Understanding Heisenberg’s ‘‘magical’’ paper of July 1925: A new look at

the calculational details

Ian J. R. Aitchisona)
Department of Physics, Theoretical Physics, University of Oxford, Oxford OX1 3NP, United Kingdom
David A. MacManus
Tripos Receptor Research Ltd., Bude-Stratton Business Park, Bude, Cornwall EX23 8LY, United Kingdom
Thomas M. Snyder
Department of Mathematics and Engineering Sciences, Lincoln Land Community College, Springfield,
Illinois 62794-9256
共Received 20 January 2004; accepted 3 June 2004兲
In July 1925 Heisenberg published a paper that ushered in the new era of quantum mechanics. This
epoch-making paper is generally regarded as being difficult to follow, partly because Heisenberg
provided few clues as to how he arrived at his results. We give details of the calculations of the type
that Heisenberg might have performed. As an example we consider one of the anharmonic oscillator
problems considered by Heisenberg, and use our reconstruction of his approach to solve it up to
second order in perturbation theory. The results are precisely those obtained in standard quantum
mechanics, and we suggest that a discussion of the approach, which is based on the direct
calculation of transition frequencies and amplitudes, could usefully be included in undergraduate
courses on quantum mechanics. © 2004 American Association of Physics Teachers.
关DOI: 10.1119/1.1775243兴

I. INTRODUCTION assumption that, having formulated a method that was ca-

Heisenberg’s paper of July 19251 on ‘‘Quantum-
mechanical reinterpretation of kinematic and mechanical
relations,’’ 2,3 was the breakthrough that quickly led to the
first complete formulation of quantum mechanics.4 – 6 Despite
its undoubtedly crucial historical role, Heisenberg’s approach
in this paper is not generally followed in undergraduate
quantum mechanics courses, in contrast, for example, to Ein-
stein’s approach in the teaching of relativity. Indeed Heisen-
berg’s paper is widely regarded as being difficult to under-
stand and of mainly historical interest today. For example,
Weinberg7 has written that ‘‘If the reader is mystified at what
Heisenberg was doing, he or she is not alone. I have tried
several times to read the paper that Heisenberg wrote on
returning from Heligoland, and, although I think I under-
stand quantum mechanics, I have never understood Heisen-
berg’s motivations for the mathematical steps in his paper.
Theoretical physicists in their most successful work tend to
play one of two roles: they are either sages or magicians ... It
is usually not difficult to understand the papers of sage-
physicists, but the papers of magician-physicists are often
incomprehensible. In this sense, Heisenberg’s 1925 paper
was pure magic.’’
There have been many discussions aimed at elucidating
the main ideas in Heisenberg’s paper of which Refs. 3 and
8 –18 represent only a partial selection.19 Of course, it may
not be possible to render completely comprehensible the
mysterious processes whereby physicists ‘‘jump over all in-
termediate steps to a new insight about nature.’’ 20 In our
opinion, however, one of the main barriers to understanding
Heisenberg’s paper is a more prosaic one: namely, he gave
remarkably few details of the calculations he performed.
In Sec. II we briefly review Heisenberg’s reasoning in set-
ting up his new calculational method. Then we present in
Sec. III the details of a calculation typical of those we con-
jecture that he performed. Our reconstruction is based on the
pable of determining the relevant physical quantities 共the
transition frequencies and amplitudes兲, Heisenberg then ap-
plied it to various simple mechanical systems, without any
further recourse to the kind of ‘‘inspired guesswork’’ that
characterized the old quantum theory. Surprisingly, this point
of view appears to be novel. For example, MacKinnon10 and
Mehra and Rechenberg11 have suggested that Heisenberg ar-
rived at the crucial recursion relations 关see Eqs. 共33兲–共36兲 in
Sec. III B兴 by essentially guessing the appropriate generali-
zation of their classical counterparts. We are unaware of any
evidence that can settle the issue. In any case, our analysis
shows that it is possible to read Heisenberg’s paper as pro-
viding a complete 共if limited兲 calculational method, the re-
sults of which are consistent with those of standard quantum
mechanics. We also stress both the correctness and the prac-
ticality of what we conjecture to be Heisenberg’s calcula-
tional method. We hope that our reappraisal will stimulate
instructors to include at least some discussion of it in their
undergraduate courses.
II. HEISENBERG’S TRANSITION AMPLITUDE
APPROACH
A. Quantum kinematics
Heisenberg began his paper with a programmatic call21,22
to ‘‘discard all hope of observing hitherto unobservable
quantities, such as the position and period of the electron,’’
and instead to ‘‘try to establish a theoretical quantum me-
chanics, analogous to classical mechanics, but in which only
relations between observable quantities occur.’’ As an ex-
ample of such latter quantities, he immediately pointed to the
energies W(n) of the Bohr stationary states, together with
the associated Einstein–Bohr frequencies23
1
␻ 共 n,n⫺ ␣ 兲 ⫽ 关 W 共 n 兲 ⫺W 共 n⫺ ␣ 兲兴 , 共1兲
ប

