a P. A. M. Dira , Pro . Roy. So . A, 112 (1926), [661 p. 674.
464
E. S hrdinger
of the a must hold:43,44
X
2it
2i X
dak1 ...kn
=
...
vk1 ...kn ,l1 ...ln al1 ...ln e h (El1 ...ln Ek1 ...kn )
dt
h
l1 =1
ln =1
[k1 . . . kn = 1, 2, 3 . . . ]. (30)
Here we have set for brevity

Ek1 + . . . + Ekn = Ek1 ...kn .
(31)
The v are onstants, indeed they are prima fa ie 3n-tuple integrals

ranging over the whole of q -spa e (Additional explanation:45 Where do
these 3n-tuple integrals ome from? They derive from the fa t that
after substituting (29) into (24) one repla es the latter equation by
the mathemati ally equivalent ondition that its left-hand side shall be
orthogonal to all fun tions of the omplete orthogonal system in R3n .
The system (30) expresses this ondition.) Writing this out one has
vk1 ...kn ,l1 ...ln =
Z 3n-fold Z
. . . dx1 . . . dzn Ve k1 (x1 , y1 , z1 ) . . . kn (xn , yn , zn ) (32)
l1 (x1 , y1 , z1 ) . . . ln (xn , yn , zn ) .
If one now onsiders the simple stru ture of Ve given in (27), one re ognises that the v an be redu ed to sextuple integrals, in fa t ea h of them
is a nite sum of some of the Coulomb potential energies dened in (23).
Indeed, if in the nite sum representing Ve , we fo us on an individual
term, for example e2 /r , this ontains only the six variables x , . . . , z .
One an thus immediately perform in (32) pre isely 3n 6 integrations
on this term, yielding (be ause of the orthogonality and normalisation
of the k ) the fa tor 1, if k = l for all indi es that oin ide neither
with nor with , and yielding instead the fa tor 0 if even just for a
single dierent from and one has: k 6= l . (One sees thus that
very many terms disappear.) For the non-vanishing terms, it is easy to
see that they oin ide with one of the p dened in (23). QEDa
a Summary of the above: Calling k and Ek the eigenfun tions and eigenvalues
of the problem for one ele tron, harge e, in the eld of a nu leus +Ze
(hydrogen problem), let us form the harge distributions (21) and (22), the
former orresponding to the existen e of a single normal mode, the latter to the
ooperation of two of them. Taken as harge densities in ordinary ele trostati s,
ea h of these would have a ertain potential energy and there would even be a
ertain mutual potential energy between two of them, assumed to oexist. These
are the onstants pkl;k l in (23). With these givens, let us atta k the problem
of the n-ele tron atom. Dividing the potential energy in the wave equation (24)
Wave me hani s
465
Let us now have a somewhat loser look at the equation system

(30), whose oe ients, as we have just seen, have su h a relatively
simple stru ture, and whi h determines the varying amplitudes of our
ansatz46 (29) as fun tions of time. We an allow ourselves to introdu e
a somewhat simpler symboli notation, by letting the string of indi es k1 , k2 , . . . , kn be represented by the single index k , and similarly
l1 , l2 , . . . , ln by l. One then has
2i X
2it
dak
=
vkl al e h (El Ek ) .
dt
h
(33)
l=1
(One must not onfuse, however, El , Ek with the single47 eigenvalues of

the hydrogen problem, whi h were earlier denoted in the same way.48 )
This is now a system of innitely many dierential equations, whi h
we annot solve dire tly: so, pra ti ally nothing seems to have been
gained. In turn, however, we have as yet also negle ted nothing: with
exa t solutions ak of (33), (29) would be an exa t solution of (24).
This is pre isely where I want to pla e the main emphasis, greater than
on the pra ti al implementation of the approximation pro edure, whi h
shall be sket hed below only for the sake of ompleteness. In prin iple
the equations (33) determine the solution of the many-ele tron problem
exa tly;49 and they no longer ontain anything multi-dimensional;
their oe ients are simple Coulomb energies of harge distributions
that already o ur in the hydrogen problem. Further, the equations
(33) determine the solutions of the many-ele tron problem a ording
to (29) as a ombination of produ ts of the hydrogen eigenfun tions.
While these produ ts (denoted above by k1 k2 ...kn )50 are still fun tions
on the 3n-dimensional q -spa e, any two of them yield in the al ulation
of the three-dimensional harge distributions in the many-ele tron atom,
as is easily seen, a harge distribution whi h if it is not identi ally zero
orresponds again to a hydrogen distribution (denoted above by kk or
kl ).
These onsiderations are the analogue of the onstru tion of the higher
atoms from hydrogen traje tories in Bohr's theory. They reinfor e the
for this problem into two terms and negle ting at rst the term Ve , due to the
mutual a tion between the ele trons, the eigensolutions would be given by (28),
that is, by the produ ts of n hydrogen fun tions. From these produ ts, taken in all
ombinations, form the series (29), whi h will yield the exa t solution of equation
(24), provided that the oe ients ak1 k2 ...kn are fun tions of time satisfying the
equations (30); (see the abbreviation (31)). The oe ients v in (30) are onstants,
dened originally by the 3n-tuple integrals (32), whi h however, thanks to the
simple form of Ve (see (27)), redu e to sextuple integrals, namely pre isely to the
onstants pk,l;k ,l (see (23)).
466
E. S hrdinger
hope that by delving more deeply one will be able to interpret and understand the results of the multi-dimensional theory in three dimensions.a
Now, as far as the approximation method is on erned, it onsists
in fa t of onsidering the ontribution Ve made to the potential energy
fun tion V by the intera tion of the ele trons with one another, to be as
far as possible small as ompared to the a tion of the nu leus. The vkl
are then onsidered small ompared to the eigenvalue dieren es El Ek ,
ex ept if El = Ek . The al will then vary slowly by omparison to the
powers of e appearing on the right-hand side of equation (33), as long as
the latter are not equal to 1, and all those terms on the right-hand side for
whi h this is not the ase will yield only small u tuations of short period
of the ak and an be negle ted in the approximation.51 Thereby, rst,
the sums on the right be ome nite, be ause in fa t always only a nite
number of eigenvalues oin ide. Se ond, the innitely many equations
separate into groups; ea h group ontains only a nite number of al and
an be integrated very easily.52 This is the rst step of the approximation
pro edure, whi h in theory an be ontinued indenitely, but be omes
more and more umbersome. We shall not enter into details.
One an also transform the untrun ated system of dierential equations (33) at a single stroke into a system of ordinary linear equations
(with innitely many unknowns!) by setting
a l = cl e
2it
h (EEl )
(34)
where the quantity E and the quantities cl are unknown onstants.

Substituting into (33) one nds
(E Ek )ck =
vkl cl ;
(k = 1, 2, 3, . . .) .
(35)
l=1
This equation system oin ides with the Heisenberg-Born `prin ipal axes
problem'. If the vkl are very small quantities, then, if not all cl are to
be very small, E must be lose to one of the El , let us say to Ek .
In the rst approximation then only ck , and all those cl for whi h
a Summary of the above: Although the system of eqs. (30) (abbreviated to (33))
does not admit a dire t solution, the number of equations as well as the number of
unknown fun tions being innite, it seems to me very interesting that the solution
to the multi-dimensional problem is provided in prin iple by a system of equations
whose oe ients have su h simple meanings in three dimensions. Further, one
realises that the harge distribution that orresponds to the solution (29) of the
n-ele tron problem turns out to be the superposition of the distributions kk
and kl that o ur already in the hydrogen problem. The hope of interpreting
and of understanding the multi-dimensional theory in three dimensions is thus
strengthened.
Wave me hani s
467
El = Ek , are dierent from zero. The problem thus separates in the rst
approximation into a denumerable set of nite prin ipal axes problems.a
a Summary of the above: One an embark on the solution of the system of equations
(33) by an approximation method. Positing (34), the onstants E and cl have to
satisfy the system (35) of ordinary linear homogeneous equations, whose number as
well as that of the unknown onstants, however, is innite. It is only by assuming
all oe ients vkl to be small that one an on lude that E has to be very lose
to one of the values El , for instan e Ek , and that [cl approximately vanishes,
unless El is equal to Ek . Sin e there is only a nite number of El that oin ide
with Ek , the problem redu es in the rst approximation to a problem of a nite
number of `prin ipal axes', or rather to an innity of su h nite problems. As a
matter of fa t, the equations (35) oin ide with the problem of an innite number
of prin ipal axes, whi h the Heisenberg-Born me hani s redu es to.
468
E. S hrdinger
Dis ussion of Mr S hrdinger's report
Mr S hrdinger.
It would seem that my des ription in terms of

a snapshot was not very fortunate, sin e it has been misunderstood.
Perhaps the following explanation is learer. The interpretation of Born
is well-known, who takes d to be the probability for the system
being in the volume element d of the onguration spa e. Distribute
a very large number N of systems in the onguration spa e, taking
the above probability as `frequen y fun tion'. Imagine these systems as
superposed in real spa e, dividing however by N the harge of ea h point
mass in ea h system. In this way, in the limiting ase where N = one
obtains the wave me hani al pi ture of the system.
Mr Bohr. You have said that from the harge distribution d
and the lassi al laws you obtain the frequen y and intensity of light,
but do the remarks about di ulties you made later indi ate that what
you had obtained was not orre t?
Mr S hrdinger.
The di ulty I mentioned is the following.

If one expands the general solution as a series with respe t to the
eigenfun tions
X
=
ck k
k
and if one al ulates the intensity of the radiation resulting from k and
l together, one nds that it be omes proportional to c2k c2l . However,
a ording to the old theory, only the square of the amplitude orresponding to the ìnitial level' should appear here; that of the `nal level'
should be repla ed by 1.
Mr Bohr. Has Dira not found the solution to the di ulty?
Mr S hrdinger. Dira 's results are ertainly very interesting
and point the way toward a solution, if they do not ontain it already.
Only, we should rst ome to an understanding in physi al terms [nous
devrions d'abord nous entendre en langage physique. I nd it still impossible, for the time being, to see an answer to a physi al question in
the assertion that ertain quantities obey a non ommutative algebra,
espe ially when these quantities are meant to represent numbers of
atoms. The relation between the ontinuous spatial densities, des ribed
earlier, and the observed intensities and polarisations of the spe tral
Dis ussion
469
rays is [too naturala for me to deny all meaning to these densities only
be ause some di ulties appear that are not yet resolved.
Mr Born. It seems to me that interpreting the quantity
as
a harge density leads to di ulties in the ase of quadrupole moments.
The latter in fa t need to be taken into a ount in order to obtain the
radiation, not only for theoreti al reasons, but also for experimental
reasons.
For brevity let us set
e2 = e2 ||2 =
and let us onsider, for example, the ase of two parti les; be omes a
fun tion of x1 and x2 , where for brevity x1 stands for all the oordinates
of the rst parti le; x2 has a similar meaning. The ele tri density is
then, a ording to S hrdinger,
Z
Z
(x) = (x, x2 )dx2 + (x1 , x)dx1 .
In wave me hani s the quadrupole moment
Z Z
x1 x2 (x1 , x2 )dx1 dx2
annot, as far as I an tell, be expressed using the fun tion (x). I

would like to know how one an, in this ase, redu e the radiation of
the quadrupole to the motion of a harge distribution (x) in the usual
three-dimensional spa e.
Mr S hrdinger.
I an assure you that the al ulation of the

dipole moments is perfe ly orre t and rigorous and that this obje tion
by Mr Born is unfounded. Does the agreement between wave me hani s
and matrix me hani s extend to the possible radiation of a quadrupole?
That is a question I have not examined. Besides, we do not possess observations on this point that ould allow us to use a possible disagreement
between the two approa hes to de ide between them.
Mr Fowler
asks for explanations regarding the method for solving

the equations in the ase of the many-ele tron problem.
Mr De Donder. Equation (24) of Mr S hrdinger's report an be

a The Fren h here reads `trop peu naturelle', whi h has the exa t opposite meaning.
The ontext would seem, however, to justify the amendment (eds.).
470
E. S hrdinger
extended to the ase in whi h the n harged parti les are dierent and
where the external a tions as well as the intera tions an be des ribed,
in spa etime, by a gravitational eld [ hamp gravique.a The quantum
equation thus obtained is the sum of the quantum equations for the
n parti les taken separately, ea h of the equations being divided by
the (rest) mass of the orresponding parti le. Thus, for instan e, the
quantum equation for the nu leus will not enter if one assumes, as a
rst approximation, that the mass of the nu leus is innitely large with
respe t to that of an ele tron.
When there is intera tion, the problem is mu h more omplex. One
an, as Mr S hrdinger indi ates, onsider the a tion of the nu leus as
an external a tion a ting on the ele trons of the loud [ ouronne, and
the (ele trostati ) a tions between the ele trons in this loud as a perturbation; but that is only a rst approximation. In order to a ount for
relativisti and ele tromagneti ee ts I have assumed that the mole ular
systems have an additive hara ter.a One an thus re over, as a spe ial
ase, the above-mentioned method of quantisation by S hrdinger.
Mr Born.
In Gttingen we have embarked on a systemati al ulation of the matrix elements that appear in perturbation theory, with
the aim of olle ting them in tables up to the prin ipal quantum number
10. Part of these al ulations, whi h are very extended, has already been
done. My oworker Mr Biemller has used them to al ulate the lower
terms of the helium atom a ording to the usual perturbation method
up to perturbations of the se ond order. The agreement of the ground
term with the empiri al value, despite the defe ts of the pro edure, is
hardly worse than in the re ently published paper by Kellner [Zeits hr.
f. Phys., 44 (1927), 91, who has applied a more pre ise method (Ritz's
pro edure).
Mr Lorentz. Do you see the out ome of this long labour as
satisfa tory?
Mr Born. The al ulation has not attained yet the pre ision of
the measurements. The al ulations we have done applying the ordinary

a Th. De Donder, L'quation fondamentale de la Chimie quantique, Comptes Rendus
A ad. S i. Paris, session of 10 O tober 1927, pp. 698700. See esp. eq. (10).
a For more details, one an onsult our note: `L'quation de quanti ation des
mol ules omprenant n parti ules le trises', published after this meeting, in
the Bull. A . R. Belg., Cl. des S ien es, session of 5 November 1927.
Dis ussion
471
perturbation method [mthode des perturbations ordinaires onsist of

a series expansion with respe t to the inverse of the nu lear harge Z ,
of the form

b
c
E = Z a + + 2 + ... .
Z
Z
The three terms shown have been al ulated. Nevertheless, in the ase
of helium (Z = 2) the pre ision is not yet as good as in the al ulations
done by Kellner using Ritz's approximation method.
Mr Lorentz. But you hope however to improve your results.

Mr Born. Yes, only the onvergen e of the series is very slow.
Mr Heisenberg. On the subje t of this approximation method,
Mr S hrdinger says at the end of his report that the dis ussion he
has given reinfor es the hope that when our knowledge will be deeper
it will be possible to explain and to understand in three dimensions
the results provided by the multi-dimensional theory. I see nothing in
Mr S hrdinger's al ulations that would justify this hope. What Mr
S hrdinger does in his very beautiful approximation method, is to
repla e the n-dimensional dierential equations by an innity of linear
equations. That redu es the problem, as Mr S hrdinger himself states,
to a problem with ordinary matri es, in whi h the oe ients an be
interpreted in three-dimensional spa e. The equations are thus `threedimensional' exa tly in the same sense as in the usual matrix theory.
It thus seems to me that, in the lassi al sense, we are just as far from
understanding the theory in three dimensions as we are in the matrix
theory.
Mr S hrdinger.
I would not know how to express more pre isely

my hope of a possible formulation in a three-dimensional spa e. Besides,
I do not believe that one would obtain simpler al ulational methods in
this way, and it is probable that one will always do al ulations using
the multi-dimensional wave equation. But then one will be able to grasp
its physi al meaning better. I am not pre isely sear hing for a threedimensional partial dierential equation. Su h a simple formulation is
surely impossible. If I am not satised with the urrent state of the
problem, it is be ause I do not understand yet the physi al meaning of
its solution.
472
E. S hrdinger
What Mr Heisenberg has said is mathemati ally unex eptionable,

but the point in question is that of the physi al interpretation. This
is indispensable for the further development of the theory. Now, this
development is ne essary. For one must agree that all urrent ways of
formulating the results of the new quantum me hani s only orrespond
to the lassi al me hani s of a tions at a distan e. As soon as light
rossing times be ome relevant in the system, the new me hani s fails,
be ause the lassi al potential energy fun tion no longer exists.
Allow me, to show that my hope of a hieving a three-dimensional
on eption is not quite utopian, to re all what Mr Fowler has told
us on the topi of Mr Hartree's approximation method.a It is true
that this method abstra ts from what one alls the èx hange terms'
(whi h orrespond, for instan e, to the distan e between the ortho and
para terms of neutral helium). But, abstra ting from that, it already
a hieves the three-dimensional aim I tend to. Should one de lare a priori
impossible that Hartree's method might be modied or developed in su h
a way as to take into a ount the ex hange terms while working with a
satisfa tory three-dimensional model?
Mr Born.
Regarding the question of knowing whether it is possible

to des ribe a many-ele tron problem by a eld equation in three dimensions, I would like to point out the following. The number of quantum
numbers of an atom rises by three with ea h additional ele tron; it is thus
equal to 3n for n ele trons. It seems doubtful that there should be an
ordinary, three-dimensional eigenvalue problem, whose eigenvalues have
a range of size 3n [dont la valeur ara tristique ait une multitude de
3n dimensions.b Instead, it follows from re ent papers by Dira and
Jordanc that one an build on a three-dimensional os illation equation if
one onsiders the eigenfun tion itself not as an ordinary number, but as
one of Dira 's q-numbers, that is, if one quantises again its amplitude as a
fun tion of time. An n-quanta os illation with this amplitude then yields
together with the three spatial quantum numbers the ne essary range
[multitude of quantum numbers. From this point of view the number of
ele trons in a system appears itself as a quantum number, that is, the
ele trons themselves appear as dis ontinuities of the same nature as the
a See the dis ussion after Bragg's report, p. 318 (eds.).
b The Fren h text here appears to make little sense, but Born is possibly referring
to the dimension of the spa e of solutions (eds.).
Cf. se tion IV of Born and Heisenberg's report (eds.).
Dis ussion
473
stationary states.
Mr S hrdinger. Pre isely the stru ture of the periodi system
is already ontained in the physi s [m anique of the three-dimensional

hydrogen problem. The degrees of degenera y 1, 4, 9, 16, et ., multiplied
by 2, yield pre isely the periodi numbers [nombres de priodes. The
fa tor 2 that I have just mentioned derives from the spin [giration (spin).
From the point of view of wave me hani s, the apparently mysterious
`Pauli a tion' of the rst two ele trons on the third (whi h they prevent
from also following an orbit with quantum number 1) means stri tly
speaking nothing other than the non-existen e of a third eigenfun tion
with prin ipal quantum number 1. This non-existen e is pre isely a
property of the three-dimensional model, or of the three-dimensional
equation. The multi-dimensional equation has too many eigenfun tions;
it is this [elle that makes the `Pauli ex lusion' (Pauliverbot) ne essary
to eliminate this defe t.a
a The Fren h text refers to the four-dimensional equation (`l'quation quatre

dimensions') as having too many solutions. This reading ould be orre t, in
the sense that the ex lusion prin iple was rst introdu ed in the ontext of the
relativisti (four-dimensional) Bohr-Sommerfeld theory of the atom, but the above
reading seems mu h more natural in ontext. Note that S hrdinger throughout his
report uses `vierdimensional' and `vieldimensional', whi h ould be easily onfused,
for `four-dimensional' and `many-dimensional', respe tively (eds.).
474
Notes to pp. 447463

1 Here and in the following, the Fren h edition omits some itali s, whi h
are quite hara teristi of S hrdinger's writing style and whi h we
ta itly restore.
2 [Energiegerade [srie des nergies
3 Bra ket added in the Fren h edition.
4 [als Bes hreibung [ omme la dnition
5 [Vollstndigkeit [perfe tion
6 [freili h [videmment
7 Printed as a footnote in the Fren h edition.
8 [sie ni ht .... bes hreibe [qu'ils ne d rivent pas
9 [Einzelsystem [systme dtermin
10 This lause is omitted in the Fren h edition.
11 [von anderer Seite vertreten [dfendue par d'autres. Footnote only in
the Fren h edition.
12 [aus einer Anzahl [d'un grand nombre
13 [Dur h die eben bespro hene Deutung [Ainsi que nous venons de le
voir
14 The equation number is missing in the Fren h edition, and the following
senten e is printed as a footnote.
15 [es
gebe
[sont donnes
16 [weitgehend [tout fait
17 [Verstndnis
[interprtation
18 [Verstndnis [signi ation
19 No ex lamation mark in the Fren h edition.

