ADVANCED
David Pines, Series Editor
AddisonWesley's Frontiers in Physics series has, since 1961, made it possible for
leading physicists to communicate in coherent fashion their views of recent
developments in the most exciting and active fields of physicswithout
having to devote the time and energy required to prepare a formal review or
monograph. Indeed, throughout its nearly fortyyear existence, the series has
emphasized informality in both style and content, as well as pedagogical clari
ty. Over time, it was expected that these informal accounts would be replaced
by more formal counterpartstextbooks or monographsas the cuttingedge
topics they treated gradually became integrated into the body of physics knowl
edge and reader interest dwindled. However, this has not proven to be the case
for a number of the volumes in the series: Many works have remained in print
on an ondemand basis, while others have such intrinsic value that the physics
community has urged us to extend their life span.
The Advanced Book Classics series has been designed to meet this demand. It
will keep in print those volumes in Frmliers in Physics or its sister series, Lecture
Notes and Suppkments in Physics, that continue to provide a unique account of
a topic of lasting interest. And through a sizable printing, these classics will
be made available at a comparatively modest cost to the reader.
These lecture notes on Richard Ft;ynnran8sCaltech course on Quantum
Electrodynamics were first published in 1961, as part of the first group of lec
ture notelreprint volumes to be included in the Frontiers in Physics series. As is
the case with all of the Feynman lecture note volumes, the presentation in this
work reflects his deep physical insight, the freshness and originality of his
approach to quantum electrodynamics,and the overall pedagogical wizardry of
Richard Feynman. Taken together with the reprints included here of
vi EDITOR" SFOREWQRD
Bavid Pines
Idrbana, Elf inois
December 2 997
Preface
idea that, as new fields of physics are opened up, students must work their way
further back, to earlier stages of the educational program. The first two terms
were the usual quantum mechanical course using Schiff (McGrawHill) as a
main reference (omitting Chapters X, XII, XIII, and XIV, relating to quantum
electrodynamics). However, in order to ease the transition to the latter part of
the course, the theory of propagation and potential scattering was developed
in detail in the way outlined in Eqs. 153 to 155. One other unusual point was
made, namely, that the nonrelativistic Pauli equation could be written as on
page 6 of the notes.
The experiment was unsuccessful. The total material was too much for one
year, and much of the material in these notes is now given after a full year grad
uate course in quantum mechanics.
The notes were originally taken by A. R. Hibbs. They have been edited and
corrected by H. T. Yura and E. R. Huggins.
R. R FEWMAN
Pasadena, California
November 1961
Editor's Foreword
Preface
P o t e n ~ aProblems
l Itn, Quantum Eleetradynamics
Pair creation and annihilation
Consewation of energy
The propagation kernel
Use of the kernel K, ( 2 , l )
Transition probablility
Scattering af an electron from a coulomb potenrial
Galccllation of the propagation kernel for a free particle
Momentum repreenration
CONTENTS
Replcints
Summary of Numerical Factors for Transition
Probabilities, Phys. Rev,, 84, 123 (1951)
The Theory of Positrons. Phys. Rev., 76, 749759 (1949)
S p a c e  x ~Approach to @anturn Electrodynamics.
Phys. Rev., 76,169189 (1949)
This page intentionally left blank
Interaction
of Light ith Matter
Second Lectuw
U W 8 OF QUANTUM EUCTRODPHAmC8
Without justification at this time the "laws of q u m t m ele,ectrdynamics9"
will be stated a s follows:
1, The amplitude that m atomic system will absorb a photon d u r i w the
process of transition from one state to another 18 emctly the same a s the
amplitude that the same transition will be made under the influence of a p
tential equal to that af a classical e l e c t r o m w e t i c wave representing that
photon, provided: (a) the classical wave i s normalized t a represent an en
ergy density equal t a b times the probabilty per cubic centinneter of find
ing the photon; @) the real classical wave is split into two complex waves
e  hU" and e*' w t , and only the e ""tart is kept; md (c) the potential acts
only once in mrturbatlon; that is, only terms to first order in the electro
magnetic fietd strength should be retained.
mplacing the word "absorbedM by 'kmit?' in rule X r s q d r e s only that
the wave represented by exp (+iut) be kept instead of exp (iwt).
2. Tbe n u m b r of states a v a i l h i e p r cubic centimeter of a given po2ar
izatlon Is
Note this ia exactly the same a s the number of normal modes per cubic cen
tirneter in classical theory.
3 , Photons obey BoeeEinstein atatistlcs, That is, the states of a csllese
tion of identical photons must be symmetric (exchmge photons, add m p l i 
tudes). Also the statistical w e i e t of a state of n identical photons i s l in
stead of the elassieal a!
Thus, in general, a photon may be represenbd by a solution of the classi
cal Maxwell equations if properly normalized.
Although many forms of expression a r e p s s i b l e i t is most convenient to
describe the electromagnetic 8eld in terms of plane waves. A plane wave
can always be represented by a vector potential only (scalar potential made
zlero by suitable gauge transformation). The vector potential representing a
real classical wave i s talqen a s
I N T E R A C T I O N QF L I G H T W I T H M A T T E R
A ZZ a e cos ( u t  f?;*x)
Thus
For emission the vector poLentiaL is the s m e except for a positive exponen
tial.
Density of states =
6 QUANTUM E LECTRQDUHAMICS
Tjzzz'rd Lecture
To determine U f i , write
Because of the rule that f i e potential acts only once, which Is the s m e
a s requiring on& firstorder terms to enter, the term in; A * A does not en
ter this problem. Making use of A = aa exp [i ( a t  XC X)] and the two
operator relations
where K@= 0 (which follows from the choice of gauge and the Mmtvell
equatiom), we may write
me matrix element i s
e

UZi =" ae Prr *
m
"f
FTG. 31
Substituting for 2,
Total prob./sec =
e2 3 1 1 x ~2l
54 G k3
iwt
U, (g) = ur, 8
and the probability of aboorption of any photon travicsling in the (B,@) direc
tion is desired, it is necesrJary to integrate aver all frequencisa, Tbis ah
sorption probability is
' 1'
Trans. prob. = 2 ~ ( % ) lulr $)dQ
where
12 Q U A N T U M ELECTROf)'YNAMPCS
This can also be written in terms of the incident iatensity (energy crossing
a unit area in unit time) by noting that
Probability of spontaneous
= 2n(1i)'(2nc)~ j ~ ulk2
~ ~ /
emission/sec
wk12 d o
Trans. prob./sec = 2 n ( ~ ) *lulk nfw,~,(13)(2nc)~
(44)
h what follows, i t will be shown that Eq. (44) is correct even when there is
a possibility of more &an. one photon per: state provided nfo,8,$)is t&en a s
the mean number of photons W r state.
Lf the initial state consists of tulo photons in the same photon state, i t will
not be possible to a s t i n w i s h W1em md the statistical weight of the initial
s t a b will be 1/2 ! However, the m p l l h d e for ab~orptionwill be Mice that
for one photon. TaXcjng the statistical weight times the square of the m p U 
tude for this groesss, the transition prabrtbility per second i s found to be
twice that for only one photon per photon state. m e n there3. a r e Lhree N o 
tons per initial photon state md one is absorbed, the f o l l ~ M n gsix proceases
(shown on Fig. 422) can occur.
Any of the *ree incident photons rnw be absorbed and, in addition, there is
the possibility that the photons which a r e not absorbed may be interchanged.
The statistical wei&t of the initial state is 113 !, the statistical weight of the
final state i s 2/21 , and the amplitude for the process is 6. Thus the t r m s i 
tion probability i s (1/3 1)(1/2 1) (6)' = 3 times that if them were one photon
p r i d t i a t state, In general, the transition probability for n. photons per
initial photon state is n times that for a s i w l e photon per photon state, s o
Eq. (44) is correct if n(cr/,@,$yj) is t&en as the mean n u r n b r of photons per
state.
14 Q U A N T U M EILECTRODUMAMICS
One photon i s incident on the atom and two indlstinwiahable photons come
off. The statistical weight of the final state I s 1/2f and the amplitude for
the process i s 2, s o the probability of emission for this process is twice
that of Eipontanems emission. For n incident photons the statistical weight
of the, initial s t a b is f[nI, the statistical weight of the final s t a k i s
l/(n + 1) f , and the a p f i t u d e for the process is (n + 1)f times the amplitude
for spontaneous emission. The probability e r second) of emission is then
n + l times the probability of spontaneous emission. The n can be said to
account for the i d u c e d part of the transition rate, while the 1 i s the spon
taneous part of the transition rats,
Since the gotentials used in computing the transition probability have
been normalized to om photon Wr cubic centimeter and the trmsition prob
ztbiiity depends on the: square of the amplimde of the potential, it i s clear
t b t when there a r e n photonsr p r photon state the eorreet transition prob
ability for abearption wodd ba obhia& by aormaliziw the potential8 ta n
pt.lotons per cubie centimeter [amplitude 6 times a s large). This i s the
basis for the validity of the socalled semiclassical ~ e o r y of radiation. h
'that theory absorption is calculated a s resultiw from the wrturbation by a
potential normalized to the actual energy In the field, that is, to energy nEw
if there a r e n photons, The correct transition pr&&ilily for e r n i ~ s i o ni s
not obtained this waly, however, because it is proportional to n + 1. The er
r o r corresponds to omitting the spontaneous part of the t r a n ~ i t i o ngrab
ability. In the semiclassical theory of radiation, the spontmeous part of the
emission probability i s arrived a t by general arguments, ixzctuding the fact
that its inclusion leads to the observed Planck distribution formula. Ein
stein first deduced these re lationshipa by sem iclassicsl reasoning.
INTERACTION O F LIGHT WITH M A T T E R
This may be seen from the fact that the coordinates X , y, and z a r e essen
tially the Legendre polynomial PI. If the o r b i k l angular momentum of the
initial sLak i s n, the wave function contains P,, But
Hence for the matrix element not ta vanish, the angular momentum of the
final stale must be n r ~ 1,
. SO that its wave function will conkin either $,?l
o r P,."$.
F o r a complex atam (more than one electron), the &miltonian i s
in the initial state do not interact, then the produet (X* + x2)iEi (xI,xZ) can be
formally regarded a s the wave bnction of a system (atom t otsject) having
possible values of J i c l, J1 , and Jz l for total m ~ l a momentum.
r There
fore the matrix element i s nonzero only if Jf , the final a.nwlar momentum,
* .
has one of the three values JI 1 o r Ji Hence the general selection rule
A J = a 1, O,
Parity, P w i t g is the prowrty of a wave function referring to i t s behaviar
upon reflection of all coordinates, That is, if
parity is even; o r if
parity 18 odd.
If" in the matrix elements involved in the dipole approximation one makes
the change of variable of integration x = X\ the result i s
Hence tlze rule that p a ~ t ymust chawe in allowed transitions. For a one
electron atom, L determines f&e gmity; tiherefore, L1 L = O would be forbid
den. h mmyelectron atoms, L, does not determine the parity (determined
by algebraic, not vector, sum of individual electron m p l a r momfsnta), s o
hL = O transition8 can occur. The a 0 transktions a r e always forbidden,
however, since a photon always c a r r i e s a m u d t of mlf;ular momentum.
All wave functions have ei&er even or d d parity. This can be seen from
the fact that the Hamiltonian (in the absenm of m external magnetic field)
is invariant under the parity operation. Then, if EIJlr(x) = E\k(x), i t i s also
true &at H@(X) = f t ; J i (  ~ ) . Therefore, if the state is nondegeneirab, it
bllows that e i a e r ?Pf X) = Jir (X) o r Q (X) = S [X). If the state is degen
erate, i t is possible that ?fr (X) ;c {X) But then a complete solution would
be one of the linear combinations

% (X) @(X) odd parity
were allowed, but not infinite, Thus, if the colliaion rate i s small enough
(collisions of the second kind ordbnarify cause deexcitation in forbidden
cases), the forbidden trmsition may have sufficient time t s occur.
In the nearly exact matrix element

that 1/2 112 and O 
In the case of electric qudrupole radiation, it is impticit in Lhs rules
X trmasitiona are forbidden (siren though &lmay
& l), since the required chawe of 2 for the vector m m l a r nclomentm i s im
' be
Actually all the implicit selection rules for A J , which become numerous
for the higher multipole orders, can be expressed explicitly by writing the
selection rule a s
where 2' is the multipole order o r 1 i s the vector change in angular mo
ment~,
XNTIERACTllOrJ O F L I G H T W I T H M A T T E R If)
"f For nuclei emitting gamma rays this d m s not seem to be true. Far an
obscure reason the magnetic radiaiLion predominates for each order of muI
tipole.
QUANTUM ELECTRODYNAMICS
not appear in Ulk. The p r d u e t A . A = (A1+ A2) (Ai + Az) will conbibale
only its c r a s s  p r d u c t term 2A1&. The second in&gral will have no con
t r i h t i o n from A . A , but will be efie sum of two terms. The f i r s t term con
tains a U1, based on p * Az and a Unk based an p .At. The second h w Ul,
based on p AI and Unk cm p *AZ. The Lime sequerzcrzs r e s u f t i w in these
two terms can ba represented schematically a s shown in Fig, 62.
The integral resulting from We first term vAlf now be developd in de
lai l.
first
t/
atom
FIG, 6  2
where L& = (EI +&W  Ek  i), and the phase angle c# i s independent of n.
  
A term. wi& the denominator given by (E, b 1 Ek)(EI + b 2 E,) has been
neglected, since previws results show that only energies such that
El + m Ek + tiwl are important. The final result e m be written
The first term under the summation comes from tlxe "first termm 're
viously referred to and t b second from the ""secand term." The last term
in the absolute brackets comes from A * A.
If l k, the scattering is incoherent, m d the result is called the '"man
effect,$8 If l = k, the s c a t b r i n g is coherent,
Further, note that if all the atoms are in, the ground state and l r k, f;hen
the energy of the atom can only i n c r e a ~ eand the frequency of the Ught ~3
can only decrease, This gives r i s e to "Stokes lines ."
The oppsaite e f h c t
gives @%tiStokes Unes .$'
FIG. 63
Einstein postulated [theory of special relativity) that the Newton laws must
be modified in such a way that they, too, a r e unchawed in form under a
Lorentz transformation.
An interestin%; consequence of the Lorentz transformation i s that clocks
appear to run slower in moving systems; that is called time dilation.. In
transforming from one coordinate system to another it i s convenient to use
tensor analysis, To this end, a fourvector will be defined a s a set of four
quantities that transforms in the same way a s x,y,z and ct. The subscript
p will be used to desimate which of the four components is being considered;
for e x m p l e ,
.lxJ
jyJ jzyP (jp ) current (and charge) density
p
$ The energy E, here, is the total energy including the r e s t energy me2.
S P E C I A L RELATXVfTU 25
It i s also inlteresting to note that the phase of a free particle wave function
 
~sxpf fi/tr)fEt p fo] i s invariant since
xnotrixy~ sttltiona xy
particle padicle
hboratarjr system Centerofmass system
QUANTUM ELECTRODYNAMICS
but
and
where use i a made of the fact that (p is the fourth component of the four
vector potential A p e From the foregoing it can be seen that B x , B,, B,, E,,
E,, and E, a r e the components of a secoadra& tensor:

This tensor i s antisymmetric (F = F,@) and the diagonal terms (Ir = v )
t" f"
a r e zero; thus there are only six ~ndewndentcomponents ( t h e e components
of E and three components of B) instead of sixteen.
This may be seen from the fact that V, operating on elk' X has the effect
of multiplying by ik, (V, does not operate on er since the coordinates a r e
rectangular). Thus,
implies that k e = 0.
where
.
is a fourvector and is called the fourvelocity up Dividing by dt2
@ves the relation between proper time and local time to be
f,  dp,/ds
f4 =
power  rate of chmge of energy 
4 3
This i s seen from the fact that m/ i s tlre total energy and also from
the o r d n a r y identity
and
QUANTUM ELECTRODYNAMICS
f, = euP F p v
Present Customary
notati on Mea~ng~ notation Value
m Mass of electron m
Energy mc2 0 '99 kev
5%
Momenhnn mc 1104 gauss ern
Frequency me2/%
Wave number mc/A
bwth (Comp li/mc 3.8625 f @'%cm
ton wave
liewtt.r)/Zr
Time
e2 Finea truetwe
constant
(dimensionless)
evm Classical radiue
of the e b ~ t r o n
2/me2 Bahr radius
This equation does not allow for "~spin'>nd therefore fails to describe
the fine structure of the hydrogen spectrum. It i s proposed now far applica
tion to the .rr meson, a particle with no spin. To demonstrate its application
to the hydrogen atom, let A =; O and @== Ze/r, then let \k= X (r) exp(iEt).
Then the equation is
Neglecting the term on the right in comparison with the f i r s t term on the
left gives the ordSnary S c h r ~ i n g e reqtlation. By using; (W  Vj2/2rn a s a
perturbation, potential, the student should obtain the finestructure? splittiw
for hydrogen and compare with the correct values.
there results
where