1370 Am. J. Phys. 72 共11兲, November 2004 http://aapt.org/ajp © 2004 American Association of Physics Teachers 1370
and noted that these frequencies, which characterize the ra-
diation emitted in the transition n→n⫺ ␣ , depend on two
variables. An example of a quantity he wished to exclude
from the new theory is the time-dependent position coordi-
nate x(t). In considering what might replace it, he turned to
the probabilities for transitions between stationary states.
Consider a simple one-dimensional model of an atom con-
sisting of an electron undergoing periodic motion, which is
the type of system studied by Heisenberg. For a state char-
acterized by the label n, the fundamental frequency ␻ (n),
and the coordinate x(n,t), we can represent x(n,t) as a Fou-
rier series
⬁
x 共 n,t 兲 ⫽ 兺
␣ ⫽⫺⬁
X ␣ 共 n 兲 e i ␣ ␻ (n)t ,
where ␣ is an integer.24 According to classical theory, the
energy emitted per unit time 共the power兲 in a transition cor-
responding to the ␣th harmonic ␣ ␻ (n) is25
冉 冊
dE e2
⫺ ⫽ 关 ␣ ␻ 共 n 兲兴 4 兩 X ␣ 共 n 兲 兩 2 .
dt ␣
3 ␲ ⑀ 0c 3
In the quantum theory, however, the transition frequency cor-
responding to the classical ␣ ␻ (n) is, in general, not a simple
multiple of a fundamental frequency, but is given by Eq. 共1兲,
so that ␣ ␻ (n) is replaced by ␻ (n,n⫺ ␣ ). Correspondingly,
Heisenberg introduced the quantum analogue of X ␣ (n), writ-
ten 共in our notation兲 as X(n,n⫺ ␣ ). 27 Furthermore, the left-
hand side of Eq. 共3兲 has to be replaced by the product of the
transition probability per unit time, P(n,n⫺ ␣ ), and the
emitted energy ប ␻ (n,n⫺ ␣ ). Thus Eq. 共3兲 becomes
2
e
P 共 n,n⫺ ␣ 兲 ⫽ 关 ␻ 共 n,n⫺ ␣ 兲兴 3 兩 X 共 n,n⫺ ␣ 兲 兩 2 .
3 ␲ ⑀ 0 បc 3
It is the transition amplitudes X(n,n⫺ ␣ ) which Heisenberg
took to be ‘‘observable;’’ like the transition frequencies, they
depend on two discrete variables.28
Equation 共4兲 refers, however, to only one specific transi-
tion. For a full description of atomic dynamics 共as then con-
ceived兲, we need to consider all the quantities X(n,n
⫺ ␣ )exp关i␻(n,n⫺␣)t兴. In the classical case, the terms
X ␣ (n)exp关i␣␻(n)t兴 may be combined to yield x(t) via Eq.
共2兲. But in the quantum theory, Heisenberg wrote29 that a
‘‘similar combination of the corresponding quantum-
theoretical quantities seems to be impossible in a unique
manner and therefore not meaningful, in view of the equal
weight of the variables n and n⫺ ␣ 关that is, in the amplitude
X(n,n⫺ ␣ ) and frequency ␻ (n,n⫺ ␣ )] ... However, one
may readily regard the ensemble of quantities X(n,n
⫺ ␣ )exp关i␻(n,n⫺␣)t兴 as a representation of the quantity
x(t)... .’’ This way of representing x(t), that is, as we would
now say, by a matrix, is the first of Heisenberg’s ‘‘magical
jumps,’’ and surely a very large one. Representing x(t) in
this way seems to be the sense in which Heisenberg consid-
ered that he was offering a ‘‘reinterpretation of kinematic
relations.’’
Heisenberg immediately posed the question: how is the
quantity x(t) 2 to be represented? In classical theory, the an-
swer is straightforward. From Eq. 共2兲 we obtain
1371 Am. J. Phys., Vol. 72, No. 11, November 2004
关 x 共 t 兲兴 2 ⫽ 兺␣ 兺 X ␣ 共 n 兲 X ␣ ⬘ 共 n 兲 e i( ␣ ⫹ ␣ ⬘ ) ␻ (n)t . 共5兲
␣ ⬘
We set ␤ ⫽ ␣ ⫹ ␣ ⬘ , and rewrite Eq. 共5兲 as
关 x 共 t 兲兴 2 ⫽ 兺␤ Y ␤共 n 兲 e i ␤ ␻ (n)t , 共6兲
where
Y ␤共 n 兲 ⫽ 兺␣ X ␣共 n 兲 X ␤ ⫺ ␣共 n 兲 . 共7兲
Thus 关 x(t) 兴 2 is represented classically 共via a Fourier series兲
by the set of quantities Y ␤ (n)exp关i␤␻(n)t兴, the frequency
共2兲 ␤ ␻ (n) being the simple combination 关 ␣ ␻ (n)⫹( ␤
⫺ ␣ ) ␻ (n) 兴 . In quantum theory, the corresponding represen-
tative quantities must be written as Y (n,n⫺ ␤ )exp关i␻(n,n
⫺␤)t兴, and the question is what is the analogue of Eq. 共7兲?
The crucial difference in the quantum case is that the fre-
quencies do not combine in the same way as the classical
harmonics, but rather in accordance with the Ritz combina-
共3兲
tion principle:
␻ 共 n,n⫺ ␣ 兲 ⫹ ␻ 共 n⫺ ␣ ,n⫺ ␤ 兲 ⫽ ␻ 共 n,n⫺ ␤ 兲 , 共8兲
which is consistent with Eq. 共1兲. Thus in order to end up with
the particular frequency ␻ (n,n⫺ ␤ ), it seems ‘‘almost nec-
essary’’ 共in Heisenberg’s words30兲 to combine the quantum
amplitudes in such a way as to ensure the frequency combi-
nation Eq. 共8兲, that is, as
Y 共 n,n⫺ ␤ 兲 e i ␻ (n,n⫺ ␤ )t ⫽ 兺␣ X 共 n,n⫺ ␣ 兲 e i ␻ (n,n⫺ ␣ )t
⫻X 共 n⫺ ␣ ,n⫺ ␤ 兲 e i ␻ (n⫺ ␣ ,n⫺ ␤ )t ,
共9兲
共4兲 or
Y 共 n,n⫺ ␤ 兲 ⫽ 兺␣ X 共 n,n⫺ ␣ 兲 X 共 n⫺ ␣ ,n⫺ ␤ 兲 , 共10兲
which is Heisenberg’s rule for multiplying transition ampli-
tudes. Note particularly that the replacements X ␣ (n)
→X(n,n⫺ ␣ ), and similarly for Y ␤ (n) and X ␤ ⫺ ␣ (n) in Eq.
共7兲, produce a quite different result.
Heisenberg indicated the simple extension of the rule
given in Eq. 共10兲 to higher powers 关 x(t) 兴 n , but noticed at
once31 that a ‘‘significant difficulty arises, however, if we
consider two quantities x(t),y(t) and ask after their product
x(t)y(t)... Whereas in classical theory x(t)y(t) is always
equal to y(t)x(t), this is not necessarily the case in quantum
theory.’’ Heisenberg used the word ‘‘difficulty’’ three times
in referring to this unexpected consequence of his multipli-
cation rule, but it very quickly became clear that the non-
commutativity 共in general兲 of kinematical quantities in quan-
tum theory was the essential new idea in the paper.
Born recognized Eq. 共10兲 as matrix multiplication 共some-
thing unknown to Heisenberg in July 1925兲, and he and Jor-
dan rapidly produced the first paper4 to state the fundamental
commutation relation 共in modern notation兲
x̂ p̂⫺p̂x̂⫽iប. 共11兲
5
Dirac’s paper followed soon after, and then the paper of
Born, Heisenberg, and Jordan.6
Aitchison, MacManus, and Snyder 1371
 The economy and force of Heisenberg’s argument in
reaching Eq. 共10兲 is remarkable, and it is at least worth con-
sidering whether presenting it to undergraduates might help
them to understand the ‘‘almost necessity’’ of non-
commuting quantities in quantum theory.
B. Quantum dynamics
Having identified the transition amplitudes X(n,n⫺ ␣ )
and frequencies ␻ (n,n⫺ ␣ ) as the observables of interest in
the new theory, Heisenberg then turned his attention to how
they could be determined from the dynamics of the system.
In the old quantum theory, this determination would have
been done in two stages: by integration of the equation of
motion
ẍ⫹ f 共 x 兲 ⫽0,
and by determining the constants of the periodic motion
through the ‘‘quantum condition’’
冖 pdq⫽ 冖 mẋ dt⫽J 共 ⫽nh 兲 ,
2
where the integral is evaluated over one period. In regard to
Eq. 共12兲, Heisenberg wrote32 that it is ‘‘very natural’’ to take
the classical equation of motion over to quantum theory by
replacing the classical quantities x(t) and f (x) by their ki-
nematical reinterpretations,33 as in Sec. II A 共or, as we would
say today, by taking matrix elements of the corresponding
operator equation of motion兲. He noted that in the classical
case a solution can be obtained by expressing x(t) as a Fou-
rier series, substitution of which into the equation of motion
leads 共in special cases兲 to a set of recursion relations for the
Fourier coefficients. In the quantum theory, Heisenberg
wrote that32 ‘‘we are at present forced to adopt this method
of solving equation Eq. 共12兲 关his Eq. 共H11兲兴 ... since it was
not possible to define a quantum-theoretical function analo-
gous to the 关classical兴 function x(n,t).’’ In Sec. III we shall
consider the simple example 共the first of those chosen by
Heisenberg兲 f (x)⫽ ␻ 20 x⫹␭x 2 , and obtain the appropriate re-
cursion relations in the classical and the quantum cases.
A quantum-theoretical reinterpretation of Eq. 共13兲 is simi-
larly required in terms of the transition amplitudes X(n,n
⫺ ␣ ). In the classical case, the substitution of Eq. 共2兲 into
Eq. 共13兲 gives
冖
⬁
mẋ 2 dt⫽2 ␲ m 兺
␣ ⫽⫺⬁
兩 X ␣ 共 n 兲 兩 2 ␣ 2 ␻ 共 n 兲 ⫽nh,
using X ␣ (n)⫽ 关 X ⫺ ␣ (n) 兴 * . Heisenberg argued that Eq. 共14兲
appeared arbitrary in the sense of the correspondence prin-
ciple, because the latter determined J only up to an additive
constant 共times h). He therefore replaced Eq. 共14兲 by the
derivative form 关Eq. 共H15兲兴
⬁
d
h⫽2 ␲ m 兺
␣ ⫽⫺⬁
␣
dn
共 ␣ 兩 X ␣ 共 n 兲 兩 2 ␻ 共 n 兲兲 .
The summation can alternatively be written as over positive
values of ␣, replacing 2 ␲ m by 4 ␲ m. In another crucial
jump, Heisenberg then replaced the differential in Eq. 共15兲
by a difference, giving
1372 Am. J. Phys., Vol. 72, No. 11, November 2004
⬁
h⫽4 ␲ m 兺
␣ ⫽0
关 兩 X 共 n⫹ ␣ ,n 兲 兩 2 ␻ 共 n⫹ ␣ ,n 兲 ⫺ 兩 X 共 n,n
⫺ ␣ 兲 兩 2 ␻ 共 n,n⫺ ␣ 兲兴 , 共16兲
which is Eq. 共H16兲 in our notation. As he later recalled, he
34
had noticed that ‘‘if I wrote down this 关presumably Eq. 共15兲兴
and tried to translate it according to the scheme of dispersion
theory, I got the Thomas-Kuhn sum rule 关Eq. 共16兲35,36兴. And
that is the point. Then I thought, That is apparently how it is
done.’’ 37
By ‘‘the scheme of dispersion theory,’’ Heisenberg re-
ferred to what Jammer38 calls Born’s correspondence rule,
namely39
⳵⌽共 n 兲
␣ ↔⌽ 共 n 兲 ⫺⌽ 共 n⫺ ␣ 兲 , 共17兲
⳵n
共12兲 or rather to its iteration to the form40
⳵ ⌽ 共 n, ␣ 兲
␣ ↔⌽ 共 n⫹ ␣ ,n 兲 ⫺⌽ 共 n,n⫺ ␣ 兲 , 共18兲
⳵n
共13兲
as used in the Kramers–Heisenberg theory of dispersion.41,42
It took Born only a few days to show that Heisenberg’s quan-
tum condition, Eq. 共16兲, was the diagonal matrix element of
Eq. 共11兲, and to guess43 that the off-diagonal elements of
x̂p̂⫺p̂x̂ were zero, a result that was shown to be compatible
with the equations of motion by Born and Jordan.4
At this point it is appropriate to emphasize that Heisen-
berg’s transition amplitude X(n,n⫺ ␣ ) is the same as the
quantum-mechanical matrix element 具 n⫺ ␣ 兩 x̂ 兩 n 典 , where 兩 n 典
is the eigenstate with energy W(n). The relation of Eq. 共16兲
to the fundamental commutator Eq. 共11兲 is discussed briefly
in Appendix A.
Heisenberg noted44 that the undetermined constant still
contained in the quantities X of Eq. 共16兲 关assuming the fre-
quencies known from Eq. 共12兲兴 would be determined by the
condition that a ground state should exist, from which no
radiation is emitted 关see Eqs. 共51兲 and 共52兲 below兴. He there-
fore summarized the state of affairs thus far by the
statement44 that Eqs. 共12兲 and 共16兲 ‘‘if soluble, contain a
complete determination not only of frequencies and energy
values, but also of quantum-theoretical transition probabili-
ties.’’ We draw attention to the strong claim here: that he has
arrived at a new calculational method, which will completely
determine the observable quantities. Let us now see in detail
how this method works, for a harmonic oscillator perturbed
by an anharmonic force of the form ␭x 2 per unit mass.45
共14兲 III. HEISENBERG’S CALCULATIONAL METHOD
AND ITS APPLICATION TO THE ANHARMONIC
OSCILLATOR
A. Recursion relations in the quantum case
The classical equation of motion is
ẍ⫹ ␻ 20 x⫹␭x 2 ⫽0. 共19兲
共15兲
We depart from the order of Heisenberg’s presentation and
begin by showing how—as he stated—Eq. 共19兲 leads to re-
cursion relations for the transition amplitudes X(n,n⫺ ␣ ).
The (n,n⫺ ␣ ) representative46 of the first two terms in Eq.
共19兲 is straightforward, being
关 ⫺ ␻ 2 共 n,n⫺ ␣ 兲 ⫹ ␻ 20 兴 X 共 n,n⫺ ␣ 兲 e i ␻ (n,n⫺ ␣ )t , 共20兲
Aitchison, MacManus, and Snyder 1372
while that of the third term is, by Eq. 共10兲,
␭ 兺␤ X 共 n,n⫺ ␤ 兲 X 共 n⫺ ␤ ,n⫺ ␣ 兲 e i ␻ (n,n⫺ ␣ )t .
The (n,n⫺ ␣ ) representative of Eq. 共19兲 therefore yields47
关 ␻ 20 ⫺ ␻ 2 共 n,n⫺ ␣ 兲兴 X 共 n,n⫺ ␣ 兲 ⫹␭ 兺␤ X 共 n,n⫺ ␤ 兲
⫻X 共 n⫺ ␤ ,n⫺ ␣ 兲 ⫽0,
which generates a recursion relation for each value of ␣ ( ␣
⫽0,⫾1,⫾2,...). For example, for ␣ ⫽0 we obtain
␻ 20 X 共 n,n 兲 ⫹␭ 关 X 共 n,n 兲 X 共 n,n 兲 ⫹X 共 n,n⫺1 兲 X 共 n⫺1,n 兲
⫹X 共 n,n⫹1 兲 X 共 n⫹1,n 兲 ⫹¯]⫽0.
No general solution for this infinite set of nonlinear algebraic
equations seems to be possible, so, following Heisenberg, we
turn to a perturbative approach.
B. Perturbation theory
To make the presentation self-contained, we need to dis-
cuss several ancillary results. Heisenberg began by consider-
ing the perturbative solution of the classical equation 共12兲.
He wrote the solution in the form
x 共 t 兲 ⫽␭a 0 ⫹a 1 cos ␻ t⫹␭a 2 cos 2 ␻ t⫹␭ 2 a 3 cos 3 ␻ t
⫹¯⫹␭ ␣ ⫺1 a ␣ cos ␣ ␻ t⫹¯ ,
where the coefficients a ␣ , and ␻, are to be expanded as a
power series in ␭, the first terms of which are independent of
␭:48
a 0 ⫽a (0)
0 ⫹␭a 0 ⫹␭ a 0 ⫹¯
(1) 2 (2)
, 共25a兲
a 1 ⫽a (0)
1 ⫹␭a 1 ⫹␭ a 1 ⫹¯
(1) 2 (2)
, 共25b兲
and
␻ ⫽ ␻ 0 ⫹␭ ␻ (1) ⫹␭ 2 ␻ (2) ⫹¯ .
We substitute Eq. 共24兲 into Eq. 共12兲, use standard trigono-
metric identities, and equate to zero the terms that are con-
stant and which multiply cos ␻t, cos 2␻t, etc., to obtain
␭ 兵 ␻ 20 a 0 ⫹ 21 a 21 ⫹ 关 ␭ 2 共 a 20 ⫹ 21 a 22 兲 ⫹¯ 兴 其 ⫽0,
共 ⫺ ␻ 2 ⫹ ␻ 20 兲 a 1 ⫹ 关 ␭ 2 共 a 1 a 2 ⫹2a 0 a 1 兲 ⫹¯ 兴 ⫽0,
␭ 兵 共 ⫺4 ␻ 2 ⫹ ␻ 20 兲 a 2 ⫹ 21 a 21 ⫹ 关 ␭ 2 共 a 1 a 3 ⫹2a 0 a 2 兲 ⫹¯ 兴 其
⫽0,
␭ 2 兵 共 ⫺9 ␻ 2 ⫹ ␻ 20 兲 a 3 ⫹a 1 a 2 ⫹ 关 ␭ 2 共 a 1 a 4 ⫹2a 0 a 3 兲 ⫹¯ 兴 其
⫽0,
where the dots stand for higher powers of ␭. If we drop the
terms of order ␭ 2 共and higher powers兲, and cancel overall
factors of ␭, Eq. 共27兲 becomes 共for ␭⫽0 and a 1 ⫽0)
␻ 20 a 0 ⫹ 21 a 21 ⫽0,
共 ⫺ ␻ 2 ⫹ ␻ 20 兲 ⫽0,
1373 Am. J. Phys., Vol. 72, No. 11, November 2004
共 ⫺4 ␻ 2 ⫹ ␻ 20 兲 a 2 ⫹ 21 a 21 ⫽0, 共28c兲
共21兲 共 ⫺9 ␻ 2 ⫹ ␻ 20 兲 a 3 ⫹a 1 a 2 ⫽0, 共28d兲
which is the same as Eq. 共H18兲. The lowest order in ␭
49
solution is obtained from Eq. 共28兲 by setting ␻ ⫽ ␻ 0 , and
replacing each a ␣ by the corresponding one with a super-
script (0) 关see Eq. 共25兲兴.
In the quantum case, Heisenberg proposed to seek a solu-
共22兲 tion analogous to Eq. 共24兲. Of course, it is now a matter of
using the representation of x(t) in terms of the quantities
X(n,n⫺ ␣ )exp关i␻(n,n⫺␣)t兴. But it seems reasonable to as-
sume that, as the index ␣ increases from zero in integer
steps, each successive amplitude will 共to leading order in ␭兲
be suppressed by an additional power of ␭, as in the classical
共23兲 case. Thus Heisenberg suggested that, in the quantum case,
x(t) should be represented by terms of the form
␭a 共 n,n 兲 , a 共 n,n⫺1 兲 cos ␻ 共 n,n⫺1 兲 t,
␭a 共 n,n⫺2 兲 cos ␻ 共 n,n⫺2 兲 t,...,
␭ ␣ ⫺1 a 共 n,n⫺ ␣ 兲 cos ␻ 共 n,n⫺ ␣ 兲 t,..., 共29兲
where, as in Eqs. 共25兲 and 共26兲,
a 共 n,n 兲 ⫽a (0) 共 n,n 兲 ⫹␭a (1) 共 n,n 兲 ⫹␭ 2 a (2) 共 n,n 兲 ⫹¯ ,
共30兲
a 共 n,n⫺1 兲 ⫽a (0) 共 n,n⫺1 兲 ⫹␭a (1) 共 n,n⫺1 兲
⫹␭ 2 a (2) 共 n,n⫺1 兲 ⫹¯ , 共31兲
and
共24兲
␻ 共 n,n⫺ ␣ 兲 ⫽ ␻ (0) 共 n,n⫺ ␣ 兲 ⫹␭ ␻ (1) 共 n,n⫺ ␣ 兲
⫹␭ 2 ␻ (2) 共 n,n⫺ ␣ 兲 ⫹¯ . 共32兲
As Born and Jordan pointed out,4 some use of correspon-
dence arguments has been made here in assuming that as ␭
→0, only transitions between adjacent states are possible.
We shall return to this point in Sec. III C.
Heisenberg then simply wrote down what he asserted to be
the quantum version of Eq. 共28兲, namely50
共26兲
␻ 20 a 共 n,n 兲 ⫹ 41 关 a 2 共 n⫹1,n 兲 ⫹a 2 共 n,n⫺1 兲兴 ⫽0 共33兲
⫺ ␻ 2 共 n,n⫺1 兲 ⫹ ␻ 20 ⫽0, 共34兲
关 ⫺ ␻ 2 共 n,n⫺2 兲 ⫹ ␻ 20 兴 a 共 n,n⫺2 兲 ⫹ 21 关 a 共 n,n⫺1 兲
共27a兲
⫻a 共 n⫺1,n⫺2 兲 ]⫽0, 共35兲
共27b兲
关 ⫺ ␻ 2 共 n,n⫺3 兲 ⫹ ␻ 20 兴 a 共 n,n⫺3 兲 ⫹ 21 a 共 n,n⫺1 兲
共27c兲 ⫻a 共 n⫺1,n⫺3 兲 ⫹ 21 a 共 n,n⫺2 兲 a 共 n⫺2,n⫺3 兲 ⫽0. 共36兲
The question we now address is how did Heisenberg arrive at
Eqs. 共33兲–共36兲?
共27d兲 We shall show that these equations can be straightfor-
wardly derived from Eq. 共22兲 using the ansatz 共29兲, and we
suggest that this is what Heisenberg did. This seems to be a
novel proposal. Tomonaga8 derived Eq. 共22兲 but then dis-
cussed only the ␭→0 limit, that is, the simple harmonic
oscillator, a special case to which we shall return in Sec.
共28a兲 III C. The only other authors, to our knowledge, who have
discussed the presumed details of Heisenberg’s calculations
共28b兲 are51 Mehra and Rechenberg.11 They suggest that Heisenberg
Aitchison, MacManus, and Snyder 1373
guessed how to ‘‘translate,’’ ‘‘reinterpret,’’ or ‘‘reformulate’’
共their words兲 the classical equation 共28兲 into the quantum
ones, Eqs. 共33兲–共36兲, in a way that was consistent with his
multiplication rule, Eq. 共10兲. Although such ‘‘inspired guess-
work’’ was undoubtedly necessary in the stages leading up to
Heisenberg’s paper,1 it seems more plausible to us that by the
time of the paper’s final formulation, Heisenberg realized
that he had a calculational method in which guesswork was
no longer necessary, and in which Eqs. 共33兲–共36兲, in particu-
lar, could be derived.
Unfortunately, we know of no documentary evidence that
directly proves 共or disproves兲 this suggestion, but we think
there is some internal evidence for it. In the passage to which
attention was drawn earlier,44 Heisenberg asserted that his
formalism constituted a complete method for calculating ev-
erything that needs to be calculated. It is difficult to believe
that Heisenberg did not realize that his method led directly to
Eqs. 共33兲–共36兲, without the need for any ‘‘translations’’ of
the classical relations.
To apply the ansatz of Eq. 共29兲 to Eq. 共22兲, we need to
relate the amplitudes X(n,n⫺ ␣ ) to the corresponding quan-
tities ␭ ␣ ⫺1 a(n,n⫺ ␣ ). We first note that in the classical
case,
X ␣ 共 n 兲 ⫽X ⫺
* ␣共 n 兲 ,
because x(t) in Eq. 共2兲 has to be real. Consider, without loss
of generality, the case ␣ ⬎0. Then the quantum-theoretical
analogue of the left-hand side of Eq. 共37兲 is X(n,n⫺ ␣ ), and
that of the right-hand side is X * (n⫺ ␣ ,n) 共see Ref. 27兲.
Hence the quantum-theoretical analogue of Eq. 共37兲 is
X 共 n,n⫺ ␣ 兲 ⫽X * 共 n⫺ ␣ ,n 兲 ,
which is nothing but the relation 具 n⫺ ␣ 兩 x̂ 兩 n 典 ⫽ 具 n 兩 x̂ 兩 n
⫺ ␣ 典 * for the Hermitian observable x̂. Although X(n,n
⫺ ␣ ) can in principle be complex 共and Heisenberg twice dis-
cussed the significance of the phases of such amplitudes兲,
Heisenberg seems to have assumed 共as is certainly plausible兲
that in the context of the classical cosine expansion in Eq.
共24兲 and the corresponding quantum terms in Eq. 共29兲, the
X(n,n⫺ ␣ )’s should be chosen to be real, so that Eq. 共38兲
becomes
X 共 n,n⫺ ␣ 兲 ⫽X 共 n⫺ ␣ ,n 兲 ,
that is, the matrix with elements 兵 X(n,n⫺ ␣ ) 其 is symmetric.
Consider a typical term of Eq. 共29兲,
␭ ␣ ⫺1 a 共 n,n⫺ ␣ 兲 cos关 ␻ 共 n,n⫺ ␣ 兲 t 兴
␭ ␣ ⫺1
⫽ a 共 n,n⫺ ␣ 兲关 e i ␻ (n,n⫺ ␣ )t ⫹e ⫺i ␻ (n,n⫺ ␣ )t 兴
2
␭ ␣ ⫺1
⫽ a 共 n,n⫺ ␣ 兲关 e i ␻ (n,n⫺ ␣ )t ⫹e i ␻ (n⫺ ␣ ,n)t 兴 ,
2 We are now ready to obtain the solutions.
using ␻ (n,n⫺ ␣ )⫽⫺ ␻ (n⫺ ␣ ,n) from Eq. 共1兲. If we assume
that a(n,n⫺ ␣ )⫽a(n⫺ ␣ ,n) as discussed for Eq. 共39兲, we
see that it is consistent to write
␭ ␣ ⫺1
X 共 n,n⫺ ␣ 兲 ⫽ a 共 n,n⫺ ␣ 兲 共 ␣ ⬎0 兲
2
and in general
␭ 兩 ␣ 兩 ⫺1
X 共 n,n⫺ ␣ 兲 ⫽ a 共 n,n⫺ ␣ 兲 共 ␣ ⫽0 兲 .
2
1374 Am. J. Phys., Vol. 72, No. 11, November 2004
The case ␣ ⫽0 is clearly special, with X(n,n)⫽␭a(n,n).
We may now write out the recurrence relations Eq. 共22兲
explicitly for ␣ ⫽0,1,2,..., in terms of a(n,n⫺ ␣ ) rather than
X(n,n⫺ ␣ ). We shall include terms up to and including
terms of order ␭ 2 . For ␣ ⫽0 we obtain
␭ 兵 ␻ 20 a 共 n,n 兲 ⫹ 41 关 a 2 共 n⫹1,n 兲 ⫹a 2 共 n,n⫺1 兲兴 ⫹␭ 2 关 a 2 共 n,n 兲
⫹ 14 共 a 2 共 n⫹2,n 兲 ⫹a 2 共 n,n⫺2 兲兲 ] 其 ⫽0. 共43兲
We note the connection with Eq. 共27a兲, and that Eq. 共43兲
reduces to Eq. 共33兲 when the ␭ 2 term is dropped and an
overall factor of ␭ is canceled. Similarly, for ␣ ⫽1 we obtain
共 ⫺ ␻ 2 共 n,n⫺1 兲 ⫹ ␻ 20 兲 a 共 n,n⫺1 兲 ⫹␭ 2 兵 a 共 n,n 兲 a 共 n,n⫺1 兲
⫹a 共 n,n⫺1 兲 a 共 n⫺1,n⫺1 兲 ⫹ 21 关 a 共 n,n⫹1 兲
⫻a 共 n⫹1,n⫺1 兲 ⫹a 共 n,n⫺2 兲 a 共 n⫺2,n⫺1 兲兴 其 ⫽0 共44兲
关see Eq. 共27b兲兴. For ␣ ⫽2 we have
␭ 兵 共 ⫺ ␻ 2 共 n,n⫺2 兲 ⫹ ␻ 20 兲 a 共 n,n⫺2 兲 ⫹ 21 a 共 n,n⫺1 兲
⫻a 共 n⫺1,n⫺2 兲 ⫹␭ 2 关 a 共 n,n 兲 a 共 n,n⫺2 兲 ⫹a 共 n,n⫺2 兲
共37兲 ⫻a 共 n⫺2,n⫺2 兲 ⫹ 21 a 共 n,n⫹1 兲 a 共 n⫹1,n⫺2 兲
⫹ 12 a 共 n,n⫺3 兲 a 共 n⫺3,n⫺2 兲 ] 其 ⫽0 共45兲
关see Eq. 共27c兲兴. For ␣ ⫽3 关see Eq. 共27d兲兴 we obtain
␭ 2 兵 共 ⫺ ␻ 2 共 n,n⫺3 兲 ⫹ ␻ 20 兲 a 共 n,n⫺3 兲 ⫹ 21 关 a 共 n,n⫺1 兲
共38兲 ⫻a 共 n⫺1,n⫺3 兲 ⫹a 共 n,n⫺2 兲 a 共 n⫺2,n⫺3 兲 ]
⫹␭ 2 关 a 共 n,n 兲 a 共 n,n⫺3 兲 ⫹a 共 n,n⫺3 兲 a 共 n⫺3,n⫺3 兲
⫹ 12 a 共 n,n⫹1 兲 a 共 n⫹1,n⫺3 兲 ⫹ 21 a 共 n,n⫺4 兲
⫻a 共 n⫺4,n⫺3 兲 ] 其 ⫽0. 共46兲
If we drop the terms multiplied by ␭ , Eqs. 共43兲–共46兲 reduce
2
to Eqs. 共33兲–共36兲. This appears to be the first published deri-
vation of the latter equations.
In addition to these recurrence relations which follow
from the equations of motion, we also need the perturbative
共39兲 version of the quantum condition Eq. 共16兲.52 We include
terms of order ␭ 2 , consistent with Eqs. 共43兲–共46兲, so that Eq.
共16兲 becomes
h
⫽a 2 共 n⫹1,n 兲 ␻ 共 n⫹1,n 兲 ⫺a 2 共 n,n⫺1 兲 ␻ 共 n,n⫺1 兲
␲m
⫹␭ 2 关 a 2 共 n⫹2,n 兲 ␻ 共 n⫹2,n 兲 ⫺a 2 共 n,n⫺2 兲
⫻ ␻ 共 n,n⫺2 兲兴 . 共47兲
共40兲
C. The lowest-order solutions for the amplitudes and
frequencies
We begin by considering the lowest-order solutions in
共41兲 which all ␭ 2 terms are dropped from Eqs. 共43兲 to 共47兲, and
all quantities (a’s and ␻’s兲 are replaced by the corresponding
ones with a superscript (0) 关compare Eqs. 共30兲–共32兲兴.53 In
this case, Eq. 共44兲 reduces to
共42兲 关 ⫺ 共 ␻ (0) 共 n,n⫺1 兲兲 2 ⫹ ␻ 20 兴 a (0) 共 n,n⫺1 兲 ⫽0, 共48兲
Aitchison, MacManus, and Snyder 1374
so that assuming a (0) (n,n⫺1)⫽0, we obtain
␻ (0)
共 n,n⫺1 兲 ⫽ ␻ 0 共49兲
a (0) 共 n,n⫺ ␣ 兲 ⫽A ␣
␤␣
␻ 2(
0
␣ ⫺1) 冑 n!
共 n⫺ ␣ 兲 !
, 共63兲