20 [innere [intimes
21 [eine prinzipielle [essentielle
22 No ex lamation mark in the Fren h edition.
24 In the Fren h edition this equation number is given to the following
equation (unnumbered in the types ript).
25 Bra ket printed as a footnote in the Fren h edition, with the addition: `
stands for the Lapla ian'.
e
e
26 [ h [e h
27 [die Ladungsdi hte aus [la densit de harge
28 [Betrag [valeur
29 Bra ket printed as a footnote in the Fren h edition.
30 The rest of the bra ket is printed as a footnote in the Fren h edition.
31 Both types ript and Fren h edition give `365' as page number.
32 The Fren h edition adds this to the footnote.
33 [treten darin in
und seinen ersten Ableitungen quadratis he Formen
auf [y gurent dans
et ses premires drives des formes
quadratiques
34 Footnote only in the Fren h edition.
35 [bewegend [par
le mouvement
36 [allerdings [ ertainement
37 [(18) [(8)
38 The Fren h edition omits the inverted ommas.
39 The equation number is missing in the printed volume.
40 [3n-dimensional [tridimensionelle
Notes to pp. 463466

41 Misprint in the Fren h edition: the
Ek
475
are not in the exponent.
42 [komplet[te [ omplexe
43 [dass fr die Abhngigkeit der
bestehen [que pour que
von der Zeit folgende Forderungen
dpende du temps les onditions suivantes
doivent tre satisfaites

44 Two misprints in the Fren h edition: the
the
ki
run to
Ek
are not in the exponent, and
n.

46 [unseres Ansatzes [de notre expression fondamentale
47 [den
einzelnen
[les
diverses
48 Bra ket printed as footnote in the Fren h edition.

49 [bestimmen die Lsung exa t [dterminent la solution
50 Misprint in the Fren h edition: `(k1 , k2 , . . . , kn )'.
51 Both the types ript and the Fren h edition read `cl ' and `ck ' instead of
àl ' and àk '.
52 Again, both the types ript and the Fren h edition read `ck ' instead of
àk '.
General dis ussion of the new ideas

presented
a

Mr Lorentz. I should like to draw attention to the di ulties one
en ounters in the old theories.

We wish to make a representation of the phenomena, to form an image
of them in our minds. Until now, we have always wanted to form these
images by means of the ordinary notions of time and spa e. These notions
are perhaps innate; in any ase, they have developed from our personal
experien e, by our daily observations. For me, these notions are lear
and I onfess that I should be unable to imagine physi s without these
notions. The image that I wish to form of phenomena must be absolutely
sharp and denite, and it seems to me that we an form su h an image
only in the framework of spa e and time.
For me, an ele tron is a orpus le that, at a given instant, is present at
a denite point in spa e, and if I had the idea that at a following moment
the orpus le is present somewhere else, I must think of its traje tory,
whi h is a line in spa e. And if the ele tron en ounters an atom and
penetrates it, and after several in idents leaves the atom, I make up a
theory in whi h the ele tron preserves its individuality; that is to say,
I imagine a line following whi h the ele tron passes through the atom.
a As mentioned in se tion 1.6, the Bohr ar hives ontain a opy of the galley
proofs of the general dis ussion, dated 1 June 1928.1 A few of the ontributions
in these proofs seem to have been still largely unedited: they ontain some gaps
and in omplete senten es, some more olloquial formulations, and in at least one
ase a senten e that was dropped from the published volume. We reprodu e in
endnotes the most substantial examples of these alternative versions. For most of
the dis ussion ontributions by Dira , we have followed his manus ript version.2 For
Bohr's dis ussion ontributions, we have used material from Bohr (1985) and from
notes taken by Ri hardson3 (also mentioned in se tion 1.6). See our notes for further
details (eds.).
476
477
Obviously, su h a theory may be very di ult to develop, but a priori

it does not seem to me impossible.
I imagine that, in the new theory, one still has ele trons. It is of ourse
possible that in the new theory, on e it is well-developed, one will have to
suppose that the ele trons undergo transformations. I happily on ede
that the ele tron may dissolve into a loud. But then I would try to
dis over on whi h o asion this transformation o urs. If one wished to
forbid me su h an enquiry by invoking a prin iple, that would trouble
me very mu h. It seems to me that one may always hope one will do later
that whi h we annot yet do at the moment. Even if one abandons the old
ideas, one may always preserve the old lassi ations [dnominations. I
should like to preserve this ideal of the past, to des ribe everything that
happens in the world with distin t images. I am ready to a ept other
theories, on ondition that one is able to re-express them in terms of
lear and distin t images.
For my part, despite not having yet be ome familiar with the new
ideas that I now hear expressed,a I ould visualise these ideas thus. Let
us take the ase of an ele tron that en ounters an atom; let us suppose
that the ele tron leaves the atom and that at the same time there is
emission of a light quantum. One must onsider, in the rst pla e, the
systems of waves that orrespond to the ele tron and to the atom before
the ollision. After the ollision, we will have new systems of waves.
These systems of waves an be des ribed by a fun tion dened in a
spa e with a large number of dimensions and satisfying a dierential
equation. The new wave me hani s will work with this equation and will
determine the fun tion before and after the ollision.
Now, there are phenomena that tea h us that there is something
else in addition to the waves, namely orpus les; one an, for example,
perform an experiment with a Faraday ylinder; one must then take
into a ount the individuality of the ele trons and also of the photons. I
think I would nd that, to explain the phenomena, it su es to assume
that the expression gives the probability that the ele trons and
the photons exist in a given volume; that would su e to explain the
experiments. But the examples given by Mr Heisenberg tea h me that I
will have thus attained everything that experiment allows me to attain.
However, I think that this notion of probability should be pla ed at
a In fa t, Lorentz had followed the re ent developments rather losely. In parti ular,
he had orresponded extensively with Ehrenfest and with S hrdinger, and had
even delivered seminars and le tures on wave me hani s and on matrix me hani s
at Leiden, Cornell and Calte h. See se tion 1.3 (eds.).
478
General dis ussion
the end, and as a on lusion, of theoreti al onsiderations, and not as

an a priori axiom, though I may well admit that this indetermina y
orresponds to experimental possibilities. I would always be able to keep
my deterministi faith for the fundamental phenomena, of whi h I have
not spoken. Could a deeper mind not be aware of the motions of these
ele trons? Could one not keep determinism by making it an obje t of
belief? Must one ne essarily elevate indeterminism to a prin iple?
Mr Bohr
expounds his point of view with respe t to the problems

of quantum theory.
The original published pro eedings add `(see the pre eding arti le)'. In the
pro eedings, the arti le pre eding the general dis ussion is a Fren h translation
of the German version of Bohr's Como le ture (Bohr 1928) (published in
Naturwissens haften).
As des ribed in se tion 1.6, this arti le was in luded
at Bohr's request, to repla e his remarks made at this point in the general
dis ussion. (In our translation of the pro eedings, we have omitted this wellknown arti le.)
The extant notes relating to Bohr's remarks at this point are parti ularly
fragmentary. Kal kar's introdu tion to volume 6 of Bohr's
Colle ted Works
(Bohr 1985) des ribes the orresponding part of notes (taken by Kramers and
by Vers haelt) in the Bohr ar hives as too in omplete to warrant reprodu tion
in that volume, but provides the following summary and omparison with the
printed versions of the Como le ture: `The notes over the wave- orpus le
aspe ts of light and matter ( orresponding to the rst se tions of the printed
le ture). The
-ray
mi ros ope is analysed, although the notes are somewhat
in omplete here (as in many other pla es), and the rle of the nite wave
trains is dis ussed in onne tion with the momentum measurement through
the Doppler ee t (as in the printed versions). After some questions .... Bohr
ontinues by dis ussing the signi an e of the phase and omments on the
Stern-Gerla h experiment and the inobservability of the phase in a stationary
state .... ' (Bohr 1985, p. 37).
Further details of what Bohr said at this point may be obtained from notes
4
on the general dis ussion taken by Ri hardson. Below, we reprodu e the
relevant parts of these notes, and omment on their relation to Bohr's paper
translated in the pro eedings.
The rst part of Ri hardson's notes relating to Bohr reads as follows:
E = h
ei2(x x+y y+z zt)
p = h
Int[erf[eren e.
xx
t
?h [?

1
1
xpx
tE
h
h
479
Fig. A.
This orresponds to part of se tion 2 of Bohr's paper translated in the pro eedings. There Bohr introdu es the on epts of energy and momentum for plane
waves, and the idea that waves of limited extent in spa etime are obtained
through the ìnterferen e' (that is, superposition) of dierent plane waves, the
resulting waves satisfying (at best) the given relations. (As a onsequen e, a
group of waves has no well-dened phase, a point Bohr takes up again below.)
This is used to justify Bohr's idea of omplementarity between a ausal pi ture
(in the sense of energy-momentum onservation for elementary pro esses) and
a spa etime pi ture.
Ri hardson's notes then ontinue as shown in Fig. A. The
-ray mi ros ope
is dis ussed in se tion 3 of Bohr (1928) (the se tion on measurement, whi h

also dis usses momentum measurements based on the Doppler ee t). Bohr
appears to have inserted a dis ussion of these experiments as an illustration
of the un ertainty-type relations above.
The next part of Ri hardson's notes returns to se tion 2 of the paper, and is
reprodu ed in Fig. B. This orresponds in fa t to the subsequent paragraphs
of se tion 2, in whi h Bohr applies the notion of omplementarity to resolve
the per eived paradoxes related to the s attering of radiation by free ele trons
(note the extended as opposed to pointlike region of s attering in the
diagram, and see Bohr's ontribution to the dis ussion of Compton's report,
p. 60) as well as the per eived paradoxes related to ollisions ( f. se tion 3.4.2).
480
General dis ussion
Fig. B.
Possibly,
harge.
stands for Ìmpuls' (that is, momentum),
for radiation,
for
The next part of Ri hardson's notes, shown in Fig. C, instead relates to part
of se tion 6 of Bohr's paper (se tions 4 and 5 of the paper are, respe tively,
a review of the orresponden e prin iple and of matrix me hani s, and a
dis ussion and ritique of wave me hani s). In se tion 6 of the published
paper, Bohr raises the following puzzle. A ording to Bohr, in any observation
that distinguishes between dierent stationary states one has to disregard the
past history of the atom, but, paradoxi ally, the theory assigns a phase to a
stationary state. However, sin e the system will not be stri tly isolated, one
will work with a group of waves, whi h (as mentioned in se tion 2) has no
well-dened phase. Bohr then illustrates this with the Stern-Gerla h experiment. The ondition for distinguishability of the eigenstates of the hydrogen
atom is that the angular spreading of the beam should be greater than that
given by dira tion at the slit (
> ),
whi h translates into the time-energy
un ertainty relation. As Bohr mentions, Heisenberg (1927) uses this as an

illustration of the un ertainty relation, while Bohr uses it as an illustration of
how knowledge of the phase is lost. (This se tion also dis usses the limit of
high quantum numbers.)
The nal se tion 7 of the paper (`The problem of elementary parti les') has
no parallel in Ri hardson's notes. The part of the notes relating to Bohr's
remarks at this point on ludes instead with the following (expli itly labelled
`Bohr'):
481
Fig. C.
1. [blank
2. Stationary states, past lost [be ause phase indetermination Stern
& Gerla h's Exp[eriment.
3. S hroedinger's , prob[ability of ele tron at a given pla e at a given
time [?, un ertainty t 1
v
c
Mr Brillouin
. Mr Bohr insists on the un ertainty of simultaneous

measurements of position and momentum; his point of view is losely
onne ted to the notion of ells in phase spa e introdu ed by Plan k a
very long time ago. Plan k assumed that if the representative point of
a system is in a ell (of size pq = h) one annot distinguish it from
another point in the same ell. The examples brought by Mr Bohr aptly
make pre ise the physi al meaning of this quite abstra t notion.
Mr De Donder
. The onsiderations that Mr Bohr has just developed are, I think, in lose relation with the following fa t: in the
482
General dis ussion
Einsteinian Gravitationa of a ontinuous system or of a pointlike system,

there appear not the masses and harges of the parti les, but entities
(m) and (e) in four dimensions; note that these generalised masses and
harges, lo alised in spa etime, are onserved along their worldlines.
Mr Born
. Mr Einstein has onsidered the following problem: A

radioa tive sample emits -parti les in all dire tions; these are made visible by the method of the Wilson loud [ hamber. Now, if one asso iates
a spheri al wave with ea h emission pro ess, how an one understand
that the tra k of ea h parti le appears as a (very nearly) straight line?
In other words: how an the orpus ular hara ter of the phenomenon
be re on iled here with the representation by waves?
To do this, one must appeal to the notion of `redu tion of the probability pa ket' developed by Heisenberg.a The des ription of the emission
by a spheri al wave is valid only for as long as one does not observe
ionisation; as soon as su h ionisation is shown by the appearan e of
loud droplets, in order to des ribe what happens afterwards one must
`redu e' the wave pa ket in the immediate vi inity of the drops. One
thus obtains a wave pa ket in the form of a ray, whi h orresponds to
the orpus ular hara ter of the phenomenon.
Mr Paulib has asked me if it is not possible to des ribe the pro ess without the redu tion of wave pa kets, by resorting to a multi-dimensional
spa e, whose number of dimensions is three times the number of all the
parti les present (-parti les and atoms hit by the radiation).
This is in fa t possible and an even be represented in a very ans hauli h manner [d'une manire fort intuitive by means of an appropriate
simpli ation, but this does not lead us further as regards the fundamental questions. Nevertheless, I should like to present this ase here as an
example of the multi-dimensional treatment of su h problems. I assume,
for simpli ity, that there are only two atoms that may be hit. One then
has to distinguish two ases: either the two atoms 1 and 2 lie on the
same ray starting from the origin (the pla e where the preparation is),
a Th. De Donder, Thorie des hamps graviques (Mmorial des s ien es
mathmatiques, part 14, Paris, 1926). See esp. equations (184), (184) and
(188), (188). One an also onsult our le tures: The Mathemati al Theory of
Relativity (Massa husetts Institute of Te hnology), Cambridge, Mass., 1927. See
esp. equations (23), (24) and (28), (29).
a Born is referring here in parti ular to Heisenberg's un ertainty paper (Heisenberg
1927) (eds.).
b Cf. Pauli's letter to Bohr, 17 O tober 1927, dis ussed in se tion 6.2.1 (eds.).
483
Fig. 1.
or they do not lie on the same ray. If we represent by the probability
that an atom will be hit, we have the following probability diagram:a
I. The points 1 and 2 are lo ated on the same ray starting from the
origin.
Number of parti les hit
Probability
0
1
2
1
0
a In the following tables, the probability for the number of parti les hit to equal 1
should be read as the probability for ea h ase in whi h the number of parti les
hit equals 1 (eds.).
484
General dis ussion
II. The points 1 and 2 are not on the same ray.

Probability
0
1
2
1 2
This is how one should express the probability of events in the ase of
re tilinear propagation.
To make possible a graphi al representation of the phenomenon, we
will simplify it further by assuming that all the motions take pla e
following only a single straight line, the axis x. We must then distinguish
the two ases where the atoms lie on the same side and on either side of
the origin. The orresponding probabilities are the following:
I. The points 1 and 2 are lo ated on the same side.
Probability
0
1
2
1
2
0
1
2
II. The points 1 and 2 are lo ated on dierent sides.