[(H e@12 (iV  * = m2%
Now the Pauli equation is also H* = E*, where
Then adding and subtracting the pair of eqationa for @,X, there r e ~ u l t s
0 0 0 0
0 0 1 0 0 X,
since crx =
P 0
1 0 0 0
which is actually four equations in feu wave functions, Then using four
vector notation, the Dirac equation is
Y
: =l = y y 2 = yz2 = 1
40 QUANTUM ELECTRODYNAMfCS
A W E B M OF THE y MATRICES
In the p r e c e d i ~ glecture the Dirac equation,
yl,(iVI,  ehI1)%
= m%
Note that with this definition of and the rule for forming a scalar prod
ucts
Other new matrices may arise by forming produets of the matrices af
ready defined. For example, the matrices of Eq. (105) are producta of y 's
taken two at a time. The matri ces
These are: the only new p r d u c t s of three. For, if two of the matrices
were equal, the produst could be recluced, thus y, y,y, = y, y, yy = yy .
The only new p r d u c t of four that can be farmed is @ven a s p c i a l name, yg,
and sat
Verify &at
0
@x,y,z
and, movirmp; the sc;cond faetor to the front, by using the cammutation rela
tionrahlips, Doiw Ghia with the first term, (bp?,) af the secoad factor produces
=g$ + 2 b e a
Exercises: (l) Show that
l"IlYlr = 4
v,&, = W
y,$#yp = 4 a e b
v,rr%h, = Z&Plc
Suppose a n o ~ e representation
r for the y @s
ifs obt&ned which satisfies the
s m e commutatio~relationships, Eq. (l03); will the form of the Dirac equa
tion, Eq, (TOI), rem%.in the s m e ? To answer this w e ~ t i o n ,m&e the fol
lowing transformation of the wave function Ji( = S*', where S i s a constant
m&rix W&& is msurned to have arz inverse S' ( ~ = l). 8 The ~ Birac
~ equa
tion becomes
=UTW1;531"IC NCE
The relativistic invarimee of the Dfrac equation may be demonstrakd by
assuming, for the moment, that y transforms similarly to a fourvector,
R E L A T I V I S T I C W A V E EQUATION
That is,
,j = : * * Q * S . R. (If l)
ahs f i e expressions for c h a g e and current density, Hence they are not true
In all representations. The Dirac equation is (wStb E, c restor&)
so that
o, = tar, a y , etc.
From analogy to classical physics, one mi&t e x p e t that the anmlar mo
mentum clpemlor is now
k and
P
The last term may be interpreted as torque. For a central force F, this
term vanishes. But then it is seen that L 0 because of the f i r s t krm;that
is, the angular molnanhtm L is eaneerved, even wl& central forces.
But consider the time derivative of Ule operator a defined as
where
But
so that
and this is the first term of iwith negative sign. Therefore it follows that
are the same as JlrZ in the PauXi eqwtion, Also, if EPHuliia the
Mnetle energy. in the PauXf equation and EBirsc = W + m i~ the msrt
plus kineWc errsrm in the?Erac ewtiont, show ~t
ltt will be assumed &at all ~ b n t i a l sare etatiomry and statimary states
will be considered. This m&es the work simpbr but is not necessary, h
&is case
T b t is,
Using the results of Leetrrre 11, given by Eqs, (116) md (117), the low

e n e r w approximation (W V) << 2m will be made, keeping terms to order
v'. Thus
(W V) l1 + (cr.
[l + (o ~)~/(8rn~]l
g

there result8 fwiGh cr. r = A and (W V) = B in the foregoiqj,
interpreted a s the angular momentum L to get fa* u/r3, the spinorbit cou
pling. This term has no effect when the electron i s in a sstate ( L = 0). On
the other hand, (5) reduces to V E = 4nZ&(r),which affects only the sstates
(when the wave funetion i s nonzero a t r = Of. So (5) and (6) together result in
a continuaus hnction for spinorbit coupling. The magnetic moment of the
electron e/2m, a p p a r s a s the cmfficient of term ( 3 ) , a d again of terms
(5) and (G), i.e., (e/2m)(1/4m2),
A classical a r g m e n t can be made to Interpret term (6). A charge mov
ing &rough m electric field vvi& velocity 'tt feels an effective magnetic

field B = v x E = ff[m)(p eA) x E, and term (6) i s just the energy (e/2m) x
(o B) in W s field. flVe get a factor 2 too much this way, however. Even, be
fore the development of" the Dirac equation, Thornas showed that tMs simple
classical armrnent i s incomplete and gave the correct term (6). The s i h a 
tion is d;ir%eread for the anomalou~morrxenh intraduced by PauU to describe
neubons and protons (aee Problem 3 below). In PauIi's mmodified equation,
the momalous moment does appeas w'rth the factor 2 vvhn multiplying terms
(5) and (6).
Problems: (1) Apply Eq. (126) Lo the Wdrogen atom and correct
the energy levels to f i r s t order. The r e s u l b should be compared to
the exact results.? Note the difference of the wave functions a t the
origin of coordinaks, This difference actualIy is too restricted in
space to have any imporbnce. Near the origin the correct solution to
the Dirac equation i~ praportioml to
in the stiandard representation, then the probability densitqy and current may
be written
To verify this, replace 5 by V*@ and note that p2 = 1 and that flyp = ap.
The ratio ug/u4 can be determined from Eq. (1398) and also from Eq.
(139d). These two value8 must agree in order that Eq. (236) be a eolurtion.
Thus
The eigenvalues of the owrator (iygfl) m w t n07ull be found (tlne i hais, been
added to nn&e eiwnvaluee come out real in w b t follows), Deaclting &ese
eiganvalues by 8 ,
Exercise: Show that the first of the wave functions, Eq. (1311), i s
We l solution and the secand is the a = I solution.
S = i
AnoWer way of obtaiaing the wave Eunction for a freely moving electron
is to p r f s r m an eqavalence tran~farmationof the wave function as in Eq.
(1012). If the electron is initially at re& w i a i& spin up o r down in the z
direction, then the spinor for an electron m ~ v with i ~a veloeity v in the
spatial direction k ies
cosh U = 1/(1 

(2m)'/2 cosh (u/2) = [ m ( ~ v5li2+ = (E + mjl/2
For the case that g i s in the xy pime, t h i ~just gives the remlt, Eq. (13XI),
with a normalization factor l / n
Noticing that for an electron at rest y,uo = uo , may be written
for
It will be conveniepll to have the matrk elemenb of all the 7% Sheen va
r l o u ~initkl and f h a l s a t e s , s o Table 131 has b e n worked out.
64 QUANTUM ELECTRODYNAMICS
1 2m F2Fi  Pi+Pz 0
YX ~ P X FzPt+ + Pa F i o
YY 2py fF2~1+"(" fP2F I 0
7% 0 0 Px*F2 + Pz+Ft
Yt 2E t + Pt * Pz 0
SQLUTXQEJ O F T H E D I R A C E Q U A T I O N
fh Lecture
Fourt~sen
(aYppu) m ( a y p ~ )
But also n&e thiat
, = m(O[ypU)
{ B ~ Y U)
beeawe Qg3 = g%l =: m& Addiw the Lvvo expreshsions, one obtains
sat
From the relation proved in the i~3xerci~ee
it is seen that
h, +?,P = W , Y, =Z
But p, ia just a number, BO it f~ilowtlthat
(ilyp~) ZP,
Xt was found &at a neeeesary condition far solution of the Dirac equation
to exist i s
E' = pZ + m'
SOLUTION Q F THE DIRAC EQUATION
The nneaniq of the positive enerw is clear but Llxat of the negative is not.
It was at one time suggested by SckS&inger &at it should be arbitrarily ex
cluded a s having no mead*. But lit waa found that &ere 858 two h n d w e n 
tal objections to tba excluaian of negative e m r m stsws, Tbe first ie Nysi
cal, &eoretically physical, that is. For the f)irae equaaon yields the result
&at starting wiLh a system in a polsiHve enerf3;Y slate &ere Is a probability
of induced trasitions into mgative e n e r a s t a b s , Heme if they were ex
oluded thias would be a coneradiction. The a e c o d objection is mathematical.
That is, excludiw the negative energy s t a k s lea& to an lineomplete set of
wave functions, It is not possible to represent an arbitrary h c t i o n a s an
expawion in functions of m incomplete aet. This sitxlation led Sehr0Mnger
into 'insurnnount&le difficulties.
positive energy
levels
+m
m
newtive energy
levels normally
FIG, 141
3 , The Pauli ernclueion principle is implied in order that the negative sea
may be full. That is, if my a m b e r rather than just one electron could oc
cupy a given state, it would be impoaaibls to fill alI. the nagaXive e n e r m
states, It i s in tMs way &at the Dirac a e o r y is sometimes consider& aa?
"pr oaf H of the exclusion principle.
hor interpretation of negaave energy s t a b s haa been proposed by
t;he prersent author. The furtdmental idea is &at the 'kncsgative en@rmt'
states represent the s b b s of electrons movz'w baekwrd in gm@,
In the c l a ~ s i c a equa~on
l of motion
X1 X2
FIG. 142

e
t only the initial electron present
the irtitial electron still present but ~omewherfs,
else an
eleclrtmposritrcm pair is formed
tr t, the initial electron, and newly arrived electron positron
a r e preerent
h%  the positran meets with the injtial electron, both of Wlem
amihilating, leaving only the previously creabd electran
t, tz only one electron presexll
Fifteenth Lecture
PAIR CREATION A m AINNXHIUTION
Two possible pass of an electron b e i w scattered between the states 91
and were &seussed in, the last lecture. m e s e are:
Case I. Both @z states of pocaitive energy, interpreted a s 9g electron
in ""past," !Ifz electron in "f~$ure.'~This is electron scattering.
Case 11, Both ?1"$, UF2 states of negative enera interpreted as 4rl posi
tron in "future," "2 positron in ""pst." This is positron scattering.
The existence of negative energy states makes two more t p s of paths
p s s i b f e . T h e ~ eare:
Case ZfX. The positive energy, @2 negative energy, i n t e r p e w d as Jrl
in ''ppa~t," %2 positron in. ""past." Both states a r e in the paat, and no%ing in
the futwre. T U s represents pair amiMlat_ion.
Case N. The negative energy, 4T12 posjiitive emrgy, inbrpreted a s Ol
positron in "future," "2 electron In "'fuh,rre," This is pair ereation.
FIG. 151
72 QUANTUM ELECTRODYNAMICS
CONSERVATION OF ENEMY
Energy relations for the scattering in case I have been established in
previous lectures. It can be seen that identical results hold for case rX. To
shaw this, recall that in case I, if the electron goes from the e n e r w El to
E2 and if the wrt;urbation potential ihs taken proportional to exp(iwt), tfien
this perhrbation b r i w e in a positive e n e r w w . To s e e t h i ~ noLe
, that the
amplibde for scattering is proportional to
Note that the kernel is defined to be Q for t2 tr. It can be shown t b t W;@
satisfies the e p a t i o n
The matrix p i s inserted in the Imt term ia order that the kermll derived
from the Hamiltonian be relativistically invariant. [Note the similarity to
the nonrelativistic e w e , Eq. (15G).] lMultiplying &is equation by B, a sim
pler Eom results:
The notation K, replace8 the & of the wmelativistic case, and Eq,(1510)
repfaees Eq. (154)a s the defining equation,
Juat aa Kv ean be expanded in the s e r i e s of Eq. (1561, s o can be
expanded m
Note that the kernel is now a fourbyfour matrix, s o that all component8
of 3 can be determined. Since this is true, the order of the terms in Eqt.
(1511) is important, The element of integration i s actually an element of
volume in fourspace,
FIG, 152
~~A(2,1)= C exp[iE,(t2tlll$n(~g)$n(~t)
m
pos. energies
t C '
exp [  i ~ , ( t l  t l ) ~ # ~ ( x ~ ) ? ~ t2
(x~)
neg. energies
for all. t2, This is, term: by term, a solution af the hamqeneouss equation
[i.e., Eq. (159) with zero righthand side). The possibility that two such
'76 Q U A N T U M ELEGlPR0DUWAMPC:S
solutions errist results from the fact &at boundary conditions have not been
definitely fixed. We shall always use K , ~ .
The kernel K + ~ defined
, by Eq. (15121, allows treatment of case 111 (pair
amihilatfon) m d caBe N @dr creation) sham at the b g i a w of a s b e 
ture. Pn each case, the got,tentiai, Te&(3), acts a t the fnbrscact;lon of positron
and electron paths.
Sixteenth Lecture
The firat term is Lfxe contribution from positive energy states at earlier
times and the sacsnd term is the contribution, from negaave energjr states
at later t h e s e mis e q r e s s i o n c m be generalized to a t a k that it is me
assary to h o w 9(xltx) on zt fawdinnensfonat surilaee, 6~urrounid;iqthe
point xz,tz (see Fig. 163):
Using the expansion of ~ , ~ ( 2 . 1in) terms of K,(2.1), Eq. (16121, and assu
ming h a t the amplitude for transition from state f to state g as a free par
ticle is zero ff and g arthogonal states), fAe firstorder ampfitude far
transition (Born appra&mtion) is
Theoe state &at &e particle has the freeparticle wave function f just prior
to scatteriw and Lbe freeparticle wave hncticsn g juet a f k r s c a t t e r i ~ and
,
that it elinninabs any computation of the moti on aist ai free particle, The am
plitude for transitfo~,ta first order, mily be written
QITAMTUM E L E C T R O D Y N A M I C S
We shall make use of the theory just presented to calculate the scattering
of an electron from an infinitely heavy nucleus of charge %et. Suppose the
incident electron has momenhm in the x direction and the scattered elec
tron has momentum in the xy plane (see Fig. 164):
FIG, 164
Thus, foy Eq. (I&$),the firstorder mplitude for transitim from state f to
s t a b g (mamentum pi to miomentm p2) i s
P R O B L E M S IN Q U A N T U M E L E C T R O D Y N A M I C S 79
Separating space and time dependence in the waive funct;ions, this becomes
Density of states = % = l
When the incoming plane wave i s normalized to one p & i c l e per cubic cen
timeter, the cross section is given in terms of the tmnsitfoxl probslbillty per
second? aa
where
Therefare, vt = p i p t .
P R O B L E M S IN Q U A N T U M E L E C T R O D Y N A M I C S
Ths criterion for validfly of the Born approximation, used in obtaining tMs
result, irsl %e2m cc 1. In tha e d r e m e relativistic limit v o. T k i beaornes
~
Z <c 137. Just as; for the aonreliktivistic ease, the scattering can actually be
calculated emotly (correct to all orders in the potential) for ttza Coulomb
pobntial. T%ia e m c t salution of the Dirac equation involves h m r g e o m e t r i e
knctians. Xt was f i r ~worked
t out by Matt and i s a l z e d Mott scattering. For
moderate ener@ee (200 b v ) there is s m e prohbility for change in spin,
Polarized eie etrons cauld h produced in this manner.
Seventeenth Lectam
a& the sum over n becomes an integral over p. The up is the spinor cor
respondiw to momrsntm p, poaitive o r negative e n e r w and spin up o r down,
a s appropriate, Then the propagation kernel for a f r e e particle i s , far t2> tl,
1
spins
The plculation will be made f i r s t for the case of t2 tl. We first calcu
late u p ~p for p ~ s i t i v eenergy, and p in the xy plane and spin up. Under
these conditions
Note that U,;, is the opposite order to that usually encountered s o that the
product i s a msttrix, not a scalar. T h t is,
W Q E+m p,ipy 0
upup = (spin dam)
0 P x  4 ~ ~ E+m 0
0 O 0 Q
11, may be verified easily dhat the sum of these malrieea for spin up and spin
down ia represented by
The sign of the energy was not used in ctbhinlrrg thia r~38ultSO i t i~ the same
for ei.f;her siw.

Now put t2 tl = t and xz  xt = X, For L > 0, the propbmtion b r n e l b
comes
The appearance of p in the form E p = (p2 + m')'/' in the time part of the
exponential makes this a difficult iuxLegraI, Note a t It may also be written
in the form
84 Q U A N T U M ELECTRODYNAMICS
where
 
where s = +(tZ x3'I2 for t > X, and i(x2 tZ}I/' for t X . & ( s ~ is
) a
delta function and ~ ~ ' ~ 'is( am ) function.? Another expression for
Hankel
the foregoing fe
FIG. 151
?See Phys. Rev., 78, 749 (1949);included in this volume,
P R O B L E M S XN QUAPi[TflrM E L E C T E T O D Y N A M I W
It i s seen that the distance along fzcf in which this &comes small i s roughly
the Campton wavelengW (recall &aL m  mcfi vvhen it represent8 a len@hv"

a s here), ss h t in reality not much of the t 1x1 h3pee oubide the light
cone i s aceessible.
The tranh~formtionto momentum representation will lilow be made, This
ie facilihted by use of the inkgral f o m u l a
But E: = p2 + m 2 s o this i s
a s before.
The fact that Eq, (171) for K(2,1) has mare &an one solution is re

flected in Eq. (172) in the fact that (d m)""' is singular if p2 = m '. We
shall have: to say Just how we a r e to k n d l e poles a r i s i q from Ghis source
in integrals. The m l e &at @electsthe particular form we wmt i s &at m h
considered as. b v i n g an infinibsimal rtemtive imagiwry m r t ,
Eighteenth Lecture
The function a(q) is interpreted as the amplitude that the ptential. con
him the momenbm (q), As an emmple, consider t l Coulomb
~ potential,
given by A = 0, p = Ze/r.
SubstiMing into Eq. (181) gives
Here the vector Q is the s m c e part of the mamerttunn. The delta Euae
Lion 6(q4) arise8 from the time demndenee of & ( X ) .
88 Q U A N T U M E LEGTRODYNAM1C8
Substituting for a free mrticle and also expressing the gotenth1 functions as
%eir Fourietr transforms by mean8 of Eq. (18g), this bcscomea
FIG. 181
wave function ui and moves from 1 to 3 a s a free particle crf momanhm Xli.
A t point 3, it is scattered by a photon of m o f a d e r the action of
the potential ie$(ql)]. bving;5 abfsorbed the um of the phoLon it Ghen
movess from 3 to 4 a s a free particle of momentum 6% + & by eonaervation
, At point 4, it i s scattered by a second photon of nnomenwnn
42[under the action of the pohntial ie$(q2) absorbing We additional momen
90 QUANTUM ELECTRODYNAMICS
tun &)l, Fimlly, it maves from 4 to 2 aa a free prartiele with. wave fune
tion u2 and momenwm & = & + d¶+ 42. It is also c h a r h a m f&e diaparn
that the in&gral need b h k n over qi ody, b c a u m when $l and gi2 a r e
even, io debrmined by d2 = pjl &.The law of conservatim of en
ergy requires p12= m', p22 = m'; but, since the intermediate state is a vir
tual stak, it is not necessary that (pl + di12 = m2. Since the operator
X/(& + dl  m) may be resolved as (fit + dl + m)/((Pt + dl)2 mZ], the impor
hnce of a virtual state is iaversely g r a p o ~ i o n at~
l the degree to which the
consemaaon law is violabd,
m e rasults @ven in Eq8. (583') &ad (186) may h summrized by the
following list of handy ruleat for computing the matrix element M = ( 5 2 ~:~ 1 )
1.. An eketron in a virGual st_ab of momenlum 6 contributtls the ampli

tude, i/(jp5 m) to PS.
2, A wtential containing the momentum q contribuks the amplihde
ie$(q) to N.
3. All indeterminate momenta qi a r e summed over d4qi//(21r)'.
Remembr, in computing the integral, the value of the integral la desired,
w i a the path of inbgration, m s s l m the sinwlarlties in a definite manner.

