James D.

Bjorken

Sidney D. Drell

' •• O(·'aC. "roJ ••• or "roJ_I.llor

S'fu"or" l.lnNr AC('e'era'flr Cerae.r Sc.anJorf' (j"lv.r.ley

MoGraw-HIII Book Company

N.M' l'orh San "-raru:l.eo Toroneo LOllf'on

Relativistic QuantuDl Mechanics

--
I .....
___ .- _,. • "II , I· /1 J
' .
•
., 'I' ,'"
.. .... , ••
.. ~ t
,
'.
(;! I
-.- _ ............ -

Relatlvlltle Quantum ftfe()hanloa

Copyright ® 1064 by McGraw-IIill, Inc. All Righte Reserved, Printed in the Unitod States of America. Thi. book, or parte thereol, may not be reproduced in any form without parmiaion of the publiahen.

Libra'il 0/ Congreu CatAlog Cord Number 63-21778 4667-MP-V87

03493

•

Preface

The propagator approach to a relativistic quantum theory pioneered in 1949 by Feynman has provided a practical, as well a8 intuitively appealing, formulation of quantum electrodynamic8 and a fertilo approach to a broad claas of problems in the theory of elementary particle8. The entire renormalization program, basio to the present confidence of theorist8 in thc predictions of quantum electrodynamics, i8 in fact dependent on a Feynman graph analysi8, u is also considerable progrc88 in the proof8 of analytic propertlee required to write diepersion relations. Indeed, one may go 80 far as to adopt the extreme view that the BOt of all Feynman graph8 i. the theory.

We do not advooate this view in this book nor in ita companion

"II

PreJac.

volume, "Rclativistic Quantum Fields," nor indeed do we advoeete any single view to the exclusion of others. The ullMtisfactory status of present-day elementary particle theory docs not allow one such a luxury. In particular, we do not wish to miuimise the importance of the progrese achieved in formal quantum field theory nor the considersble understanding of low-energy meeon-nuelecn prOCe88e8 given by dispersion theory. However, we give first ernphaeis to the development of the !i'eynman ruloa, proceeding directly from a particle wave equation for the Dirac electron, integrated with hole-theory boundary conditions.

Three main convictions guiding us in this approach were the prinlary motivation for undertaking this book (later to become books) :

1. The Feynmau graphs and ruloa of calculation aunuuarise quantum field theory ill a form in close contact with the cxperlmental numbers one wante to understand. Although the statement of the theory in Wins of grapbs may imply perturbation thcory, use or graphical methods in the many-body problem shows that this Iormalism is flexible enough to deal with phenomena of nonperturbative character (for example, superconductivity and the herd-sphere B080 Bas).

2. Bome modification of the Feynmau rules of calculation may well outlive the elaborate mathematical structure of local canonical quantum field theory, based as it is Oil such idealizatioll8 &8 fields defined at pointe in epace-tiiue. Therefore, let us develop thcac rulea first, independently of the field tht'ory Iormalism which in time lnay come to be viewed more &8 a supeestrueture than as a foundation.

3. Such a development, more direct Dud leas formal-if leea compelling-than a deductive field theoretic approach, should briug quantitative calculation, analysis, and understanding of Feynman graphs into the bag of tricks of a mueh larger community of physicist8 than the specialized narrow one of second quantized theoriste, In particular, we have in mind our experimental colleaguee and studenta intcrested in particle physics. 'Ve believe this would be a healthy development.