Probability
0
1
2
0
1
2
Now, these relations an be represented by the motion of a wave pa ket

in a spa e with three dimensions x0 , x1 , x2 . To the initial state there
orresponds:
In ase I, the point x0 = 0, x1 = a x2 = b
In ase II, the point x0 = 0, x1 = a x2 = b
where a and b are positive numbers. The wave pa ket at rst lls the
spa e surrounding these points and subsequently moves parallel to the
axis x0 , dividing itself into two pa kets of the same size going in opposite
dire tions. Collisions are produ ed when x0 = x1 or x0 = x2 , that is to
say, on two planes of whi h one, P1 , is parallel to the axis x2 and uts
the plane x0 x1 following the bise tor of the positive quadrant, while the
se ond, P2 , is parallel to the axis x1 and uts the plane x0 x2 following
the bise tor of the positive quadrant. As soon as the wave pa ket strikes
485
the plane P1 , its traje tory re eives a small kink in the dire tion x1 ; as
soon as it strikes P2 the traje tory re eives a kink in the dire tion x2
(Fig. 1).
Now, one immediately sees in the gure that the upper part of the
wave pa ket, whi h orresponds to ase I, strikes the planes P1 , P2 on
the same side of the plane x1 x2 , while the lower part strikes them5
on dierent sides. The gure then gives an ans hauli h representation
of the ases indi ated in the above diagram. It allows us to re ognise
immediately whether, for a given size of wave pa ket, a given state, that
is to say a given point x0 , x1 , x2 , an be hit or not.
To the `redu tion' of the wave pa ket orresponds the hoi e of one
of the two dire tions of propagation +x0 , x0 , whi h one must take as
soon as it is established that one of the two points 1 and 2 is hit, that
is to say, that the traje tory of the pa ket has re eived a kink.
This example serves only to make lear that a omplete des ription of
the pro esses taking pla e in a system omposed of several mole ules is
possible only in a spa e of several dimensions.
Mr Einstein.
Despite being ons ious of the fa t that I have not

entered deeply enough into the essen e of quantum me hani s, nevertheless I want to present here some general remarks.b
One an take two positions towards the theory with respe t to its
postulated domain of validity, whi h I wish to hara terise with the aid
of a simple example.
Let S be a s reen provided with a small opening O (Fig. 2), and P a
hemispheri al photographi lm of large radius. Ele trons impinge on S
in the dire tion of the arrows. Some of these go through O, and be ause of
the smallness of O and the speed of the parti les, are dispersed uniformly
over the dire tions of the hemisphere, and a t on the lm.
Both ways of on eiving the theory now have the following in ommon.
There are de Broglie waves, whi h impinge approximately normally on S
and are dira ted at O. Behind S there are spheri al waves, whi h rea h
the s reen P and whose intensity at P is responsible [massgebend for
what happens at P.c
a
a The extant manus ript in the Einstein ar hives6 onsists of the rst four paragraphs
only, whi h we have translated here (footnoting signi ant dieren es from the
published Fren h) (eds.).
b The published Fren h has: Ì must apologise for not having gone deeply into
quantum me hani s. I should nevertheless want to make some general remarks'
(eds.).
In the published Fren h, the German expression ìst massgebend' is misrendered
486
General dis ussion
Fig. 2.
We an now hara terise the two points of view as follows.
1. Con eption I. The de Broglie-S hrdinger waves do not orrespond
to a single ele tron, but to a loud of ele trons extended in spa e. The
theory gives no information about individual pro esses, but only about
the ensemble of an innity of elementary pro esses.
2. Con eption II. The theory laims to be a omplete theory of
individual pro esses. Ea h parti le dire ted towards the s reen, as far
as an be determined by its position and speed, is des ribed by a pa ket
of de Broglie-S hrdinger waves of short wavelength and small angular
width. This wave pa ket is dira ted and, after dira tion, partly rea hes
the lm P in a state of resolution [un tat de rsolution.
A ording to the rst, purely statisti al, point of view ||2 expresses
the probability that there exists at the point onsidered a parti ular
parti le of the loud, for example at a given point on the s reen.
2
A ording to the se ond, || expresses the probability that at a given
instant the same parti le is present at a given point (for example on the
s reen). Here, the theory refers to an individual pro ess and laims to
des ribe everything that is governed by laws.
as `donne la mesure' [gives the measure instead of as ìs responsible'. This is of
some signi an e for the interpretation of Einstein's remarks as a form of the later
EPR argument; see se tion 7.1 (eds.).
487
The se ond on eption goes further than the rst, in the sense that all
the information resulting from I results also from the theory by virtue
of II, but the onverse is not true.a It is only by virtue of II that the
theory ontains the onsequen e that the onservation laws are valid for
the elementary pro ess; it is only from II that the theory an derive
the result of the experiment of Geiger and Bothe, and an explain the
fa t that in the Wilson [ loud hamber the droplets stemming from an
-parti le are situated very nearly on ontinuous lines.
But on the other hand, I have obje tions to make to on eption II.
The s attered wave dire ted towards P does not show any privileged
2
dire tion. If || were simply regarded as the probability that at a ertain
point a given parti le is found at a given time, it ould happen that the
same elementary pro ess produ es an a tion in two or several pla es
2
on the s reen. But the interpretation, a ording to whi h || expresses
the probability that this parti le is found at a given point, assumes an
entirely pe uliar me hanism of a tion at a distan e, whi h prevents the
wave ontinuously distributed in spa e from produ ing an a tion in two
pla es on the s reen.
In my opinion, one an remove this obje tion only in the following
way, that one does not des ribe the pro ess solely by the S hrdinger
wave, but that at the same time one lo alises the parti le during the
propagation. I think that Mr de Broglie is right to sear h in this dire tion. If one works solely with the S hrdinger waves, interpretation II of
2
|| implies to my mind a ontradi tion with the postulate of relativity.
I should also like to point out briey two arguments whi h seem to me
to speak against the point of view II. This [view is essentially tied to a
multi-dimensional representation ( onguration spa e), sin e only this
2
mode of representation makes possible the interpretation of || pe uliar
to on eption II. Now, it seems to me that obje tions of prin iple are
opposed to this multi-dimensional representation. In this representation,
indeed, two ongurations of a system that are distinguished only by the
permutation of two parti les of the same spe ies are represented by two
dierent points (in onguration spa e), whi h is not in a ord with the
new results in statisti s. Furthermore, the feature of for es of a ting only
at small spatial distan es nds a less natural expression in onguration
spa e than in the spa e of three or four dimensions.
Mr Bohr.
I feel myself in a very di ult position be ause I don't
a The Fren h has Ì' and ÌI' ex hanged in this senten e, whi h is illogi al (eds.).
a These remarks by Bohr do not appear in the published Fren h. We have reprodu ed
488
General dis ussion
understand what pre isely is the point whi h Einstein wants to [make.
No doubt it is my fault.
....
As regards general problem I feel its di ulties. I would put problem
in other way. I do not know what quantum me hani s is. I think we
are dealing with some mathemati al methods whi h are adequate for
des ription of our experiments. Using a rigorous wave theory we are
laiming something whi h the theory annot possibly give. [We must
realise that we are away from that state where we ould hope of des ribing things on lassi al theories. Understand same view is held by
Born and Heisenberg. I think that we a tually just try to meet, as in
all other theories, some requirements of nature, but di ulty is that we
must use words whi h remind of older theories. The whole foundation for
ausal spa etime des ription is taken away by quantum theory, for it is
based on assumption of observations without interferen e. .... ex luding
interferen e means ex lusion of experiment and the whole meaning of
spa e and time observation .... be ause we [have intera tion [between
obje t and measuring instrument and thereby we put us on a quite
dierent standpoint than we thought we ould take in lassi al theories.
If we speak of observations we play with a statisti al problem. There
are ertain features omplementary to the wave pi tures (existen e of
individuals). ....
....
The saying that spa etime is an abstra tion might seem a philosophi al triviality but nature reminds us that we are dealing with something
of pra ti al interest. Depends on how I onsider theory. I may not have
understood, but I think the whole thing lies [therein that the theory
is nothing else [but a tool for meeting our requirements and I think it
does.
Mr Lorentz. To represent the motion of a system of n material
points, one an of ourse make use of a spa e of 3 dimensions with

n points or of a spa e of 3n dimensions where the systems will be
represented by a single point. This must amount to exa tly the same
thing; there an be no fundamental dieren e. It is merely a question of
them from Bohr's Colle ted Works, vol. 6 (Bohr 1985, p. 103), whi h ontains a
re onstru tion of Bohr's remarks from notes by Vers haelt (held in the Bohr
ar hive). The tentative interpolations in square bra kets are by the editor of Bohr
(1985), J. Kal kar (eds.).
489
knowing whi h of the two representations is the most suitable, whi h is

the most onvenient.
But I understand that there are ases where the matter is di ult. If
one has a representation in a spa e of 3n dimensions, one will be able to
return to a spa e of 3 dimensions only if one an reasonably separate the
3n oordinates into n groups of 3, ea h orresponding to a point, and I
ould imagine that there may be ases where that is neither natural nor
simple. But, after all, it ertainly seems to me that all this on erns the
form rather than the substan e of the theory.
Mr Pauli. I am wholly of the same opinion as Mr Bohr, when
he says that the introdu tion of a spa e with several dimensions is only
a te hni al means of formulating mathemati ally the laws of mutual
a tion between several parti les, a tions whi h ertainly do not allow
themselves to be des ribed simply, in the ordinary way, in spa e and
time. It may perfe tly well be that this te hni al means may one day
be repla ed by another, in the following fashion. By Dira 's method one
an, for example, quantise the hara teristi vibrations of a avity lled
with bla kbody radiation, and introdu e a fun tion depending on the
amplitudes of these hara teristi vibrations of unlimited number. One
an similarly use, as do Jordan and Klein, the amplitudes of ordinary
four-dimensional material waves as arguments of a multi-dimensional
fun tion . This gives, in the language of the orpus ular pi ture, the
probability that at a given instant the numbers of parti les of ea h spe ies present, whi h have ertain kinemati al properties (given position or
momentum), take ertain values. This pro edure also has the advantage
that the defe t of the ordinary multi-dimensional method, of whi h Mr
Einstein has spoken and whi h appears when one permutes two parti les
of the same spe ies, no longer exists. As Jordan and Klein have shown,
making suitable assumptions on erning the equations that this fun tion
of the amplitudes of material waves in ordinary spa e must satisfy,7 one
arrives exa tly at the same results as by basing oneself on S hrdinger's
multi-dimensional theory.
To sum up, I wish then to say that Bohr's point of view, a ording
to whi h the properties of physi al obje ts of being dened and of being
des ribable in spa e and time are omplementary, seems to be more
general than a spe ial te hni al means. But, independently of su h a
means, one an, a ording to this idea, de lare in any ase that the
mutual a tions of several parti les ertainly annot be des ribed in the
ordinary manner in spa e and time.
490
General dis ussion
To make lear the state of things of whi h I have just spoken, allow me
to give a spe ial example. Imagine two hydrogen atoms in their ground
state at a great distan e from ea h other, and suppose one asks for their
energy of mutual a tion. Ea h of the two atoms has a perfe tly isotropi
distribution of harge, is neutral as a whole, and does not yet emit
radiation. A ording to the ordinary des ription of the mutual a tion
of the atoms in spa e and time, one should then expe t that su h a
mutual a tion does not exist when the distan e between the two neutral
spheres is so great that no notable interpenetration takes pla e between
their harge louds. But when one treats the same question by the multidimensional method, the result is quite dierent, and in a ordan e with
experiment.
The lassi al analogy to this last result would be the following: Imagine
inside ea h atom a lassi al os illator whose moment p varies periodi ally. This moment produ es a eld at the lo ation of the other atom whose
periodi ally variable intensity is of order E rp3 , where r is the distan e
between the two atoms. When two of these os illators a t on ea h other,
a polarisation o urs with the following potential energy, orresponding
to an attra tive for e between the atoms,
1
1
1
E 2 p2 6 ,
2
2
r
where represents the polarisability of the atom.

In speaking of these os illators, I only wanted to point out a lassi al
analogy with the ee t that one obtains as a result of multi-dimensional
wave me hani s. I had found this result by means of matri es, but Wang
has derived it dire tly from the wave equation in several dimensions. In
a paper by Heitler and London, whi h is likewise on erned with this
problem, the authors have lost sight of the fa t that, pre isely for a large
distan e between the atoms, the ontribution of polarisation ee ts to
the energy of mutual a tion, a ontribution whi h they have negle ted,
outweighs in order of magnitude the ee ts they have al ulated.
Mr Dira .
I should like to express my ideas on a few questions.

The rst is the one that has just been dis ussed and I have not mu h
to add to this dis ussion. I shall just mention the explanation that the
a
a Here we mostly follow the English version from Dira 's manus ript.8 (The Fren h
translation may have been done from a types ript or fairer opy.) We generally
follow the Fren h paragraphing, and we uniformise Dira 's notation. Interesting
variants, an ellations and additions will be noted, as will signi ant deviations
from the published Fren h (eds.).
491
quantum theory would give of Bothe's experiment.9 The di ulty arises
from10 the inadequa y of the 3-dimensional wave pi ture. This pi ture
annot distinguish between the ase when there is a probability p of a
light-quant being in a ertain small volume, and the ase when there is
a probability 12 p of two light-quanta being in the volume, and no probability for only one. But the wave fun tion in many-dimensional spa e
does distinguish between these ases. The theory of Bothe's experiment
in many-dimensional spa e would show that, while there is a ertain probability for a light-quantum appearing in one or the other of the ounting
hambers, there is no probability of two appearing simultaneously.
At present the general theory of the wave fun tion in many-dimensional
spa e ne essarily involves the abandonment of relativity.11 One might,
perhaps, be able to bring relativity into the general quantum theory in
the way Pauli has mentioned of quantising 3-dimensional waves, but this
would not lead to greater Ans hauli hkeit12 in the explanation of results
su h as Bothe's.
I shall now show how S hrdinger's expression for the ele tri density
appears naturally in the matrix theory. This will show the exa t signi ation of this density and the limitations whi h must be imposed on
its use. Consider an ele tron moving in an arbitrary eld, su h as that
of an H atom. Its oordinates x, y, z will be matri es. Divide the spa e
up into a number of ells, and form that fun tion of x, y, z that is equal
to 1 when the ele tron is in a given ell and 0 otherwise. This fun tion
of the matri es x, y, z will also be a matrix.a There is one su h matrix
for ea h ell whose matrix elements will be fun tions of the oordinates
a, b, c of the ell, so that it an be written A(a, b, c).
Ea h of these matri es represents a quantity that if measured experimentally must have either the value 0 or 1. Hen e ea h of these matri es
has the hara teristi values 0 and 1 and no others. If one takes the two
matri es A(a, b, c) and A(a , b , c ), one sees that they must ommute,13
sin e one an give a numeri al value to both simultaneously; for example,
if the ele tron is known to be in the ell a, b, c, it will ertainly not be in
the ell a , b , c , so that if one gives the numeri al value 1 to A(a, b, c),
one must at the same time give the numeri al value 0 to A(a , b , c ).
We an transform ea h of the matri es A into a diagonal matrix A
a The published version has: `Divide the spa e up into a large number of small ells,
and onsider the fun tion of three variables , , that is equal to 1 when the
point , , is in a given ell and equal to 0 when the point is elsewhere. This
fun tion, applied to the matri es x, y , z , gives another matrix' (eds.).
492
General dis ussion
by a transformation14 of the type

A = BAB 1 .
Sin e all the matri es A(a, b, c) ommute,15 they an be transformed

simultaneously into diagonal matri es by a transformation of this type.
The diagonal elements of ea h matrix A (a, b, c) are its hara teristi
values, whi h are the same as the hara teristi values of A(a, b, c), that
is, 0 and 1.
Further, no two A matri es, su h as A (a, b, c) [and A (a , b , c ),
an both have 1 for the same diagonal element, as a simple argument
shows that A (a, b, c)+A (a , b , c ) must also have only the hara teristi
values 0 and 1. We an without loss of generality assume that ea h A
has just one diagonal element equal to 1 and all the others zero. By
transforming ba k, by means of the formula
A(a, b, c) = B 1 A (a, b, c)B ,
we now nd that the matrix elements of A(a, b, c) are of the form
1
A(a, b, c)mn = Bm
Bn ,
i.e. a fun tion of the row multiplied by a fun tion of the olumn.
It should be observed that the proof of this result is quite independent of equations of motion and quantum onditions. If we take these
1
and Bn are apart from onstants just
into a ount, we nd that Bm
S hrdinger's eigenfun tions m and n at the point a, b, c.
Thus S hrdinger's density fun tion m (x, y, z)m (x, y, z) is a16 diagonal element of the matrix A referring to a ell about the point x, y, z .
The true quantum expression for the density is the whole matrix. Its
diagonal elements give only the average density, and must not be used
when the density is to be multiplied by a dynami al variable represented
by a matrix.
I should now like to express my views on determinism and the nature of
the numbers appearing in the al ulations of the quantum theory, as they
appear to me after thinking over Mr Bohr's remarks of yesterday.17 In the
lassi al theory one starts from ertain numbers des ribing ompletely
the initial state of the system, and dedu es other numbers that des ribe
ompletely the nal state. This deterministi theory applies only to an
isolated system.
But, as Professor Bohr has pointed out, an isolated system is by
denition unobservable. One an observe the system only by disturbing
it and observing its rea tion to the disturban e. Now sin e physi s is
493
on erned only with observable quantities the deterministi lassi al

theory is untenable.18
In the quantum theory one also begins with ertain numbers and
dedu es others from them. Let us inquire into the distinguishing hara teristi s19 of these two sets of numbers. The disturban es that an
experimenter applies to a system to observe it are dire tly under his
ontrol, and are a ts of freewill by him. It is only the numbers that
des ribe these a ts of freewill that an be taken as initial numbers for a
al ulation in the quantum theory. Other numbers des ribing the initial
state of the system are inherently unobservable, and do20 not appear in
the quantum theoreti al treatment.
Let us now onsider the nal numbers obtained as the result of an
experiment. It is essential that the result of an experiment shall be a
permanent re ord. The numbers that des ribe su h a result must help to
not only des ribe the state of the world at the instant the experiment is
ended, but also help to des ribe the state of the world at any subsequent
time. These numbers des ribe what is ommon to all the events in a
ertain hain of ausally onne ted events, extending indenitely into
the future.
Take as an example a Wilson loud expansion experiment. The ausal
hain here onsists of the formation of drops of water round ions, the
s attering of light by these drops of water, and the a tion of this light on
a photographi plate, where it leaves a permanent re ord. The numbers
that form the result of the experiment des ribe all of the events in this
hain equally well and help to des ribe the state of the world at any time
after the hain began.
One ould perhaps extend the hain further into the past.21 In the
example one ould, perhaps, as ribe the formation of the ions to a
-parti le, so that the result of the experiment would be numbers des ribing the tra k of a -parti le. In general one tries with the help
of theoreti al onsiderations to extend the hain as far ba k into the
past as possible, in order that the numbers obtained as the result of
the experiment may apply as dire tly as possible to the pro ess under
investigation.22
This view of the nature of the results of experiments ts in very well
with the new quantum me hani s. A ording to quantum me hani s the
state of the world at any time is des ribable by a wave fun tion ,
whi h normally varies a ording to a ausal law, so that its initial value
determines its value at any later time. It may however happen that at a
494
General dis ussion
ertain time t1 , an be expanded in the form

X
=
cn n ,
n
where the n 's are wave fun tions of su h a nature that they annot
interfere with one another at any time subsequent to t1 . If su h is the
ase, then the world at times later than t1 will be des ribed not by but
by one of the n 's. The parti ular n that it shall be must be regarded
as hosen by nature.23 One may say that nature hooses whi h n it is
to be, as the only information given by the theory is that the probability
of any n being hosen is |cn |2 .24 The value of the sux n that labels
the parti ular n hosen may be the result of an experiment, and the
result of an experiment must always be su h a number. It is a number
des ribing an irrevo able hoi e of nature, whi h must ae t the whole
of the future ourse of events.a
As an example take the ase of a simple ollision problem. The wave
pa ket representing the in ident ele tron gets s attered in all dire tions.
One must take for the wave fun tion after the pro ess not the whole
s attered wave, but on e again a wave pa ket moving in a denite
dire tion. From the results of an experiment, by tra ing ba k a hain
of ausally onne ted events one ould determine in whi h dire tion the
ele tron was s attered and one would thus infer that nature had hosen
this dire tion. If, now, one arranged a mirror to ree t the ele tron
wave s attered in one dire tion d1 so as to make it interfere with the
ele tron wave s attered in another dire tion d2 , one would not be able
to distinguish between the ase when the ele tron is s attered in the
dire tion d2 and when it is s attered in the dire tion d1 and ree ted
ba k into d2 . One would then not be able to tra e ba k the hain of ausal
events so far, and one would not be able to say that nature had hosen
a dire tion as soon as the ollision o urred, but only [that at a later
time nature hose where the ele tron should appear. The25 interferen e
between the n 's ompels nature to postpone her hoi e.
a The last two senten es appear dierently in the published version: `The hoi e,
on e made, is irrevo able and will ae t the whole future state of the world. The
value of n hosen by nature an be determined by experiment and the results of
all experiments are numbers des ribing su h hoi es of nature'.
Dira 's notes ontain a similar variant written in the margin: `The value of n
hosen by nature may be determined by experiment. The result of every experiment
onsists of numbers determining one of these hoi es of nature, and is permanent
sin e su h a hoi e is irrevo able and ae ts the whole future state of the world'
(eds.).
Mr Bohr.
495
Quite see that one must go into details of pi tures,

if one wants to ontrol or illustrate general statements. I think still
that you may simpler put it in my way. Just this distin tion between
observation and denition allows to let the quantum me hani s appear
as generalisation. What does mean: get re ords whi h do not allow to
work ba kwards. Even if we took all mole ules in photographi plate one
would have losed system. If we tell of a re ord we give up denition of
plate. Whole point lies in that by observation we introdu e something
whi h does not allow to go on.
....
a
Mr Born
. I should like to point out, with regard to the onsiderations of Mr Dira , that they seem losely related to the ideas expressed
in a paper by my ollaborator J.26 von Neumann, whi h will appear
shortly. The author of this paper shows that quantum me hani s an be
built up using the ordinary probability al ulus, starting from a small
number of formal hypotheses; the probability amplitudes and the law of
their omposition do not really play a role there.
Mr Kramers
. I think the most elegant way to arrive at the results