mlts r e p l a ~ em by m i~ in the inlegrsnd; then in the solution @lice the
limit a s r 0,
For relativistic work, only a few termrs in the pertrxrbiltion series a r e
necessary for eompuhtion, To assume that fast electrons (and positrons)
i ~ k r a ewl i t h tt p h ~ t i a only
l once (Born appro&mtfort) Is often, odffeiently
aceumte .
A f b r the matrix element is debrxnined, the probabilit;y of trawltioa per
second is given by
P = 2n/(n N) J MX ~
(density of final stabs)
This makes any matrix element cornpub$ with Uzeacz pobntials invariant,
but to obhin the correct tramition probability in a @ven coordinate sgsbm,
i t ie necessary to reinsert a factor (2w)""fsr each photon in the initial and
final s k t e s . This becomes part of the normalization factor RN, which con
k i m a similstr factor for each electron in the initial and final @&Le%.
92 QUANTUM ELECTRODYNAMICS
MDUTION FROM A T O m
The transition. prokbility per second is
x ew(fkerz)] d vot
hobn f (incamin
X
recoil electron
FIG. 191
I M T E R A C T I O M OF" P A R T I C L E S W I T H L I G H T 93
For the incoming photon take as a potential A l p = alp exp (iq! X) and for
the outgoing photon take Az = exp (iq2 * X). The light is polarized per
p n d i c u h r to the direction af propgation (see Fig. 191). Thus,
et ql = O
S eg q g = O
91 41 = 9t2 = 0
g and q2 q, = q12 = 0
$2 = ~2 e x p (ip2 * X)
The l a t k r two eqwtioxrs follow from the fact that, for a photon, the e n e r m
and momentum a r e both equal to the frequency (in units In which c = X). The
momentum has been resolved into components, The incoming photon barn
can be resolved into two t y m s of polarization, which will h designated tym
A and tyw B:
Type A has the electric vector in the z direction and type B has the elec
tric vector in the y direction. SinrtiZarly the outgoing photon beam can be
resolved into two t y w s of polsllrization:
QUANTUM ELECTRODYNAMICS
= m 2 +o+O+Zmwl2mw22wiw2 (I  cos 8)
where the last line was obtained from the preceding line by u s i w Eqs, f 1931,
(1941, and (196a, c, d). This can be written
~ ( WI02)  cos 8)
This i s the wellhown formuh for the Campton shift in wavelength (or fie
qaeney) .
DlGREBJON ON THE DENSITY OF FmAL 8TATLS
By the method discussed in the earlier part of the course, the following
final state densities (per unit e m r g y inrterval) can h obdained.
k m of t o h l energy E and total linear moment= g disintegrahs into a two
particle fiml state,
dGI
Density of states = ( 2 n 1  ~
 Ej(P0Pl) (B1)
W? day
Dens3ity af states = ( 2 v 3c~ $2
iln woreung out the nrr%trix element M, there a r e Wo ways in which the scat
tering can b p m n : (R) the incoming photon is absorbed by the electron and
emits the outgoing photon; (8)the electron emits a photon
then the e b ~ t r o n
and subsequently absorbs the incident photon. m e a e two procesees are
shown diagrammatically in Fig, 192.
h momentum represenktion, the matrix element M for the first proe
ess R i s
mived momentum 13fl from the potential the electron travels a8 a free elee
tron with momentum gig + &; (d) the electron emits a photon. of pohrization
6,; and (e) we now ask for the amplitude, t b t the electron i s in a @Lateug .
Exercise: Write down the mtrix; element for h e secand process
5, The t o b l matrix element is lthe e r n of thte;se two. Ratf onalize
these matrix elements and, u s i w the: table of matrix elenrrents
(TaMe 131) work out 1 Mfz,
Twentieth Lecture
The complete m a t r k element is the sum of these, s o that the eross section
bcomes
The matrix elements for the various spin and polarizatlorr eombimtion8 can
be calculated straightforwardly from &is paint, But c e r h i n prelimimry
manipulations wiZX reduce the h b a r involved. Using the identity
it i s seen t b t
and thig is the matrix element of the ffrst term of R. Xt i s also f i e nemtive
of the matrix element of the last term of R, so R may be replaced by the
equivalexlC
98 QUANTUM ELECTRODYNAMICS
Substituting & = w  
 Y,) and dz = w 2(y, Y , cos B yy sin 8 ) and bans
posing the 21x1 factor, the complete matrix m;ly be writbn
Using this f o m of the matrix, the matrix ebmrsts may be compubd easily.
For example, consider the case for polarization: 8; = y,, #2 = yy cos B  y,
sin 8. This eorreawnds to cases (A) and (B" )of h c t w e 19 and will be de
noted by (AB". The matrix i s

2m(R+ S) = my, (yy cos 6 y, sin B)[y,(l cos B)  yy sin 81
since ez m e i = 0. Expanded this becomes
2m(R+ S) = y,[yyyx cos 6 ( 1 cos 8) + caa @ sin B c sin @(l  COS 8 )
 y, sin 8
where the antieommubtion of t;he y 'S h a b e n used. IXn the ease of spin
up for the incoming particle and spin d o m for tke: outgctiw particb (st = I),
sz =: 11, the matrlx elements
may be found by reference to Table 13I. But note that in, this prciblenn pi,
= p,g + i ~ y =i O s i w e particle 1 is a t reat, Hence L;ke final ntatrix elennent
for this ease, polarization (AB"), spin sg = +l, sz = 1, is
Q U A N T U M ELECTROLZYNAMrGS
2m ( F ~ F ~ } ' / ~ ( ; ~+(s)u~)
R = (l  C08 
@)iF1p2+ sin 6 Pz+ Fi
and
the squarre magnitude8 of the matrix elements for the various easea redurn,
a&er eorresfderable amount of afmbra, to Lhe expressions given In Table
202,
AB" [(@l ~ 2 1 ~ / ~ i @ 2 1
BB' [(&li  w $ ~ / w ~ w+ ~4 ]~ 0 68 ~
This is the meinNfshim formula for polarized light, For unpolarized light
this c r o s s section must be averaged over all polartzalions.
It is noted t b t diagram eases s u c h a s Fig. 201 Mve been included In
Twentyfirst Leetgre
the total c r o s s section for s c i a t k r i x through any angle i s obbfrted. So, from
Eq, (21X),
where the last five terms replace sin2 B = cos2@ 1 using Eq. (21F).
Sf mpf e integrations fleldS
$1 = IEy,  p,. Y)
where the emrgies E and E, of the electron and pasitran a r e both posl
tive n u m b r s . The conservation law gives
(just cta for Gompton s c a t k r i q , but the direction of gfi reveraczd), s o the
matrix element for this ineraction i s
me&, i s oblafned from the first by inbrcbaaging the turo photons (see Fig.
2 122); again no& similarity to Compton seatbrlng.
Immediately, the matrfx element is
FIG. 212
The sum of the two m a t r h elements 13md the density of final states gives
the e m s s section
This reduces to
106 QUANTUM E LECTRQDYNAMICS

The formula for positronelectron amihilation derived in betrxre 2 l di
v e q e s as the positron velocity approaches zero (cr l/v; this i~
other c r o s s sections when a process involves abhfo~tion
true for
of the, irmcoming
particle, and Is the wellhown l/v law). To ealculab the positron lzelfme
in m electron density p (recil~llthat the p m c e d i w cm@@
section ws for a n
electron density of one per cubic centimeter) as v+ 0, we use
+

plus the fact t h e , a s v, 0, E, m and ol  
w2 m (when the electron
and positran a r e btch.approximably a t rest, momentum and energy can be
earnerved only with tvvo phobna of mornentra e q u d in magnitude but opposih
in direction). Thus
electron 2
J Coulomb field
of the nucleus
photon
Notice tfnat there i s a, spectrum of phoans; t b t is, the photon energy i s not
d e b m i n e d (as i t was in the Compton effect, for emmple). Letting "S1Z =
(G2MU$).
where the f i r s t term comes from Fig, 222a and the second term from Fig.
222b, The explaimtion of tkm factors in the f i r s t term, for example, is,
reading from right to left, that an electron initially in state u l i s scattered
by the Coulomb potential acquiring an additional momentum 5& , the electron
moves a s a free w r t i c l e with momentum 6%+ @ until it emits a photon of
polarization gd, We t h ~ nask: 1s the eletctron in sbtf? u29 F o r the Coulomb
pokntial
The outgoing photon can h polarized in either of t w directions, and the in
earning and ouQofng electron each have two possible spin sbhs. The vari
s u s m a l r k elements can be worked out using Table 331. exactly a s w s
d o in~d ~ r i v i n gthe m e i n  M i s ~ n aeross section in Eczeture 20, Nothiw new
is involved, so we omit the delafls. M k r (l) summing over phoWn polari
z a t i o n ~ ,(2) summing over outgoing efeetrsn spin stabs, and ( 3 ) averagf%
over h c o m h g electron a p h sdates, the following dgferential eross section
is obWned:
Q U A N T U M EZ1E:CTRORYNAMId:S
using the relationskipa gOPlz = $z$ + 2~3* p2 , $14 =: 661 + Ze PI,and neglecting
4 in the numerator, since it i s small, this becomes
where use is made of the fact thatJhe ma_trix element of M between states
u2 and u l is to km ealeulakd and uz$z= dtui = mule
The cross section f a r photon emission can then writbn
The firat bracket is the probability of transition for elastic scattering (see
Lecture 161, so the last bracket may b i n b w r e k d as t k pmbability of
phabn ernissian in frequency interval dw and solid a w l e da, if tficlre i s
elastic seatteriw from momentum pi Lo pz,
XMTERACTXQN O F PARTXCLES W I T H L I G H T 115
FIG. 223
$l i s the positron momenlum and yJ2 i s the electron momentum. Notice that,
with respect to the directions that the arrows point f a d without mgard to
direction of increasim time), these d h g r a m s look ercactfy like those for the
brsmsstmkjluq pmeess: S t a f i i q with in case (a), the particle i s f i r s t
s c a t k r e d by the Coulomb potential and then by the phown; in case m)the
order of the events is reversed. The differernce b t w e e n pair prduction. and
bremastrafilung, when the direction sf tim is taken into account, is {X) is
a positron state (melectron traveliw h c k w a r d in time), and (2) t b photon
4 is absorbed rather Lbn emftbd, As a result, the bremsstraMung m a t r h
elements can b w e d for thia process if 6%i s replaced by $, and 4 by 4,
122 Q U A N T U M ELlEGTROTZcYNAMICS
= ( 1 / 2 a ) ( ~ e ~ / ~ 'e2
) ' @,p sin 8 , d B , sin 8 d8 d+/w3)
F i m r e 224 defines the angles (rS, = angle between electronphobn plane and
pohsitronphoton ghne) .
positron
one needs
@  1
C
2 spins I
C
spins 2
I(u2~uill~
where mean8 the sum over final spin states for only one s i m of the
spins 2
the e n e r n , that is, over only two of the four possible eigenakles. Similarly,
is the sum over initial spins for one sign of the energy. The purpose
spins I
now is to develop a simple m e t h d for obkining these sums.
h a c e o r k n e e with t b u s w l rule f o r m a l r h multiplieatioa, the fallowing
i s true:
) I ~ u 2=) 2 r n ( G 2 ~ ~ u 2 )
( G 2 ~ u (I U
all Y
where A and B a r e any operators o r matrices, the 2m factor on the right
a r i s e s from the normlization uu = 21x1, and the sum i~ over all eigenstatas
represen;ted by ul, But the shbs u, which we want in Eq, (231) a r e nat all
shks, just those satisfying IzJiut = mu$. T b t is, they belong the eigen
value + m of the operator &. Since 612= rnZF$l also has the eigenvalue m,
t h t is, there a r e twa more solutions 05: &U = mu wkich, together with the
two we wish in Eqi, (231)bring the t o k l to four. Let us call the l a t b r
'"negative eigenvalue" sslisks,
Mow, !if in Eq. (232) the m a t r k elements of B were zero in negative
eigenvalue states, this would be the same a s , that is, just over posi
tive eigenvalue s k h s . So consider spins 1
C (;2~u,)(;l@f+m)~uz) = (:2~(dt+m)~u2)2m
all U,
But
rV
C G 2 ~ u l ) a m ( GBuz)
f
spins 1
CaneeXliw the 2m factors, this gives
C
spins 1
(Gz~u!)(';l Buz) = (G~AMI+ m)Buz)
spins 2 all U,
where the last niohtion means the spur of the m a t r k in the brackeds. It is
true whether &, &12 represent electrons o r positrons.
Tke followiw list af the spurs of several frequently e n c o m b r e d matrices
may km verffied easily:
It i s also true that the spur of the product of any add n u m b r of daggerad
o p e r a b r s i s zero.
I H T E R A C T I O N O F P A R T I C L E S WITH LXCHT I l5
The spur can bt3 evaluabd irnmedia&ty from Eq. (235) with m2 = m4 5; O and
g4 = p14 = yt. 
Anotht?r way is: Since ytdI = 2Ei ~ % y tit,i s Been t h t
The cmas sections for the paf r production and brernsstrahlung processes
contained tbe factor [v(Q)I',
where V(&) i s the momentum representation
of the pobntial; t b t 18,
Clearly V(Q) gets large a s Q gets small. The minimum value of Q oe
c u r s when all three moxrrenk a r e I l n d up (F&, 2811:
Pr Pa
g
FIG,231
so t h t in this ease
From this i t is seen that Qdn  O a s Er  00 . This ahows clearly why the
then s e r e e n h g effect wtll be i m p o a n t , and vice versa for the uppsi* in
equalities, E from this estimate s e r e e n i w would appear to b imporknl,
one should use the screened Coulomb podential. It gives the result
P~oblrsm:h d i s c u s s i ~
bremslsptrahlung it was found t d t the c r o s s
section for ernksion czf a lowenergy pll~tQacan IM approximakd as
FIG, 241.
Show that the cross section for this effect 18 given, by Eq. (24l), with
the KleinNisfiina formula r e p l m i ~ m e m e m b r to assume q
small ,)
potential
region
Interaction of
Several Electrons
Even though the Dirac equation describes the motion of one particle only,
we can o b h b the amplitude for the inbraetion of two o r more particles
from the principles of quantum eleetradywmics [so long a s nuclear forces
a r e not involved),
F i r s t consider two electrons n n o v i ~throu& a region where a p a h n t h l
is present and assume t b t they do not interact with one another (see Fig,
242). The amplitude f o r electron a moving from 1 3, wMle electron b
moves from 2 4 i s given the symbol 33;(3,4; U ) .If it i6a assumed t b t no
inkractian betwem electrons takes place, then K can be written a s the
product of kernels ~,(')(3,1) K,(& )(4,2), where the superscript means that
K
,(
'
) operates only on those variables describing particle a, and similarly
for K , ( ~ ) .
A second type of interaction gives a, result indistfngufshable from the
f i r s t by any measurement in aceo&nce with the Pauli principle. This dif
f e r s from the first caae by the i n t e r c h q e of w r t i c l e s k t w e s n gossitfana 3
and, 4 (see Fig, 243). Hour the Pauli principle says that the wave function of
a system compssctd of several electrons i s suck tbat the inbrchirnge of space
variables for two pa@icles results in a c b n g e of alw for the wave function,
Thus the amplitude (Including both possibilities) i s K = K,(%) ($,l) K , ' ~ ) (4,2) 
1) ~ , ( ~ " ( s , 2 ) ,
A similar sitwtioa a r i s e s in the follovving occurrence. Initially, one elec
tran moves into a r e a o n where ra potenaal is prewnt. The potential creates
a pair. Finally one positron and two electrons emrge from the region,
There a r e two possibilities for this occurrence, a s shown in Fig. 244.
Again, the total amplitude for the occwrenccz is the digerence b t w e e n the
amplitude8 for the two possibilltiea.
IPJTERACTXOM 01F S E V E R A L E L E C T R O N S 119
potentkl
region
2 2
TABLE 241
and
Separate variables may be used for tBand t 6 , if the fmction 6(t5  t6) i s
ineluded aa a factor, Then
where the differential d~ includes both spa= and #me variables. It is con
ceivable t h t the rektivistie h r n e l couM be obtained by s u b s t i t u t i ~K + for
K@, and introducing the i&a of a retarded potenLial by replaciw 6(tb t6) 
 
by 6(tS t g r6,@).Nowever this 6 function i s not quite right. Its Fourier
transform cmtains both positive and negaave frequencies, whereas a photon
has only positive energy. Thus
+
c Q

4 (X) = lim ( l / ~ i ) ( X i ~ )

Abbreviating tg ts 3 t and r g , ~S T, and taking account of the fact that both
, retarded potential i s
t 5 5 tg and t 5 =P t6 a r e ~ a s i b l e the
XHTERAGTXQN O F S E V E R A L E L E C T R O N S

&fining t2 r2 as a relativistic invariant, the potential is
e26, (gS,2). Another term h i c h must be included i s the magnetic in
t e r a d o n , proportism1 to V, * Vb. In the n e t i o n used for the Dirac

eqmtion, this prduct i s a, orb. Zt will km found emvedent to ex
p r e s s this in the equivalent form (Be), * ( P c u ) ~ and
, in & i notstt;ion
~
the retarded Coulomb potential. is proporZ;iomf to P The~c:P '8
come from the use of the relativistic b r n e l . Thus Lhe complete po
tentfal for the interaction becomes
FIG. 245
tween the electrons. The virtual photon can be polarized in any one of
the four directions, t, X, y, z. Summation over these four possibilities
is indicahd by the repeatttd index af The integral expression for
 
tke kernel, Eq. (242), implies t k t t b amplitude f o r a photon to go
from 5 6 (or from 6 5 depending on timing) i s b , ( s s , t ) Equation
(242) can b taken a s another a?&bment of t b fundamen&l laws of
qumtum electr&ynamics.
(2) Show t h t
2
in momentum representation. It is interesting to note that 6 , (si, ) 18 the
same as I, (sI8$), the quantity appearing in the derivation of the propagation
kernel of a free particle, with m, t b particle mass, s e t e q w l to zero, A
more direct comeetion with the MaweXl equations ean km seen by writing
the wave equation. 0 klI, = 47r JP in momentum representation,
Comparing this vvith the msult of the last lecture, i t must b that
This can be written in a form which makes the spacetime symmetry evl
dent by using the Faurier t r a n s h r m
exp ( iKlt/ ) = i 1 2 i ~ / ( w
~ K'+ i ~ ) exp
] (iwt) d w / 2 r
and cornwring this with the result of the last problem of Lect;ure 24 estab
Zisbs that the rules given in Leeturs 2 a r e consistent with relativistic elec
trodywmiics develop4 in the last lecture.
The f i r s t term comes from diagram R, the second from diagram S , and the
summation over g Is implied. h the eenterofmass s y s k m , the prob~bility
of transition per second i s
where y9 i s the y matrix for the Q diractlons and y,, represents the y m&
trix in either of the transverse directions, The matrix element of 4 =
 i s zero in general (from the argument f a r gaugr;. hvariance). t Thus
y~ can be replaced by y, with the result
Now l . / ~
represents
~ a Coulomb field in momentum space and yt ie the
fourth component of the current density or* ctrarge, so t h t the first &mm
represents a Coulomb ineraelfon white the second term conkins the i n b r 
action tbough transverse waves,
ELECTRONELECTRON mTEMCTIQ;N
The simplest diagrams f o r the interaction a r e shown in Fig. 281. The
amplituide for the process h a b e n found Co be praportioml, in momentum
d.t
FIG. 261
128
representation, to
From this identity it was deduced in the fast lecture that the amplitude far
the process a s just given is equivalent to
and
but
Also, since
where the cornmuation relations for the a's have been used. From this it
is. easily seen that the amplitude for the correction ta the Coulomb potentllal
may km w i e e n altogether in the form
But since crx Q is z e r s w k n Q and $ have the same direction, the sum
could a s well be over all three directions and then it is equivalent to a dot
product. That is, this term of the correction i s
‘COR3"ZE:CTION" T E R M S 13 1.