Our original idea of one book has grown in time to two volumes, In the first book, "Relativistie Quantum Mechanlce," we develop a propagator theory of Dirac particles, photons, and Klein-Gordon mesons and perform a series of caloulations designed to illustrate various useful teehniq uee and concepts in electromagnetic, weak, and strong interactions. 'fhcae include defining and implementing the renormalization program and evaluating effoots of radiative eorree-

tione, such &8 the Lamb shift. in low-order calculations. The neceesary background for this book is provided by a course in 1I0nreiativistic quantum mechanics at the general level of SchitT·s text "Quentum Ivl00 hanics. "

In the second book, "Relativistic Quantuln Flelds," we develop canonical field theory, and after constructing closed expressions for propagators and for ecattcring amplitudes with the LS~ reduction technique, rcturn to the Iteynnlan graph expansion. The perturbation expansion of the scattering amplitude constructed by canonical field theory is shown to be identical with the II'cynman rules in the first book. \Vith further graph analysis we study analyticity properties of It'eynman amplitudes to arbitrary orders in the coupling parameter and illustrate dispersion relation methods. Finally, we prove the finiteness of renormalized quantum electrodynamics to each order of

the interaction. •

\Vithout dwelling further on what we do, we may list the major topics we omit from diecueaiou in these books. The development of action principles and a formulation of quantum field theory from a variational approach, spearheaded largely by Schwinger. are on the whole ignored. \Ve refer to action variations only in search of symmetries. There is no detailed dlecuseion of the powerful development! in axiomatic field theory on the one hand and the purely S-lnatrix approach. divorced from field theory, on the other. Aside from a dieeussion of tho Lam b shift. and the hydrogen atom spectrum in the firet book, the bound-state problem is ignored. Dynamieal applications of the dispersion relations arc explored only minimally, A formulation of a quantum field theory for massive vector mesona is not gh'en-nor is a formulation of any quantum field theory with derivative couplings. I'·inally, we have not prepared a bibliography of all the significant originul papers underlying many of the developments recorded in these books. Among the Iollowing recent excellent books or monographs is to be found the remedy for one or more of these deficiencies :

Selt"ebor, s.: UAn Introdurtion to nalativiatie Quantum Fiald 'l'hoory," Na" York. H ar~r " Ro", Pu blillh.rs, J ne., 1061.

Jauah, J. M .• and F. Rohrlieh: "The Theory 01 Photons and Elootrona," Cambridle. M .... AddilOn-WaliflY Publishinl Company, Inc .• 1066.

BOIOliubov. N. N .• and D. V. Shirkov: Ulntroduetion to the Theory of Qua.ntiled Flolds." No" York. Int.arlClenoo Publishers. Inc .• 1060.

AkhielOr. A., and V. B. Berelt.atald: uQuantum Electrodynamic.," 2d ed., Ne" York. JOhD Wiloy" Sons. Ina .• 106.1.

Umelawa. H.: "Quantum Fiald Theory," Amlltemam. North Holland PubtJshlnl Company. 10156.

Ix

lIamiltoo, J.: uThoory 01 Elemeowy Partialea," Loodoo, Oxford Uoivoralty Prall, 1060.

Mandl, F.: Ulotroduation to Quant.um Field Theory," New York, Interecience l'ubUaharl, Ina., 1060.

Roman, P.: "Thoory 01 Elamentary Particlol," Amaterdam, North Hollaod Publilhinl Company, 1960.

Weotlol, G.: uQuantum 'rbeory 01 Flold," New York, 10torIOience l'ubUahen, Ina., 1040.

Schwinlor, 8.: uQuaotum Electrodyoamlcl," New York, Dover Publicat.lool, Ine., 10M.

Feyomao, n. P.: uQuantum Eleot.rodyoamlel," New York, \V. A. BeoJamln, Ina., 1062.

Klein, L. (ad.): uDlapclIloo Relationl aod the Abttraat Approach to Fiold Theory," Now York, Oordon aod Breach, Scianae Publilhen, loa., 1061.

Screaton, G. It. (od.): uDilperlioo Ralatiolll; Scottiah Unlvenitlel Summer 8ohool," New York, Inteneiooee Publilhen, Iae., 1061.

Chew, G. P.: uS-Matrix Thaory 01 Btrool 10teractJone," New York, W. A.

DanJamio, Iae., 1962.