for all n. If we substitute Eq. 共49兲 into the lowest-order ver- where A ␣ is a numerical factor depending on ␣; Eq. 共63兲 is
sion of Eq. 共47兲, we find equivalent to Eq. 共H21兲.
h It is instructive to comment on the relation of the above
⫽ 关 a (0) 共 n⫹1,n 兲兴 2 ⫺ 关 a (0) 共 n,n⫺1 兲兴 2 . 共50兲 results to those that would be obtained in standard quantum-
␲m␻0 mechanical perturbation theory. At first sight, it is surprising
The solution of this difference equation is to see nonzero amplitudes for two-quantum 关Eq. 共61兲兴, three-
quantum 关Eq. 共62兲兴, or ␣-quantum 关Eq. 共63兲兴 transitions ap-
h pearing at lowest order. But we have to remember that in
关 a (0) 共 n,n⫺1 兲兴 2 ⫽ 共 n⫹constant兲 , 共51兲
␲m␻0 Heisenberg’s perturbative ansatz, Eq. 共29兲, the ␣-quantum
amplitude appears multiplied by a factor ␭ ␣ ⫺1 . Thus, for
as given in Eq. 共H20兲.53 To determine the value of the con-
stant, Heisenberg used the idea that in the ground state there example, the lowest order two-quantum amplitude is really
can be no transition to a lower state. Thus ␭a (0) (n,n⫺2), not just a (0) (n,n⫺2). Indeed, such a transi-
tion is to be expected precisely at order ␭ 1 in conventional
关 a (0) 共 0,⫺1 兲兴 2 ⫽0, 共52兲 perturbation theory. The amplitude is 具 n⫺2 兩 x̂ 兩 n 典 where, to
and the constant in Eq. 共51兲 is determined to be zero. Equa- order ␭,
tion 共51兲 then gives 共up to a convention as to sign兲
0 具 k 兩 x̂ 兩 n 典 0
3
1
a (0)
共 n,n⫺1 兲 ⫽ ␤ 冑n, 共53兲 兩 n 典 ⫽ 兩 n 典 0 ⫹ m␭
3 兺
k⫽n 共 n⫺k 兲 ប ␻ 0
兩k典0 . 共64兲