repreeenbtion of the interaction bdvveen two dipoles as was sfaled.
go% t b t the approxinnatian q4 (v/c)Q used a b v e applies anly liretween
positive e n e r n sbt8s. For, if one of the s h h s represents a, positron, then
FIG. 262
V"j3
FIG, 263
in this case tq2 C 4m2(Q = O in the C.G. system), and all matrix elements
a r e 1 o r O (regading parEicles a s essentially at r e s t in the positroniurn),
s o the result is just a n u r n h r , This means that t a k i w the Fourfer transform
one gets a (S function of the relative eoodinrzte of the electron and positron
for the interaction in real space, For tMs reason it is sometimes referred
to a s the '%sfro&mnge'"nteraetlon of the electron and positron.
T b combined finestructure splitting due to the effects already outlined
turns out to be represented by
FIG. 264
and reabsorb f i a own photon. Takiw this into account, the finestructure
splitting in positroniurn i s given by f
having a value of 2.0337 l@ Mc. The experimental value for the positro
nium fine atructum is 2,035& QA00Mc, so i t ia seen that this last correction,
though of orcfer a amaller t b n the main k m s , i s necessary to abbin
agmement with ewerfmenl, It is referred to bath in pasitronium and in hy
drogen as the Lambshgt correction because of its experimenbf observa
tion by Lamb a s the source of the small splitting between the 'glj2 and Z~1i2
levels in hydrogen, In general, i t comes under Ibe h e a d i q of selfaction of
the electron, to be treated in more dehf2. later,
o r in momenhm representation it i s
FIG, 286
where
(see Fig. 2661, Thus it i s possible to determine Ifi and Ih, in terms sf eaeh
other but not indewndently; that is, the momentum may be s h r e d in any
ratio b t w e e n the two photons, It ia for this reason that the integral over gI
a r i s e s in the e q r e s s i o n for the amplitude.
The integral over d3xs gives a V which cancels with the denominator, and
the inkgral, over dt3 gives a T which cancels with the lefthand side, s o
f imlly
using $u = mu and the relations of Lecture 10. Then Eq. (276) becomes
(l/k2)c(k2) instead of just {l/k2), where c(k2) i s s o chosen that c(O) = 1 and
c(k9 0 as k2 +W. In space representation the madlffcation t a b s the form
this value a2, and the general behavior of f, i s shown by Fig. 272. Thus a 2
i s sort of a "mean width" of f, . If a 2<< 1, as assumed, then when
which is the size of the interval. The significance of the form of f,(s2) can
be understood from the following. The original function, b+(s2)differs from
zero only when a' = t2  rZ= 0. That i s to say, an electromagnetic signal can
reach a point a t distance r only a t a time t such that tZ r2= O or t = r
(i.e., the speed of light is 1). This i s no longer true for f,(s2). The depar
ture i s obtained by a measure of 1  r, But, by Eq. (2781, for all values of
r a this measure is negligible. Thus, depending on A ~the , laws will be
found undfected over any practical dislanee,
Choosing h2>> m2, a practical (and general) representation of c(k2) i s
which ean be written as the sum of two inkgrals, wMch differ only by having
m o r g in the numemtor, t k t is, m o r ka (since g = k,y,).
APPEAIRmQ D
METHOD OF m1C'EaMTION CIF m"1C"EGMU
QUANTUM ELECTRODmAMICS
We shall need to do many integrals of a form similar to the preceding
one, A method k s k e n worked out to do these fairly efficiently. We now
stop to d e s e r i b this method of inkgration,
EveryLhing will be based on the fallotviryi: two integrals:?
In Eq, (27XI), to write a little more compactly, we use the rrohtion fl;k,)
to mean tbat either X o r k, i s in the numerator, in which case, on the right
hand side the (1;0) i s I o r 0, respeelively. 410 grove the first of these, n o k
?R, P, Feynman, Phys, Rev,, 76, 769 (1949); included in this volume. N o b
that i n the article d% i s equivalent to 4$[d'k/(2~)~1 in our notation.
t k t , if kk, i s in the numerator, the integrmd i s an oc3d function. Thus the
integral i s zero, With 1 in the numerator, contour integration i s employed.
Write the integral
Then for E L + k2, there a r e poles a t w = & [(L + k2)'/2  it], and contour
integration of w gives
with the contour in the upper halfplane, Two dgferentiratious with respect
to L give
which proves Eq, (271 1). E k  p fs substituted for the variab2e of' integra
tion in Eq. 627113, the result i s
The integral over k can be done by u s i w Eq. (283) with the substitutions
 X) for h , giving
xp f o r p and L ( l
4
I = 2(32n2 i)""'
f
dx (l;xp,) In [(l  x)A? + m2x2/m2x2]
When h2 >> m2, it is legitimate to neglect m2x 2 in the numerator [it i s true
that when x m I, (X  x)h2 i s l z o t much larger than m"', but the Lnbrval
over which this IS true i s so small, for h2 a m2, Chat the e r r o r i s snzarl],
so that, wlren the x integration i s performed, t
Q U A N T U M ELEeTROD"YbJAM1CcS
Now (ev1271) i s about 10", so that oven if h, i s many times m, the fraction
c b g e in mass will not be large, The intewretation of this result is a s fol
lows. There is a shift in mass which d e p e d s on h and hence camot be de
termined theoretically. One can imagine an experimenkl mass and a theo
retical mass which am r e l a h d by
All our measurements a r e of m,xp, that is, selfaction is included, and mth,
the mass without selfaction, cannot be dehrmined, Mare aecurably s b b d ,
a theory using m e p , 8 1 % ~
A. t h e 0 0 u s i w mtt, and e21fic self action minas
is equivalent to
e 2 h c self action Am a s carnguted for a
h e particle S
When the electron is free, the ez/gc seffaction k r m emctfy cancels the
Am term and a theory using meXpi s exactly correct. When the electron i s
not free, e 2 h c selfaction is not q u f h equaX to the Am term and there i s
.
a small correction to a theory using meXp This effect leads to the Lamb
shift in the! hydrogen atoonn, and, in order to calculate! s w h effects, we s h l l
now consider the effect of selfaction on the s e a t k r i n g of an electron by an
e x b r m l pobntial ,
11
FIG. 282,
144 QUANTUM ELECTRaDYNAM.ICS
e2 m 4