In conclusion, we owe thanks to the many students and colleagues who have been invaluable eritics and sounding boards &8 our books evolved. from lectures into chaptere, to Prof. Leonard I. Schiff for important initial encouragement and support to undertake the writing of these books, and to Rosemarie StOlnpfcl and Ellen Mann for marvelously cooperative 8CCretarial help.

J ames D. Bjorken Sidnl'l/ D. DreU.

Content

preface "U

cllapter 1 The Hirac Equation 1-/

1.1 Formulation 01 a Relativistic Quantum Theory • 1.2 Early Attempta .f-

1.8 The Dirac Equation B

1.4 Nonrolativiatic CcrreepondeDce 10

chapter J Lorentz Covariance of the Dirac Equation 15 ../

2.1 Covariant Form 01 the Dirac Equation 18

2.2 Prool 01 CovariaDce 18

2.8 Space Reflection '4 2.4 Dilinear Covarianta U

chalJISr 3 Solutions 10 tbe Illroc I~quatlon (or a I'·ree Porti£le 3.1 Plane Wavo Solutions .8

8.2 I'rojection Operatore for Energy and Spin 3.

3.3 I'hysicallnterpretntion of Free-partide Solutions and Packets

~haptsr 4 The Foldy-"louthuyeen Tran.Cormation 4S

4.1 Introduction 46

4.2 Il'reo-particle Transrormation 48

4.3 The General Transrormation 48

4.4 The lIydrogen Atom 6.

cl.npter S lIole "heory 68

6.1 'fhe Problem of Negative-energy Solutions 64

6.2 Charge Conjugation 00

6.8 Vacuum Polarization 70

6.4 TIme Ilcver8l11 and Other SymmetriCR 71

cllaptsr 6 JlrOI)llgator Theory 77

6.1 Introduction 78

6.2 The N onrelativistic Propaaator 78

6.8 Formal DefinitioD.l and Properties of the Groen'. Functions 83

6.4 The Propagator in Posltron Theory 80

cl.apter 1 Alll,lIcation. 99

7.1 Coulomb Scattering or Electrons 1 ()()

7.2 Some 1"raoe Theorems; the Spin-averaaed Coulomb

Crose Section 108

7.3 Coulomb Seattering of Positrons 100

7.4 Electron Scattering from a Dirac Proton 108

7.6 Hiaher-order Corrections to Electron-I'roton Scattering 116

7.6 BremutrahluDK 110

7.7 Compton Scattering 117

7.8 Pair Annihilation into Gamma Raye 16.

7.9 Electron-Electron and Electron-POIltron Scattering 136

7.10 Polarilation in Electron Scattering 140

"c/loptsr 8 IIiMber-order Correction. to the Scattering Matrls 8.1 Electron-Poeitron Scattering in Fourth Order 148

8.2 Vacuum Polarization 163

8.8 Renormalization or Externall'hoton Linea 161 8.4 Selr-mau or the Electron 16.

8.6 Renormalization or the Electron Propagator 164 8.6 The Vertex Correction 186

8.7 The Lamb Shift 177

.. cllapter 9 The 1{lcln-Gordon Equlltlon 183 9.1 Introduotion 184

9.2 The Propagator ror Klein .. Oordon Particlea 186 9.8 Introduction or Electromagnetic Potentials 188

9.4 Scattering Amplitudea 190

9.6 Low-order Scattering l'roceBles 101 9.6 Ilighcr.ordcr Processes 106

- 9.7 Nonrelativistic Reduction and Interpretation

or the Klein-Gordon EquaUcin 108

(·lltlIJler 10 Non('lectromagnetic Ilitera(~tlon. 209 10.1 Introduction .10

10.2 St.rong Interactions .11

10.3 Isotopic Spin Formalism ••• 10.4 ConllOrvod Curronts "8

10.6 Approximate Calculationa; Nucleon-Nucleon Scattering •• 7 10.6 ~fC!SOn-Nucleon Scattering .81

10.7 Projection Operaton ror Isotopic Spin

and Angular Momentum .34

10.8 Cross Sections for Pi-nucleon Scattering 130

10.9 .~lectromagnetic Structure or y.,·[eeons and Nucleona '41 10.10 \Veak Interactionl '48