where
␤ ⫽ 共 h/ ␲ m ␻ 0 兲 . 1/2
Equations 共49兲 and 共53兲 were Heisenberg’s first results,
and they pertain to the simple 共unperturbed兲 oscillator. We
can check Eq. 共53兲 against the usual quantum mechanical
calculation via
a (0) 共 n,n⫺1 兲 ⫽2X (0) 共 n,n⫺1 兲 ⫽2 0 具 n⫺1 兩 x̂ 兩 n 典 0 ,
where the states 兩 n 典 0 are unperturbed oscillator eigenstates.
It is well known that54
The operator x̂ 3 connects 兩 n 典 0 to 兩 n⫹3 典 0 , 兩 n⫹1 典 0 , 兩 n
共54兲 ⫺1 典 0 , and 兩 n⫺3 典 0 , and similar connections occur for 0 具 n
⫺2 兩 , so that a nonzero O(␭) amplitude is generated in 具 n
⫺2 兩 x̂ 兩 n 典 .
It is straightforward to check that Eq. 共61兲 is indeed cor-
rect quantum-mechanically, but it is more tedious to check
Eq. 共62兲, and distinctly unpromising to contemplate checking
共55兲 Eq. 共63兲 by doing a conventional perturbation calculation to
order ␣ ⫺1. For this particular problem, the improved pertur-
bation theory represented by Eq. 共29兲 is clearly very useful.