2n
fn
X min
+@tan@+=
[see Eq. (71)I. Reference to page 54 shows that the effect of a particle'^
having an anomalous magnetic moment is to subtract a potential p FP,
from the ordinary potential d = ypAp appearing in the Dirac equation. Since
this is precisely what the f i r s t k r m of Eq. (2810) does, one c m say that
this part of the selfaction correction looks like a correction to t k elec
tron" magnetic moment, s o that
N o b that this result [and (289)md (28%@)ldoss not depend on the cutoff
h, and hence h em now be &ken to be. infinity, f
rdr I
FIG. 282
when considered along with the effect of the mass correction, i s another
correction k r m ,
But the part of the amplitude for diagram TX. (Fig 282) eonbined in the pa
rentheses is just Awnul, s o that XI and 11' ssem to cancel, A similar result
applied for d i a p a m s m d El" This is a n e r r o r , however, arising from
the fact t k t both of these amplitudes a r e i d i n i k , owing to the factor 4  m
in the denomimtor. Hence their difference i s i n d e b m i m t e , But by suh
t r a c t i w them p r o p r l y it will be found that their difference does not vanish.
The method proposed to accomplish this subtraction will, in fact, give
the combhed effect of the selfaction and mass correction of both diagrams
11 arxd Kf and Ifband 1311'. It iia based on the fact t h t an electron is never
actually free. An electron's M~storywill have alwaya fnvolved a s e r i e s of
scatterings, a s will its future, These scatterings will h conaidered as oc
e u r r i w a t long but finite time intervals, Xt will be sdficient to calculate the
effect of seEaction and the mass: correction k t w e e n any t w of these scat
b r i w s , since the result will evidently be the same bdween each pair of
them. T k n , the effect will accounkd for simply by r e g a r d i q a, eerrrec
tf on, q u a l to that calculated for om of the i n b r v a l s b t w e e n scatterings,
a s being assochtc3d with the pobntial a t eaeh s c a t e r i n g ( n u r n b r of inter
vals e q u d s number of s c a t t e r i q s ) . Then, considering a s i w l e s e a t t e r i w
event a s here, this correction to the pobntial represents all the effects of
diagrams II[, In, B*,and 111"
F o r an electron which is not quite free, pa .:rn2 exactly, but insbad
QUANTUM ELECTRODYNAMICS
up to brms of order E With the perturbations of self action and mass cor
rection, this m a t r h is
This i s just k2/2a)r times the matrix for no perarbation. Hence the cor
reetion 2erm due to diagrams 11, fII, E ' , and 111' i s obhined simply by re
placing the s c a t h r i w potential $ by (e2/2r)r$, a s was slated earlier.
It should be noted t k t the dZfleulity in obtaining the proper subtraction of
the selfaction and mass corrections just clarified does not represent a
""dvergence " v a b l e m of qtrmtum eleetrodymmics, It i s a tsiczal problem
which could as well a r i s e in nanrelativistic quantum mechanics if, for ex
ample, one c b s e same n o m e m value at3 a reference of potential, that is,
regarded a free electran as m o v u in a uniform nomero pokntial. It may
be easily verified that this would givr;. r i s e h an "energy correction'9or
the f r e e e l e e t r o ~amlogous ta the ss correction hvotved here, Then in
computing the anrrplibde for a s c a t t e r i q process where one used ;a "theo
retical emrgy" a d subtracted the effect of the "energy correction," the
difference of idinite terms would appear if one used freeelectron wave
functions.. In this simple ease the i d i n i h term would, indeed, caneel upon
proper subtmction but in principle the problem is the same as the present
one.
F inally, the complete correction &run a r ising from self =lion and rnaas
correction i s
4
+8bn6+
tan 28
e2 28
sk + 
8nm Mid  dd) sin
728
where a. i s the eross section for the potential 6 only. This c r o s s section
diverges 1ol;~arithicallya s h ,in .+ 0, and it i s this divergence wMch was
formerly referred do a s the "i d r a r e d c a t a ~ t r o p h e . ~ ~
This result, however, a r i s e s from the physical fact t k t it i s impossible
to sea%er an electron, with the emission of 7ao photons, When the electron is
scattered, t h electromagnetic field rnust c h n g e from t b t of a ehrrrge mov
ing with momentum pi to that for momentum pz. This c h n g e of the field i s
met3ssarily accompanied by radiation,
h the theory of brehmsstraMuw, it was s h a m t k t the e r o s s section for
emission of one lowenergy photon is
PrclbEem: Show that the integral over all directions and the sum
over plarizations of the foregoing c r o s s section i s
+ (terms inde
of s r d e r e 4 )
terms independent
This does not depelld on hmin and hence resolves the " i d r a r e d c a b ~ t r o p h e . ? ~
It has been rshown by Bloch and Nordsieck that the same idea applies to all
orders.
It i s interesting that the largest k r m in the quantumelectrodynamic
corrections to the scal;t;eriq cross section, namely,
af free electrons, The best way tn handle the problem i s to calculate every
thing in t e r m s of the Xmi, and then to ask only questiana which can Izave a
sensible answer a s verified by the evenha1 elimiwtion of"the hmia.
The term t o be s u b t m c k d Is
Then
The f i r s t integml fst written down from Fig. 301. The second is the free
particle effect a s noted in previous lectures. The kernel i s not well
FIG. 301
= C +fbp
+D
(iK * R)lon (E, + K  ~ ~ 1(aP
 l exp (iK R)],@
*
This cormetion i s of greatest inzporbnce for the s state, since with a Gou
lomb potential V ~ = V 4nze26m), and only in the s states is differentm(R)
from 0 a t 1R = 0.
The c b f e e of fq is dekrmined by the inequalities m ko >> R y d b r g . A
satisfactory value i s = 137 Ryd. With such a h,the effect of photons of
k c ko must b included, This wil b done by separating the effect into the
sum of t b e e cantrilbutiw effects. It will. be seen that two of these effects
'VGQRRECTJON" TERMS
probable improbable
FIG. 302
The matrix element for cx4 i s very small, and only the elements for at need
be considered. Then the sum over negative s k t e s i s
If this sum is continued for +n, a negligible term of order vZ/c2 i s added.
Thus the sum i s approximately
This brm is f n d e ~ n d e n of
t V, a d thus is also eaneele
tfon.
(3) Tramverse p0sitive energy s h t e s : Since ko is la
size of the atom, the dipole approximation can km used. t
in the sum of Eq, (307) &comes
w r iting
U s h g the relation
This cancels with the In of Eq, (3077, leaving the final correction as
This sum has b e n carried out n u m e r i ~ a l l yto be compared with the observed
Lamb shift,
i4m2 C
spin states
( ~ Z Y ~ ~ I ) $ J U rrn
1
Y~
1
(l
where u is the spinor pax% of the closedloop wave function. The f i r s t pa
renthesis is the ampffbde for the electron to be scattered by the photon;
1/q2 i s the photon propagation factor; and the second parenthesis i s the am
plitude for the closedloop process which produces the photon. The expres
sion i s i n h g r a h d over p &cause the amplitude for a positron of mo
m e n b is desired. b the sum over four apin s k t e s of U, two s h t e s take
c a r e of the processes of dlagram I and two s k t e s take c a r e of the proc
e s s e s of diagram XI. No projection opemtors a r e required, s o the method of
spurs may be used directly to give
This integral also diverges, but a phobn convergence factar, as used in the
previous lectures, Is of no v d u e b e c a u s e now the integral i s over p, the mo
mentum of the positron. 1x2 the intermediate step, The method which. has h e n .
used to circumvent the divergence d@fi.culLyis to subtract from ttris integral,
a similar integral with m replaced by M. M i s taken to be much: larger
11
FIG. 311
(4m2 + ~ ~ "+ 1
/ 1/91
3 ~ ~ (3 f 3)
?See R. P. Feynman, Phys. Rev,, 76, 769 (1949); included in this volume,
"CORRECTION" TERMS
Notice that (GzYpul) = (G2$u1),SO that, considering only the divergent part
of the correction, the effective potential i s
The 1 comes from the theory without radiative corrections, while the e2
term fs the correctian due to processee of the type just d e s c r i b d , Thus
the correction can be interpreted a s a small reduction in the effect of all
potentials, and one can introduce an experimental charge eeXpand a theo
retical c k r g e eth related by
The amplitude f a r the creation and annihilation for two pairs is a factor L
f o r each, but, to avoid counting each twice when integrating over all d q and
d ~ it ~i s ,IJ2/2, For three pairs the amplitude is lL3/3!. The total amplitude
for a vacuum to remain a vacuum is, then,
where the I comes from the amplitude to remain ai vacuum with nothing
happening. The use of minus signs for the amplitude for an odd number of
pairs can be given the following justification in k r m s of the Pauli principle,
Suppose the diagram for t ti i s a s s b m in Fig, 314. The completion of
this proeess cart occur in two ways, however (see Fig, 325) The second
way can be thought of a s obkfnod by the inbrchmgc? of the two electrons,
hence the amplikde of the second must be subtracted from that of the first,
FIG. 314
P A W L 1 PRIEMCIPLE A N D DXRAG E Q U A T I O N S 163
FIG, 315
according to the Pauli principle. But the second process i s a oneloop proc
e s s , whereas the first process is a twoloop p m c e ~ s s, o i t can b~ concluded
that amplitudes for an odd number of loops must h subtraete.d. The prob
ability for a vacuum to remain. a vacuum is
P robability of vaeuum
= Pv
remaining a vacuum
Probabiltty OF vacuum
to 1 pair =Pv 23
a 1 p&rs I ~ l ~ a i r I '
Probgbility of vacuum
to 2 pairs PV C IK~
a i pairs
~
IKI pair/'
164 QUANTUM E L E C T R B D Y N A M I C S
However, it is a p r o ~ d of
y the Dirac kernel that
because we should not allow the case that the electron in Ifre pair is in the
same sbte as the electron a t b.
F o r t k k e r m l sf the Klefnaorbon equation, i t turns out that the sign of
the ineqmli+ in Eq, (31IQ) is reversed, Therefore, f a r a spinzero pa&f
cle neither F e m i  D i r a c statistics nor independent particles a r e possfbfe.
If the wave f"urmctiions a r e h k e n symmetric (charges reversed add ampli
tudes, ElnsbinBose sbtfsties), the inquality Eq. (3111) ia also reversed.
h symmetrical statistics the? preeence of a pa&ieXe in a sate (say 6 ) ert
hances the chance t h t another i s created in the same s h t e , So the mein
Gordon q u a t i o n requf reh~Base statistiee ,
t i q to t r y to s h a r p n , these a r s m e n t s to show t b t t b
[~,@,a)l' db and 1 is quantitatively exactly compen
sabd f o r by the exclusion principle, Such a f m h r n e n h i relaaon ought to
h v e rt cleat: md simple exposition,
R CALCULUS 123
The problem of the khavior of positram and electrons in given in time (po~trona t t e r i w ) or fomard (pair produc6on). For
nrtrrna1 potentkb, nqlwting their mutml inwaction, is m d y d such a parricle tfte amplitude for transition from an ioitial to a
by repking the &mv of holes by rt, reink~retrttianof the solu hnal s a t e is analysed to any ordw in the poten&f by mn&derlng
tions of the Dlrac equation. It is possible to write down a complete it to undergo a q u e n c e of such scattierings.
slution of tL problem; in terms of: baundary conditions on the The amplitude for a process involving m n y such particles is
wave function, and thi salution contajw ;rutamarialky all the the product of the transition amplitad= for each prLicle. The
pcrsibilities of virtuat (and real) pair f o r m h n rrnd annihiiaGon mclu&on principte requires that antisymmetric mmbinatiuns of
rolfether with the ordinary scattering procem, including the amplitudes be c b m n for these mmplete p r o c m wbjch diifer
w r r s t relative signs of the various terms. only by exchaage of partid. It e m s that a congatent interpre
In this saiution, the ""negative energy sbtegB'appfzbr in a form ~arionis only posJIbk if thr! =elusion pdaciple is adopted, The
wbieh may be pictured (as by Stiickelberg) in wacetime as waves exclusion principle n d not he &km h t o m u n t in intermediate
traveEng away from the external potmtiaf badwards in time. sitam. Vacuum prAlenrs do not azise for chsrgw which do not
E ~ r i m e n k l l ysuch
, a wave carrebponds to a positran apprh i n m t with one another, but these are anaimd nererthelm in
in$ the potential and annihilating the electron. A particle moving mticiparion of application to quantum etatrodynadcs.
forward in d a e (electron) in a potential may be scatered forward The resnlts are also e r & in momentumener@ vdables,
in time (ordinary mtcclring) or backward (pair anni2lilatioian). kuivafence to the wcond quanthtion t h a r y of bletj is proved
When moving backward (positron} it may be wattered M w a d in arm ~pacia;.
$(X., t.)= S K . ( X ~ . 1%;X,, laCbl.)$&~,t.+d.ldh%, Fm. L. The SEhrodiner (md Dim) njuatian a n be visualid
W describing the fact that pfane waves are mttered srrawvely
bp a potential, Figtlre t (a) illustrsrtm the situation in 5rst &er.
since after It+dla the partick is again free, Therefore Rd.&3) is rfie am~fitudefar a free prtide rarting at point 3
the change in the wave function at 2 brought about by 2 $g:r:p; ~ & a ~ ~ ~ & ~ ~ "
the potenthi is (suhtitutc (7) into (8) and f8) into cm*, (Eq, (9)). In (b) is ilfustratd the secmd order r w a s
the equation for $(X*,In)): the waves mitered I 3 are %atfeteda a i n at P H ~ W 
(Eq !10)),,
ever In Dirae oneelectron theory K44 3) wenld rqrmat elcc
~$(2) iSK0(2, troni both of poGiive and of nmlive inergkm proceding from
3) (1(3)Ko(3,2) ~ ( l ) @ ~ , @ ~ & ~3K,(4,3),
, to 4 %isE%. is2,mrdied by chmdng 8 different ~atteringkernd
In the case that the potential exists for an extended was, Qne can in this way obvbusly write down any of
time, it may be booked upon tls a sum of effects from the terms of the expansion ((i),6
each interval &a! so that the total eEfect is obtain& by
integrating over 13 ccswe11 as Xa. From the definition (2) 3. TmflNIBET OR THE D m C EQVATIOPI
of K then, we find We shall now extend the methcrd of the last section
to apply to the D h c equation. AII that would =em
to be necessry in the previous equations is to consider
EZ aa the Birac Hamiltonian, J, as a symbol with four
where the integral can now be extend& over all space indica (fer each prarlic;te). Then Ko can still be de&n&
and time, d7ad8r&a. Automatically them wilt be no by (3) or (4) and is mow a 4  4 matrix which operating
contribution if !a is outside the mnge 18 to ka b a u s e of on the initial wave function, gives the firm1 wave func
our definition, Ko(2, 1)=0 for tzS11. tion. In (10), U(3) can be generafizd to A 43) aaA(3)
We can undersund the result (B), (9) this way. We where A 6, A. are the scalar and vector potentkg (timm e,
can imitgine that a prticle traveb as a free particle the elmtron charge) and a are D k c matrices.
fram point to point, but is scattered by the patential U. To discuss this we s b l l define a convenient r e C
Thus the total ampiitude for a r ~ v a al t 2 from l can tivistic notation. We represent fourvwtom like x, 1 by
be considered as the sum of the amplitudes for variaus a symbol S,, where p= l, 2,3,4 and 2qJra l h real. Thus
alternative routes. It may go directly from I to 2 the vector and scalar potential ( t h e s e} A, A , isi A,.
(amplitude K&?, I), giving the zero order term in (6)). The four tnat~ces@a,B tan be comidered m trmform
Or (see Fig. I(&)) it may go from I to 3 (amplitude ing: as a four vcrctor r, (our y, digem from Paufi's by a
&(St l)), get scattered there by the potential (scatter factor i for P==l , 2,3). We use the summation conven
ing amplitude z'U(3) per unit volume and time) and tion &,h,= a&& atb~dsada= a.6. In pzlrticukr if
then go from 3 to 2 [amplitude K&!, 3)). This ntay a, is any four vector (but not a matrk) we write
occur for any point 3 so that summing over these a=afiy, so that a is a matrk with a vector
alternatives gives (9). (a wit1 often be used in place ymbol for the
&&in, it may be scattered twice by the potential vector). The satisfyr p 7 * 3  . ~ p ~226,
f i, = where &M= f l,
(Fig. l(b}). It goes f r ~ m1 to 5 ( K @ @I)),
, gets s c a t t e d
there (iU(3)) then prmeeds ta =me other pint, 4,
in space time (amplitude Re(&3)) is scattered w i n and S,,= 4, Note that ab+ba= 2ctb and that &=G@@,
(ilZ(4)) and then proceds to 2 (&(;l, 4)). Summing = @ . aisa pure number. The sym'boI a/8z, will m a n
over all pssible places and times for 3 , 4 fuld that the a/a6 for p ~ 1 . 4and
~  
a/az, a/ay, a/& for a = l,
scond order contribution to the total aqlitltde 2,3. Gall V= y,a/d+= Ba/attisa" V. We shall im@ne
K""(Z, 1) is
We are rintpky wlving by suamive apprdmaticns an in1
quation (deducible dimctly from (1) WE& l j l  H ~ + U and (8
with R=&@>,
whae the first inte@ exleads aver ciU s w e and all times
This urn be radily verifid directfy fram ( I f just as (9) peater than the 6, alpwring h the m n d term, and k>rb
expmion of the intern1 equation
K+cA'(Z,1) l])
wKch it a h satbges.
We would now expet to choo*, for the spttciaf solu
tion of (12), II+mXe whem Ke(2,I) w n a w for 12<it
and for t3>& is given by (3) where 4, and E, are the
eigenfwctians and energy valum of a parGcle satis
f h BkacJsequation, and @*, is by $n.
The formuk arising from this choice, however, s&er
keta from the drawback that they apply to the one electron
W DW P OIIMB, LQ, (84) theory of Dirac rather than to the hob theory of the
F%@ 2. The D i m equsrtien wrmits anaiher mlutien K+(2 1) witran. For example, consider its in Fig. 1(a) an
if oat k n ~ d e r tbat
s wava ~~ibttemd by the ptential can electron after kkg z a t t e r d by a potential io a small
M w m d a in drne as ia Fig. 2 (a). This is intcpreted in the wand
orda rw (b) (c) by nortiag that there is new the pad region 3 of space t h e . The one elecbon ays
M t y Pcf of virtuadlpai: Probuction at 4 Ihe positron goin to 3 (as does (3) with K+= KO)that the m t b r d ampEtude
to b. aoi&tied. ?his o n be pirrvd as dmilar to o r $ l i ~ ~ yat another poitlt 2 will p r o c d toward mitive t h e s
w t t e fb) a m t that the ektron ts mttclled bwkwwds in
time horn J to 4
tht mibiitit 2 a
TL wavm sfatkred mm 3 to in (4 r rmat with both positive and nqatlve energh, that is with
arrivfngat 3 fram 2' and a n z k a n g k t h p i t i v e and negative rabs of c k w e of phase, PITo
the dmtron &mX. %%is view is proved equlvdent to h k h t y :
electram tmveLing bekwards in time are rmmizad m etrons. wave 19 mtLerd to times previous. to the time of
mtkring. n rnare just the prowrtim of R0(2,3).
herdter, purely for relativistic convenience, that +,* OR the other haad, according to the p i t r a n
in (3) is rephced by its d j o k t &= , cbn*fl. neptive e n e w stittes are not availabk to the electron
Thus the I)& equation for a partick, mms m,in sn after the mtteriw, Therefore the choice K+=Ko is
exteml field A =A ,y, is unsa&facby. But there are other mlutions of (12).
We shall choose the solutrion d e h h g Kc(2, 1) so that
K+(Z, 1)for tt> 1% is 1h I(!%= u j (3) m podioe w g y
md m, (G determining the propgation of a free stales o ~ l yNow . this new solu~onmmt =My (12) for
ilU tima in orda that the rwrewnk&a be cornpieb.
prticle bwoma
I t must *erefore dBer from the old wlution K@by a
(iVam>K+(2, 1)= i6(2, l), f12) mlution of the homogenaus Dirsc eqwGon. It is c k
the hdex 2 on 17% indiating dgerentiation with rmpct from &e deftnition t h t the difference K@K+ is the
to the coordhtes sss which are reprewated as 2 in sum of (3) over all neetive enerp;y statm, ars long as
12>11. But this gaerence m a t be a wiuGoa of the
IE+(2,1) and W , 1).
The function 11=,(2,1) is d&ed in the a b n m of a homogenmus Dhae equa~onfor all t i m s and m a t
field. If a ptentiilf A is acting a s h i k r function, ay &erefore h rreprwnteci by the =me sum over nega~ve
K+(h",(2,1)m be d e h d , It differs from K+(2,1) by a e n e w s k t a aka for Ca<ll. Since K B ~ in O S k t cam,
%mt order comeetion given by the analoee of (9) it fobws that our new kernel, I),for In<lx i s the
namely mrg&iw Ql the sunr (3)ouer wgolifx?m g y states. That is,
the electron assumed originauy in a positive enerw With ehi inteqretation real pair grduction is aho
state propagates in that state (by f(+(3, X)) to posihn dmribed carretly (W Fig. 3). For example in (13) if
3 where it gets scatter& (A(3)). f t then prmeds to 4, tl<la<ta the equatbn gives the amplitude that if a t
which it must do as a positive energy ekdron. Thh is time $1 one electron is present at 1, then at time just
correctly descGbed by (l+ for fC+(4, 3) contains only one ctl~tronwilt be present (having been scattered a t 3)
positive energy compnents in its expansion, as tr>t,, and it will be a t 2. On the other hand if IS is less than h,
After being scattered at 4 It then p r w d s on to 2, for examph, if t2= tI<t g , the same expression gives the
a e i n n e c m s ~ l yin a positive energy sCate, as It> tr. amplitude that a pair, electron at 1, gositran a t 2 will
In positron theory there is an additional contribution annihihte at 3, and subvequently no particles wijl he
due to the possibiiity d virtual pair prduction (Fig. prewnt. Likewis if 6% and I 1 exceed ta we have (minus)
2fc)). A pair could be created by the potential A(4) the amplitude for Snding a single pair, electron at. 2,
at 4, ihe electron of which is that fwnd later at 2. The psitxon at 1 crezrtad by A(3) from a vacuum, If
positron (or rather, the hob) PEW& to 3 where it ta>ta>ls, ( $ 3 ) dacribes the wattering of a psitron,
annihilates the electron which has arfived there from l. A'fl these amphttldes are rehtive to the ampiitude that
This alternative is alrmdy included in (14) as; con a vwuum will remain a vacuum, which is taken as
tributions far which l,< La, and its study will lead us to unity. (This will be discussed more fully later.)
an inltrvretation of K+(4,3) for Ir<la. The fmtor The analsue of (2) can be easily worked our.@Xt is,
K+(2, 4) de~ribesthe electron (after the pair prduc
tion at 4) proceeding from 4 to 2. Ckewiss: K+@, It)
r e p m n t s the electron prweding from 1to 3. K+(4,3)
must therefore represent the propagation of the positron
where d8iVr is the volunte element of the c l a d 3
or hole from 4 to 3. That it d m SO is clear. The fact dirmensioml surf;tcseof a re&on of space time contaking
that in hole theay the hok prweeds in the mnner CA
and electron of negative energy is r e k t e d in the fact
that K+(4,3) far ( P < ~ J is fmhus) the sum of only
negative energy componenb. In hale theaq the real
e n e r a of these intermdiate s b m is, of coum,
positive. Thb is true here too, since in the phasa
ap(iEn(lrls)) d e s m g K+(& 3) in (17), ERis nega
tive but so is Irlg. That is, the contributions vary with
11 as e:p(i\E,/ (to14)) as they would if the enera
of the mtermediate state were t E,/.The fact that the
entire sum is taken as negrttive in computing IC+(4,3)
is reflected in the fact that in hale theary the amplitude
has its sign reversed in accordance with the Pauli
principle and the fact that the electron arriving at 2
has been exchanged with one in the %a.@To this, and
to higher orders, all procesw involving virtual p i r s Fto. 3. Several dilferent prw can be demibed by the =me
form& de ndiag on the time relations of the variables 11, It,
are correctly dwritrted in this way. Thus P,]GA~(~, I)/* is the probattUity that: (a) An electron at
1 will be ~catteredat 2 (and 00 other psirs form in vacuum).
The eqressions such as (14) can still be d e w h W as
a pawge of the ekctron from 1 to 3 (K+(3, I)), scatkr
ing at 3 by A@), proceeding to 4 (K+(4,3)), scatrering
(3
(b Electron at 1 and pasitmu at 2 annihilate leaving noFmg.
A &n& pair at I md 1 is rreaM frrm rmm. (d) A
at 2 io metered ta 1. (~+(41(2,1) is tbe sum of the e e k 2
agin, A(4), arriving finally at 2. The scatterin@ may, sflttteh in the potential to all orders. P, is a normalizing
mtanttt)
bowwer, be toward both future and past times* an
electron prapagakimg backwards in time king recag
&ed ats a positron.
This tkerefore sul~geststhat negative energy corn
ponents created by scattering in a p t e n t k l be con
sider& as waves propgating from the scattering mint
toward the p t , and that such waves r q m n t the
propgation of a positron annihilating the eltt~tronin
tfie p ~ t e n t h i . ~
"t has often boen noted &at the oneelectron theory s patently
gives the m e matrix e h e n t s for lhi pr as dam hofe theory.
The pr&lern is one of iniepretadon a p d I y in a way that will
alsa 've urrreet r a d t s for other r&, e.g., selfenmgy.
7 &e idea that p i t m m a n & repiewnted as elrrvoos with
proper time r e v e d retative to tme em &S been k w d b
the author and othms, particuisrly by StBckeikr~E. c, C!
paint 2, and N(1) h N,(l)r, where N,(I) is the i ~ w e ~ dtivistic calculations, om be removed irs follows. Instead
drawn unit nomaI to the s d a c e a t the paint X. That of dehing a s k t e by the wave function f ( x ) , which it
is, the wave function #(2) (m this case for a free par h a at a given time tt=O, we define the state by the
ticle) is determind at any point inside a fourdimen function $(l) of four varirrblm XI, t.1 Ilvhi& is a satution
sbnal region if its values on the surface of that re@on of the free particle e q u a t h for all 4 and is f(xr) for
are sp&ed, k=Q. The h f state is m e w i ~defined by a functian
To intewret this, cornider the case that the 3surfam g(2) overall s p ~ e  t h e Then
. aur sudace integrah can
comkts m n t k l l y of all space at =me time say t = O be p r f o r m d since JK+(3, l)@j[xl)@x~= f(3) and
previous to h, and of aU s p c e a t the time T> t g . The JQ(s;~)B@x&+(2,3) =#(3), There results
q l h d e r connecting these to wmpIete the chsure of the
s u d a e may be very distant from X* so that it gives no
appreciable contribution (as R+(2, l ) deamws expo
nentially in spce&L ddir~tbns),Hence, if yr= @, since the i n t e p l now biryq averaU swetime, The transi
the inward &awn norm& N will be B and 8,
tion amplitude to wwnd order (from (14)) is
vacuum (unity), or one pair is created, or two p k s , etc. h ddiition to these si_ngle loops wa have the posrsi
The sum is P,l.) Put in this form the theory is cam bilily that two independent p i r s may be cratted and
plete and there are no divergence problems. Reai proc each pair m y annihilate itself again. That is, there m y
emes are completely indepndent of what gaes on in be famed in the vacuum two c l o d h p s , and the
the vacuum. conbibution in amplitude from this alternative is just
When we c a m , in the succeeding paper, to deal with the prduct of the contribution f r m each of the laops
interactions between charges, however, the situation is considered singly. The total contribution from all such
not so simph, There is the possibility that virtunl elec pairs of loops (it is still consistent to dkregard the
trons in the vacum may interact efectrompetially exclusion principle for these virtwl stzLtm] is L2/2 for
with the real electrons. For that reaon procesw %cur in Ls we count every pair of b p twice, The total
ing in the vacuum are analyzeil in the next section, in vacuumvaeuum amplitude is then
which an independent method of abtsining P, is
discussed, C=
, 1L+L2/2LB/6+ . exp(L), (30)
the succemive terms representing the amplitucle from
zero, one, two, etc., hops. The fact that the contribu
An alternative way of obtaining absolute amplitudes tion to c, of single imps h L is a coweqnence of the
is to multiply aU amplituctes by C,, the vacuum to Pauli principle. For exawler consider a. situation in
vacuum ampLitude, that. is, the absolute amplitude that which two pairs of prticfm are creatd. Then these
there be no particles both initially and finally, We can wirs later destroy themselves m that we have two
wsume C,= l if no potential is present during the h o p , The elwtrons ~ u M at , a given t h e , be inter
internal, and otherwise we compute it as follows. It ckanged fanning a kind of figure eight which is a single
dsem from unity because, for example, a pair could be Emp. The fact that the interchange must change the
created which eventually annihilates itself q a h . Such sign of the contribution requires that the terms in C,
a path would appear as a c l o d loop on s spacetime apwar with alternate signs, (The exclusion principle k
d i q a m , The sum of the amplitudes resulting from all also rapnsibfe in in similar way for the fact that the
such dngle clo*d loops we tall, L, To a first approxima amplitude for a pair creation is K+ rather than +K+.)
tion L is Syxnmebical statistks would l e d to
as expect4 in accordance with the r m n i n g of the previous sec we therdare obtain for the derivative of CSthe exprmien
tions (i.e., (20) with replacing K,),
The prmf is rmdify atended to the m m gmeml e x p r d o n R,
(# .lcan
which ), be a m l y d by induction. First one replwa Fa*
by s relation such as (41) obGning two terns
R=(xo**.Qp*Qzf*..G&lFgpgSr*SI;(P*.. .PIPS* . ex@)
(X$ 
.Qs*Qtu* GLGISP*~,,'"*FI*." 'PLPP' * 'WO), which will be reduced to a simple factor times C,ife) by methods
In the first term the order of Pt,,"* and C;r is then inter&angd, anartogous to those wed in rerlucing R. The operator % can be
prducing an additional term Jg~*(x)fi~~r'(xjd"rtima an exprm imagined to be split into ttrvo pieces Wm. and Jt,,, operating on
sinn with one tess eletron in iniriaX and ftnal state. Nett it is pasitive and negative energy states rmpectively. The iy,,, on
=changed with C*prociuetng an addition Jg~*(x)ftpo7"(~)$% gives zero so we are left with t w terns ~ in the current density,
times a similar term, etc. WnaHy on rmching the Q* with w&ch *,m*@ABaeI and S,$SA?ir,,, The latter V.neg*BA*aeg is just
it anticommutes it can be &mpfy m v e d over to juxawktion .the q e e t a t i o n value of BA taken over aH negative energy states
with X@* where i t gives zero. The s e n d term is imihrly b d l e d ( d n u s 'Vma,flAwnO,* which gives zero acdng on X*). This is the
by moving Ptm,,*" thrmgh anti commuting PI", etc, until i t effect of the vacuum ewectatian current of the e l ~ t r o n sin the
r w h m Pt. Then ir is exchanged with PI to produce an addi sea which we should hsve subtract4 from our original Hamif
tional simpler term with a factor 7"Jpl*jx)Jtn,,"'(x)dt or ronian in the c n s b m a ~way.,
FJPi*(xs)K+tA){2, I)J~~~(xI)$'x~QBXO from (49),with taixi.&=O (the The mmaining term 0,,,*@A~,,,, or its quiva!ent *p,,*@AIV.
extra fi(xg) in (49j gives a r a as it is orthogonal to pg(x1)). This can be considered as Jt*(x)f,,,(x) where fw(x) is written far the
dmcriba in the expected manner the annlhiiation of the pair, p i l i v e energy a m w e n t of the m r a t a r @A*(X). Mow this
electron fi, W t r o n pt. The P,,'"* is moved in this way accerr opmtor, \t+fx)fPo,(x),or more precisely just the **(X) part of it,
gively through the P's untif it. gives when acting on X* '&us a n be pushed through the =p(i&erEdfl) in a manner exmtly
R is rrducd, with the q e c t e d facltlrs (and with attwoating $gm analogous LO (47) when f is a function. (An altermtive derivation
requird by the a l u s i o u prindpfef, to dmpler terms mnlaining results from the congideratian that the operator *(X, l ) which
two bss operatm which may in turn he furher rerluced by u i n g satisfies the Dirac equarion also satis& the linear integral equa
h*in a fimitar maner, etc. After all the F* are used the @*'S tions which are quivaient to it,) That is, (581 can be written
can be r e d u d in a similar manner. They are movd through the by Wh ((50,
S in the opposite direction in such a manner as to p r d u ~ ae purely
neg;ttive enerw aperator a t time 0, using relstions amlogow to
(45) to (49). After all this is done we an? left *ply with the ex
pected factor times Cr (amming the net cha~geis the same In
initial and final state.)
Xn this way we hsve written the Ilofution to the general problem
of the mntion of electrons in given potentih. The &tor CVIs
obtain& by normalization. However for photon fie& it L deair
able to h v e an emlicit form for C. in terms of the wtenentiafs. I
This is &ven by (3Cj)m d (29) a d it is readity demonsirat& that
this also is corrmt m w d i n g to m o n d quaotilation. where in the ftxslt term h==T , and in the m n d l p t e = . t t , The
(A) in &cA) refers to that part of the potential A after The
ht term va&hes for it involvm (from the Kc(A"2, l)) only
b. Aodyds sf the Vacam Rablem WCive $ n e w components of W , wMch give zero operating into
We &aft almlate C, from second quantizrltion by induction xe'. In the m e n d term onIy nqative comwnmts of **(X$)
consideidering a series of problems each wntaining 8 patentiaf die appear. U, then ifr"fx8) is interchanged in order with %(X$)it will
ttibutiou more nearly like the one we wish. sap^ we know C, give zero oprating an X@, and only the term,
for a prohhm &e the one we want and baving the =me ptentials
far time t ktween Jame k and T, hut having pawnlid zero for
times from Q to Ip. Cal! this C,(@, the w r m p n d i n g Eliamiltonian
Bto and the sum of contrihtions for all dngle loops, Ufe).Then
for lo=IX" we have zero potential a t slt times, no pairs a n be will renrain, from the usual commutation rdarion of o*and *,
praduced, I;(T)O and CI(T)=I, For 1030 we have the mm m e factor of C,(&) i (52) times &a is, according to (29)
pie& prohiem, w that C,(@ is what is d e 6 d ar, Cp.in (38). (mfe~nce101, just L(t~&)l;(la) i a a this ddigermce arises
h e r a l l y we faave, from the tlrcra p o t m m AAA during the s h r t time interval
AteB Hen= dG,(rd)/dCo +(&L(IoI/&DZC~(IO) so that integration
from T 0 JoO ebhti&es (30).
Starting from the theory of the eleictromagnetic W d in secand
q m b G o n , a c f d u c ~ o nof the quations fw quantum electro
d m m i a W&& a p w r in tae succetding paper may be w r k d
since B& is identicat to the c o n a b t vamnm Bmiltonian Br for out uru'ng very rjnikr principles. Tlre PaulWei&opf Ebwv of
;<h sad X@is an eigenfuncgon of HT with an eigwvalue (enwgy the RIeinhrdon q m t i o n can a m e n t I y be a n a l y d in emn
of vam) which we m fake as zero. tiallv the =me way as that U& here for Dirac electrow.
PHYSICAL REVIEW VOLUME 76, N U M B E R 6 SEPTIBABER $ 5 , 1941)
R. P, F
13epat.lntd 4 Pkyska, C d Usimsicy, Iikft New Fwk
(Receivd May 9,1949)
Xa this paper two things are done. (1) It is s b w n that a con and p r m w h i y consistent, metbad is therefare availabie for the
sidwable rjmptilteation can be attained in writing down matrix involving electrons and p h t o m ,
ekments for mmplen p r w w s in electrodynamics. Further, a The siroplifiaaon in writing the e r c p h o n s results from on
phydcai point of view is availabte which perraits thew to be ecmphmis sn the overabl @acetime view rmlting from a study
wrirtwt down directly for any swiltc probtem. Being simply a of the 3101ution of the equations of dmaadynamics. The rela(ion
rmhtement of mnventional elwtrodynamia, howevw, the matrix of this to the mnre mnventional Narniltoniau point of view is
elements diverge fox comp1ex process. (2) EIectr&ynamics is dwmd. Pt would be very d a c u f t to mske the modifLcatian
modified by altering the interaction of electrons at short distanw. which is p r o p a d if one insisted on having the quatinns in
All matrix efements are new finite, with the e x ~ t i a nof tb* Eamiltoniarn farm.
relating ta problems of vacuum pofarizstion. The Latter are The methods appiy as well to chrges obeying the Klein&don
maEwted in a mnner suaested by Pauti and Bethe, which gives equaGon, and to the variow m m n t h m ~ e sof nuclear farces.
finite rerrults for these matrkes atso. The only eEccts ~ n G r i v ete IUustrative amples are given. Althowh a mdifreation like that
used in elecrrodynadcs a n make alt matrices finite for all of the
the m d i b t i o n are changa in m s s and charge of the electron%, meson &wries, for m e of the theories i t is no longer tme that
Such e k n p cauid not be dirmtly abservd. Phmomem directly a! d i r e l y obwmbie phenomma am inansicive to t6e d e ~ oEh
obwvable, are insnsitive to the deltlils of the modification used the nrMfifiCation d.
(exucpt at extrme energies). For such phmomena, a lmit can The actual ewluation of integrals a p p r i n g in the matrix
be takm as the ran@ of the mdificatlon goes to zero. Tke rmults ehrrrents m y be faciliktect, in the simpler cam, by m e t M s
then agree with these of Schwlnger. A complete, unambiguous, dwribed in the qpmdix.
h g real quanta, while the interaction view is best for far difftzrent observers in relative motion the instan
the discussion of the virtual F a n & involved. We shall hneous p r w n t is dlgerent, and corresponds to a
emphasize the interaction viewpoint in this paper, first dserent 34imensional cut of spamtime, Thus the
because it is less familiar and &erefore requires more tempsral analyses of diaerent observers is different and
discussion, and second because the important pest in their Hamiltonian equations are developing the process
the probiernsj with which we shall deal is the effect of in different ways, These digerences are irrelevant, how
virtual quanta, ever, for the mlutian is the same in any space time
The Hamilttonian method is not well adapted to frame. By foraking the Hamiltonlan method, the
represent the direct action at a distance between charges wedding of rehtivity and quantum mechanics can k
because that action is dekyed. The Hamiltonian method accomplished mast naturally.
represents the future as developirtg out of the present, We UIustrate these points in the next wction by
If the values of a complete set of quantitim are known studying the solution of Schrtidinger's equation for non
now, their values can be computed a t the next instant relativistic particles interacting by an instantaneous
in time. If particles interact through a delayed inter Coufomb potential (Eq. 2). When the salution is mrdi
actioil, however, one cannot predict the future by hed to include the effects of delay it1 the interaction
simply knowing the present motion of the particles. and the relativistic properties of the electronswe obtain
One would also have to know what the motions of the an expression of the laws af qumtum. el~trodynamics
particles were in the past in view of the interaction this (Eq. 4).
may have an the future motions, This is done in the
Namiltonian electrodynamics, of course, by requirhg
that one specify ksides the present motion of the We study by the same methods as in I, the interaction
particles, the values of a host of new variables (the of two particles using the same notation as X. We start
cmrdinates of the held oscillators) to keep track of that by considering the nonrelativisticca* descrikd by the
aspect ol the past motions of the particles which de Schr6dinger equation (I, Eq. X). The wave function a t
termines their future bhavior. The use of the Hamil a given time is a functbn $(X,, X*, 1) of the cwrdinates
tonian farces one to c h o w the field viewpaint rather X, and X&of each particb. Thus call K(x,, xt,, t ; xe", xt,', 6')
than the interaction viewpoint;. the amplitude that particle a at X,,' at time 1' wiH get
Xn many problems, for example, the close collisions to X, at .iwhife particle b at X; at 1' gets to rl,at i . Xf the
of particles, we sre not interested in the precise tem particbs are free and do not interact this is
sequence of events. I t is not of interest to be able
to say how the situation would look at each instant of
time during a collision and how it progresses from in where Kffeis the KOfunction for particle a consikred
stant to instant. Such ideas are only useful for events as free. In this case we can obviously d e h e a quantity
taking a long time and for which we can readily obtain like K", but for which the time 6 need not tze the same
informatian during the interveningperid, For collisions for partides a and 6 (likewise for 19; e.g.,
it is much easier to treat the process as a whole.' The
Mpcller interaction matrix for the the coltision of two elec
trons is not essentially more complicated than the non can be thought of as the anrpliitude that particle a gaes
relativistic Rutherford formula, yet the mathemtical from x l at 11 to ~8 at fa and that particle b goes from X*
machinery used to obtain the former from quantum at 12 to ~4 at 16.
electrodynamics is vastly more complicated than When the particles do i n t e r ~ tone
, can only define
Schrijdinger" equation with the 8/rl* interaction the quantity K@,4; l , 2) precisely if the interaction
needed to obtain the latter, The dBerence is only that vanishes ktween tt and 12 and also between ta and I,.
in the latter the action is instantaneow so &at the In a real physical system such is not the case. There is
Hamiltonian method requires no extra variables, while such an enormous advantage, however, to the conapt
in the formr relativistic case it is delayed and the that we shalt continue to use it, imgining that we can
Hamittonian method is very cumkrsom. neglwt the effect of interactions between 11 and t s and
We shall be dixussing the solutions of equatiow between ta and t 4 . For practical problem this means
rather than the time digerential equations from which choosing such Iong time internals tolr and t d  1 2 that
they come, We shall discover that the solutions,because the extra interactions near the end pints have small
of the overall spacetime view that they prmil, are as relative egects. As an example, in a mttering prcsblm
easy to understand when interactions are delwed as it may wet1 be that the particle are so well sparat&
when they are instanbneous. initially and finally that the interaction at these times
As a further point, rebtivistic invarhce will be self is nqligihle. Again energy values can be defined by the
eviclent. The &miltonian form of the equations de avmage rate of change of phase over such bng iime
velops the future from the instanhneous presnt. But intewals that errors initially and finally can be neg
"his is. t
k view~ointof the thfiary af the S matrix of Heisen lected, Inasmuch as any physical proMem can be defined
tterg. in terms of scnttering procresses we cto not lose much in
This turns out to be not quite right: for when this
interaction is repraented by photons they must be of
only positive energy, while the Fuurier transform of
S(Ebbr5~fcontains frequencim of both signs. Xt should
instead be replaced by S + f f p ~  r Lwhere
~)
l0 1 fbf
FIG,5. Comptan scattering, Erl, (151.
FIG.3, Ittirttarrvt correction to scattering, momentum space
trated in Fig. &fa),find the ntktrix:
I, Eq. (34). The F means ( K e R )  I or more precisely
the limit as M of (kk+i&)L Further @R means
(2~)~dkldFE2dkadkp.If we imagine that quanta are par
ticles of zero mass, then we can make the genera1 rule
that all poles are to be resolved by considering the For in this case, firstf2a quantum of momentum k is
Inasses of the particles and quanta to have inffnitesimal emitted (r,), the electron then having momentum
tregative imaginary parts, PSk, and hence propagating with factor @I k?)',
Using these results we see that the xlfenerm (9) is Next ~tis scattered by the potential (matrix a) receiving
the matrix element between a and 16 of the matrix additional mamntum q, propagating on then (factor
@ km)q with the new mommtum until the quan
turn is reabsorbed (7,). The quantum propagate front
emission to absorption (k+) and we integrate over all
quanta (&K), and sum on polarization p. When this is
integrated on R#, the result can be shown to be exactfy
where we have used the expr~sion(I, Eq. (31)) for the equal to the expressions (16) and (17) given in B for
burier transform of K+. This form far the selfenerm the same prceess, the various terms coming from resi
is easier to work with than is (9) dues of the pies of the interand (12).
The equation can be understood by imagining (Fig. 3) Or again if the quantum is both emitted and re
that the electron of momentum p emits (7,)a quantum absorbed &fore the scattering takes place one finds
of momentum R, and makes its way now with mo (Fig. .Q<b)l
mentum P R to the next event (factor wkm)i)
which is to abfitrrb the quantum (another 7,). The
amplitude of propagation of quanta is k2. (There is a
f&tor b/*i for each virtual quantum). One integrates
aver a11 quanta. The reawn an electron of momentum P or if both emission and absorption occur after the
propagates as l j ( p  m ) is that this operator is the re scattering, (Fig. 4(4)
ciprocal of the Dirac equation operator, and we are
simply solving this equation, Likewise light goes as
l/&, for this is the reciprocal D'AAiembertian operator
of the wave equation of light. The first y, reprexnts
the current which gneratm the vector gatential, while These terms are discussed in detai! below.
the second is the velocity operator by which this poten We have now achieved our simplification of the form
tial is multiplierl in the Birac equation tvhen an external of writing matrix elements arising from v~rtualproc
field acts on an electron. esses. Procesm in which a number of real quanh is
Using the same tine of rmoning, other probfems may given initially and finally offer no problem (assuming
be set up directly in momentum space. Fbr example, carret normalization). For example, consider the
conider the scattering in a potential A = A,?, varying Gompton eflFect (Fig. S(a)) in which an electron in state
in space and time as a eq(iqz). An eliectron initiebtly p, abmrbs a quantum of momentum q,, plarizatian
in stare of momentum p~=pl,r,will be defiected to vector e t so that its interaction i s e~,y,=rz"~, and emits
state $2 where p t = p ~ C q . The zeroder answer is a screond quantum of momentum q?, polariation
simpty the matrix element of a between states l and 2. ta arrive in final state of momentum p*.The matrix for
W e next ask for the first order (in 8)radiative comec "S First, next, etc., here refer not to the order in true time hut to
tian due to virtual radiation of one quantum. There are is, the suatsion of events alan the trajectory of the electron. Tbst
more prwkky, to the or%r af qpearance of the mat+ in
several ways this can happn, First for the case illus the exprmians.
this prKess is ez@l+qlns)%el,
The total matrix for for prol~agationof quanta of mrnenlum k is
the Compton effect is, then,
@ z C P ~ + q l  mj'@z+@~(P~+(lam)1@,, (15)
the second &m arising b a u s e the ernhion of sz may
also precede the abmqtion of el (Pig. Sfb)). One takm rather than W .That is, writing F+{#) .  ~  ~ k C ( k ~ ) ,
matrix elements of this between initial and find electron
states @ 1 + ~ ~ = # 2  ~ 2 ) , to obtain the Klein Nishina
formula. Pair annihgation with emission of two quanta,
etc., are given by the =me matrk, p i t r o n statw being
t h w with negrttive time component of B, Wether E v q intal over an intemediate quantum which
quanta are abmrbed or emittd depends on whether the prwiousty invofved a fczctor b k / R Zh now suppiied with
time compnent of q is positive or negative. a convergence factor C(#) where
5, TEE G Q ~ X C i E N G EOF PROCESSES WTE
VIRTUAL QUANTA
Thew expressions are, as has been indicated, no more
than a reexpresion of conventional qusntum electro The poles are defined by rephcing k2 by @+is in the
dynamia. As a consquence, many of them are mean limit 64. That is X2 may be assumed to have an infxni
ingkm. For example, the &fenergy exprasim (9) or teshal negative imaginary p a t ,
(ll) gives an infinite result when waluatd*The infinity The function f+(s1z2) m y still b v e a discontinuity
ariscrs, apparendy, from the coincidence of the Bfunction in value on the Ii@t cone, This is of no influence for the
singufarities in K+($, 3) and 4(s4a2f.Only at this point Dirm electron. For a particle satisfying the Rlein
is it necasary to m&e a real dqarture from conven Gordon equation, however, the interaction involves
tional electrodynamics, a departure other thain simply g d i e n t s of the potential which reinstates the d func
rewriting ezpressions in a simpler form. tion if f has discontinuities. The condition that f is to
We desire to make a modificationof quantum eleetro have no discontinuity in value on the light cone impties
dmmics analogous to the modification of classical k2C(ka) appraaches a r o as k2 approaches infinity, In
electrodynamics described in a previous article, A, terms of G(k)the condition is
There the &(slag)appearing in the action of interaction
was rephced by where J(s)is a function of small
width and great height,
The obviom comespondin@; mdif"ication in the quan
tum theary is ta =$ace the &+(ss) appsring the %'his condition will also be U& in discusing the con
quantum mechanical interaction by a new function vergence of vvacuun? polarization integmIs,
f+(s2). We can postulate that if the Fourier trans The exprasion for the sezenergy matrix is now
form of the classical J(st22) is the integral over all R of
F(k2)exp(ik.xtz)d4k, then the Fourier trwform of
f+.($\s") is the same inlegxa1 t&en over only positive fre
quenciies for &> tl and over only negative ones for which, since e(kz) faus off at temt as rapidly as 1/k2,
I.r<l~in analop to the relation of &(ss) to S($), The converges. For practical pup we sfxall suppose
function l(.@)==j(zz)can be written* as herafter that Cfk2) is simply X2/( k2.kg) implying
that =me average (with weight G(X)dX) over values of
X may be taken afterwards. Since in all procmes the
quantum momentum will be cantabed in at teat one
extra factor 01 the form CpR.m)' representing
X cos(]R ~)dkrEKg(K. R), propagation of an electron while that quantum is in
where g(KeK)is h" times the density of oscillators and the held, vve can expect aU such integrals with their
d pwitive RI as (A, Q. (26))
may be a p r e ? r ~fox convergence hcrors to converge and that the result of
all such pracessrts will now be finite and detinite {ex
cepting the p Itb c l o ~ loop,
d discas& below,
in which the inlegr%Isarsover the moments
of the electrons rather than the quanta),
where &"G(X)dX= l and G involves values of X large The integral of (IQ) with C(P) 
X2(kZ h2)koting
compared to m. This sixndy means that the ampfitude and dropphg terms of order d h ,
is (see Appendix A)
*This reIalion is given incarrecdy in A, equiitian just pre
ceding lb.
QUANTUM ELECTRODYNAMICS 177
When applied to a state of an electron of Illamenturn P Vile mwt now study the remaining terns (13) and
satisfying pzl=mu, it gives for the change in mass (as (14). The integrar on k in (13) can be performed (aftw
h B, Eq. (9)) mdtiplimtion by C(R3) since it involves nothing but
the i n t e p l (19) for the selfenergy and the reult is.
&low& to operate on the initial s b t e SE,(so that
Ptlkr=ml). Hence the factor folloGring af#Imf%iii
be just Am, But, if one ROW tries 10 expand l/(P1m)
We can now cornpteb the discussion of the radiative =(P~+m)/(nPtlf~) one obtains an infinite result,
corrections to sattering. In the integrails we include the since pla=mz. This is, however, just what is expected
convergence factor C(kZ),so that they converge for physically. For the quantum a n be emitted and ab
large k. Xntqral (12 1 is also not eonverpnt because of sorbed a t any time previous to the scattering. Such a
the wellknown infrared cabstrophy. For this reason proceps has the e8ect of a change in mass of the electron
we catculate (as disrzusd in B) the vatue of the integml in the state 1. It therefore changes the energy by dE
meuming the photons ta have a smaff mass Xm;,<<m<X, and the amplitude to first order in AE by iAB.! where
The integral (12) becomes 1 is the time it is acting, which is infinite, T b t is, the
major effect of this term wuld be eanceted by the e @ ~ t
of ckange of mass 6%.
The situation can be analyzed in the following
manner, We suppofe that the electron approaching the
scrtttering potential a has not been free for an infinite
which when integrated (see Appendix B) gives (e2/2r) time, but at =me time far past suffered a scattering by
times a potential b. 11 we limit our discusion to the egects
of Am and of the virtual radiation of one quantum be
tween two such scatterings each of tlre eEfects will be
finite, though larp, and their Blgerence is determinate.
The propagation from b to a is represented by a matrix
not unity and its dwiation from unity arise from the
imaginary part of J,,, Again, with closed Imps ex
cludd, a pair of electrom once created cannot annihi
late one another again, the xatiering of light by fight
This is b m u the~ potential praducm the pair with would be zero, etc. Atthough we are not exprimeatally
amplitude proprtionai to a,y,, the electrons of mo sure of these phenomena, this d a s seem to indicate
menta p, and M,+@ proced from there to annihi that the closed loops are necessaxy, To be sure, it is
late, prohcing a quantum (factor y,) which propagates always passibk that these mlnatters of probabiliity con
(fa~toxTV($)> over to the other eiedron, by which sewstion, etc., will work themselves out as dmply i t 1
it absorbed (matrix element of 7, between states 1 the case of interacting particIes as for thow in a fixed
an&2 of the ori&naI electron (%qy,%,)). AI1 momenta P, potenthi. Lacking such ii. demonstration the prmump
and spin slates of the virtual. electron are admitted, tion is that the diaculties of vacuum polarlzsltion axe
which m a n s the spur and the inte@ on d4p, are not sia easily
ealculatitd. An alternative prmedure discusrjed in B is to msume
One can imagine that the ctosd loop path of the that the function K+(2,1) used above i s i n c a r r ~ and
t
pitroneiectron produces a current is to be replaced by a modified function K+' having no
shgularity on the light cone. The eEect of this is to
provide a convergence fmtar C(pm*) for every inte
which is the source of the quanta which a t on the grsI over electron momenta.lS This will multiply the
~ c o a ed l ~ t r o nThe
. qmndty inteeand of (32) by G(pm2)C((P+ q)2ntZ), since the
integral was originally S@,$a+~)d~p&'pb and both
p, and f i b get wnvergence factom. The integrat now
converges but the result is unsatisfa&~ry.~
One e x p c a the current (31;) to be conserved, that is
q,j,=O or qJ,,=O. Also one expects no current if a,
is then chmacteristic for this prablem of ~ h r i m t i o n is a gradient, or @,=g, times a constant. This leads to
of the vacuum. the condition J,g,.=O which is equivalent to qJ,,==O
One sees at once that J,, divergm badly. The modifi dnce I,, Is symmetrical. But when the eqnpresion (32)
fation of S to f afters the amplitude with which the is integatd with such canvergmce f ctors it does not
cumnt j,, will a E e t the scattered e l ~ t r o n but
, it can satisfy this condition. I& altering the kernel from K to
do nothing to prevent the divergence of the integral (32) another, K', which does not satisfy the Dirac equation
and of its edects. we have iost the gauge invariance, ib consequent cur
One way to avoid such difffmjties is apparent. From rent canservation and the general mnsistency of the
one p i n t of view we are consideriry~dl routes by which thmry.
a given elelron can get from one lugion of spa=tim One can st?e this best by calculating J,d, directly
to another, i.e., from the source of electrons to the from (32). The eqrasion within the spur Fcomes
apparatus which measuretr t h m . From this point of (P+qrn)lg(irCPm)Iy, which can be written as the
view the clmeif loop path I d i n g to (32) is unnatural. 
difference of two terms: (Pm)%# &+ @ m)LyL'
It might be: assumed that the only wths of meaning are Each of these terns would give the same result if the
those which start Irorn the source arid work their way integration d4$ were 6thout a convergence fxtor, far
in a cantinuous path [pssibly containing m n y time
reveals) to the d e t ~ t o r .Closed Ioops would be ex U Et would be very intcrmting to calculate the Lamb shiR
wmrately enou h to be sure that the 20 memyclm CM
eluded, We have alrady found that this may be done from vacuum p%xbtion are actually present.
for electrom moving in a fixed potentid. isW s techrii ue also makm selfenergy and rdiationlm rcat
Such a suggestion must m e t era1 qurrstions, howW tering integrals even without t k mdifieacion of 6, taJs for
the diation (and the conseqjuent convergeace factar C(ltn) for
ever. The d o 4 lwps are a conNquence of the w a l the quant;l). Ziee B.
hole theory in electrodynmicts. Among o t k r things, "Added tci the terms givcn k i a w (33) there is a tern
they are require$ ta keep probability conmwed. The 4{k"2@'+.fqrjbr for C(kJ) Xt(&
5 X'))', which is not gaugg
invariant, (In irddician tfte chsrge rmonnslizationh 7/6 irddd
pmbbility that no p k is prduced by a potential is to the iomithm.)
the first can be converted into the hecond by a shift of I t is zero for a free light quantum (qZ=Q).For smlt $
the origin of 8, namely lyP"=p+q.This does not reult it behavm as (2/1S)q2 (adding 8 to the logitrithm in
in cancehtion in (32) however, for the convergenca: the Lamb effect). For qz> (2mj2it is mmplex, the
factor is aaltered by the substitution, imaginary part representing the Ioss in amplitude re
A method of making (32) convergent without spiling quired by the fact that the probicbility that no quanta
the gauge invariance has been found by Bethe and by are produe& by a potential able to produce p i r ~
Pauli. The convergence factor for light can be lmked ((@)b> 2m) decrea~eswith time. {To make the neces
upan as the resuft of supwsition of the e@ects of sary analytk continuation, imagine m to have a smdl
quanta of mrious masses (same contributing nega negative imaginary part, so that ( I  q 2 / h t ) f becomes
tiveiCy), Likewise if we take the factor G P  m 2 ) i(qz/4m2 l)t as q2 g w from below to above 4ma,
=  hz(bzl nz'XZ)l sa that (pns2)Vwnt2) +
Then 8 lr/2+iu where sinhu= (@/4m2 l]$ and
 p  m 2 }   L m2X*)+e are taking the dig=  l/tanB i tanhg= +i(@4m2)1fqS)b.)
enm of the resdt for electrons of mass ns and mass Closed loops containinga number of quanta or poten
(Xz+m2n2)k.But we have taken this digerence for each tial interactions larger than two produce no troubte.
propagation betwecm internetions with photons. They Any Imp with an odd numkr of interactions gives zero
suggest instead that once created with a certain mass (1,reference 5)) Four or moPe potential interactio~give
the ekectron should continue to propagate with this integals which are wavergent even without a con
m a s through all t h potential illteractions until It vergence factor as is well Emown, The situation is
closes its loop. That is if the quantity f32), integrated analogous to that for selfcsnergy. Once the simple
over some finite range of p, is ealied J,,(m22)and the problem of a single closed loop is solved there are
corresponding quantity over the =me rmge of p, but no further divergence diacutties for more complex
with m repbced by (m2+Xz)k is J,,(d+X2) we should proce~.~
calculate
This unpolar* y S; aU in vlrtue of the fact that the many fompnent wave
tight is a r e ~ t i v i s t i a ~ ~&S
invaknt EOncept, afld permiB some shPr&atiom in functio~ 1(" (5 components for spin O 10 for in I) mtisfies
(6Vn)+=?@ which ir lormaliy idmtikl to the %irac W u a l i w
computing cross e i o n s for such Eght. See W. Pauh, Rev. Mod. Fhya. 13, 203 f1W},
The important kernel is now I+@,1) defined in (1,Fq. m t s the msibility of the simultanwus emission and
(32)). For a free particle, the wave function $42) satisfies
absorption of the same virtual quantum. This integral
+DJIm2$=O, At a pint, 2, inside a space tine region without the C(k2)diverge quadratically and would not
it is given by converge if CfE ) =  XZ/(k2 F), Since the interaction
occurs through the gradients of the ptential, we must
use a stronger convergence factor, for exam@€ C<kf?
 X4(R? X2)*, or in general (27) with &XzC;(X)dX.= 0,
In this ase the slfenergy converges but depends
qmdratically on the cutoff X and is not necesmrily
(as is readily shown by the usual method of demon small compared to m. The radiative corrections to
stratiqg Green" theorem) the inteeal being over an scattering after mass renormatization are insensitive to
entire &surface boundary of the region (with nomal the cutod just as for the Birac equation.
vector M,). Only the positive frequency companents of IVhen there are several particles one can obtain Bose
JI contribute from the surface preceding the time corre statistics by the rule that if two processes lead to the
sponding to 2, and only negative frequencies from the same state but with two e l ~ t r o n sexchanged, their
sudace future to 2. These can be interpreted as elmtrons ampritudres are to be added (rather than subtracted as
and positrons in direct analogy to the Uirac case. for Fermi statistics). Xn this case equivalence to the
The righthand side of (35) a n be considered as a second quantization treatment of Pauli and Weisskopf
source ofnew waves and a series of t e r m written down should be demonstrabke in a way very much like that
to represent matrix elements for processes of increasing given in I (appendix) for B i r u electrons, The Base
order. There is only one new point here, the term in statistics mean that the sign of contributian of a closed
A&, by which two quanta can act at the same time. Imp to the vacuum polarination is the ogpofite of what
As an example, suppose three quanta or ptentials, it is far the Femi case (see I). Xt is ftta=ft,+g)
a, expfiq,.~], It, exp(iqb~)~ and G, exp(iqc.12) are
to act in that order on a particle of original momentum
pep SO that. fiaflo+qa and Pb=fiD+qb; the final mo
mentum being P,@b+qe. The matrix element is the
sum of three term w=p,p,) (illustrated in Fig, 7)
giving,
it?. APPLECATXQR TO MESON T R B O m S factors, the resutts then are sensitive to the method used
The theories which have been cfeveioped to dmribe for convergence and the size of the cut*@ values of X,
mesons and the interaction of nucieons can be easily For IOW order prwesses p"r,q is equivalent to the
expressd in the languag us& kre. Calculations, to pseuctosalar interaction ZMF"& beaus9 i f taken be
fowet order in the interactions can be made very emily tween free prticle wave functions of the nucleon of
for the various theories, but agreement with expri momenta fit and &=fi~+q, we have
mental results is not obbined. Most likely alI of our
present formuIatio~lsare quantitatively unsatisfactory.
We shall content oumlva therefore with a brief sum
mary of the methotis which can be used, since 7 5 anticommutes with p? and p3 ovrating on the
The nucleons are usualiy assum& to eiatisfy Dirac's state 2 equivalient to K as is fix on the state l, This
equation so that the factor for propagation of a nuclean shows that the inleractiotl is unusua!)iy are& in tire
of momentum j5 is where M is the mass of the nonrelativistic limit (for example the expected value
nuciwn (which implies that nucleons can tie created in of 78 for a free nucleon is zero), but since ysZ==1 is not
pairs). The nucleon is then asumed to interact with small, peudoscalar theory gives a more important inter
mesons, the various theories digering in the form as action in second order than it does in first. Thus the
sumed for this interaction. pseudomlar coupling comtsnt should be chosen to fit
First, we consider the case of neutral mmns. The nuclear forces illcluding these imprtant second order
theo:ory closest to electrodynamics is the theory of vector prmesm.%The equivalence of psudoscaIar and psudo
m w n s with vector coupling. Were the factor for emis vector coupling rvhich holds for low order processes
sion or ahsolrfrtion of a mmon is gy, when this mefan is therefore does not hold when the pseudoseatar theory
""palarized" in the p, direction, The factor g, the is giving its most importa~tteffects. These theories will
"mesonic charge," replaces the electric charge e. The therefore give quite different rmults in the majority of
amplitude for propagation of a meson of momentum q prcrctical problems,
in intemedhte s t a t e is (q2 $]"rather than q* as it Xn alcuhting the mrrectia~lsto scattering of a nu
is for light) where p is the mass of the meson. The n e c e cleon by a neutral vector rneson field (7,) due to the
sary intqrals are made finite by convergence factors effecls of virtual mmons, the situation is just as in
C ( ~ %  B $as) in eiectroctynamics, For scalar msons with eletrodynamics, in that the result: converges without
scalar coupling the only change is that one replaces the need for a cutoff and depends only on gradients of the
y, by 1 in emission and absorption. There is no loner meson potential. With scaIar (I) or pseudoscalar (7s)
a direction of polarization, p, to sum upon, f i r pseudo neutral mmns the result. diverges loerithmically and
scalar mesom, pseudoscalar coupling repftrce 7, by so must be cut off. The part sensitive to the cutoff,
~ ~ = i y , yFar ~ ~emmple,
~ ~ ~ .the selfenerw mtrix of hovirever, is diret!y proportional to the mcmn poten
a nucleon of momentum in this theory is tiai. It may thereby be rmoved by a renormalization
of mesonic char@ g, After this renormalization the re
sults depend only on gradients of the meson potential
and. are essentiaiiy iindepndent of cut08. This is in
addition to the mesonic chars rmorrnaIizatian coming
Other t y p s of maon theov result from the reptace from the prdudion of virtual nuclwn pairs by a meson,
ment of y, by other expressions (for example by anaiogous to the vacuum wlarization in electro
k ( ~ ~ ~ e  r p 7 ~ with
) a subwquent sum over all p and v dynamim. But here there is a lurtlrer difference from
for virtual m a n s ) . Scalar mesons with vector coupling electrodynamics for scalar or pseudmalar mewas in
result from the replacement of r, by y"where q is the that the polarizaf;ion also gives a term in the induced
final mornentum of the nucleon minus its initial mo current proportional to the meson potential representing
mentum, that is, it is the momentum of the m w n if therefore an additional renormalization of the mass clf
absorbd, or the negative of the mornentum of a meson tlte mesoa wlxich usuatly de~xndsquadratimlly on the
emitted. As is well known, this theory with neutral cutoff.
msans gives zero for all procam, as is proved by our Next consihr charged mesons in the absnce of an
discussion on Iongitudinal wavm In efectrodynamics. electromagnetic field. One can introduce imtopic spin
Pseudoscalar mesons with peud+vector coupiing carre opratrtrs in an obvious way, (SpecifimIly repface the
swn& to 7, king replaced by p'~raq while vector neutral y ~ ,say, by r,yr and sum over i= X, 2 where
mesons with tensor coupling comapond ta using 7++ T  ~ r p i(r+ 7) and T+ change neutron to
(2fi)q(rr~517u). These extra gradients invoive the proton (T+ on proton=O) and T chngm proton to
danger of producing higher divergencies for real proc neutron.) It is just as easy for practical problems simply
e e s . For example, y& gives a logarithmially divergent to keep track of whether the particle is a protan or a
interaction of neutron and electr~n?.~ Although t h w neutron on a &=gram drawn to help write down the
divergencies tan be held. by strong enough convergence
=H. AA,Wethe, Bull. Am. E"11)rs. Soc. 24, 3, 2
2 (Wahiqtan,
M.S b ~ &and W, HeitIer, Phya. Rev. 75, 1645 (1949). 1949).
matrix ekment, This exciuda certain procBses. For method of second quantization of meson fielids over the
example in the scattering of a negative meson from present formulation. There such errors of sign are obvi
to qz by a neutron, the mesan 48 must be emittect first ous while here we seem to be able to write seemingly
(in arder of operators, not time) for the neutron cannot innocent expressions which can give absurd results.
absorb the negative meson ql until it beclomes a protan. Ptseudovectxlr mesans with pgudovector coupling corre
That is,in comparison to the Klein Nishina formula (IS), spond to using YJ(Y,~~~,Q) far absorption and
only the analogue of s ~ o n dterm (see Rg. ii(b)) would for emission for bath charged and neutrnl mesons.
appear in the scattering of negative means by neu In the presencce of an electromagnetic field, whenever
trons, and onIy the first term (Fig, 5(a)) in the neutron the nucleon it; a proton it interacts wi* the field in the
scattering of positive mmns, w;di~described far electrons. The meson interacts in the
The source of mesons of a $ven charge is not con scalar or pseudoscalar c m as a particle obeying the
sented, for a neutron capable of emitting negative me KleinSordon equation, It is important here to use the
sons may (on emitting one, wy) become a proton no method of calculation of Bethe and Pauli, that is, a
longer able to do so. The proof that a perturbatim q virtual meson is assumed to have the same 'Lmass'Vur
gives zero, dirjcussed for longitudinal electromagnetic ing all its intemctions with the etectromagnetic field.
waves, fails. This has the consequencl?that vector me The result far mass p and for (fi2+X") are subtract&
sons, if represented by the interaction v, would not and the difference intqrated over the function G(X)dX.
satisfy the condition that the divergence of the poteen A separate convergence factor is not provided for each
tial is zero, The interaction is to be takenn as 7, fZg,q meson propagation between etectromagnetic interac
in emission and as =y, in abmqtion if the real ensrssxon tions, a&erwise gauge invariance is not insured, When
of m m n s with a nonz@rodivergence of potential is to the coupling involves a gradient, such as rsq where q is
be avoided. (The correstion term p%q,g gives zero in the final minus the initial moanturn of the nucleon,
the neutral cm.) The asymmetfy in emission and ab the vector potential A must be subtracted from the
w ~ t i o nb only apparent, as this is clearly the same momentum of the p t o n . That is, there is an additional
thing as subtracting from the original 7,. .y,, a term coupling f r b A (plus when going from proton to neu
~  ~ q.q, That k, if the tern plu2q,q is omitted the tron, minus for the reverse) reprwnting the new possi
raulting theory dexribe a combination of mesans of bility of a simulti~neousemission (or a b m ~ t i o n )of
spin one and spin zero. The spin zero mesons, coupled meson and photon.
by vector coupling 4, are removed by subtracting the Emission of positive or absorption of negative virtual
term p*q. . 4. memns are represented in the %me term, thesign of the
.
Tibe two extra gradients q. .q make the problem of charge being determined by temporal relations a for
diverging intepals still more serious (for example the electrons and positrons.
interaction ktween two protons correspnding to the Calculations are very easily caded out in this way
exchange of t w charged vector mesons depnds q u d  to lowest order in $ for the various theories for nucleon
ratically on the cutoff if catculateb in a straightforward interaction, smttering of mesons by nucleons, meson
way). One is tempted in this formulatkn to choose production by nuclear cotiisions and by gammarays,
simply 7,.. and accept the admixture of spin zero
e r ,
nuclear magnetic momnts, neutron ekctron xattering,
mesons, But it appears that this f a d s in the conven etc., However, no good apeement with experiment re
tional formalism to negative energia for the spin zero sults, when these are available, is obtained, Probably
component. This shows one of the advantages of the aU of the formulations are incorrect. An uncertainty
arises since the calculatians are only to fint order in g%,
The vector meson field pokntiats ortisfy and are not valid if g v / L is large.
The author is particularty indebted to Professor H,
where s the source for such m w u h is the matrix element of A. Betk for his expknation of a method of obtaining
betw&n states of neutron and protan, By taking the divergence finite and gauge invari~ntresults for the problem of
a/d+ of both sides, conclude that tttp,/ax, s.4ari?as,fdz, so that
the giginat equation can be rewritten as vmuum polarization, He ia also gratttful far Professor
r j s ~ f i  c ; " ~ p .P~fr,+ci"a/a.r(ds~/ax~)~.