10.11 Beta Deca.y I '47

10.12 Two-component. Neutrino Theory .67 10.13 ~fu-mClOn Decay .61

10.14 Pi-meson Decay 184

10.16 Two Neutrinos .68

10.16 Conacrved Vector Current Hypothesis 170

10.17 uParti&Jly CoDIOrvod" Axial Vector Coupling .73

Appendl% A Notation 281

Appentli% II Rulce for Fcynman Gral)". 285 Inde% 291

1

The Dirac Equation

•

1 Formulation of a Rclativh.ltlc Quanluill Theory

Since the principles or special relativity are generally accepted at this time, a correct quantum theory should aatisfy the requirement of relativity: laws of motion valid in one inertial system must be true in all inertial systems. Stated mathematically. relativistic quantum theory must be fonnuleted in a Lorentz covariant form.

In making the transition Irom nonreletivistie to relativistic quantum mechanics, we shall endeavor to retain the principles underlying the nonrelativistic theory. ,\\fe review them briefly:1

•

I. It~or n given physicalsystcm there exists a state runction ~ that summarizes all that we can know about the system. In our initial development of the relativistic one-particle theory, we usually deal directly with a coordinate realization of the state function, the wave function t/I(qe • • • ,8e • • • ,I). t/I(q,8,t) is a complex function of all the cla88ical degrees of freedom, ql · · . q", of the time t and 01 any additional degroos of freedom, such as spin 84, which are intrinsically quantum-mechanical. The wave function has no direct physical interpretation, however, 1"'(q1'" q",81 . • • 8",1)" ~ 0 is interpreted 88 the probability or the system having values (ql . · . 8n) at time t. Evidently this probability interpretation requires that the sum of positive contributions 1",11 for all velues of q1 • • • I" at time 1 be finite for all physically acceptable wave functions t/I.

o 2. Every physical observable is represeuted by a linear hermitian operator. In particular, for the canonical momentum P4 the operator correspondence in a coordinate realization is

Ii a

P4-+'" r

, ",q4

3. A physical system is in an eigenstate of the operator n if

(1.1)

where cI>" is the nth eigenstate corresponding to the eigenvalue w". For a hermitian operator, w" is real. In a coordinate realization the equation corresponding to (1.1) is

n(q,8.0t/l,,(q.B,t) - w""',,(q,8,t)

•

I See, tor oxample, W. Pauli, ItHandbuch tier PhYllik,u 2d ed., vol. 24, p. 1, J. BprlnKer, Berlin, 1933. L. I. SohifT. "Quantum Mechanlea," 2<1 ed., !\{oGraw. Hili Book Company, Ine., New York. 1066. P. A. ),1. J)lrac, uTbo J'rinciplll of Quantum ~leohanicl," 4th ed., Oxtord Univol'llity PI'OII, London, U168.

2

Dirac quoUon

4. The expansion postulate states that an arbitrary wave Iunction, or state function, (or a physical systenl can be expanded in a complete orthonormal set or eigenfunctions ~,. or a complete ect or commuting operators ({t.). We write, then,

whcrc the statement of orthonormality is

2 / (dql • • ')"':(91 • • • ,a . • • Jt)~",(ql ' • • " • • • ,t) - a"",

,

(atll1 records the probability that the Bystcm is in the nth eigenstate.

o. The result or a measurement of a physical observable is any one or its eigenvalues, In particular, for a pbysical systcm described by the wave function '" - %0,,"',., with o~" - w,."'" , measurement or a physical observable 0 results in the eigenvalue WIt with a probability In·"II, The average of many messurementa of the observable n on identically prepared systcms is given by

(o)* • 2 / ",-(ql • · • ,8 . , • ,t)U"'(ql • • • ,If • • • ,t)(dql • • .)