冉 冊
After having calculated the amplitudes for this problem to
ប 1/2

0 具 n⫺1 兩 x̂ 兩 n 典 0 ⫽ 冑n, 共56兲

lowest order, Heisenberg next considered the energy. Unfor-
2m ␻ 0 tunately he again gave no details of his calculation, beyond
saying that he used the classical expression for the energy,
which agrees with Eq. 共53兲, using Eq. 共54兲. A similar treat- namely
ment of Eq. 共43兲 leads to
␤2 W⫽ 12 mẋ 2 ⫹ 21 m ␻ 20 x 2 ⫹ 31 m␭x 3 . 共65兲
a (0)
共 n,n 兲 ⫽⫺ 共 2n⫹1 兲 . 共57兲
4 ␻ 20 It seems a reasonable conjecture, however, that he replaced
Turning next to Eq. 共45兲, the lowest-order form is each term in Eq. 共65兲 by its corresponding matrix, as dis-
cussed in Sec. II A. Thus x 2 , for example, is represented by
共 ⫺ 关 ␻ (0) 共 n,n⫺2 兲兴 2 ⫹ ␻ 20 兲 a (0) 共 n,n⫺2 兲 a matrix whose (n,n⫺ ␣ ) element is
⫹ 12 a (0) 共 n,n⫺1 兲 a (0) 共 n⫺1,n⫺2 兲 ⫽0. 共58兲
Because the combination law Eq. 共8兲 must be true for the 兺␤ X 共 n,n⫺ ␤ 兲 X 共 n⫺ ␤ ,n⫺ ␣ 兲 e i ␻ (n,n⫺ ␣ )t , 共66兲
lowest-order frequencies, we have
according to his multiplication rule, Eq. 共10兲. A similar re-
␻ (0) 共 n,n⫺2 兲 ⫽ ␻ (0) 共 n,n⫺1 兲 ⫹ ␻ (0) 共 n⫺1,n⫺2 兲 ⫽2 ␻ 0 , placement is made for x 3 , and ẋ 2 is replaced by
共59兲
where we have used Eq. 共49兲, and in general 兺␤ i ␻ 共 n,n⫺ ␤ 兲 X 共 n,n⫺ ␤ 兲 e i ␻ (n,n⫺ ␤ )t
␻ (0) 共 n,n⫺ ␣ 兲 ⫽ ␣ ␻ 0 共 ␣ ⫽1,2,3,... 兲 . 共60兲
If we use Eqs. 共53兲, 共59兲, and 共60兲, we obtain ⫻i ␻ 共 n⫺ ␤ ,n⫺ ␣ 兲 X 共 n⫺ ␤ ,n⫺ ␣ 兲 e i ␻ (n⫺ ␤ ,n⫺ ␣ )t

␤2
a (0) 共 n,n⫺2 兲 ⫽
6 ␻ 20
冑n 共 n⫺1 兲 . 共61兲 ⫽ 兺␤ ␻ 共 n,n⫺ ␤ 兲 ␻ 共 n⫺ ␣ ,n⫺ ␤ 兲 X 共 n,n⫺ ␤ 兲
A similar treatment of Eq. 共46兲 yields ⫻X 共 n⫺ ␤ ,n⫺ ␣ 兲 e i ␻ (n,n⫺ ␣ )t , 共67兲
␤3 using ␻ (n,m)⫽⫺ ␻ (m,n). The total energy is represented
a (0) 共 n,n⫺3 兲 ⫽ 冑n 共 n⫺1 兲共 n⫺2 兲 . 共62兲
48␻ 40 by the matrix with elements
Consideration of the lowest-order term in Eq. 共22兲 leads to W 共 n,n⫺ ␣ 兲 e i ␻ (n,n⫺ ␣ )t . 共68兲