of Mr Dira 's onsiderations is given to us by the methods he presented
in his memoir in the Pro . Roy. So ., ser. A, vol. 113, p. 621. Let us
onsider a fun tion of the oordinates q1 , q2 , q3 of an ele tron, that is
equal to 1 when the point onsidered is situated in the interior of a
ertain volume V of spa e and equal to zero for every exterior point, and
let us represent by (q, ) and (, q) the transformation fun tions that
allow us to transform a physi al quantity F , whose form is known as a
matrix (q , q ), into a matrix ( , ), 1 , 2 , 3 being the rst integrals
of the equation of motion. The fun tion f , written as a matrix (q , q ),
will then take the form f (q )(q q ), where (q q ) represents
Dira 's unit matrix. As a matrix ( , ), f will then take the form
Z
, q )dq f (q )(q q )dq (q , )
f ( , ) = (
Z
, q )dq (q , ) ,
(
=
V
the integral having to be extended over the whole of the onsidered

volume. The diagonal terms of f ( , ), whi h may be written in the
a Again, these remarks do not appear in the published Fren h and we have
reprodu ed them from Bohr's Colle ted Works (Bohr 1985, p. 105) (eds.).
496
General dis ussion
form
f () =
,
dq
will dire tly represent, in a ordan e with Dira 's interpretation of the
matri es, the probability that, for a state of the system hara terised by
given values of , the oordinates of the ele tron are those of a point
situated in the interior of V . As is nothing other than the solution of
S hrdinger's wave equation, we arrive at on e at the interpretation of
the expression under dis ussion.
Mr Heisenberg
. I do not agree with Mr Dira when he says

that, in the des ribed experiment, nature makes a hoi e. Even if you
pla e yourself very far away from your s attering material, and if you
measure after a very long time, you are able to obtain interferen e by
taking two mirrors. If nature had made a hoi e, it would be di ult
to imagine how the interferen e is produ ed. Evidently, we say that this
hoi e of nature an never be known before the de isive experiment
has been done; for this reason, we an make no real obje tion to this
hoi e, be ause the expression `nature makes a hoi e' then implies no
physi al observation. I should rather say, as I did in my last paper,
that the observer himself makes the hoi e,a be ause it is only at the
moment where the observation is made that the ` hoi e' has be ome a
physi al reality and that the phase relationship in the waves, the power
of interferen e, is destroyed.
Mr Lorentz
. There is then, it seems to me, a fundamental dieren e of opinion on the subje t of the meaning of these hoi es made by
nature.
To admit the possibility that nature makes a hoi e means, I think,
that it is impossible for us to know in advan e how phenomena will
take pla e in the future. It is then indeterminism that you wish to
ere t as a prin iple. A ording to you there are events that we annot
predi t, whereas until now we have always assumed the possibility of
these predi tions.
a From Heisenberg's publi ation re ord, it is lear that he is here referring to his
un ertainty paper, whi h had appeared in May 1927. There we nd the statement
that àll per eiving is a hoi e from a plenitude of possibilities' (Heisenberg 1927,
p. 197). When Heisenberg says, in his above omment on Dira , that the observer
`makes' the hoi e, he seems to mean this in the sense of the observer bringing
about the hoi e (eds.).
Photons
Mr Kramers
Photons
497
. During the dis ussion of Mr de Broglie's report, Mr

Brillouin explained to us how radiation pressure is exerted in the ase
of interferen e and that one must assume an auxiliary stress. But how
is radiation pressure exerted in the ase where it is so weak that there
is only one photon in the interferen e zone? And how does one obtain
the auxiliary tensor in this ase?
Mr de Broglie. The proof of the existen e of these stresses an
be made only if one onsiders a loud of photons.
Mr Kramers
. And if there is only one photon, how an one

a ount for the sudden hange of momentum suered by the ree ting
obje t?
Mr Brillouin. No theory urrently gives the answer to Mr Kra-
mers' question.
Mr Kramers
. No doubt one would have to imagine a ompli ated

me hanism, that annot be derived from the ele tromagneti theory of
waves?
Mr de Broglie
. The dualist representation by orpus les and asso iated waves does not onstitute a denitive pi ture of the phenomena.
It does not allow one to predi t the pressures exerted on the dierent
points of a mirror during the ree tion of a single photon. It gives only
the mean value of the pressure during the ree tion of a loud of photons.
Mr Kramers
. What advantage do you see in giving a pre ise value

to the velo ity v of the photons?
Mr de Broglie. This allows one to imagine the traje tory followed
by the photons and to spe ify the meaning of these entities; one an thus
onsider the photon as a material point having a position and a velo ity.
Mr Kramers. I do not very well see, for my part, the advantage
that there is, for the des ription of experiments, in making a pi ture
where the photons travel along well-dened traje tories.
498
Mr Einstein
General dis ussion
. During ree tion on a mirror, Mr L. de Broglie

assumes that the photons move parallel to the mirror with a speed c sin ;
but what happens if the in iden e is normal? Do the photons then have
zero speed, as required by the formula ( = 0)?
Mr Pi ard. Yes. In the ase of ree tion, one must assume that
the omponent of the velo ity of the photons parallel to the mirror is
onstant. In the interferen e zone, the omponent normal to the mirror
disappears. The more the in iden e in reases, the more the photons
are slowed down. One thus indeed arrives at stationary photons in the
limiting ase of normal in iden e.a
Mr Langevin
. In this way then, in the interferen e zone, the

photons no longer have the speed of light; they do not then always have
the speed c?
Mr de Broglie. No, in my theory the speed of photons is equal
to c only outside any interferen e zone, when the radiation propagates

freely in the va uum. As soon as there are interferen e phenomena, the
speed of the photons be omes smaller than c.
Mr De Donder
. I should like to show how the resear h of Mr L.

de Broglie is related to mine on some points.
By identifying the ten equations of the gravitational eld and the
four equations of the ele tromagneti eld with the fourteen equations
of the wave me hani s of L. Rosenfeld, I have obtainedb a prin iple of
orresponden e that laries and generalises that of O. Klein.c
In my prin iple of orresponden e, there appear the quantum urrent
and the quantum tensor. I will give the formulas for them later on; let
it su e to remark now that the example of orresponden e that Mr de
Broglie has expounded is in harmony with my prin iple.
Mr L. Rosenfeldd has given another example. Here, the mass is onserved and, moreover, one resorts to the quantum urrent. We add that
a Note that here the wave train is ta itly assumed to be limited longitudinally. Cf.
our dis ussion of the de Broglie-Pauli en ounter, se tion 10.2 (eds.).
b Bull. A . Roy. de Belgique, Cl. des S . (5) XIII, ns. 89, session of 2 August
1927, 5049. See esp. equations (5) and (8).
Zeits hr. f. Phys. 41, n. 617 (1927). See esp. equations (18), p. 414.
d L. Rosenfeld, `L'univers inq dimensions et la m anique ondulatoire (quatrime
ommuni ation)', Bull. A . Roy. Belg., Cl. des S ., O tober 1927. See esp.
paragraphs 4 and 5.
Photons
499
this model of quantisation is also in luded, as a parti ular ase, in our

prin iple of orresponden e.
Mr Lorentz has remarked, with some surprise, that the ontinuity
equation for harge is preserved in Mr de Broglie's example. Thanks
to our prin iple of orresponden e, and to Rosenfeld's ompatibility27
theorem, one an show that it will always be so for the total urrent
(in luding the quantum urrent) and for the theorem of energy and
momentum. The four equations that express this last theorem are satised by virtue of the two generalised quantum equations of de BroglieS hrdinger.
One further small remark, to end with. Mr de Broglie said that relativisti systems do not exist yet. I have given the theory of ontinuous
or holonomi systems.a But Mr de Broglie gives another meaning to
the word system; he has in mind intera ting systems, su h as the Bohr
atom, the system of three bodies, et . I have remarked re entlyb that the
quantisation of these systems should be done by means of a (ds)2 taken
in a onguration spa e with 4n dimensions, n denoting the number of
parti les. In a paper not yet published, I have studied parti ular systems
alled additive.
Mr Lorentz
. The stresses of whi h you speak and whi h you all

quantum, are they those of Maxwell?
Mr De Donder. Our quantum stresses must ontain the Maxwell
stresses as a parti ular ase; this results from the fa t that our prin iple
of orresponden e is derived (in part, at least) from Maxwell's equations,
and from the fa t that these quantum stresses here formally play the
same role as the stresses of ele trostri tionc in Einsteinian Gravity. Let
us re all, on this subje t, that our prin iple of orresponden e is also
derived from the fundamental equations of Einsteinian Gravity. Mr de
Broglie has, by means of his al ulations, thus re overed the stresses of
radiation.
a C. R. A ad. S . Paris, 21 February 1927, and Bull. A . Roy. Belgique, Cl. des
S ., 7 Mar h 1927.
b Bull. A . Roy. Belgique, Cl. des S ., 2 August 1927. See esp. form. (22).
For more details, see our Note: `L'le trostri tion dduite de la gravique
einsteinienne', Bull. A . Roy. Belgique, Cl. des S ., session of 9 O tober 1926,
6738.
500
General dis ussion
Mr Langevin makes a omparison between the old and modern stati-
sti s.
Formerly, one de omposed the phase spa e into ells, and one evaluated the number of representative points attributing an individuality to
ea h onstituent of the system.
It seems today that one must modify this method by suppressing the
individuality of the onstituents of the system, and substituting instead
the individuality of the states of motion. By assuming that any number
of onstituents of the system an have the same state of motion, one
obtains the statisti s of Bose-Einstein.
One obtains a third statisti s, that of Pauli-Fermi-Dira ,28 by assuming that there an be only a single representative point in ea h ell of
phase spa e.
The new type of representation seems more appropriate to the on eption of photons and parti les: sin e one attributes a omplete identity
of nature to them, it appears appropriate to not insist on their individuality, but to attribute an individuality to the states of motion.
In the report of Messrs Born and Heisenberg, I see that it results
from quantum me hani s that the statisti s of Bose-Einstein is suitable
for mole ules, that of Pauli-Dira for ele trons and protons. This means
that for photons29 and mole ules there is superposition, while for protons and ele trons there is impenetrability. Material parti les are then
distinguished from photons30 by their impenetrability.31
Mr Heisenberg
. There is no reason, in quantum me hani s, to

prefer one statisti s to another. One may always use dierent statisti s,
whi h an be onsidered as omplete solutions of the problem of quantum
me hani s. In the urrent state of the theory, the question of intera tion
has nothing to do with the question of statisti s.
We feel nevertheless that Einstein-Bose statisti s ould be the more
suitable for light quanta, Fermi-Dira statisti s for positive and negative
ele trons.a The statisti s ould be onne ted with the dieren e between
radiation and matter, as Mr Bohr has pointed out. But it is di ult to
establish a link between this question and the problem of intera tion. I
shall simply mention the di ulty reated by ele tron spin.
Mr Kramers reminds us of Dira 's resear h on statisti s, whi h

a That is, for protons and ele trons (eds.).
501
has shown that Bose-Einstein statisti s an be expressed in an entirely

dierent manner. The statisti s of photons, for example, is obtained by
onsidering a avity lled with bla kbody radiation as a system having
an innity of degrees of freedom. If one quantises this system a ording
to the rules of quantum me hani s and applies Boltzmann statisti s,
one arrives at Plan k's formula, whi h is equivalent to Bose-Einstein
statisti s applied to photons.
Jordan has shown that a formal modi ation of Dira 's method allows
one to arrive equally at a statisti al distribution that is equivalent to
Fermi statisti s. This method is suggested by Pauli's ex lusion prin iple.
Mr Dira
points out that this modi ation, onsidered from a

general point of view, is quite arti ial. Fermi statisti s is not established
on exa tly the same basis as Einstein-Bose statisti s, sin e the natural
method of quantisation for waves leads pre isely to the latter statisti s
for the parti les asso iated with the waves. To obtain Fermi statisti s,
Jordan had to use an unusual method of quantisation for waves, hosen
spe ially so as to give the desired result. There are mathemati al errors
in the work of Jordan that have not yet been redressed.
a
Mr Kramers
. I willingly grant that Jordan's treatment does not

seem as natural as the manner by whi h Mr Dira quantises the solution
of the S hrdinger equation. However, we do not yet understand why
nature requires this quantisation, and we an hope that one day we will
nd the deeper reason for why it is ne essary to quantise in one way in
one ase and in another way in the other.
Mr Born
. An essential dieren e between Debye's old theory, in

whi h the hara teristi vibrations of the bla kbody avity are treated
like Plan k os illators, and the new theory is this, that both yield quite
exa tly Plan k's radiation formula (for the mean density of radiation),
but that the old theory leads to inexa t values for the lo al u tuations
of radiation, while the new theory gives these values exa tly.
Mr Heisenberg
. A ording to the experiments, protons and ele trons both have an angular momentum and obey the laws of the statisti s of Fermi-Dira ; these two points seem to be related. If one takes
two parti les together, if one asks, for example, whi h statisti s one
a On this riti ism by Dira , f. Kragh (1990, pp. 12830) (eds.).
502
General dis ussion
must apply to a gas made up of atoms of hydrogen, one nds that the
statisti s of Bose-Einstein is the right one, be ause by permuting two
H atoms, we permute one positive ele tron and one negative ele tron,a
so that we hange the sign of the S hrdinger fun tion twi e. In other
words, Bose-Einstein statisti s is valid for all gases made up of neutral
mole ules, or more generally, omposed of systems whose harge is an
even multiple of e. If the harge of the system is an odd multiple of e,
the statisti s of Fermi-Dira applies to a olle tion of these systems.
The He nu leus does not rotate and a olle tion of He nu lei obeys
the laws of Bose-Einstein statisti s.
Mr Fowler
asks if the ne details of the stru ture of the bands of

helium agree better with the idea that we have only symmetri states
of rotation of the nu lei of helium than with the idea that we have only
antisymmetri states.
Mr Heisenberg
. In the bands of helium, the fa t that ea h se ond

line disappears tea hes us that the He nu leus is not endowed with a
spinning motion. But it is not yet possible to de ide experimentally,
on the basis of these bands, if the statisti s of Bose-Einstein or that of
Fermi-Dira must be applied to the nu leus of He.
Mr S hrdinger
. You have spoken of experimental eviden e in

favour of the hypothesis that the proton is endowed with a spinning
motion just like the ele tron, and that protons obey the statisti al law
of Fermi-Dira . What eviden e are you alluding to?
Mr Heisenberg. The experimental eviden e is provided by the
work of Dennisona on the spe i heat of the hydrogen mole ule, work
whi h is based on Hund's resear h on erning the band spe tra of hydrogen.
Hund found good agreement between his theoreti al s heme and the
experimental work of Dilke, Hopeld and Ri hardson, by means of the
hypotheses mentioned by Mr S hrdinger. But for the spe i heat, he
found a urve very dierent from the experimental urve. The experimental urve of the spe i heat seemed rather to speak in favour of
Bose-Einstein statisti s. But the di ulty was elu idated in the paper
by Dennison, who showed that the systems of `symmetri ' and àntia That is, we permute the two protons, and also the two ele trons (eds.).
a Pro . Roy. So . A 114 (1927), 483.
503
symmetri ' terms (with regard to protons) do not ombine in the time
ne essary to arry out the experiment. At low temperature, a transition
takes pla e about every three months. The ratio of statisti al weights
of the systems of symmetri and antisymmetri terms is 1 : 3, as in
the helium atom. But at low temperatures the spe i heat must be
al ulated as if one had a mixture of two gases, an òrtho' gas and a
`para' gas. If one wished to perform experiments on the spe i heat
with a gas of hydrogen, kept at low temperature for several months, the
result would be totally dierent from the ordinary result.
Mr Ehrenfest wishes to formulate a question that has some relation
to the re ent experiments by Mr Langmuir on the disordered motion of

ele trons in the ow of ele tri ity through a gas.
In the well-known Pauli ex lusion (Pauliverbot), one introdu es (at
least in the language of the old quantum theory) a parti ular in ompatibility relation between the quantum motions of the dierent parti les of
a single system, without speaking expli itly of the role possibly played
by the for es a ting between these parti les. Now, suppose that through
a small opening one allows parti les that, so to speak, do not exert for es
on ea h other, to pass from a large spa e into a small box bounded by quite rigid walls with a ompli ated shape, so that the parti les en ounter
the opening and leave the box only at the end of a su iently long time.
Before entering the box, if the parti les have almost no motion relative
to one another, the Pauli ex lusion intervenes. After their exit, will they
have very dierent energies, independently of the weakness of the mutual
a tion between the parti les? Or else what role do these for es play in the
produ tion of Pauli's in ompatibility ( hoi e of antisymmetri solutions
of the wave equation)?
Mr Heisenberg
. The di ulty with Mr Ehrenfest's experiment is

the following: the two ele trons must have dierent energies. If the energy
of intera tion of the two ele trons is very small, the time 1 required for
the ele trons to ex hange an appre iable amount of energy is very long.
But to nd experimentally whi h state, symmetri or antisymmetri ,
the system of the two ele trons in the box is in, we need a ertain time
2 whi h is at least 1/ , if h is the [energy dieren e between the
symmetri and antisymmetri states. Consequently, 1 2 and the
di ulty disappears.
Mr Ri hardson. The eviden e for a nu lear spin is mu h more
504
General dis ussion
omplete than Mr Heisenberg has just said. I have re ently had o asion
to lassify a large number of lines in the visible bands of the spe trum
of the H2 mole ule. One of the hara teristi features of this spe trum is
a rather pronoun ed alternation in the intensity of the su essive lines.
The intensities of the lines of this spe trum were re ently measured by
Ma Lennan, Grayson-Smith and Collins. Unfortunately, a large number
of these lines overlap with ea h other, so that the intensity measurements
must be a epted only with reservations.
But nevertheless, I think one an say, without fear of being mistaken,
that all the bands that are su iently well-formed and su iently free of
inuen es of the lines on ea h other (so that one an have onden e in
the intensity measurements) have lines, generally numbered 1, 3, 5, ...,
that are intrinsi ally three times more intense than the intermediate
lines, generally numbered 2, 4, 6, ... . By intrinsi intensity, I mean that
whi h one obtains after having taken into a ount the ee ts on the
intensity of temperature and quantum number (and also, of ourse, the
ee ts of overlap with other lines, where it is possible to take this into
a ount). In other words, I wish to say that the onstant c of the intensity
formula
m+ 1 2 h2

( 2)
1
e 8KkT
J =c m+
,
2
where m is the number of the line and K the moment of inertia of the
mole ule, is three times bigger for the odd-numbered lines than for the
even-numbered ones. This means that the ratio 3 : 1 applies, with an
a ura y of about 5%, for at least ve dierent vibration states of a
three-ele tron state of ex itation. It also applies to another state, whi h
is probably 31 P if the others are 33 P . It is also shown, but in a less
pre ise way, that it applies to two dierent vibration states of a state of
ex itation with four ele trons.
At present, then, there is a great deal of experimental eviden e that
this nu lear spin persists through the dierent states of ex itation of the
hydrogen mole ule.
Mr Langmuir
. The question has often been raised of a similarity

in the relation between light waves and photons on the one hand, and de
Broglie waves and ele trons on the other. How far an this analogy be
developed? There are many remarkable parallels, but also I should like
to see examined if there are no fundamental dieren es between these
relations. Thus, for example, an ele tron is hara terised by a onstant
505
harge. Is there a onstant property of the photon that may be ompared

with the harge of the ele tron? The speed of the ele tron is variable;
is that of the photon also? The ele tromagneti theory of light has
suggested a multitude of experiments, whi h have added onsiderably to
our knowledge. The wave theory of the ele tron explains the beautiful
results of Davisson and Germer. Can one hope that this theory will be
as fertile in experimental suggestions as the wave theory of light has
been?32
Mr Ehrenfest
. When one examines a system of plane waves of

ellipti ally polarised light, pla ing oneself in dierently moving oordinate systems, these waves show the same degree of ellipti ity whatever
system one pla es oneself in. Passing from the language of waves to
that of photons, I should like to ask if one must attribute an ellipti al
polarisation (linear or ir ular in the limiting ases) to ea h photon?
If the reply is armative, in view of the invarian e of the degree of
ellipti ity in relativity, one must distinguish as many spe ies of photons
as there are degrees of ellipti ity. That would yield, it seems to me, a
new dieren e between the photon and the spinning ele tron. If, on the
other hand, one wishes above all to retain the analogy with the ele tron,
as far as I an see one omes up against two di ulties:
1. How then must one des ribe linearly polarised light in the language of
photons? (It is instru tive, in this respe t, to onsider the way in whi h
the two linearly polarised omponents, emitted perpendi ularly to the
magneti eld by a ame showing the Zeeman ee t, are absorbed by a
se ond ame pla ed in a magneti eld with antiparallel orientation.)
Mr Zeeman, to whom I posed the question, was kind enough to perform the experiment about a year ago, and he was able to noti e that
the absorption is the same in parallel and antiparallel elds, as one ould
have predi ted, in fa t, by onsiderations of ontinuity.
2. For ele trons, whi h move always with a speed less than that of
light, the universality of the spin may be expressed as follows, that
one transforms the orresponding antisymmetri tensor into a system
of oordinates arried with the ele tron in its translational motion (àt
rest'). But photons always move with the speed of light!
Mr Compton
. Can light be ellipti ally polarised when the photon

has an angular momentum?
506
General dis ussion
Mr Ehrenfest
. Be ause the photons move with the speed of

light, I do not really understand what it means when one says that ea h
photon has a universal angular momentum just like an ele tron.
Allow me to remind you of yet another property of photons. When
two photons move in dire tions that are not exa tly the same, one an
say quite arbitrarily that one of the photons is a radio-photon and the
other a -ray photon, or inversely. That depends quite simply on the
moving system of oordinates to whi h one refers the pair of photons.
Mr Lorentz. Can you make them identi al by su h a transfor-
mation?
Mr Ehrenfest
. Perfe tly. If they move in dierent dire tions.