,

6. The time development of a physical system is expreesed by tho SchrOdillger equation

itt : - II~

(1.2)

where thc hamiltonian II is a linear hermitian operator. It haa no explicit time dependance for a closed physical systeln, that is, aillat - 0, in which case its eigenvalues are the possible stationary statcs of the systcm. A superposition prineiplo follows from the linearity or II and a statement of conservation of probability from the hermitian property of II:

; 2 / "'*~(dql • • .) - k 2 / (dql • • ')[(II"')*~ - ",*(111/1»)

, ,

- 0 (1.3)

Vle strive to maintain these ramiliar six principlce 8S underpinnings of a rclativistic quantum theory.

Relotlvl.tdc quantum meclaorl'C".

2 Early Attempt.

The simplest physieal 8yBtoln i8 that of an Isolated free particle, for which the nonrelativistic hamiltonian is

pi 11 - --- 2m

•

(1.4)

The tran8ition to quantum meehalli08 i8 achieved with the tranecription

11-+ ilt :t It p-+-:-v ,

(1.5)

which leade to the nonrelativietic SchrOdinger equation "It a';(q,t) _ _ItIVI .,.( t)

"at 2m" q,

,

(1.6)

Equation8 (1.4) and (1.6) are noncovariant and therefore unsatiefactory. The left- and right-hand sides transform difTerently under Lorentz tran8fornlations. According to the theory of speelal relativity, the total energy E and momente (P.,P",P.) transform 88 components of a contravariant four-vector

'P" - (1",p',pI,p') - (~, 1'., P .. p. ) of invariant length

(1.7)

m i8 the rest m88S of the particle and c the velocity of light in vacuo. The covariant notation used throughout this book is dlscuesed in more detail in Appendix A. Here we only note that the operator tranICription (1.5) i8 Lorentz covariant, since it i8 a eorrespondenee between two contravariant four-vectors I p. -+ ih a/azlI-'

Following thi8 it i8 natural to take a8 the hamiltonian of a relativistic free particle

II - V pici + mlc· I We define ~ - (d,s) and V" - ~/~z ...

(1.8)

Dirac equation

and to write for a relativistic quantum analogue of (1.6) iii ~ - V - 'hic't" + m'04 ~

at

(1.9)

I mmediatcly we are laced with the problem of interpreting the squareroot operator on the right in Eq. (1.9). If we expand it, we obtain an equation containing all powers of the derivative operator and thereby a nonlocal theory. Such thoorice are very difficult to handle and present an unattractive version of the SchrOdinger equation in which the space and time coordinates appear in unsymmetrical form,

In the interest of mathelnaticalsimplicity (though perhaps with a lack of complete physical cogency) we remove the square-root operator in (1.9), writing

(1.10)

Equivalently, iterating (1.9) and using the faot thatl if [A,B) - 0, A~ - B", impliee A"It - Bty" we have

a'

.-~, Cfti '" - (-~'V'c' + m'cl)",

This is recognized as the cla88ical wave equation

a a o • 8z,. 8ztl •

Before looking rurther into (1.11), we note first that in squaring

the energy relation we have introduced an extraneous negative-energy root

•

where

(1.11)

In order to gain a slmple equation, we have MCrificod p08itivo definite energy and introduced the difficulty 01 "extra" negative-energy solutions. This difficulty is eventually surmounted (as we shall study in Chap. 5), and the negative-energy solutions prove capable or physical interpretation. In particular, they are 88sociated with antiparticfN. and the existence of antiparticl08 in nature lends strong: oxperimental support for this procedure. So let UB for a moment consider Eq. (1.10) and the inferred wave equation (1.11). Our first task ill to construct a conserved current, since (1.11) is a seeond-order

a Throulhout, we ule the notation- (A,B] _ AB - BA for commutator bracketa And [A,Bt _ AB + BA lor antioommutator bruketa.