Bethe's criticisms of the m a n u z ~ p t and
, for innumer
able discussions during the development of this work.
The right hand side gives in momentum representation 7,
~**~,p,r, the left yieids the (@&g and finally the interaction He wishes to thank Profesmr J. Ashkin for his careful
s , ~an the hgran&n gives the 7, on absoiorption. reading of the manuscript.
~ find generaliy that arueIes of spin one
droceedlng in t h way
can be represented by a l o u r  v m ~ I
r (
jwhieE,
, for a free particle
of momentum p satisfia q.rr0). The promation of vixtuili
prticles of momentum q from state v to C is reprmated by In this appn&x a method will be iltustrated by which the
mllltiplicatian by the 44 mtrix (or tensclr) P,,=(b,,#*q.&,)
X( m e 6rst+rder intmmtion (front the Proea equaann) simpkr integrals appearing in problems in electrodynamics can
w i t f i n rlectmmagnetie potendai o rrp(ih.s) corresponds to he directly evaluated. The integmk arl%ng in more cornpiear
nuttiplieation by the matrix I f , , = ( q z  a + q ~  a ) 8 ~ ~  q ~ p ~  p 1prm
~~ lead to rather eamplicated functions, but the study of
w b e p; and qg=qt+k are the mommta before and after the the relatias of one intqrat to another and their exprminn in
with m&k 
interaction. Finally, two tentiak a, B m y act simulmmaaty, terms uZ simpler inlegrats m y be f d t i t a t d by the m e a d s
(abR,+bIraI. given here.
QUANTUM ELECTRODYNAMICS 78.5
As a tpimi problem eonsider the integral (12) appearing in We Ben have to do inkmats of the farm
tb first order radiatianlem mttering problem:
Jfl; ka; bL7ld%(VL)*(A"2pl.k At)"
~ 8  km}IrPk"d4RC(kl),
j ~ ~ (km)Ia&~ (la) X ( V  Z p z  R  A ~ )  ~ , (98)
where we shall take C(RI)to be lypicaliy A2(8"Xyi and where by (1; R,; kOkJ we m a n that in the place of this symbol
d" (2r)*dRldk&b&k4. we first rationJEze factors either ' 8 Or may s(;ind in diaTerenr In
@ k.r)i U k+Itl){(PR}*rn*]l obtaining, compkicatcrd problems &ere may be more factors f V 2p6.Kd,i*
or other powers of these factors (the (&L)""may Fte considered
ss a special case of such a factor with #,=a, &,=L) and further
S"vP@kfmlailtt
1 k  t d ~ , k " d * k ~ ( @ l factors R e k,k,R,. in the numerator. The potes in all the factors
X((p8A]*m""@2k)~m33~1 (24 am d e degnite by the assumptian that L, and the 4%have
infinitesimal negative i r n a g i ~ r yparts.
m e matrix e ~ r d o may n be simpiifid, f t a P W r s to h to We shall do the integrals of sucmgve complexity by inducGen,
do so a j t e the integraeons are pertormed. Since. AB==ZA.BBA W., start with hesimplest convergent one, and show
where A .BA,B, is a, numhr mmmuting with all matricerr, Bnd,
if R is any e x p r e e n , snd A a vector, since r,A. Ar,,+2A,, . f ; l % ~ ( ~  ~ )  s(8iL)1,
= (loaf
For this integral is J(2x)*dR&E(k4# K  KL)a where the
Expmeons betwmn two ymBs can be thereby redue& by induc vector R, of magnitude K==(W;.K)b is RE, Rn, ka. The integral m
tion, Particularly useful are kc s h w s third order polea a t k4= +(P+rl)b and k.5 (hq+Ljb,
fmwining, in accordance with our defmitions, that I, has B small
negative ima&nary part only the first i s below the real axis. %c
contour can be closed by an infinite semicircle below tbto axis,
without change of the value of the integral since the ~ontribution
from the micircle vanish= in the b i t . Thus the contour can
be sbmnk about the pole R4 +(k"+L)f and the resulting R+ inte
where A, B, C are any three vectormatricm (i.e., linear com gral is  2 r i times the residueai this pole, Writing kc= (ICVLL))fe
binations of the four 7's). and emandin8 (kr*lc'"L)~~*(e+2(Iri."s+L]~)~ in powm of
In order to crrllculate the integral in (2a) the inkgrat may be t, the residue, being the toeacitnt of the term cl, is seen to be
written as tbe sum of three t e r n (since k==k s r J , 6(2(E+L)4)6 SO O U integd ~ is
 (3i/32s)jeer 4r@dK(RI+ I;)M (318i)(1/3L)
estahkishing (10a).
We akw have fkod"R(VL)a=O from the symmetry in the
k space. We write t h e results as
1
(8i ] (l; k,)dik(&L)ei. (l; O)&l;i, (Ifa)
1: is far I; the (S; k,; k,k,) is rqtaced by 1, for JSby h,, and where in the hrackeu (f; k d and ( 1 ; 0) corresponding entries are
for 1 8 by k,k,. to be U&.
More complex p r m s w of the Ernt order involve mare facton Substituting R= @P in (1Xa),md calling Lpa=d shows that
fike (@a k)' m Z , Fand a eorraponding increase in the number
of R's which m y appear in the numerator, as kOk,k,... Higher
order prmmses involving two or more virtual quanta involve
simitar integrals but with factors possihiy involving k+# instead By digerentiating both sides of (f2a) with respat to A, or with
of just A, and the intqrsl eztending on kgd4kC(Wk""d%'6flk"). r a p t to p* there follows directly
T h q can k sinlpliged by metb& s ~ i o g m to s t h o s U& on
th h s t order integralri, wi)j '(1,, k U, RQR. )d'b(R"2p.kA)'
The factors @k)*m' may be writ& . U; p,; p,p,j~,,w+a)>w+&)+. ff~a)
Further diBerentiations give dirertty succasive integrals in
cluding more k factors in the numerator and higher powers of
whsxe h=&Bf, dt=mIP1: etc., and we can consider dmlinp (BP 2p. k  d ) in the denominator.
with cases of greater generality in that the diEerent denomimtors The integrals so far only contain one factor in the denornilla.tor,
need not have B e %me value of the msss m. In our specific prob To obtain results for two factors we make use of tbe identity
lem (64, pp.& so that Al..@, but we desire to work with greater
gmesality.
Now for the factor C(&}/@ we shall use K(& ha)Wa.
'This am he written as ( s u ~ a r by
d some work of Sehwinger's involving Gaussian inte
grals). This r q r e n t s the prduet of two recipro~llsas a para
metric integral over one a& wilI therefore permit integrals with
two factors to be expressed in terms of one, For other powers of:
l"hufi we can q l a c e by (&l;)* and ot the end inte a, b, we make use of all of tbe identiem, such as
grate the result with respect to L from m0 to X% We can for
many p r ~ t i wpl u w s consider As very I s s e retative t~ m* or p.
When the ~ d d n a lintegral anverges even without the con
vergence factor, it will be obvious since the & fntemtian wili then deducible from f14a) by successjve dilferentiahns with r a p t
wavergent to infutity, EX an infrered catastrophe exists in the to a or 6.
b & g d one can imply m m e quantei have a smaE mass X,ta To perfom an integral, such aa
and ertend tfie integral on L from X#, to h', rather t h n frnm (8@J(1; ka)d%(kr2pl.RAf)*(iP12p~~hA~~, (l&)
zero ta h'.
write, using (IS*), ~4(iz)dxln(x(lz)*)= (1/4) find
(kf2@,. ~A,]"(Y  2 p z z ~  d z )  g = x2lcdx(k)2&*k&~,
where
px=x#t't.f lz)#~ and A,xA~+(l xj.Laz, (17s)
(note that d, is not equal ta m'fir? so that the expraian (164 sv that substitution into (19) (after the Ukm)' in (19) is
is (81)&tM~$(l;k,)d*kR()rZZp~*krl~)~which may now be replac& hy (p k+m)(Re 2p.k)'") gives
evaluated by (IZa) and is
(1% ~&/8rlrpt@+nrlt2In(hg/=a1+2)
p(in(ht/='3 #)l%
.(@/8r)C8rn(tn(h2imZ)+1)#(2 In(~~/nr")+5)7,
where#,, A, are given in f17a). The integral in (18a) is etementary,
being the integral of ratio of plymmials, the denominator of m.
using (4a:ita)tr, remove the 7,'s. ThisagrecPs with (201 of t
k text,
w n d degrw in x. The general expra&on aftbough rmdily ab and gives the selfeneru (21) whea ft is r ~ l g by d m.
tained is a rather complicated camhination of rootsand lagarithms.
Other integrals can he obtained again by pramet& differentia
tion, For exampie difierentlation of (L&), (18a) with respect to
ba or p*, gives The term (12) in the radiationlms mttering, after rationa2idng
the matrix denominators and using Pf =##r)i" requira the
integrals (Ba), as we have dirscumd, This is an integrat with
three denaminatare which we do in tw stagea. First h e fwtars
(g2pl.R) and ( l  2 p ~  k )are cornbind by a Wrameter y;
(@2prkj."l2ps* k]'i~f6 & ( @  2 h * h ) T
again lading to elementary integrals.
As an erample, codder the case that the second factor is just
(@L)* and in the first put #*=P, &#=it, Then f"lrp,
&rh+(f x)L. There results
from (14a) where