1375 Am. J. Phys., Vol. 72, No. 11, November 2004 Aitchison, MacManus, and Snyder 1375
It follows that if energy is to be conserved 共that is, time-
independent兲 the off-diagonal elements must vanish:
W 共 n,n⫺ ␣ 兲 ⫽0. 共 ␣ ⫽0 兲 .
The term ␣ ⫽0 is time-independent, and may be taken to be
the energy in the state n. The crucial importance of checking
the condition Eq. 共69兲 was clearly appreciated by Heisen-
berg.
To lowest order in ␭, the last term in Eq. 共65兲 may be
dropped. Furthermore, referring to Eq. 共29兲, the only
␭-independent terms in the X-amplitudes are those involving
one-quantum jumps such as n→n⫺1, corresponding in low-
est order to amplitudes such as X (0) (n,n⫺1)⫽ 21 a (0) (n,n
⫺1). It then follows from Eqs. 共66兲 and 共67兲 that the ele-
ments W(n,n), W(n,n⫺2) and W(n,n⫹2), and only these
elements, are independent of ␭ when evaluated to lowest
order. In Appendix B we show that W(n,n⫺2) vanishes to
lowest order, and W(n,n⫹2) vanishes similarly. Thus, to
lowest order in ␭, the energy is indeed conserved 共as Heisen-
berg noted兲, and is given 关using Eq. 共66兲 and Eq. 共67兲 with
␣ ⫽0 and ␤ ⫽⫾1] by
W 共 n,n 兲 ⫽ 21 m 关 ␻ (0) 共 n,n⫺1 兲兴 2 关 X (0) 共 n,n⫺1 兲兴 2
⫹ 12 m 关 ␻ (0) 共 n⫹1,n 兲兴 2 关 X (0) 共 n⫹1,n 兲兴 2
⫹ 12 m ␻ 20 关 X (0) 共 n,n⫺1 兲兴 2 ⫹ 21 m ␻ 20
⫻ 关 X (0) 共 n⫹1,n 兲兴 2
⫽ 共 n⫹ 21 兲 ប ␻ 0 ,
where we have used Eqs. 共49兲, 共53兲, and 共54兲. Equation 共70兲
is the result given by Heisenberg in Eq. 共H23兲.
These lowest order results are the only ones Heisenberg
reported for the ␭x 2 term. We do not know whether he car-
ried out higher-order calculations for this case or not. What
he wrote next55 is that the ‘‘more precise calculation, taking
into account higher order approximations in W, a, ␻ will
now be carried out for the simpler example of an anharmonic
oscillator ẍ⫹ ␻ 20 ⫹␭x 3 ⫽0.’’ This case is slightly simpler be-
cause in the expression corresponding to the ansatz 共29兲 only
the odd terms are present, that is, a 1 ,␭a 3 ,␭ 2 a 5 , etc.
The results Heisenberg stated for the ␭x 3 problem include
terms up to order ␭ in the amplitudes, and terms up to order
␭ 2 in the frequency ␻ (n,n⫺1) and in the energy W. Once
again, he gave no details of how he did the calculations. We
believe there can be little doubt that he went through the
algebra of solving the appropriate recurrence relations up to
order ␭ 2 in the requisite quantities. As far as we know, the
details of such a calculation have not been given before, and
we believe that it is worth giving them here, as they are of
both pedagogical and historical interest. In the following sec-
tion we shall obtain the solutions for the ␭x 2 term 共up to
order ␭ 2 ) which we have been considering, rather than start
afresh with the ␭x 3 term. The procedure is the same for both.
Before leaving the lowest order calculations, we address a
question that may have occurred to the reader. Given that, at
this stage in his paper, the main results actually relate to the
simple harmonic oscillator rather than to the anharmonic
one, why did Heisenberg not begin his discussion of toy
models with the simplest one of all, namely the simple har-
monic oscillator? And indeed, is it not possible to apply his
procedure to the simple harmonic oscillator without going
1376 Am. J. Phys., Vol. 72, No. 11, November 2004
through the apparent device of introducing a perturbation,
and then retaining only those parts of the solution that sur-
vive as the perturbation vanishes?
共69兲
For the simple harmonic oscillator, the equation of motion
is ẍ⫹ ␻ 20 x⫽0, which yields
关 ␻ 20 ⫺ ␻ 2 共 n,n⫺ ␣ 兲兴 X 共 n,n⫺ ␣ 兲 ⫽0 共71兲
for the amplitudes X and frequencies ␻. It is reasonable to
retain the quantum condition, Eq. 共16兲, because this condi-
tion is supposed to hold for any force law. If we assume that
the only nonvanishing amplitudes are those involving adja-
cent states 共because, for example, in the classical case only a
single harmonic is present56兲, then because X(n,n⫺1)
⫽ 12 a(n,n⫺1), Eqs. 共16兲 and 共71兲 reduce to Eqs. 共50兲 and
共48兲, respectively, and we quickly recover our previous re-
sults. This is indeed an efficient way to solve the quantum
simple harmonic oscillator.57 For completeness, however, it
would be desirable not to have to make the adjacent states
assumption. Born and Jordan4 showed how this could be
done, but their argument is somewhat involved. Soon there-
after, of course, the wave mechanics of Schrödinger and the
operator approach of Dirac provided the derivations used
ever since.
D. The solutions up to and including ␭ 2 terms
We now turn to the higher order corrections for the ␭x 2
term. Consider Eq. 共44兲 and retain terms of order ␭. We set
关see Eqs. 共25兲 and 共26兲兴
共70兲 ␻ 共 n,n⫺1 兲 ⫽ ␻ 0 ⫹␭ ␻ (1) 共 n,n⫺1 兲 , 共72兲
a 共 n,n⫺1 兲 ⫽a (0) 共 n,n⫺1 兲 ⫹␭a (1) 共 n,n⫺1 兲 , 共73兲
and find
2␭ ␻ 0 ␻ (1) 共 n,n⫺1 兲 a (0) 共 n,n⫺1 兲 ⫽0, 共74兲
so that
␻ (1) 共 n,n⫺1 兲 ⫽0. 共75兲
If we consider Eq. 共44兲 up to terms of order ␭ 2 and employ
Eqs. 共53兲, 共57兲, and 共61兲 for the zeroth-order amplitudes, we
obtain the O(␭ 2 ) correction to ␻ (n,n⫺1) 关see Eq. 共26兲兴:
5␤2
␻ (2) 共 n,n⫺1 兲 ⫽⫺ n. 共76兲
12␻ 30
The corresponding corrections to a(n,n⫺1) are found
from the quantum condition Eq. 共16兲. To order ␭ we set
a 共 n⫹1,n 兲 ⫽a (0) 共 n⫹1,n 兲 ⫹␭a (1) 共 n⫹1,n 兲 , 共77兲
as in Eq. 共73兲, and find
冑n⫹1a (1) 共 n⫹1,n 兲 ⫺ 冑na (1) 共 n,n⫺1 兲 ⫽0. 共78兲
Equation 共78兲 has the solution a (1) (n,n⫺1)⫽constant/ 冑n,
but the condition a (1) (0,⫺1)⫽0 关see Eq. 共52兲兴 implies that
the constant must be zero, and so
a (1) 共 n,n⫺1 兲 ⫽0. 共79兲
In a similar way, we obtain to order ␭ 2
11␤ 3
冑n⫹1a (2) 共 n⫹1,n 兲 ⫺ 冑na (2) 共 n,n⫺1 兲 ⫽ 共 2n⫹1 兲 ,
72␻ 40
共80兲
which has the solution
Aitchison, MacManus, and Snyder 1376
 11␤ 3 for this term in the energy. If we combine Eqs. 共85兲, 共86兲,
a (2) 共 n,n⫺1 兲 ⫽ n 冑n. 共81兲 and 共88兲, we obtain the energy up to order ␭ 2 ,
72␻ 40
We now find the higher order corrections to a(n,n) by
considering Eq. 共43兲. We obtain a (1) (n,n)⫽0 and
冉 冊
W 共 n,n 兲 ⫽ n⫹
1
2
ប ␻ 0⫺
5␭ 2 ប 2 2
12m ␻ 40
共 n ⫹n⫹11/30兲 , 共89兲

a result59 that agrees with classical perturbation theory when

␤4
a (2) 共 n,n 兲 ⫽⫺ 共 30n 2 ⫹30n⫹11兲 . 共82兲 n is large,60 and is in agreement with standard second-order
72␻ 60 perturbation theory in quantum mechanics.61
As mentioned, Heisenberg did not give results for the ␭x 2
Similarly, we find from Eq. 共45兲 a (1) (n,n⫺2)⫽0 and
term beyond zeroth order. He did, however, give the results
3␤4 for the ␭x 3 term up to and including ␭ 2 terms in the energy,
a (2) 共 n,n⫺2 兲 ⫽ 共 2n⫺1 兲 冑n 共 n⫺1 兲 , 共83兲 and ␭ terms in the amplitudes. By ‘‘the energy’’ we mean, as
32␻ 60
usual, the (n,n) element of the energy matrix, which as
where we have used noted in Sec. III C is independent of time. We also should
check that the off-diagonal elements W(n,n⫺ ␣ ) vanish 关see
␻ (2) 共 n,n⫺2 兲 ⫽ ␻ (2) 共 n,n⫺1 兲 ⫹ ␻ (2) 共 n⫺1,n⫺2 兲 Eq. 共69兲兴. These are the terms that would 共if nonzero兲 carry a
5␤2 periodic time-dependence, and Heisenberg wrote62 that ‘‘I
⫽⫺ 共 2n⫺1 兲 . 共84兲 could not prove in general that all periodic terms actually
12␻ 30 vanish, but this was the case for all the terms evaluated.’’ We
These results suffice for our purpose. If n is large, they agree do not know how many off-diagonal terms W(n,n⫺ ␣ ) he
with those obtained for the classical ␭x 2 anharmonic oscil- evaluated, but he clearly regarded their vanishing as a crucial
test of the formalism. In Appendix B we outline the calcula-
lator using the method of successive approximations.58
As an indirect check of their quantum mechanical validity, tion of all off-diagonal terms for the ␭x 2 term up to order ␭,
we now turn to the energy evaluated to order ␭ 2 . Consider as an example of the kind of calculation Heisenberg probably
did, finishing it late one night on Heligoland.63
first the (n,n) element of 21 m ␻ 20 x̂ 2 . This matrix element is
given to order ␭ 2 , by
1
2
m ␻ 20
1
4 再
关共 a (0) 共 n,n⫺1 兲兲 2 ⫹ 共 a (0) 共 n,n⫹1 兲兲 2 兴
IV. CONCLUSION

We have tried to remove some of the barriers to under-

␭2 standing Heisenberg’s 1925 paper by providing the details of
⫹ 关 4 共 a (0) 共 n,n 兲兲 2 ⫹2a (2) 共 n,n⫺1 兲 a (0) 共 n⫺1,n 兲 calculations of the type we believe he performed. We hope
4
that more people will thereby be encouraged to appreciate
⫹2a (2) 共 n,n⫹1 兲 a (0) 共 n⫹1,n 兲 ⫹ 共 a (0) 共 n,n⫺2 兲兲 2 this remarkable paper.

冎 冋 冉 冊
The fact is that Heisenberg’s ‘‘amplitude calculus’’ works,
1 ␤2 1 at least for the simple one-dimensional problems to which he
⫹ 共 a (0) 共 n,n⫹2 兲兲 2 兴 ⫽ m ␻ 20 n⫹
2 2 2 applied it. It is an eminently practical procedure, requiring no

册
sophisticated mathematical knowledge to implement. Be-
5 ␤ 4␭ 2 cause it uses the correct equations of motion and incorpo-
⫹ 共 n 2 ⫹n⫹11/30兲 . 共85兲
12␻ 40 rates the fundamental commutator, Eq. 共11兲, via the quantum
condition, Eq. 共16兲, the answers obtained are correct, in
Similarly, using Eq. 共67兲 up to order ␭ 2 , with ␣ ⫽0, the agreement with conventional quantum mechanics.
(n,n) element of 21 mx̂˙ 2 is found to be We believe that Heisenberg’s approach, as applied to

冋 冉 冊 册
simple dynamical systems, has much pedagogical value, and
1 ␤2 1 5 ␤ 4␭ 2 2 could usefully be included in undergraduate courses on quan-
m ␻ 20 n⫹ ⫺ 共 n ⫹n⫹11/30兲 . 共86兲 tum mechanics. The multiplication rule, Eq. 共10兲, has a con-
2 2 2 24␻ 40
vincing physical rationale, even for those who 共like Heisen-
Finally we consider the (n,n) element of the potential energy berg兲 do not recognize it as matrix multiplication. Indeed,
3 m␭x̂ . To obtain the result to order ␭ , we need to calculate
1 3 2 this piece of quantum physics could provide an exciting ap-
the (n,n) element of x̂ only to order ␭. If we use
3 plication for those learning about matrices in a concurrent
mathematics course. The simple examples of Eq. 共10兲, in
equations such as Eq. 共22兲 or the analogous one for the ␭x̂ 3
x̂ 3 共 n,n 兲 ⫽ 兺␣ 兺␤ X 共 n,n⫺ ␣ 兲 X 共 n⫺ ␣ ,n⫺ ␤ 兲 term, introduce students directly to the fundamental quantum
idea that a transition from one state to another occurs via all
⫻X 共 n⫺ ␤ ,n 兲 , 共87兲 possible intermediate states, something that can take time to
emerge in the traditional wave-mechanical approach. The so-
we find that there are no zeroth-order terms, but twelve terms
lution of the quantum simple harmonic oscillator, sketched at
of order ␭ 关recall that amplitudes such as X(n,n) and the end of Sec III D, is simple in comparison with the stan-
X(n,n⫺2) each carry one power of ␭兴. We evaluate these dard methods. Finally, the type of perturbation theory em-
terms using Eqs. 共53兲, 共57兲, and 共61兲, and obtain ployed here provides an instructive introduction to the tech-
5m␭ 2 ␤ 4 nique, being more easily related to the classical analysis than
⫺ 共 n 2 ⫹n⫹11/30兲 共88兲 is conventional quantum-mechanical perturbation theory
24␻ 20 共which students tend to find very formal兲.