One an then give them the same olour by adopting a suitable frame of
referen e. It is only in the ase where their worldlines are exa tly parallel
that the ratio of their frequen ies remains invariant.
Mr Pauli
. The fa t that the spinning ele tron an take two orientations in the eld allowed by the quanta seems to invite us at rst to
ompare it to the fa t that there are, for a given dire tion of propagation
of the light quanta, two hara teristi vibrations of bla kbody radiation,
distinguished by their polarisation. Nevertheless there remain essential
dieren es between the two ases. While in relativity one des ribes waves
by a (real) sextuple ve tor Fix = Fxi , for the spinning ele tron one has
proposed the following two modes of des ription for the asso iated de
Broglie waves: 1. One des ribes these waves by two omplex fun tions
, (and so by four real fun tions); but these fun tions transform
in a way that is hardly intuitive during the hange from one system
of oordinates to another. That is the route I followed myself. Or else:
2. Following the example of Darwin, one introdu es a quadruple ve tor
with generally omplex omponents (and so eight real fun tions in total).
But this pro edure has the in onvenien e that the ve tor involves a
redundan y [indtermination, be ause all the veriable results depend
on only two omplex fun tions.
These two modes of des ription are mathemati ally equivalent, but
independently of whether one de ides in favour of one or the other, it
seems to me that one annot speak of a simple analogy between the
polarisation of light waves and the polarisation of de Broglie waves
asso iated with the spinning ele tron.
507
Another essential dieren e between ele trons and light quanta is this,
that between light quanta there does not exist dire t (immediate) mutual
a tion, whereas ele trons, as a result of their arrying an ele tri harge,
exert dire t mutual a tions on ea h other.
Mr Dira .
I should like to point out an important failure in the

analogy between the spin of ele trons and the polarisation of photons.
In the present theory of the spinning ele tron one assumes that one an
spe ify the dire tion of the spin axis of an ele tron at the same time as its
position, or at the same time as its momentum. Thus the spin variable
of an ele tron ommutes33 with its oordinate and with its momentum
variables. The ase is dierent for photons. One an spe ify a dire tion of
polarisation for plane mono hromati light waves, representing photons
of given momentum, so that the polarisation variable ommutes with
the momentum variables. On the other hand, if the position of a photon
is spe ied, it means one has an ele tromagneti disturban e onned to
a very small volume,34 and one annot give a denite polarisation, i.e.
a denite dire tion for the ele tri ve tor, to this disturban e. Thus the
polarisation variable of a photon does not ommute with its oordinates.
a
Mr Lorentz
. In these dierent theories, one deals with the probability . I should like to see quite learly how this probability an
exist when parti les move in a well-dened manner following ertain
laws. In the ase of ele trons, this leads to the question of motions in
the eld (de Broglie). But the same question arises for light quanta.
Do photons allow us to re over all the lassi al properties of waves? Can
one represent the energy, momentum and Poynting ve tor by photons?
One sees immediately that, when one has an energy density and energy
ow, if one wishes to explain this by photons then the number of photons
per unit volume gives the density, and the number of photons per se ond
that move a ross a unit surfa e gives the Poynting ve tor.
The photons will then have to move with a speed dierent from that of
light. If one wished to assign always the same speed c to the photons, in
some ases one would have to assume a superposition of several photon
urrents. Or else one would have to assume that the photons annot be
used to represent all the omponents of the energy-momentum tensor.
Some of the terms must be ontinuous in the eld. Or else the photons
are smeared out [fondus.
a Again, here we follow Dira 's original English (eds.).
508
General dis ussion
A related question is to know whether the photons an have a speed

dierent from that of light and whether they an even be at rest. That
would altogether displease me. Could we speak of these photons and of
their motion in a eld of radiation?
Mr de Broglie
. When I tried to relate the motion of the photons

to the propagation of the waves of the new me hani s, I did not worry
about putting this point of view in a ord with the ele tromagneti on eption of light waves, and I onsidered only waves of s alar hara ter,
whi h one has normally used until now.
Mr Lorentz
. One will need these waves for photons also. Are

they of a dierent nature than light waves? It would please me less to
have to introdu e two types of waves.
Mr de Broglie
. At present one does not know at all the physi al

nature of the -wave of the photons. Can one try to identify it with the
ele tromagneti wave? That is a question that remains open. In any ase,
one an provisionally try to develop a theory of photons by asso iating
them with waves .
Mr Lorentz. Is the speed of the wave equal to that of light?

Mr de Broglie. In my theory, the speed of photons is equal to
c, ex ept in interfering elds. In general, I nd that one must assign to

a moving orpus le a proper mass M0 given by the formula
r
h2 a
,
M0 = m20 2 2
4 c a
the fun tion a

a being al ulated at the point where the moving body
is lo ated at the given moment (a is the amplitude of the wave ). For
photons, one has
m0 = 0 .
Thus, when a photon moves freely, that is to say, is asso iated with an
ordinary plane wave, M0 is zero and, to have a nite energy, the photon
must have speed c. But, when there is interferen e, a
a be omes dierent
from zero, M0 is no longer zero and the photon, to maintain the same
energy, must have a speed less than c, a speed that an even be zero.
Mr Lorentz. The term
would be ome imaginary.
a
a
509
must be negative, otherwise the mass
Mr de Broglie
. In the orpus ular on eption of light, the existen e of dira tion phenomena o uring at the edge of a s reen requires
us to assume that, in this ase, the traje tory of the photons is urved.
The supporters of the emission theory said that the edge of the s reen
exerts a for e on the orpus le. Now, if in the new me hani s as I develop
it, one writes the Lagrange equations for the photon, one sees appear
on the right-hand side of these equations a term proportional to the
gradient of M0 .
This term represents a sort of for e of a new kind, whi h exists only
when the proper mass varies, that is to say, where there is interferen e.
It is this for e that will urve the traje tory of the photon when its wave
is dira ted by the edge of a s reen.
Furthermore, for a loud of photons the same Lagrange equations lead
one to re over the internal stresses pointed out by Messrs S hrdinger
and De Donder.a One nds, indeed, the relations

ik
T + ik = 0 ,
k
x
where the tensor T ik is the energy-momentum tensor of the orpus les

T ik = 0 ui uk .
The tensor ik , whi h depends on derivatives of the amplitude of the

wave and is zero when this amplitude is onstant, represents stresses
existing in the loud of orpus les, and these stresses allow us to re over
the value of the radiation pressure in the ase of ree tion of light by a
mirror.
The tensor T ik + ik is ertainly related to the Maxwell tensor but,
to see learly how, one would have to be able to larify the relationship
existing between the wave of the photons and the ele tromagneti
light wave.
Mr Pauli.
It seems to me that, on erning the statisti al results

of s attering experiments, the on eption of Mr de Broglie is in full
agreement with Born's theory in the ase of elasti ollisions, but that
it is no longer so when one also onsiders inelasti ollisions. I should
b
a Cf. S hrdinger (1927b) and De Donder's omments above (eds.).

b Cf. se tion 10.2 (eds.).
510
General dis ussion
like to illustrate this by the example of the rotator, whi h was already
mentioned by Mr de Broglie himself. As Fermia has shown, the treatment
by wave me hani s of the problem of the ollision of a parti le that moves
in the (x, y) plane and of a rotator situated in the same plane, may be
made lear in the following manner.b One introdu es a onguration
spa e of three dimensions, of whi h two oordinates orrespond to the
x and y of the olliding parti le, while as third oordinate one hooses
the angle of the rotator. In the ase where there is no mutual a tion
between the rotator and the parti le, the fun tion of the total system
is given by35
(x, y, ) = Ae2i[ h (px x+py y+p )t] ,
1
where one has put

p = m
h
2
(m = 0, 1, 2, ...) .
In parti ular, the sinusoidal os illation of the oordinate orresponds

to a stationary state of the rotator. A ording to Born, the superposition
of several partial waves of this type, orresponding to dierent values of
m and by onsequen e of p ,36 means that there is a probability dierent
from zero for several stationary states of the rotator, while a ording to
the point of view of Mr de Broglie, in this ase the rotator no longer has
a onstant angular velo ity and an also exe ute os illations in ertain
ir umstan es.
Now, in the ase of a nite energy of intera tion between the olliding
parti le and the rotator, if we study the phenomenon of the ollision by
means of the wave equation in the spa e (x, y, ), a ording to Fermi
the result an be interpreted very simply. Indeed, sin e the energy of
intera tion depends on the angle in a periodi manner and vanishes
at large distan es from the rotator, that is to say from the axis , in
the spa e (x, y, ) we are dealing simply with a wave that falls on a
grating and, in parti ular, on a grating that is unlimited in the dire tion
of the axis . At large distan es from the grating, waves ome out only in
xed dire tions in onguration spa e, hara terised by integral values
of the dieren e m m . Fermi has shown that the dierent spe tral
orders orrespond simply to the dierent possible ways of transferring
the energy of the olliding parti le to the rotator, or onversely. Thus to
a Zeits hr. f. Phys. 40 (1926), 399.
b See se tion 10.2 for a dis ussion of Fermi's argument (eds.).
511
ea h spe tral order of the grating orresponds a given stationary state

of the rotator after the ollision.
It is, however, an essential point that, in the ase where the rotator is
in a stationary state before the ollision, the in ident wave is unlimited
in the dire tion of the axis. For this reason, the dierent spe tral orders
of the grating will always be superposed at ea h point of onguration
spa e. If we then al ulate, a ording to the pre epts of Mr de Broglie,
the angular velo ity of the rotator after the ollision, we must nd that
this velo ity is not onstant. If one had assumed that the in ident wave
is limiteda in the dire tion of the axis , it would have been the same
before the ollision. Mr de Broglie's point of view does not then seem
to me ompatible with the requirement of the postulate of the quantum
theory, that the rotator is in a stationary state both before and after the
ollision.
To me this di ulty does not appear at all fortuitous or inherent in the
parti ular example of the rotator; in my opinion, it is due dire tly to the
ondition assumed by Mr de Broglie, that in the individual ollision pro ess the behaviour of the parti les should be ompletely determined and
may at the same time be des ribed ompletely by ordinary kinemati s
in spa etime. In Born's theory, agreement with the quantum postulate
is realised thus, that the dierent partial waves in onguration spa e,
of whi h the general solution of the wave equation after the ollision
is omposed, are appli able [indiques separately in a statisti al way.
But this is no longer possible in a theory that, in prin iple, onsiders it
possible to avoid the appli ation of notions of probability to individual
ollision pro esses.
Mr de Broglie
. Fermi's problem is not of the same type as that

whi h I treated earlier; indeed, he makes onguration spa e play a part,
and not ordinary spa e.
The di ulty pointed out by Mr Pauli has an analogue in lassi al
opti s. One an speak of the beam dira ted by a grating in a given
dire tion only if the grating and the in ident wave are laterally limited,
be ause otherwise all the dira ted beams will overlap and be bathed in
the in ident wave. In Fermi's problem, one must also assume the wave
to be limited laterally in onguration spa e.
a The Fren h reads ìllimite' [unlimited, whi h we interpret as a misprint. Pauli
seems to be saying that if, on the other hand, the in ident wave had been taken as
limited, then before the ollision the rotator ould not have been in a stationary
state and its angular velo ity ould not have been onstant (eds.).
512
General dis ussion
Mr Lorentz
. The question is to know what a parti le should do

when it is immersed in two waves at the same time.
Mr de Broglie
. The whole question is to know if one has the right

to assume the wave to be limited laterally in onguration spa e. If
one has this right, the velo ity of the representative point of the system
will have a onstant value, and will orrespond to a stationary state
of the rotator, as soon as the waves dira ted by the -axis will have
separated from the in ident beam.
One an say that it is not possible to assume the in ident beam to
be limited laterally, be ause Fermi's onguration spa e is formed by
the superposition of identi al layers of height 2 in the dire tion of the
-axis; in other words, two points of onguration spa e lying on the
same parallel to the -axis and separated by a whole multiple of 2
represent the same state of the system. In my opinion, this proves above
all the arti ial hara ter of onguration spa es, and in parti ular of
that whi h one obtains here by rolling out along a line the y li variable
.
Mr De Donder
. In the ourse of the dis ussion of Mr L. de

Broglie's report, we explained how we obtained our Prin iple of Corresponden e; thanks to this prin iple, one will havea
c X h an
.n 2 e a ,
.n
g
gK 2
e n 2i
c

X X a b
= g

. . + . . ab L
(e) ua + a =
(m) ua ub + ab
(a, b, n = 1, ..., 4; , = 0, 1, ..., 4) .
The rst relation represents the total urrent ( ele troni urrent
+ quantum urrent) as a fun tion of and of the potentials g an , a .
a I adopt here L. Rosenfeld's notation, so as to fa ilitate the omparison with his

formulas, given later.
513
Re all that one has set
XX

1
. . + k 2 2
,
2
ab g ab , 0a a , 00 2 a a
2 2 ,
8G
,
c2
1
,
G 6.7 108 c.g.s.
We have already mentioned the examples (or models) of orresponden e found respe tively by L. de Broglie and L. Rosenfeld. To be able
to show learly a new solution to the problem relating to photons that
Mr L. de Broglie has just posed, I am going to display the formulas
on erning the two above-mentioned models.a
a L. Rosenfeld, `L'univers inq dimensions et la m anique ondulatoire', Bull. A .

Roy. Belgique, Cl. des S ., O tober 1927. See respe tively the formulas (*38),
(*31), (*27), (21), (1), (8), (35), (28), (29), (*35).
514
General dis ussion

Model of L. de Broglie.
Quantum urrent a 0.
Model of L. Rosenfeld.
Quantum urrent a = 2K 2 A2 C.a ,
where A is the modulus of
and where the potential C
S S . The fun tion S satises
the lassi al Ja obi equation; the
fun tion S satises the modied
Ja obi equation; one then has
1
,
2
1
A
= 2
+ 2 .
2 K A
S. S. = 2
S.
S.
The quantum potential C produ es

the dieren e between physi al
quantisation and geometri al quantisation.
2
Re all that m0ec , where m0 and
e are respe tively the mass (at rest)
and harge of the parti le under
onsideration. We have also put
k iK i
2 e
.
h c
Charge density (e) = 2K 2 A2 , Charge density (e) = 2K 2 A2 .

where we have put
Here then one retains, at the same
time,
the mass m0 and the harge e.
A
2 = 2 + 2 ,
K A
whi h, retaining the harge e,
redu es to substituting for the
mass m0 the modied mass of
L. de Broglie:
r
h2 A
M0 m20 + 2 2 .
4 c A
Let us respe tively apply these formulas to the problem of the photon
pointed out by Mr L. de Broglie. The proper mass m0 of the photon is
zero; in the model of Mr L. de Broglie, this mass must be repla ed by the
modied mass M0 ; on the ontrary, in the model of Mr L. Rosenfeld, one
515
uses only the proper mass m0 0. In the two models, the harge density
(e) is zero. Finally, in the rst model, the speed of the photon must
vary; in ontrast, in the se ond model, one an assume that this speed is
always that of light. These on lusions obviously speak in favour of the
model of L. Rosenfeld, and, in onsequen e, also in favour of the physi al
existen e of our quantum urrent a (a = 1, 2, 3, 4). This urrent will
probably play a dominant role in still unexplained opti al phenomena.a
Mr Lorentz
. Let us take an atom of hydrogen and let us form

the S hrdinger fun tion .37 We onsider as the probability for the
presen e of the ele tron in a volume element. Mr Born has mentioned all
the traje tories in the lassi al theory: let us take them with all possible
phases,a but let us now take the orresponding to a single value Wn of
energy and then let us form . Can one say that this produ t n n
represents the probability that the ele trons move with the given energy
Wn ? We think that the ele tron annot es ape from a ertain sphere. The
atom is limited, whereas extends to innity. That is disagreeable.38
Mr Born. The idea that
represents a probability density has

great importan e in appli ations. If, for example, in the lassi al theory
an ele tron had two equilibrium positions separated by a onsiderable
potential energy, then lassi ally, for a su iently weak total energy only
one os illation ould ever take pla e, around one of the two equilibrium
positions. But a ording to quantum me hani s, ea h eigenfun tion extends from one domain into the other; for this reason there always exists
a probability that a parti le, whi h at rst vibrates in the neighbourhood
of one of the equilibrium positions, jumps to the other. Hund has made
important appli ations of this to mole ular stru ture. This phenomenon
probably also plays a role in the explanation of metalli ondu tion.
Mr de Broglie. In the old theory of the motion of an ele tron in
the hydrogen atom, an ele tron of total energy

e2
m 0 c2
W = p
r
1 2
a On this subje t, Mr L. Brillouin has kindly drawn my attention to the experiments

by Mr F. Wolfers: `Sur un nouveau phnomne en optique: interfren es par
diusion' (Le Journal de Physique et le Radium (VI) 6, n. 11, November 1925,
35468).
a The `phases' of lassi al traje tories seems to be meant in the sense of a tion-angle
variables (eds.).
516
General dis ussion
annot es ape from a sphere of radius