vP11(1 rPz.
We therefore need the integrah
(21a)
since (P k)tmS= Rs 2p.k, as p";.. me.This is af the form (164
with &$=l;, $#=Q, hsd?, PI=$ so that (18a) gives, since
(XX)~, Asm~Lp
The integrals on y give,
These i n t e ~ a i son y were performed as fotlovrs, Since &=#tf q where we assume X*>>& and have put some terms into the arbi
where q is the momentum carried by the potential, it follows from trary eonsbnt C' which is indepndent of Xvhut in principfecould
p$);2=mz that Q=# so that sinct? @,=P~+g(ly), depend on 92) and which drops out in the inkegrat on G(X)dh. We
pu~=~m2g~yjf y). The substitution 2y l=ana/tan@ where B have set 4m' sin%.
is dehnetl hy 4mx sin2@=q" is useful forlr nawc*a/secV En a very similar way the integral with rtp" in the numerator can
and @,Sdy = j n ~ gsin2B)Va where n goes from e to 48,  be worked out. It is, of coum, n e c e m y ta dlEerentiak his m*
Thew rrtsultsare subtituted into the original swtexing formula also when calculating I' and I f i 3 h e r e mulls
(2a), giving (22). X t has been simpiified by frequent use of the
fact that p, oprating on the initial s a t e is m, and likewise #I
when it appears at the left is replacable by m. (Thus, to Smplify:
yltps~jb8~~ ~PIQ#O by with another unimprtant conshnt C'. The mmplek problem re
= 2@2qfa@3+ql = 2fmg)a(m+g). quires the furlher integral,
A tern, like qaq= q1u+2(a.q)q is equivalent to just @a ince
mM has zero matrix element.) The renormalization
term requ~resthe correqonding integrab for the special case
40,
The value of the integral (31.a) times d differs fram (33a), of
C, Vacuum Palmization cuum, because the resuits an the right are not actually the ink
grals on the left, hut rather equal their actual value minus their
The expressions (32) and (32') for J,, in the vanrum polariza value for nzP ?n2+P.
ikon lxrobicm require the caIculation of the integral, Combining tbesa quanti.ties, as required by (32), dropping the
J.,(iifl=  5SSPEY.@ t4+n)rp@+1~+r~>~4~
constants 6, C" and evaluating the spur gives (33). The spun are
maluated in the usual way, noting that the spur of any odd
number of y matrices vanishes and Sp(AB)=S#(BA) for arbi
X (Q &@lz m?*)"@+ f 1 ~ 1 7  " , (32)
trary A , B. The S#(I)=B anti we aim k v e
Itbere ue have repiaad P by B tq to simplify the cakculation
mmewhat. We shall indicate the method of caieulstion by studying  
the integral,
I ( 4= ~ P , P P S ~ C P iq)2ma)h((P+ t~)"f&~l~. 
ts@f@%+mr)@s~z~@at~a)@~mt)f
(Pt.#smm~a)(ptP4"mrc)
 (~z.~amrmd(P~.Pe~@O
The factors in the denminator, p@,gmP+jqs and fP+p.g +(Pt.#~mlmd(@~.lbams~z~~ (S@
  f itre combined as usual by f8a"l
~ t 1 ~ 4q2 for symmetry we where m,, m, are arbitrary fourvecbrs and constmts,