1377 Am. J. Phys., Vol. 72, No. 11, November 2004 Aitchison, MacManus, and Snyder 1377
 It is true that many important problems in quantum me- will contribute, because the amplitudes X(n,n⫺ ␣ ) are sup-
chanics are much more conveniently handled in the wave- pressed by increasing powers of ␭ as ␣ increases. In fact, for
mechanical formalism: unbound problems are an obvious ex- ␣ ⭓2 the leading power of ␭ in W(n,n⫺ ␣ ) is ␭ ␣ ⫺2 , which
ample, but even the Coulomb problem required a famous arises from terms such as X(n,n⫺1)X(n⫺1,n⫺ ␣ ) and
tour de force by Pauli.64 Nevertheless, a useful seed may be ␭X(n,n⫺1)X(n⫺1,n⫺2)X(n⫺2,n⫺ ␣ ). Thus to order ␭,
sown, so that when students meet problems involving a finite we need to calculate only W(n,n⫺1),W(n,n⫺2),W(n,n
number of discrete states—for example, in the treatment of
⫺3).
spin—the introduction of matrices will come as less of a
shock. And they may enjoy the realization that the somewhat 共a兲 W(n,n⫺1). There are four O(␭) contributions to the
mysteriously named ‘‘matrix elements’’ of wave mechanics (n,n⫺1) element of 21 m ␻ 20 x̂ 2 :
are indeed the elements of Heisenberg’s matrices. 1
4 m ␻ 20 ␭ 兵 a (0) 共 n,n 兲 a (0) 共 n,n⫺1 兲 ⫹a (0) 共 n,n⫺1 兲
APPENDIX A: THE QUANTUM CONDITION, EQ. ⫻a (0) 共 n⫺1,n⫺1 兲 ⫹ 21 关 a (0) 共 n,n⫹1 兲 a (0) 共 n⫹1,n⫺1 兲
„16…, AND x̂p̂Àp̂x̂Äiប
⫹a (0) 共 n,n⫺2 兲 a (0) 共 n⫺2,n⫺1 兲兴 其
Consider the (n,n) element of (x̂x̂˙ ⫺x̂˙ x̂), which is
⫽⫺ 5
24 m␭ ␤ 3 n 冑n. 共B1兲
兺␣ X 共 n,n⫺ ␣ 兲 i ␻ 共 n⫺ ␣ ,n 兲 X 共 n⫺ ␣ ,n 兲 There are two O(␭) contributions to the (n,n⫺1) element
of 12 mx̂˙ 2 :
⫺ 兺␣ i ␻ 共 n,n⫺ ␣ 兲 X 共 n,n⫺ ␣ 兲 X 共 n⫺ ␣ ,n 兲 . 共A1兲
⫺ 18 ␭m 兵 ␻ (0) 共 n,n⫹1 兲 ␻ (0) 共 n⫹1,n⫺1 兲 a (0) 共 n,n⫹1 兲
In the first term of Eq. 共A1兲, the sum over ␣ ⬎0 may be ⫻a (0) 共 n⫹1,n⫺1 兲 ⫹ ␻ (0) 共 n,n⫺2 兲 ␻ (0) 共 n⫺2,n⫺1 兲
rewritten as
⫻a (0) 共 n,n⫺2 兲 a (0) 共 n⫺2,n⫺1 兲 其 ⫽ 1
12 m␭ ␤ 3 n 冑n. 共B2兲
⫺i 兺 ␻ 共 n,n⫺ ␣ 兲 兩 X 共 n,n⫺ ␣ 兲 兩
␣ ⬎0
2
共A2兲
There are three O(␭) contributions to the (n,n⫺1) element
using ␻ (n,n⫺ ␣ )⫽⫺ ␻ (n⫺ ␣ ,n) from Eq. 共1兲 and X(n of 13 m␭x̂ 3 :
⫺ ␣ ,n)⫽X * (n,n⫺ ␣ ) from Eq. 共38兲. Similarly, the sum over 1
m␭ 兵 a (0) 共 n,n⫺1 兲 a (0) 共 n⫺1,n 兲 a (0) 共 n,n⫺1 兲
24
␣ ⬍0 becomes
⫹a (0) 共 n,n⫺1 兲 a (0) 共 n⫺1,n⫺2 兲 a (0) 共 n⫺2,n⫺1 兲
i 兺
␣ ⬎0
␻ 共 n⫹ ␣ ,n 兲 兩 X 共 n⫹ ␣ ,n 兲 兩 2 共A3兲 ⫹a (0) 共 n,n⫹1 兲 a (0) 共 n⫹1,n 兲 a (0) 共 n,n⫺1 兲 其

on changing ␣ to ⫺ ␣ . Similar steps for the second term of ⫽ 18 m␭ ␤ 3 n 冑n. 共B3兲

Eq. 共A1兲 lead to the result
共 x̂x̂˙ ⫺x̂˙ x̂ 兲共 n,n 兲 ⫽2i
The sum of Eqs. 共B1兲–共B3兲 vanishes, as required.
共b兲 W(n,n⫺2). The leading contribution is independent
兺
␣ ⬎0
关 ␻ 共 n⫹ ␣ ,n 兲 兩 X 共 n⫹ ␣ ,n 兲 兩 2
of ␭. From the term 21 m ␻ 20 x̂ 2 , it is

⫺ ␻ 共 n,n⫺ ␣ 兲 兩 X 共 n,n⫺ ␣ 兲 兩 2 兴 1
8 m ␻ 20 a (0) 共 n,n⫺1 兲 a (0) 共 n⫺1,n⫺2 兲 , 共B4兲
⫽2ih/ 共 4 ␲ m 兲 , 共A4兲
which is canceled by the corresponding term from 21 mx̂˙ 2 .
where the last step follows from Eq. 共16兲. We set p̂⫽mx̂˙ and The next terms are O(␭ 2 ), for example from the leading
find term in the (n,n⫺2) element of 31 ␭mx̂ 3 .
共c兲 W(n,n⫺3). There are two O(␭) contributions from
共 x̂p̂⫺p̂x̂ 兲共 n,n 兲 ⫽iប 共A5兲
2 m ␻ 0 x̂ :
1 2 2

for all values of n. Equation 共A5兲 was found by Born43

shortly after reading Heisenberg’s paper. In further develop-
1
8 m ␻ 20 ␭ 兵 a (0) 共 n,n⫺1 兲 a (0) 共 n⫺1,n⫺3 兲
ments the value of the fundamental commutator x̂p̂⫺ p̂x̂,
⫹a (0) 共 n,n⫺2 兲 a (0) 共 n⫺2,n⫺3 兲 其
namely iប, was taken to be a basic postulate. The sum rule
in Eq. 共16兲 is then derived by taking the (n,n) matrix ele- ⫽ 241 m␭ ␤ 3 冑n 共 n⫺1 兲共 n⫺2 兲 . 共B5兲
ment of the relation 关 x̂, 关 Ĥ,x̂ 兴兴 ⫽ប 2 /m.
There are two O(␭) contributions from 21 mx̂˙ 2 :
APPENDIX B: CALCULATION OF THE OFF- ⫺ 18 m␭ 兵 ␻ (0) 共 n,n⫺1 兲 a (0) 共 n,n⫺1 兲 ␻ (0) 共 n⫺1,n⫺3 兲
DIAGONAL MATRIX ELEMENTS OF THE ENERGY
W„n,nÀ ␣ … FOR THE ␭x 2 TERM ⫻a (0) 共 n⫺1,n⫺3 兲 ⫹ ␻ (0) 共 n,n⫺2 兲
We shall show that, for ␣ ⫽0, all the elements (n,n⫺ ␣ ) ⫻a (0) 共 n,n⫺2 兲 ␻ (0) 共 n⫺2,n⫺3 兲 a (0) 共 n⫺2,n⫺3 兲 其
of the energy operator 12 mx̂˙ 2 ⫹ 21 m ␻ 20 x̂ 2 ⫹ 31 ␭mx̂ 3 vanish up to ⫽⫺ 121 ␭m ␤ 3 冑n 共 n⫺1 兲共 n⫺2 兲 . 共B6兲
order ␭. We begin by noting that at any given order in ␭, only
a limited number of elements W(n,n⫺1),W(n,n⫺2), . . . There is only one O(␭) contribution from m␭x̂ 3 : 1
3