R=
e2
W m 0 c2
be ause the value of the term m0 c
1 2
has m0 c2 as a lower limit.
In my on eption one must take
e2
M 0 c2
,
W = p
r
1 2
as the expression for the energy, where M0 is the variable proper mass
whi h I have already dened. Cal ulation shows that the proper mass
M0 diminishes when r in reases, in su h a way that an ele tron of energy
W is no longer at all onstrained to be in the interior of a sphere of radius
R.
Mr Born
. Contrary to Mr S hrdinger's opinion, that it is nonsense to speak of the lo ation and motion of an ele tron in the atom, Mr
Bohr and I are of the opinion that this manner of speaking always has
a meaning when one an spe ify an experiment allowing us to measure
the oordinates and the velo ities with a ertain approximation.
Again in Ri hardson's notes on the general dis ussion ( f. p. 478), the following
text together with Fig. D (both labelled `Bohr'), and a similar gure with the
shaded region labelled `B', appear immediately after notes on De Donder's
lengthy exposition just above, and learly refer to remarks Bohr made on the
topi being addressed here:
B[ohr says it has no point to worry about the paradox that the ele tron in the atom is in a xed path (ellipse or ir le) and the probability
that it should be found in a given pla e is given by the produ t whi h
is a ontinuous fun tion of spa e extending from zero to . He says if
we take a region su h as B a long way from the atom in order to nd
if the ele tron is there we must illuminate it with long light waves and
the frequen y of these is so low that the ele tron is out of the region by
reason of its motion in the stationary state before it has been illuminated
long enough for the photoele tri a t to o ur. I am really not sure if
this is right. But, anyway, it is no obje tion to pulling it out with an
intense stati ele tri eld & this appears to be what is happening in
the W experiments.
517
Fig. D.
Mr Pauli
. One an indeed determine the lo ation of the ele tron

outside the sphere, but without modifying its energy to the point where
an ionisation of the atom o urs.
Mr Lorentz
. I should like to make a remark on the subje t of

wave pa kets.a
When Mr S hrdinger drew attention to the analogy between me hani s and opti s, he suggested the idea of passing from orpus ular
me hani s to wave me hani s by making a modi ation analogous to
that whi h is made in the passage from geometri al opti s to wave
opti s.39 The wave pa ket gave a quite striking pi ture of the ele tron,
but in the atom the ele tron had to be ompletely smeared out [fondu,
the pa ket having the dimensions of the atom. When the dimensions of
the wave pa ket be ome omparable to those of the traje tories of the
lassi al theory, the material point would start to spread; having passed
this stage, the ele tron will be ompletely smeared out.
The mathemati al di ulty of onstru ting wave pa kets in the atom
is due to the fa t that we do not have at our disposal wavelengths
su iently small or su iently lose together. The frequen ies of stable
waves in the atom (eigenvalues) are more or less separated from ea h
other; one annot have frequen ies very lose together orresponding to
states diering by very little, be ause the onditions at innity would not
a Cf. also the dis ussion of the Lorentz-S hrdinger orresponden e in se tion 4.3
(eds.).
518
General dis ussion
be satised. To onstru t a pa ket, one must superpose waves of slightly

dierent wavelengths; now, one an use only eigenfun tions n , whi h
are sharply dierent from ea h other. In atoms, then, one annot have
wave pa kets. But there is a di ulty also for free ele trons, be ause in
reality a wave pa ket does not, in general, retain its shape in a lasting
manner. Lo alised [limits wave pa kets do not seem able to maintain
themselves; spreading takes pla e. The pi ture of the ele tron given by
a wave pa ket is therefore not satisfying, ex ept perhaps during a short
enough time.
What Mr Bohr does is this: after an observation he again lo alises
[limite the wave pa ket so as to make it represent what this observation
has told us about the position and motion of the ele tron; a new period
then starts during whi h the pa ket spreads again, until the moment
when a new observation allows us to arry out the redu tion again. But
I should like to have a pi ture of all that during an unlimited time.40
Mr S hrdinger
. I see no di ulty at all in the fa t that on orbits

of small quantum number one ertainly annot onstru t wave pa kets
that move in the manner of the point ele trons of the old me hani s.
The fa t that this is impossible is pre isely the salient point of the
wave me hani al view, the basis of the absolute powerlessness of the old
me hani s in the domain of atomi dimensions. The original pi ture was
this, that what moves is in reality not a point but a domain of ex itation
of nite dimensions, in parti ular at least of the order of magnitude of a
few wavelengths. When su h a domain of ex itation propagates along a
traje tory whose dimensions and radii of urvature are large ompared
with the dimensions of the domain itself, one an abstra t away the
details of its stru ture and onsider only its progress along the traje tory.
This progress takes pla e following exa tly the laws of the old me hani s.
But if the traje tory shrinks until it be omes of the order of magnitude
of a few wavelengths, as is the ase for orbits of small quantum number,
all its points will be ontinually inside the domain of ex itation and one
an no longer reasonably speak of the propagation of an ex itation along
a traje tory, whi h implies that the old me hani s loses all meaning.
That is the original idea. One has sin e found that the naive identi ation of an ele tron, moving on a ma ros opi orbit, with a wave
pa ket en ounters di ulties and so annot be a epted to the letter.
The main di ulty is this, that with ertainty the wave pa ket spreads
in all dire tions when it strikes an obsta le, an atom for example. We
know today, from the interferen e experiments with athode rays by
519
Davisson and Germer, that this is part of the truth, while on the other
hand the Wilson loud hamber experiments have shown that there must
be something that ontinues to des ribe a well-dened traje tory after
the ollision with the obsta le. I regard the ompromise proposed from
dierent sides, whi h onsists of assuming a ombination of waves and
point ele trons, as simply a provisional manner of resolving the di ulty.
Mr Born
. Also in the lassi al theory, the pre ision with whi h the
future lo ation of a parti le an be predi ted depends on the a ura y
of the measurement of the initial lo ation. It is then not in this that
the manner of des ription of quantum me hani s, by wave pa kets, is
dierent from lassi al me hani s. It is dierent be ause the laws of
propagation of pa kets are slightly dierent in the two ases.
520
Notes to pp. 476500

1 Mi rolmed in AHQP-BMSS-11, se tion 5.
2 AHQP-36, se tion 10.
3 These notes are to be found in the Ri hardson olle tion in Houston,
in luded with the opy of Born and Heisenberg's report (mi rolmed in
AHQP-RDN, do ument M-0309).
4 In luded in AHQP-RDN, do ument M-0309.
5 Fren h edition: `les' is misprinted as `le'.
6 AEA 16-617.00 (in German, with trans ription and ar hival omments).
7 The Fren h text has ` ' instead of `', and `doit satisfaire dans l'espa e
ordinaire' instead of the other way round. Note that
is a fun tional of
`material' waves whi h themselves propagate in ordinary spa e.

8 AHQP-36, se tion 10.
9 The Fren h adds: `d rite par M. Compton'.
10 The Fren h reads: `provient uniquement de'.
11 Dira 's manus ript omits Àt present'.
12 The Fren h reads ìntuitivit'.
13 Instead of ` ommute' the Fren h has `permuter leurs valeurs'.
14 The Fren h reads: `transformation anonique'.
15 Instead of ` ommute' the Fren h has ` hangent de valeur'.
16 Dira 's manus ript reads `the'.
17 In Dira 's manus ript, the words `determinism and' are an elled and
possibly reinstated. They appear in the Fren h, whi h also omits òf
yesterday'.
18
<therefore
unsatisfa tory>
<untenable>,
the latter seems reinstated.
The Fren h has ìndfendable'.

19 Instead of `the distinguishing hara teristi s' the Fren h has `l'essen e
physique'.
20
<do>,
{would} appears above the line, { an} below. The Fren h reads
`ne gurent pas'.

21 This senten e does not appear in the Fren h.
22 In the Fren h, this senten e appears at the beginning of the paragraph.
23 This senten e does not appear in the Fren h.
2
24 Dira 's manus ript has `cn '.
25
<Thus <a
possibility> {the existen e} of> .
26 The Fren h has `F.'.

27 Misprinted as ` omptabilit', despite having been orre ted in the galley
proofs.
28 In the printed text, the word `Dira ' is mispla ed to later in the
paragraph.
29 Misprinted as `protons'.
30 Again misprinted as `protons'.
31 The version of this ontribution in the galley proofs reads as follows:
Mr Langevin makes a omparison between the old and modern

statisti s.
Formerly, one de omposed the phase spa e, into ells and one evaluated
the representative points.
Notes to pp. 504518
521
It seems that one must modify this method by suppressing the

individuality of the representative points and [blank
Third method: that of Pauli.
This type of representation seems more appropriate to the on eption of
photons and parti les [blank attribute identity of nature, attribute at the
same time individuality representing a state.
In the report of Messrs Born and Heisenberg, I see that it results from
quantum me hani s that the statisti s of Bose-Einstein is suitable for
mole ules, that of Pauli-Dira , instead, is suitable for ele trons. This
means that for [blank there is superposition, while for photons and
ele trons there is impenetrability.
32 The galley proofs ontain the following version of this ontribution:
Mr Langmuir would like to see established learly a parallel between

ele trons and photons. What hara terises an ele tron? A well-dened
harge. What hara terises the photon? Its velo ity, perhaps? What is
the analogy, what are the dieren es? Ele tron: de Broglie waves; photon:
ele tromagneti waves. For ertain respe ts, this parallelism is lear, but
perhaps it an be pursued to the end? What are the suggestions in the
way of experiments?
33 The Fren h renders ` ommute' throughout with ` hanger'.
34 The Fren h adds: ` un instant donn'.
35 `h' misprinted as `'.
36 `p ' misprinted as `'.
37 ` ' missing in the original, with a spa e instead.
38 Here the galley proofs in lude an additional senten e:
If one took the integral extended over the whole of this spa e, the
exterior part would be omparable.
39 The original mistakenly reads `geometri al me hani s' and ` orpus ular
opti s'.
40 The version in the galley proofs reads as follows. (Note that in the ase of
this and the pre eding ontribution by Lorentz in the galley proofs, the
published version was learly not edited by him, sin e he had died at the
beginning of February.)
Mr Lorentz. I should like to make a remark on the subje t of wave

pa kets.
When Mr S hrdinger drew attention to the analogy between me hani s

and opti s, he suggested the idea of passing from geometri al me hani s
to wave me hani s by making a modi ation analogous to that whi h is
made in the passage from orpus ular opti s to wave opti s. The wave
pa ket was a quite striking pi ture, but in the atom the ele tron is
ompletely smeared out, the pa ket being of the dimensions of the
522
Notes to pp. 518518
atom [blank, material point that would start to spread [blank, passed,
these ele trons are ompletely smeared out.
Mathemati al di ulty, wave pa kets in the atom, more or less
distinguished frequen ies (eigenvalues), but you ould not have
frequen ies very lose together by states diering by mu h or little [par
des tats tant soit peu dirants, be ause one would not have the
onditions at innity. To onstru t a pa ket, one must superpose waves of
slightly dierent wavelengths; now, one an use only eigenfun tions n ,
whi h are sharply dierent from ea h other. Thus one does not have the
waves with whi h one ould build a pa ket. In atoms, then, one annot
have the wave pa kets; it is the same for free ele trons. All these wave
pa kets will end up dissolving.
In reality a wave pa ket does not last; wave pa kets that would remain
lo alised [limits do not seem to maintain themselves; spreading takes
pla e; the pi ture is therefore not satisfying [blank, short enough time
perhaps [blank.
What Mr Bohr does is this [small blank after an observation we have
again lo alised [limit [blank; a new period starts [blank. But I should
like to have a pi ture of all that during an indenite time.
Bibliography
Aharonov, Y., and Vaidman, L. (1996). About position measurements whi h

do not show the Bohmian parti le position. In Cushing, Fine and
Goldstein (1996), pp. 14154.
Ba iagaluppi, G. (2005). The role of de oheren e in quantum theory. In
Stanford En y lopedia of Philosophy,
The
Summer 2005 edn., ed. E. N. Zalta.
http://plato.stanford.edu/ar hives/sum2005/entries/qm-de oheren e .

Ballentine, L. E. (1970). The statisti al interpretation of quantum me hani s.
Reviews of Modern Physi s, 42,
35881.
Ballentine, L. E. (1972). Einstein's interpretation of quantum me hani s.
Ameri an Journal of Physi s, 40,
176371.
Ballentine, L. E. (1986). Probability theory in quantum me hani s.
Journal of Physi s, 54,
Ameri an
8839.
Ballentine, L. E. (1987). Resour e letter IQM-2: Foundations of quantum

me hani s sin e the Bell inequalities.
Ameri an Journal of Physi s, 55,
78592.
Ballentine, L. (2003). The lassi al limit of quantum me hani s and its
Quantum
impli ations for the foundations of quantum me hani s. In
Theory: Re onsideration of Foundations 2,
ed.
A.
Khrennikov,
Mathemati al modelling in physi s, engineering and ognitive s ien e,

vol. 10. Vxj: Vxj University Press, pp. 7182.
Barbour, J. B. (1994a). The timelessness of quantum gravity, I. The eviden e
from the lassi al theory.
Barbour,
J. B.
(1994b).
Classi al and Quantum Gravity, 11,
The
timelessness
of
appearan e of dynami s in stati ongurations.
Gravity, 11,
285373.
quantum gravity,
II. The
Classi al and Quantum
287597.
Bell, J. S. (1964). On the Einstein-Podolsky-Rosen paradox.
Physi s, 1,
195
200. Reprinted in Bell (1987), pp. 1421.

Bell, J. S. (1966). On the problem of hidden variables in quantum me hani s.
44752. Reprinted in Bell (1987), pp. 1
13.
Bell, J. S. (1980). De Broglie-Bohm, delayed- hoi e double-slit experiment,
and density matrix.
International Journal of Quantum Chemistry, 14,
1559. Reprinted in Bell (1987), pp. 11116.

Bell, J. S. (1984). Beables for quantum eld theory. CERN-TH. 4035/84.
Reprinted in Bell (1987), pp. 17380.
Bell, J. S. (1986). Interview. In
The Ghost in the Atom,
523
eds. P. C. W. Davies
524
Bibliography
and J. R. Brown. Cambridge: Cambridge University Press, pp. 4557.
Bell,
J.
S.
(1987).
Speakable and Unspeakable in Quantum Me hani s.
Cambridge: Cambridge University Press.

Bell, J. S. (1990). Against `measurement'. In Miller (1990), pp. 1731.
Beller, M. (1999).
Quantum Dialogue: The Making of a Revolution.
Chi ago:
University of Chi ago Press.

Belousek, D. W. (1996). Einstein's 1927 unpublished hidden-variable theory:
its
ba kground,
ontext
and
Studies in History and
signi an e.
Philosophy of Modern Physi s, 27,
43761.
Berndl, K., Drr, D., Goldstein, S., and Zangh, N. (1996). Nonlo ality,
Lorentz invarian e, and Bohmian quantum theory.
53, 206273.
Berry, M. (1997). Slippery as an eel.

412.
Bjorken, J. D., and Drell, S. D. (1964).
York: M Graw-Hill.
Bhme, K. (ed.) (1975).
Weltkrieg.
Physi al Review A,
Physi s World, 10
(De ember 1997),
Relativisti Quantum Me hani s.
New
Aufrufe und Reden deuts her Professoren im ersten
Stuttgart: Re lam.
Bohm, D. (1952a). A suggested interpretation of the quantum theory in terms

of `hidden' variables, I.
Physi al Review, 85,
16679.
Bohm, D. (1952b). A suggested interpretation of the quantum theory in terms

of `hidden' variables, II.
18093.
Bohm, D. (1953). Proof that probability density approa hes

interpretation of the quantum theory.
||2
in ausal
45866.
Bohm, D., Dewdney, C., and Hiley, B. J. (1985). A quantum potential
Nature, 315, 2947.

The Undivided Universe: An Ontologi al
Interpretation of Quantum Theory. London: Routledge.
approa h to the Wheeler delayed- hoi e experiment.

Bohm, D., and Hiley, B. J. (1993).
Bohm, D., and Vigier, J. P. (1954). Model of the ausal interpretation of

quantum theory in terms of a uid with irregular u tuations.
Review, 96,
20816.
Bohr, N. (1925). ber die Wirkung von Atomen bei Sten.
Physik, 34,
Physi al
Zeits hrift fr
14257.
Bohr, N. (1928). Das Quantenpostulat und die neuere Entwi klung der
Atomistik.
Die Naturwissens haften, 16,
24557.
Bohr, N. (1949). Dis ussion with Einstein on epistemologi al problems in

atomi physi s. In S hilpp (1949), pp. 20141.
Bohr, N. (1984).
Niels Bohr: Colle ted Works,
vol. 5, ed. K. Stolzenburg.
Amsterdam: North-Holland.
Bohr,
N.
(1985).
Niels Bohr: Colle ted Works,
vol.
6,
ed.
J.
Kal kar.
Amsterdam: North-Holland.
Bohr, N., Kramers, H. A., and Slater, J. C. (1924a). The quantum theory
Philosophi al Magazine, 47, 785802. Reprinted in Sour es

of quantum me hani s, ed. B. L. van der Waerden. Amsterdam: North-
of radiation.
Holland, 1967, pp. 15976.

Bohr, N., Kramers, H. A., and Slater, J. C. (1924b). ber die Quantentheorie
der Strahlung.
Zeits hrift fr Physik, 24,
6987.
Studies in the
History and Philosophy of S ien e, 25, 191209.
Born, M. (1924). ber Quantenme hanik. Zeits hrift fr Physik, 26, 37995.
Born, M. (1926a). Zur Quantenme hanik der Stossvorgnge. Zeits hrift fr
Bonk, T. (1994). Why has de Broglie's theory been reje ted?
Bibliography
Physik, 37,
525
8637.
Born, M. (1926b). Quantenme hanik der Stossvorgnge.
38, 80327.
Zeits hrift fr Physik,
Born, M. (1926 ). Das Adiabatenprinzip in der Quantenme hanik.
fr Physik, 40,
Born,
M.
Zeits hrift
16792.
(1926d).
Problems of Atomi Dynami s.
Cambridge,
Mass.:
Massa husetts Institute of Te hnology. Reprinted Cambridge, Mass.: MIT

Press, 1970.
Probleme der Atomdynamik.
Born, M. (1926e).
Berlin: Springer.
Born, M. (1927). Physi al aspe ts of quantum me hani s.

Born, M. (1969).
Atomi Physi s,
8th edn. Glasgow:
Nature, 119, 3547.

Bla kie and Son.
Reprinted New York: Dover, 1989.