sul>strtute .v g(1frl), (12) 4 U 1)) and integrate s from I t is interesting that the terms of order X9nh* go out, so that
l tafl: the charge renormalization depnds onky fogaritbmically an ha.
~(~a') Jy : I P . P ~ ' P ~~p 4 m".+i@l"d@iZ.
' C3Oa)
This is not true for Borne of the m w n rhm~.es Electrdynamics
.
h saviriowly unique in the miidnas of its divergeaa.
Btrt the integral on ) will not be found in our list for it is badly
divergent, However, as d i s c u d in Section 7, Eq. (32') we do not L), More C@mpi@8:
PrabXenrs
wish I($+) but rather J6"[l(m*f l(mD+ h))s(h}dhh We can
calculate the difference I(="I(mr+XP) by first cakulating the Matrix compla prohiems
derivative l'(m9+L) of I with respect to m%t &+L and later =hnus that ad for lhe
L from eero to dE~erenlialinlJ(%l, three iilweatjons; higher order carrections to the MQIHhr Katter
respect ta N10 find,
=Jiff
I ~ c ~ papde@w
~ ~ R ?.pg+~t ~:p~~d~,
This still diverges, but we can cfigerentiste again to get
I " < ~ ~
+3
Bcp . p ~ " a ~  @ p . gd ~+ip2)"drr
::fllgi ( t ~ l q ~ q ~ ~ it (cip, , ~ ~ f d n
(where D $(gf)@+~g+L),which now convergesand Itas been
(3 1 a)