1378 Am. J. Phys., Vol. 72, No. 11, November 2004 Aitchison, MacManus, and Snyder 1378
 1
m␭a (0) 共 n,n⫺1 兲 a (0) 共 n⫺1,n⫺2 兲 a (0) 共 n⫺2,n⫺3 兲
28
Conventional notation, subsequent to Ref. 4, would replace n⫺ ␣ by a
24
second index m, say. We prefer to remain as close as possible to the
⫽ 1
24 ␭m ␤ 3 冑n 共 n⫺1 兲共 n⫺2 兲 . 共B7兲 29
notation of Heisenberg’s paper.
Reference 2, p. 264.
The sum of Eqs. 共B5兲–共B7兲 vanishes, as required.
30
Reference 2, p. 265.
31
Reference 2, p. 266.
32
a兲 Reference 2 p. 267.
Electronic mail: i.aitchisonl@physics.oxford.ac.uk 33
1 This step apparently did not occur to him immediately. See Ref. 11, p. 231.
W. Heisenberg, ‘‘Über quantentheoretische Umdeutung kinematischer und 34
Actually not quite. We have taken the liberty of changing the order of the
mechanischer Beziehungen,’’ Z. Phys. 33, 879– 893 共1925兲.
2
This is the title of the English translation, which is paper 12 in Ref. 3, pp. arguments in the first terms in the braces; this 共correct兲 order is as given in
261–276. We shall refer exclusively to this translation, and to the equa- the equation Heisenberg wrote before Eq. 共H20兲.
35
tions in it as 共H1兲, 共H2兲, ... . W. Thomas, ‘‘Über die Zahl der Dispersionelektronen, die einem station-
3
Sources of Quantum Mechanics, edited by B. L. van der Waerden 共North- ären Zustande zugeordnet sind 共Vorläufige Mitteilung兲,’’ Naturwissen-
Holland, Amsterdam, 1967兲. A collection of reprints in translation. schaften 13, 627 共1925兲.
36
4
M. Born and P. Jordan, ‘‘Zur Quantenmechanik,’’ Z. Phys. 34, 858 – 888 W. Kuhn, ‘‘Über die Gesamtstärke der von einem Zustande ausgehenden
共1925兲, paper 13 in Ref. 3. Absorptionslinien,’’ Z. Phys. 33, 408 – 412 共1925兲, paper 11 in Ref. 3.
37
5
P. A. M. Dirac, ‘‘The fundamental equations of quantum mechanics,’’ W. Heisenberg, as discussed in Ref. 11, pp. 243 ff.
Proc. R. Soc. London, Ser. A 109, 642– 653 共1926兲, paper 14 in Ref. 3.
38
Reference 9, p. 193; ⌽ is any function defined for stationary states.
6
M. Born, W. Heisenberg, and P. Jordan, ‘‘Zur Quantenmechanik II,’’ Z.
39
M. Born, ‘‘Über Quantenmechanik,’’ Z. Phys. 26, 379–395 共1924兲, paper
Phys. 35, 557– 615 共1926兲, paper 15 in Ref. 3. 7 in Ref. 3.
40
7
S. Weinberg, Dreams of a Final Theory 共Pantheon, New York, 1992兲, pp. Reference 9, p. 202.
41
53–54. Weinberg goes on to say that ‘‘Perhaps we should not look too H. A. Kramers and W. Heisenberg, ‘‘Über die Streuung von Strahlen durch
closely at Heisenberg’s first paper ... .’’ We will not follow his suggestion Atome,’’ Z. Phys. 31, 681–707 共1925兲, paper 10 in Ref. 3.
42
here. For considerable further detail on dispersion theory, sum rules, and the
8
S.-I. Tomonaga, Quantum Mechanics: Old Quantum Theory 共North- ‘‘discretization’’ rules, see Ref. 8, pp. 142–147, 206 –208, and Ref. 9, Sec.
Holland, Amsterdam, 1962兲, Vol. 1. 4.3.
43
9
M. Jammer, The Conceptual Development of Quantum Mechanics See Ref. 3, p. 37.
44
共McGraw–Hill, New York, 1966兲. Reference 2, p. 268.
45
10
E. MacKinnon, ‘‘Heisenberg, models and the rise of matrix mechanics,’’ For an interesting discussion of the possible reasons why he chose to try
Hist. Stud. Phys. Sci. 8, 137–188 共1977兲. out his method on the anharmonic oscillator, see Ref. 11, pp. 232–235;
11
J. Mehra and H. Rechenberg, The Historical Development of Quantum and Ref. 3, p. 22. Curiously, most of the commentators—with the notable
Theory 共Springer-Verlag, New York, 1982兲, Vol 2. exception of Tomonaga 共Ref. 8兲—seem to lose interest in the details of the
12
J. Hendry, The Creation of Quantum Mechanics and the Bohr-Pauli Dia- calculations at this point.
logue 共Reidel, Dordrecht, 1984兲.
46
The (n,n⫺ ␣ ) matrix element, in the standard terminology.
13
T.-Y. Wu, Quantum Mechanics 共World Scientific, Singapore, 1986兲.
47
Equation 共22兲 is not in Heisenberg’s paper, although it is given by To-
14
M. Taketani and M. Nagasaki, The Formation and Logic of Quantum monaga, Ref. 8, Eq. (32.20⬘ ).
Mechanics 共World Scientific, Singapore, 2002兲, Vol. 3. 48
Note that this means that, in x(t), all the terms which are of order ␭ p arise
15
G. Birtwistle, The New Quantum Mechanics 共Cambridge U.P., Cambridge, from many different terms in Eq. 共24兲.
49
1928兲. Except that Heisenberg relabeled most of the a ␣ ’s as a ␣ (n).
16
M. Born, Atomic Physics 共Dover, New York, 1989兲. 50
Because of an oversight, he wrote a 0 (n) in place of a(n,n) in Eq. 共33兲.
17
J. Lacki, ‘‘Observability, Anschaulichkeit and abstraction: A journey into 51
MacKinnon 共Ref. 10兲 suggests how, in terms of concepts from the ‘‘virtual
Werner Heisenberg’s science and philosophy,’’ Fortschr. Phys. 50, 440– oscillator’’ model, Eq. 共28兲 may be transformed into Eqs. 共33兲–共36兲. We
458 共2002兲. do not agree with MacKinnon 共Ref. 10, footnote 62兲 regarding ‘‘mistakes’’
18
J. Mehra, The Golden Age of Theoretical Physics 共World Scientific, Sin- in Eqs. 共35兲 and 共36兲.
gapore, 2001兲, Vol. 2. 52
In the version of the quantum condition that Heisenberg gave just before
19
Of these the most detailed are Ref. 3, pp. 28 –35, Ref. 8, pp. 204 –224, Eq. 共H20兲, he unfortunately used the same symbol for the transition am-
Ref. 10, pp. 161–188, and Ref. 11, Chap. IV. plitudes as in 共H16兲–see Eq. 共16兲—but replaced 4 ␲ m by ␲ m, not ex-
20
Reference 7, p. 53. plaining where the factor 1/4 came from 关see Eq. 共42兲兴; he also omitted the
21
Reference 2, p. 262. ␭’s.
22
All quotations are from the English translation of Ref. 2. 53
Heisenberg omitted the superscripts.
23
We use ␻ rather than Heisenberg’s ␯. 54
See, for example, L. I. Schiff, Quantum Mechanics 共McGraw–Hill, New
24
We depart from the notation of Refs. 1 and 2, preferring that of Ref. 8, pp. York, 1968兲, 3rd ed., p. 72.
204 –224. Our X ␣ (n) is Heisenberg’s a ␣ (n). 55
Reference 2, p. 272.
25
The reader may find it helpful at this point to consult Ref. 26, which 56
This is the justification suggested in Ref. 4, p. 297 of paper 13 in Ref. 3.
provides a clear account of the connection between the classical analysis 57
It is essentially that given by L. D. Landau and E. M. Lifshitz, Quantum
of an electron’s periodic motion and simple quantum versions. See also J. Mechanics 共Pergamon, Oxford, 1977兲, 3rd ed., pp. 67– 68.
D. Jackson, Classical Electrodynamics 共Wiley, New York, 1975兲, 2nd ed., 58
L. D. Landau and E. M. Lifshitz, Mechanics 共Pergamon, Oxford, 1976兲,
Sec. 9.2. 3rd ed., pp. 86 – 87.
26
W. A. Fedak and J. J. Prentis, ‘‘Quantum jumps and classical harmonics,’’ 59
This equation corresponds to Eq. 共88兲 of Ref. 4, in which there appears to
Am. J. Phys. 70, 332–344 共2002兲. be a misprint of 17/30 instead of 11/30.
27
The association X ␣ (n)↔X(n,n⫺ ␣ ) is generally true only for non- 60
See Ref. 4, p. 305.
negative ␣. For negative values of ␣, a general term in the classical Fou- 61
Reference 57, p. 136.
rier series is X ⫺ 兩 ␣ 兩 (n)exp关⫺i␻(n)兩␣兩t兴. If we replace ⫺ ␻ (n) 兩 ␣ 兩 by 62
Reference 2, pp. 272–3.
⫺ ␻ (n,n⫺ 兩 ␣ 兩 ), which is equal to ␻ (n⫺ 兩 ␣ 兩 ,n) using Eq. 共1兲, we see that 63
See W. Heisenberg, Physics and Beyond 共Allen & Unwin, London, 1971兲,
X ⫺ 兩 ␣ 兩 (n)↔X(n⫺ 兩 ␣ 兩 ,n). The association X ⫺ 兩 ␣ 兩 ↔X(n,n⫹ 兩 ␣ 兩 ) would not p. 61.
be correct because ␻ (n,n⫹ 兩 ␣ 兩 ) is not the same, in general, as ␻ (n 64
W. Pauli, ‘‘Über das Wasserstoffspektrum vom Standpunkt der neuen
⫺ 兩 ␣ 兩 ,n). Quantenmechanik,’’ Z. Phys. 36, 336 –363 共1926兲, paper 16 in Ref. 3.

1379 Am. J. Phys., Vol. 72, No. 11, November 2004 Aitchison, MacManus, and Snyder 1379