Born, M., and Jordan, P. (1925). Zur Quantenme hanik, I.
Physik, 34,
Zeits hrift fr
85888.
Born, M., Heisenberg, W., and Jordan, P. (1926). Zur Quantenme hanik, II.
557615.
Born, M., and Wiener, N. (1926a). Eine neue Formulierung der Quantengesetze
fr periodis he und ni htperiodis he Vorgnge.
17487.
Born,
of
M.,
and
Wiener,
quantization
of
N.
(1926b).
periodi
and
Mathemati s and Physi s M.I.T., 5,
new
formulation
aperiodi
of
phenomena.
the
laws
Journal of
8498.
Bose, S. N. (1924) Wrmeglei hgewi ht im Strahlungsfeld bei Anwesenheit

von Materie.
38492.
Bothe, W., and Geiger, H. (1924). Ein Weg zur experimentellen Na hprfung
der Theorie von Bohr, Kramers und Slater.
Zeits hrift fr Physik, 26, 44.
Bothe, W., and Geiger, H. (1925a). Experimentelles zur Theorie von Bohr,
Kramers und Slater.
4401.
Bothe, W., and Geiger, H. (1925b). ber das Wesen des Comptoneekts: ein
experimenteller Beitrag zur Theorie der Strahlung.
32, 63963.
Zeits hrift fr Physik,
Brillouin, L. (1922). Diusion de la lumire et des rayons X par un orps

transparent homogne; inuen e de l'agitation thermique.
Physique, 17,
88122.
Broglie, L. de (1922). Rayonnement noir et quanta de lumire.
Physique et le Radium,
(6)
3,
Annales de
Le Journal de
4228.
Comptes Rendus Hebdomadaires des

San es de l'A admie des S ien es (Paris), 177, 50710.
Broglie, L. de (1923b). Quanta de lumire, dira tion et interfren es. Comptes
Rendus Hebdomadaires des San es de l'A admie des S ien es (Paris),
Broglie, L. de (1923a). Ondes et quanta.
177,
54850.
Broglie, L. de (1923 ). Les quanta, la thorie intique des gaz et le prin ipe
Comptes Rendus Hebdomadaires des San es de l'A admie

des S ien es (Paris), 177, 6302.
Broglie, L. de (1923d). Waves and quanta. Nature, 112, 540.
Broglie, L. de (1924a). A tentative theory of light quanta. Philosophi al
Magazine (6), 47, 44658.
de Fermat.
Broglie, L. de (1924b). Sur la dnition gnrale de la orrespondan e entre
Comptes Rendus Hebdomadaires des San es de

l'A admie des S ien es (Paris), 179, 3940.
Broglie, L. de (1924 ). Sur un thorme de M. Bohr. Comptes Rendus
onde et mouvement.
526
Bibliography
Hebdomadaires des San es de l'A admie des S ien es (Paris), 179,
6767.
Broglie, L. de (1924d). Sur la dynamique du quantum de lumire et les
Comptes Rendus Hebdomadaires des San es de l'A admie

des S ien es (Paris), 179, 103941.
interfren es.
Broglie, L. de (1924e). Re her hes sur la thorie des quanta. Ph.D. Thesis,
University of Paris.
Broglie, L. de (1925). Re her hes sur la thorie des quanta.
Physique
(10),
3, 22128.
Annales de
Broglie, L. de (1926). Sur la possibilit de relier les phnomnes d'interfren es

et de dira tion la thorie des quanta de lumire. Comptes Rendus
Hebdomadaires des San es de l'A admie des S ien es (Paris), 183, 447
8.
Broglie, L. de (1927a). La stru ture atomique de la matire et du rayonnement
et la m anique ondulatoire. Comptes Rendus Hebdomadaires des San es
de l'A admie des S ien es (Paris), 184, 2734.
Broglie, L. de (1927b). La m anique ondulatoire et la stru ture atomique de
Le Journal de Physique et le Radium
la matire et du rayonnement.
8, 22541.
Broglie, L. de (1930).
An Introdu tion to the Study of Wave Me hani s.
(6),
New
York: E. P. Dutton and Company.
Physi s and Mi rophysi s. New York: Pantheon Books.

Une tentative d'interprtation ausale et non linaire
de la m anique ondulatoire (la thorie de la double solution). Paris:
Gauthier-Villars.
Broglie, L. de (1974). Beginnings of wave me hani s. In
First Fifty Years,
Wave Me hani s: The
eds. W. C. Pri e, S. S. Chissi k and T. Ravensdale.
London: Butterworths.
Broglie, L. de (1999). The wave nature of the ele tron. Nobel le ture, 12
De ember 1929. In
Nobel Le tures in Physi s (19011995),
Singapore: World S ienti .

Broglie, L. de, and Brillouin, L. (1928).
CD-Rom edn.
Sele ted Papers on Wave Me hani s.
London: Bla kie and Son.

Brown, L. M. (ed.) (2005).
Theory.
Feynman's Thesis: a New Approa h to Quantum
New Jersey: World S ienti .
Studies in
History and Philosophy of Modern Physi s, 37, 298315.
Campbell, N. R. (1921). Atomi stru ture. Nature, 107, 170.
Campbell, N. R. (1926). Time and han e. Philosophi al Magazine (7), 1,
Camilleri, K. (2006). Heisenberg and the wave-parti le duality.
110617.
Chevalley,
C.
(1988).
Physi al
reality
Heisenbergs early papers. In
and
losed
theories
Theory and Experiment,
in
Werner
eds. D. Batens
and J. P. van Bendegem. Dordre ht: Reidel.

Clifton, R., Bub, J., and Halvorson, H. (2003). Chara terizing quantum theory
in terms of information-theoreti onstraints.
Foundations of Physi s, 33,
156191.
Compton, A. H. (1928). Some experimental di ulties with the ele tromagneti theory of radiation.
Journal of the Franklin Institute, 205,
15578.
Compton, A. H., and Simon, A. W. (1925). Dire ted quanta of s attered Xrays.
Physi al Review, 26, 28999.

Quantum Me hani s: Histori al Contingen y and the
Cushing, J. T. (1994).
Bibliography
Copenhagen Hegemony.
527
Chi ago: University of Chi ago Press.
Cushing, J. T. (1996). The ausal quantum theory program. In Cushing, Fine

and Goldstein (1996), pp. 119.
Bohmian Me hani s
and Quantum Theory: An Appraisal. Dordre ht: Kluwer.
Darrigol, O. (1992). From -numbers to q-numbers: The Classi al Analogy in
the History of Quantum Theory. Berkeley: University of California Press.
Cushing, J. T., Fine, A., and Goldstein, S. (eds.) (1996).
Darrigol, O. (1993). Strangeness and soundness in Louis de Broglie's early
Physis, 30,
works.
30372. Reprinted with typographi al orre tions,
1994.
Davisson, C., and Germer, L. (1927). The s attering of ele trons by a single
rystal of Ni kel.
Debye,
P.
(1909).
Nature, 119,
Das
55860.
Verhalten
von
Brennpunktes oder einer Brennlinie.
Li htwellen
in
der
Nhe
Annalen der Physik, 30,
eines
75576.
de Regt, H. (1997). Erwin S hrdinger, Ans hauli hkeit, and quantum theory.
Studies in History and Philosophy of Modern Physi s, 28B,
46181.
de Regt, H. (2001). Spa e-time visualisation and the intelligibility of physi al

theories.
Studies in History and Philosophy of Modern Physi s, 32B, 243
65.
Ernest Solvay et son temps.
Despy-Meyer, A., and Devriese, D. (1997).

Bruxelles: Ar hives de l'ULB.
Deuts h, D. (1999). Quantum theory of probability and de isions.
of the Royal So iety of London A, 455,
Pro eedings
312937.
Dewdney, C., Hardy, L., and Squires, E. J. (1993). How late measurements of
quantum traje tories an fool a dete tor.
Physi s Letters A, 184,
611.
DeWitt, B. S. (1967). Quantum theory of gravity, I. The anoni al theory.
111348.
DeWitt, B. S., and Graham, N. (1973).
Quantum Me hani s.
The Many-Worlds Interpretation of
Prin eton: Prin eton University Press.
Dira , P. (1926a). Quantum me hani s and a preliminary investigation of the

hydrogen atom.
Pro eedings of the Royal So iety A, 110,
56179.
Dira , P. (1926b). Relativity quantum me hani s with an appli ation to

Compton s attering.
Pro eedings of the Royal So iety A, 111, 40523.

Pro eedings of the
Dira , P. (1926 ). On the theory of quantum me hani s.
Royal So iety A, 112,
Dira ,
P.
(1927a).
The
66177.
physi al
interpretation
Pro eedings of the Royal So iety A, 113,
Dira , P.
(1927b).
The
of
quantum
dynami s.
62141.
quantum theory of
emission and absorption of
Pro eedings of the Royal So iety A, 114, 24365.

Dira , P. (1927 ). The quantum theory of dispersion. Pro eedings of the Royal
So iety A, 114, 71028.
Dira , P. A. M. (1933). The Lagrangian in quantum me hani s. Physikalis he
Zeits hrift der Sowjetunion, 3, 6472.
radiation.
Drr, D., Fusseder, W., Goldstein, S., and Zangh, N. (1993). Comment
on `Surrealisti Bohm traje tories'.
Zeits hrift fr Naturfors hung, 48a,
12612.
Drr, D., Goldstein, S., and Zangh, N. (1992). Quantum equilibrium and the
origin of absolute un ertainty.
Journal of Statisti al Physi s, 67, 843907.

Dymond, E. G. (1927). On ele tron s attering in Helium.

43341.
E kart, C. (1926). Operator al ulus and the solution of the equations of
528
Bibliography
quantum dynami s.
Ehrenfest,
P.
(1917).

Adiabati
Philosophi al Magazine, 33,
invariants
71126.
and
the
theory
of
quanta.
50013.
Einstein, A. (1909). ber die Entwi kelung unserer Ans hauungen ber das
Physikalis he Zeits hrift, 10,

The Colle ted Papers of Albert Einstein,
Wesen und die Konstitution der Strahlung.

81726. English translation in
vol. 2, eds. A. Be k and P. Havas. Prin eton: Prin eton University Press,
1989, pp. 37998.
Sitzungsberi hte der Preussis hen Akademie der Wissens haften, Physikalis hmathematis he Klasse (1924), 2617.
Einstein, A. (1924). Quantentheorie des einatomigen idealen Gases.
Einstein,
A.
(1925a).
Quantentheorie
des
einatomigen
idealen
Gases,
Sitzungsberi hte der Preussis hen Akademie der

Wissens haften, Physikalis h-mathematis he Klasse (1925), 314.
Einstein, A. (1925b). Quantentheorie des idealen Gases. Sitzungsberi hte der
Preussis hen Akademie der Wissens haften, Physikalis h-mathematis he
Klasse (1925), 1825.
2.
Abhandlung.
Einstein, A. (1949). Remarks on erning the essays brought together in this

o-operative volume. In S hilpp (1949), pp. 66588.
Einstein,
A.,
and
Besso,
M.
Correspondan e: 19031955,
(1972).
ed.
P. Speziali. Paris: Hermann.

Einstein, A., Podolsky, B., and Rosen, N. (1935). Can quantum-me hani al
des ription of physi al reality be onsidered omplete?
47, 77780.
Physi al Review,
Elsasser, W. (1925). Bemerkungen zur Quantenme hanik freier Elektronen.
Englert, B.-G.,
Surrealisti
S ully, M.
Bohm
O.,
711.
Sssmann, G.,
traje tories.
and Walther,
H. (1992).
Zeits hrift fr Naturfors hung, 47a,
117586.
Everett, H. (1957). `Relative state' formulation of quantum me hani s.
of Modern Physi s, 29,
Fnyes,
I.
(1952).
Eine
wahrs heinli hkeitstheoretis he
Interpretation der Quantenme hanik.

Fermi, E. (1926). Zur Wellenme hanik
Physik, 40,
Reviews
45462.
Begrndung
und

des Stovorganges. Zeits hrift fr
399402.
Fermi, E. (1927). Sul me anismo dell'emissione nella me ani a ondulatoria.
Rendi onti Lin ei, 5, 795800.

Einstein and the Generations of S ien e. New York: Basi
Feuer, L. S. (1974).
Books.
Feynman, R. P. (1942). The prin iple of least a tion in quantum me hani s.

Ph.D. thesis, Prin eton University.
Feynman, R. P. (1948). Spa e-time approa h to non-relativisti quantum
me hani s.
36787.
Feynman, R. P., Leighton, R. B., and Sands, M. (1965).
on Physi s, vol. III: Quantum Me hani s.
The Feynman Le tures
Reading,
Massa husetts:
Addison-Wesley.
Fine, A. (1986).
The Shaky Game: Einstein, Realism and the Quantum Theory.
Chi ago: University of Chi ago Press.

Fine, A. (1999). Lo ality and the Hardy theorem. In
Philosophy,
From Physi s to
eds. J. Buttereld and C. Pagonis. Cambridge: Cambridge
University Press, pp. 111.
Bibliography
529
Fu hs, C. A. (2002). Quantum me hani s as quantum information (and only

a little more), quant-ph/0205039.
Gell-Mann,
M.
and
Hartle,
J.
B.
light of quantum osmology. In
of Information,
(1990).
Quantum
me hani s
in
the
Complexity, Entropy, and the Physi s
ed. W. H. Zurek. Reading, Mass.:
Addison-Wesley,
pp. 42558.
Ghirardi, G.C., Rimini, A., and Weber, T. (1986). Unied dynami s for
mi ros opi and ma ros opi systems.
Goldstein, H. (1980).
Classi al Me hani s.
Physi al Review D, 34,
47091.
Reading: Addison-Wesley.
Gordon, W. (1926). Der Comptoneekt na h der S hrdingers hen Theorie.

The Elegant Universe: Superstrings, Hidden Dimensions,
and the Quest for the Ultimate Theory. London: Vintage.
Greene, B. (2005). The Fabri of the Cosmos: Spa e, Time, and the Texture
of Reality. New York: Vintage.
Greene, B. (2000).
Griths, R. B. (1984). Consistent histories and the interpretation of quantum

me hani s.
Journal of Statisti al Physi s, 36, 21972.

Consistent Quantum Theory. Cambridge:
Griths, R. B. (2002).
Cambridge
University Press.
Guth, A. H. (1981). Inationary universe: A possible solution to the horizon
and atness problems,
34756.
Hardy, L. (1995). The EPR argument and nonlo ality without inequalities
for a single photon. In
Fundamental Problems in Quantum Theory,
eds.
D. M. Greenberger and A. Zeilinger. New York: New York A ademy of

S ien es, pp. 60015.
Hardy, L. (2001). Quantum theory from ve reasonable axioms. quantph/0101012.
Hardy, L. (2002). Why quantum theory? In
Non-lo ality and Modality,
eds.
T. Pla ek and J. Buttereld. Dordre ht: Kluwer, pp. 6173.

Hardy, L. (2005). Probability theories with dynami ausal stru ture: A new
framework for quantum gravity. gr-q /0509120.
Hartle,
J.
B.
(1995).
Spa etime
quantum me hani s
and the
quantum
Gravitation and Quantizations: Pro eedings of

the 1992 Les Hou hes Summer S hool, eds. B. Julia and J. Zinn-Justin.
me hani s of spa etime. In
Amsterdam: Elsevier, pp. 285480.

Heisenberg, W. (1925a). ber eine Anwendung des Korrespondenzprinzips
auf die Frage na h der Polarisation des Fluoreszenzli htes.
Physik, 31,
Zeits hrift fr
61728.
Heisenberg, W. (1925b). ber quantentheoretis he Umdeutung kinematis her

und me hanis her Beziehungen.
87993.
Heisenberg, W. (1926a). S hwankungsers heinungen und Quantenme hanik.
Heisenberg,
W.
(1926b).
5016.
Mehrkrperproblem
und
Resonanz
in
der

Heisenberg, W. (1926 ). Quantenme hanik. Die Naturwissens haften, 14, 989
Quantenme hanik, I.
94.
Heisenberg, W. (1927). ber den ans hauli hen Inhalt der quantentheoretis hen Kinematik und Me hanik.
17298.
Die
Naturwissens haften, 14, 4906.
Heisenberg, W. (1930a). Die physikalis hen Prinzipien der Quantentheorie.
Heisenberg, W. (1929). Die Entwi klung der Quantentheorie 19181928.
530
Bibliography
Leipzig: Hirzel. Reprinted Mannheim: BI Wissens haftsverlag, 1991.
Heisenberg, W. (1930b).
The Physi al Prin iples of the Quantum Theory,
translated by C. E kart and F. C. Hoyt. Chi ago: University of Chi ago

Press. Reprinted New York: Dover, 1949.
Heisenberg, W. (1946). Der unans hauli he Quantensprung.
s he Bltter, 2,
Neue physikali-
46.
Heisenberg, W. (1948). Der Begri der Àbges hlossenen Theorie' in der
Diale ti a, 2, 3316.
Physi s and Philosophy: The Revolution in Modern
modernen Naturwissens haft.

Heisenberg, W. (1962).
S ien e.
New York: Harper and Row.
Niels Bohr:
His Life and Work as seen by his Friends and Colleagues, ed. S. Rozental.
Heisenberg, W. (1967). Quantum theory and its interpretation. In

New York: John Wiley and Sons, pp. 94108.
Der Teil und das Ganze. Mn hen: Piper.

Gesammelte Werke, eds. W. Blum, H.-P.
Drr and
H. Re henberg. Berlin: Springer.

Hendry, J. (1986).
Field.
James Clerk Maxwell and the Theory of the Ele tromagneti
Bristol: Adam Hilger.
Hilbert, D., von Neumann, J., and Nordheim, L. (1928). ber die Grundlagen
Mathematis he Annalen, 98, 130.

The Quantum Theory of Motion: An A ount of the de
Broglie-Bohm Causal Interpretation of Quantum Me hani s. Cambridge:
der Quantenme hanik.
Holland, P. R. (1993).
Cambridge University Press.

Holland, P. R. (2005). What's wrong with Einstein's 1927 hidden-variable
interpretation of quantum me hani s?
Foundations of Physi s, 35,
177
96.
Howard, D. (1990). `Ni ht sein kann was ni ht sein darf ', or the prehistory of
EPR, 19091935: Einstein's early worries about the quantum me hani s
of omposite systems. In Miller (1990), pp. 61111.
Jammer, M. (1966).
The Con eptual Development of Quantum Me hani s.
New York: M Graw-Hill.
M.
(1974).
The Philosophy of Quantum Me hani s: The
Interpretations of Quantum Me hani s in Histori al Perspe tive. New
Jammer,
York: John Wiley and Sons.

Jordan, P. (1926a). ber kanonis he Transformationen in der Quantenme hanik.
3836.
Jordan, P. (1926b). ber kanonis he Transformationen in der Quantenme hanik, II.
51317.
Jordan, P. (1927a). ber quantenme hanis he Darstellung von Quantensprngen.
6616.
Jordan, P. (1927b). ber eine neue Begrndung der Quantenme hanik, I.
80938.
Jordan, P. (1927 ). ber eine neue Begrndung der Quantenme hanik, II.
125.
Jordan, P. (1927d). Zur Quantenme hanik der Gasentartung.
Physik, 44,
Zeits hrift fr
47380.
Quantenme hanik.
Die
Jordan, P. (1927f ). Kausalitt und Statistik in der modernen Physik.
Die
Jordan,
P.
(1927e).
Die
Entwi klung
Naturwissens haften, 15,
der
neuen
61423, 63649.
10510.
Jordan, P. (1927g). Philosophi al foundations of quantum theory.
Nature, 119,
Bibliography
531
5669.
Kiefer, C., Polarski, D., and Starobinsky, A. A. (1998). Quantum-to- lassi al
transition for u tuations in the early universe.
Modern Physi s D, 7,
45562.
Kirsten, C., and Treder, H. J. (1979).
International Journal of
Albert Einstein in Berlin 19131933.
Berlin: Akademie-Verlag.
Klein, O. (1926). Quantentheorie und fnfdimensionale Relativittstheorie.
895906.
Koopman, B. O. (1955). Quantum theory and the foundations of probability.

In
Applied Probability,
ed. L. A. Ma Coll. New York: M Graw-Hill,
pp. 97102.
Kramers, H. (1924). The quantum theory of dispersion.
Nature, 114,
31011.
Kramers, H., and Heisenberg, W. (1925). ber die Streuung von Strahlung
dur h Atome.

Dira : A S ienti Biography. Cambridge:
Kragh, Helge (1990).
Cambridge
University Press.
Kuhn, W. (1925). ber die Gesamtstrke der von einem Zustande ausgehenden
Absorptionslinien.
40812.
La ki, J. (2004). The puzzle of anoni al transformations in early quantum
Studies in History and Philosophy of Modern Physi s, 35,
me hani s.
31744.
Lan zos,
K.
(1926).
Quantenme hanik.
ber
eine
feldmssige
Darstellung
81230.
Laue, M. von (1914). Die Freiheitsgrade von Strahlenbndeln.
Physik, 44,
11971212.
Liddle, A. R. and Lyth, D. H. (2000).
Stru ture.
der
neuen
Annalen der
Cosmologi al Ination and Large-S ale
Lo hak, G. (1992).
Louis de Broglie.
Paris: Flammarion.
London, F. (1926a). ber die Ja obis hen Transformationen der Quantenme hanik.
915-25.
London, F. (1926b). Winkelvariable und kanonis he Transformationen in der
Zeits hrift fr Physik, 40, 193-210.

Wave Me hani s. Oxford: Pergamon Press.
Undulationsme hanik.
Ludwig, G. (1968).
Madelung,
E.
(1926a).
S hrdinger.
Eine
ans hauli he
Deutung
der
Glei hung
1004.
Madelung, E. (1926b). Quantentheorie in hydrodynamis her Form.
fr Physik, 40,
3226.
Margenau, H. (1939). Van der Waals for es.

135.
von
Zeits hrift
J. (1975). The Solvay Conferen es on Physi s: Aspe ts of the

Development of Physi s sin e 1911. Dordre ht and Boston: Reidel.
Mehra, J., and Re henberg, H. (1982a). The Histori al Development of
Quantum Theory, vol. 1, part 2. New York: Springer.
Mehra, J., and Re henberg, H. (1982b). The Histori al Development of
Quantum Theory, vol. 2. New York: Springer.
Mehra, J., and Re henberg, H. (1982 ). The Histori al Development of
Quantum Theory, vol. 3. New York: Springer.
Mehra, J., and Re henberg, H. (1987). The Histori al Development of
Mehra, J., and Re henberg, H. (2000). The Histori al Development of
Mehra,
532
Bibliography
Miller, A. I. (ed.) (1990).
Sixty-two Years of Un ertainty.
New York: Plenum
Press.
Millikan, R. A. (1916). A dire t photoele tri determination of Plan k's h .
Physi al Review, 7, 35588.

S hrdinger: Life and Thought.
Moore, W. (1989).
Cambridge: Cambridge
University Press.
Muller, F. (1997). The Equivalen e Myth of Quantum Me hani s, parts I and
II and addendum.
28B, 3561
Studies in History and Philosophy of Modern Physi s,

ibid., 30B, 5435.
and 21947;
Nelson, E. (1966). Derivation of the S hrdinger equation from Newtonian

me hani s.
107985.
Neumann, J. von (1927). Mathematis he Begrndung der Quantenme hanik.
Na hri hten der Akademie der Wissens haften in Gttingen. II,

Mathematis h-Physikalis he Klasse, 20 May 1927, 157.
Newton, I. (1730). Opti ks: Or, a Treatise of the Ree tions, Refra tions,
Ine tions and Colours of Light, 4th edn. London. Reprinted as I. Newton,
Opti ks. New York: Dover, 1979.
Ni olai, G. F. (1917). Die Biologie des Krieges. Zri h: Orell Fssli.
Norsen, T. (2005). Einstein's boxes. Ameri an Journal of Physi s, 73, 16476.
Nye, M. J. (1997). Aristo rati ulture and the pursuit of s ien e: the de
Isis, 88,
Broglies in modern Fran e.
397421.
Omns, R. (1992). Consistent interpretations of quantum me hani s.
of Modern Physi s, 64, 33982.

Omns, R. (1994). The Interpretation of Quantum Me hani s.
Reviews
Prin eton:
Prin eton University Press.

Oseen, C. W. (1999). Presentation spee h, 12 De ember 1929. In
Le tures in Physi s (19011995),
S ienti .
Padmanabhan, T. (1993).
CD-Rom
edn.
Singapore:
Stru ture Formation in the Universe.
Nobel
World
Cambridge:
Cambridge University Press.

Pais, A. (1982).
Subtle is the Lord: The S ien e and the Life of Albert Einstein.
Oxford: Oxford University Press.

Patterson, A. L. (1927). The s attering of ele trons from single rystals of
ni kel.
Nature, 120,
467.
Pauli, W. (1925). ber die Intensitten der im elektris hen Feld ers heinenden
Kombinationslinien.
Matematisk-fysiske Meddelser, 7,
320.
Pauli, W. (1927). ber Gasentartung und Paramagnetismus.
Physik, 41,
Zeits hrift fr
81102.
W. (1979). Wissens haftli her Briefwe hsel mit Bohr, Einstein,

Heisenberg u.a., Teil I: 19191929, eds. A. Hermann, K. v. Meyenn and
Pauli,
V. F. Weisskopf. Berlin and Heidelberg: Springer.

Pearle, P. (1976). Redu tion of the state ve tor by a nonlinear S hrdinger
equation.
85768.
Pearle, P. (1979). Toward explaining why events o ur.
of Theoreti al Physi s, 18,
International Journal
489518.
Pearle, P. (1989). Combining sto hasti dynami al state-ve tor redu tion with
spontaneous lo alization.
Physi al Review A, 39,
227789.
Pearle, P., and Valentini, A. (2006). Quantum me hani s: generalizations. In
En y lopaedia of Mathemati al Physi s,
eds. J.-P. Franoise, G. Naber
and T. S. Tsun. Amsterdam: Elsevier, pp. 26576.

Perez, A., Sahlmann, H., and Sudarsky, D. (2006). On the quantum origin
Bibliography
of the seeds of osmi stru ture.
533
Classi al and Quantum Gravity, 23,
231754.
Perovi , S. (2006). S hrdinger's interpretation of quantum me hani s and
the relevan e of Bohr's experimental ritique.
27597.
Philippidis, C., Dewdney, C., and Hiley, B. J. (1979). Quantum interferen e
Nuovo Cimento B, 52, 1528.

Quantum Theory: A Very Short Introdu tion. Oxford:
and the quantum potential.

Polkinghorne, J. (2002).
Oxford University Press.

Popper, K. R. (1982).
Quantum Theory and the S hism in Physi s.
Unwin Hyman.
Przibram,
K.
(ed.)
(1967).
London:
Letters on Wave Me hani s. New York:

Briefe zur Wellenme hanik.
Philosophi al Library. Originally published as

Wien: Springer, 1963.
Rovelli, C. (2004).
Quantum Gravity. Cambridge: Cambridge University Press.

Synthese,
Saunders, S. (1995). Time, de oheren e and quantum me hani s.
102,
23566.
Saunders, S. (1998). Time, quantum me hani s, and probability.
114,
Synthese,
373404.
S heibe, E. (1993). Heisenbergs Begri der abges hlossenen Theorie. In
Heisenberg: Physiker und Philosoph,
Werner
eds. B. Geyer, H. Herwig and
H. Re henberg. Heidelberg: Spektrum, pp. 2517.
Quantum Me hani s. New York: M Graw-Hill.

Albert Einstein: Philosopher-S ientist, The Library
S hi, L. I. (1955).
P. A. S hilpp (ed.) (1949).
of Living Philosophers, vol. VII. La Salle, Ill.: Open Court.

S hrdinger, E. (1922). Dopplerprinzip und Bohrs he Frequenzbedingung.
3013.
S hrdinger, E. (1924a). ber das thermis he Glei hgewi ht zwis hen Li htuns S hallstrahlen.
S hrdinger,
E.
Energiesatz.
S hrdinger,
E.
E.
Mitteilung).
S hrdinger,
E.
Mitteilung).
S hrdinger,
Bohrs
neue
(1926a).
Zur
7204.
Einsteins hen
Gastheorie.
und
der
Physikalis he
95101.
(1926b).
Quantisierung

(1926 ).
(1926d).
ber
das
als
Eigenwertproblem
(erste
36176.
Quantisierung
E.
8994.
Strahlungshypothese
Zeits hrift, 27,
S hrdinger,
(1924b).
als
Eigenwertproblem
(zweite
489527.
Verhltnis
Jordans hen Quantenme hanik zu der meinen.
der
Heisenberg-Born-
von
der
73456.
S hrdinger,
E.
(1926).
Makrome hanik.
S hrdinger,
E.
Der
stetige
bergang
(1926f ).
Quantisierung
als
Mikro-
zur
6646.
Eigenwertproblem
(dritte
Mitteilung: Strungstheorie, mit Anwendung auf den Starkeekt der

Balmerlinien).
S hrdinger,
E.
Mitteilung).
(1926g).
Quantisierung
43790.
als
Eigenwertproblem
(vierte
109139.
S hrdinger, E. (1926h). An undulatory theory of the me hani s of atoms and

mole ules.
10491070.
S hrdinger, E. (1926i). Vorwort zur ersten Auage. In S hrdinger (1928),

p. iv.
534
Bibliography
S hrdinger, E. (1927a). ber den Comptoneekt.
25764.
S hrdinger, E. (1927b). Der Energieimpulssatz der Materiewellen.
der Physik, 82,
26572.
S hrdinger, E. (1927 ). Energieaustaus h na h der Wellenme hanik.
der Physik, 83,
Annalen
Annalen
95668.
S hrdinger, E. (1928).
Abhandlungen zur Wellenme hanik, 2nd, enlarged

Colle ted Papers on Wave Me hani s.
edn. Leipzig: Barth. Translated as
London and Glasgow: Bla kie and Son, 1928.

S hrdinger, E. (1929a). Was ist ein Naturgesetz?
17, 911.
S hrdinger,
E.
(1929b).
Die
kontinuierli he Funktionen.
Erfassung
der
Die Naturwissens haften,

Quantengesetze
dur h
4869.
S hrdinger, E. (1935). Die gegenwrtige Situation in der Quantenme hanik.
80712, 8238, 844-9.
S hrdinger, E. (1936). Probability relations between separated systems.
Pro eedings of the Cambridge Philosophi al So iety, 32, 44652.

E. (1984). Gesammelte Abhandlungen, eds. sterrei his he
S hrdinger,
Akademie der Wissens haften. Brauns hweig: Vieweg.

S hrdinger, E. (1995).
The Interpretation of Quantum Me hani s.
Ed. and
with introdu tion by M. Bitbol. Woodbridge, Conn.: Ox Bow Press.

S hulman,
L.
S.
(1997).
Time's Arrows and Quantum Measurement.

Senftleben, H. A. (1923). Zur Grundlegung der `Quantentheorie', I.
fr Physik, 22,
Zeits hrift
12756.
Shankar, R. (1994).
Prin iples of Quantum Me hani s,
2nd edn. New York:
Springer-Verlag.
Shimony,
A.
(2005).
Comment
on
Norsen's
defense
of
Einstein's
`box
Ameri an Journal of Physi s, 73, 1778.

Slater, J. C. (1924). Radiation and atoms. Nature, 113, 3078.
Smekal, A. (1923). Zur Quantentheorie der Dispersion. Die Naturwissens haften, 11, 8735.
Taylor, G. I. (1909). Interferen e fringes with feeble light. Pro eedings of the
Cambridge Philosophi al So iety, 15, 11415.
argument'.
Thomas, W. (1925). ber die Zahl der Dispersionselektronen, die einem

stationren
Zustande
zugeordnet
sind
(vorluge
Mitteilung).
Die
627.
Unruh, W. G. and Wald, R. M. (1989). Time and the interpretation of

anoni al quantum gravity.
25982614.
Valentini, A. (1991a). Signal-lo ality, un ertainty, and the subquantum

theorem, I.
511.
Valentini, A. (1991b). Signal-lo ality, un ertainty, and the subquantum

theorem, II.
HH-
18.
Valentini, A. (1992). On the pilot-wave theory of lassi al, quantum and

subquantum physi s. Ph.D. thesis, International S hool for Advan ed Studies, Trieste, Italy. http://www.sissa.it/ap/PhD/Theses/valentini.pdf .
Valentini, A. (1997).
dynami s.
On Galilean
and Lorentz invarian e in pilot-wave
21522.
Valentini, A. (2001). Hidden variables, statisti al me hani s and the early

universe. In
Chan e in Physi s: Foundations and Perspe tives,
eds. J.
Bri mont, D. Drr, M. C. Galavotti, G. Ghirardi, F. Petru ione and N.
Bibliography
535
Zangh. Berlin: Springer-Verlag, pp. 16581.

Valentini, A. (2002a). Signal-lo ality in hidden-variables theories.
Letters A, 297,
2738.
Valentini, A. (2002b). Subquantum information and omputation.
Journal of Physi s, 59,
Physi s
Pramana
26977.
Valentini, A. (2006). Inationary osmology as a probe of primordial quantum

me hani s. hep-th/
Valentini, A. (2007). Cambridge: Cambridge University Press. Forth oming.
Valentini,
A.
and
Westman,
H.
(2005).
Dynami al
origin
of
quantum
Pro eedings of the Royal So iety of London A, 461, 25372.

Walla e, D. (2003a). Everett and stru ture. Studies in History and Philosophy
of Modern Physi s, 34, 87105.
probabilities.
Walla e, D. (2003b). Everettian rationality: defending Deuts h's approa h

to probability in the Everett interpretation.
41539.
Wallstrom, T. C. (1994). Inequivalen e between the S hrdinger equation and

the Madelung hydrodynami equations.
Physi al Review A, 49, 161317.

Zeits hrift fr
Wentzel, G. (1926). Zur Theorie des photoelektris hen Eekts.
Physik, 40,
Wentzel,
G.
57489.
(1927).
ber
die
Wheeler,
J.
A.
(1978).
experiment. In
Ri htungsverteilung
der
Photoelektronen.
82832.
The
`past'
and
the
`delayed- hoi e'
double-slit
Mathemati al Foundations of Quantum Me hani s,
ed. A.
R. Marlow. New York: A ademi , pp. 948. Reprinted in Wheeler and

Zurek (1983), pp. 182213.
Wheeler, J. A. (1986). Interview. In
The Ghost in the Atom,
eds. P. C. W.
Davies and J. R. Brown, Cambridge: Cambridge University Press, pp. 58

69.
Wheeler, J. A., and Zurek, W. H. (eds.) (1983).
Measurement.
Wien,
W.
(1923).
Radiologie,
Quantum Theory and
Prin eton: Prin eton University Press.
Kanalstrahlen,
2nd
edn.
Vol.
IV.1
of
Handbu h der
ed. E. Marx. Leipzig: Akademis he Verlagsgesells haft.
Wigner, E. P. (1961). Remarks on the mind-body question. In
Spe ulates,
The S ientist
ed. I. J. Good. London: Heinemann, pp. 284302. Reprinted
in E. P. Wigner
Symmetries and Ree tions.
Bloomington: Indiana
University Press, 1967, pp. 17184. Also reprinted in Wheeler and Zurek
(1983), pp. 16881.
Some Strangeness in
the Proportion: a Centennial Symposium to Celebrate the A hievements of
Albert Einstein, ed. H. Woolf. Reading, Massa husetts: Addison-Wesley,
Wigner, E. P. (1980). Thirty years of knowing Einstein. In
pp. 4618.
Wootters, W. K., and Zurek, W. H. (1979). Complementarity in the doubleslit experiment: quantum nonseparability and a quantitative statement of
Bohr's prin iple.
47384.
Zurek, W. H. (1991). De oheren e and the transition from quantum to

lassi al.
Physi s Today, 44,
n. 10 (O tober 1991), pp. 3644.

Quant

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Quant

Uploaded by

Copyright:

Available Formats

Quantum Theory at the Crossroads

arXiv:quant-ph/0609184v2 24 Oct 2009

Re onsidering the 1927 Solvay Conferen e

Cambridge University Press

To the memory of James T. Cushing

1.1 Ernest Solvay and the Institute of Physi s

De Broglie's pilot-wave theory

2.5 Signi an e of de Broglie's work from 1923 to 1927

From matrix me hani s to quantum me hani s

Summary of Born and Heisenberg's report

S hrdinger's wave me hani s

Planning of S hrdinger's report

Quantum theory and the measurement problem

What is quantum theory?

Interferen e, superposition, and wave pa ket ollapse 168

Measurement as a physi al pro ess: quantum

Probability and interferen e

Lo ality and in ompleteness

Time, determinism, and the spa etime framework

Guiding elds in 3-spa e

Einstein's 1927 argument for in ompleteness

Einstein's early attempts to formulate a dynami al

S attering and measurement in de Broglie's pilot-wave

Pilot-wave theory in retrospe t

Beyond the Bohr-Einstein debate

Histori al mis on eptions

12.1 The standard histori al a ount

The intensity of X-ray ree tion (W. L. Bragg )

The lassi al treatment of X-ray dira tion phenomena

Disagreements between experiment and the ele tromagneti

The new dynami s of quanta (L. de Broglie )

I.  Prin ipal points of view

Quantum me hani s (M. Born and W. Heisenberg )

Wave me hani s (E. S hrdinger )

General dis ussion of the new ideas presented

Causality, determinism, probability

And they said one to another: Go to, let us build us a

Anyone who has taken part in a debate on the interpretation of quantum

dieren e being that de Broglie's dynami s (like the form of pilot-wave

To James T. Cushing, physi ist, philosopher, historian and gentleman,

Lake Maggiore, August 2006

Note on the bibliography

Lin ei Rend.: Rendi onti Lin ei.

Note on the bibliography

Permissions and opyright noti es

Permissions and opyright noti es

Quantum re on iliation very [added, deleted unpleasant

The onferen e was surely the most interesting s ienti

The early Solvay onferen es were remarkable o asions, made possible

1.1 Ernest Solvay and the Institute of Physi s

Histori al introdu tion

1.1 Ernest Solvay and the Institute of Physi s

the ollaboration of Lorentz (who was eventually asked to preside), of

Lorentz set up a ommittee to onsider questions relating to the

Histori al introdu tion

o ially established on 1 May 1912, whi h predates by several years the

1.2 War and international relations

1.2 War and international relations

Kulturwelt!', denying German responsibilities in the war.a Among the

Histori al introdu tion

In the meantime, after the treaty of Lo arno of 1925, Germany was

1.2 War and international relations

Histori al introdu tion

2.5 Signi an e of de Broglie's work from 1923 to 1927

Guiding elds in 3-spa e

The intensity of X-ray ree tion (W. L. Bragg )

The lassi al treatment of X-ray dira tion phenomena

I. Prin ipal points of view

dieren e being that de Broglie's dynami s (like the form of pilot-wave

The onferen e was surely the most interesting s ienti

o ially established on 1 May 1912, whi h predates by several years the

1.3 S ienti planning and ba kground

1.3 S ienti planning and ba kground

1.3 S ienti planning and ba kground

1.3 S ienti planning and ba kground

Darwin, Smekal (is indeed a very deserving onnoisseur of quantum nesses,

At the April meeting (as listed in the report by Vers haelt55 ) it

Tuesday 25 O tober, starting 9:00 a.m.: re eption oered by the ULB.

Per eptions of the signi an e of the onferen e diered from ea h other,

Notes to pp. 39

Notes to pp. 918

Notes to pp. 1819

Notes to pp. 1924

for a `pilot wave' (x1 , x2 , ..., xN , t) in onguration spa e, and the de

is dira ted by the edge of a s reen.

2.2 A new approa h to parti le dynami s: 192324