5

R,,'ad,II.dc ql,ant"m me",han'c,

wave equation and is altered Irom the SchrOdingcr form (1.2) upon which the probability interpretation in the nonrelativistic theory is based. This we do ill analogy with the Schr6<linger equation, taking .p. times (1.11), t/! thnce the complex conjugate equation, and eubt.racting:

~. [0 + (TY]~- ~[D + (TrJ~· - 0

VI' (~·V,.t/! - ",V".p·) - 0

or

a [ ~ ("'. at/! _ '" a.p.)] + div !. [",.(-Vt/!) _ t/!'v",.») _ 0 (1.12)

at 2mcl at at 2tm

We would like to interpret (ill/2mc") (~. ~~ - ~ ~.) B8 a probability dellsity p. However, this ie impossible, since it is not a positive definite expression. For this reason we follow the pat.h or history· and temporarily discard Eq. (1.11) in the hope of finding an equation or first order in the time derivative which adrnlta a straightforward probability interpretation ae ill the SchrOdinger case. \\r e shall return to (1.11), however. Although we shall find a first-ordcr equation, it still proves impossible to retain a positive definite probability density ror a single particle while at the Mme time providing a physical in terpretatlon or the negatlve-energy root of ,1.10). ThereIore Fq, (1.11), also referred to fl'CQuent.ly B8 the Klein-Gordon equation, remains an equally strong candidnte for n relntiviatic quantum meehanlea B8 tho one which we now discu88.

S The lliroC! Equotion

\Ve follow the historic path taken ill 1028 by Dirac' in BOOking a relativistically covariant equation or the form (1.2) with positive definlte probability density. Sinoo such an equatlon ie linear in the time derivative, it is natural to attempt to form a hamiltonian linear in the space derivatives B8 well. Such an equation might assume a form

ill ~ - ~(at:' + at ~ + a.:') + /Jmcl~ • l/~ (1.13)

J E. 8chr6dinler, Ann. PA7/Iik, 81, 100 (1026); \V. Gordon, Z. PA7/,ik, 40, 117 (1026); O. Klein, Z. PhU,ik, 41, 407 (1927).

I P. A. M. Dirac, Proe. 1107/. Soc. (London), All?, 610 (1028); ibid., A1I8, 861 (1028); "Tbe PrineiplOi 01 Quantum MeohaniCII," Opt cil,

Dlrac equation

The coefficicnts a, here cannot aimply be numbers, aince the equation would not be invariant even under a spatial rotation. Also, ir we wish to proceed at this point within the Iramework stated in Sec. 1.1, the wave function", cannot be a simple scala.". III ract, the probability density p - "'.'" should be the time component or a conserved four-vector if ita integral over all epaee, at fixed t, is to bc an Invariaut,

To f.'oo (1.13) Irom these limitations, Dirac proposed that it be considered Q8 a matlix equation .. The wave Iunetiou ~, in analogy with the spin wave functions or nonreletlvlstle qusutum mechanics, ia written Q8 a column matrix with N eomponents

... " ..

"' .

"'-

•

"'N

and the constant eoeffielcnte ai, (J are N X N mau"ices. In effect then, Eq, (I.I:J) is replaced by N coupled first-order equationa

a", r. ~c ~ (a a a ) ~ \

iii 0; - T ~ aa ax1 + al ax1 + a. oz. "tV, + ~ (J.,mcl"" ,

,-I ,-I

N

- lll.r~r

~-I

(1.14)

Hcreafter we adopt matrix notation and drop summation indices, in which case Eq. (1.14) appcare as (l.la), to be now interpreted as a metrix equation.

If this equation is to acrve &8 a eat.isfactory point of departure firet, it must give the correct energy-momentum relation

ror a rree particle, second, it must allow a continuity equation and a probnbility lnterprotntlon for the wave function tV, and third, it.. must be Lorentz covariant. We now diseuse the first two or these requirements.

In ordCl" that the correct energy-momentum relation Cll1erge from Eq. (1.13), each component tV. of '" must eatitlry the Klein-Gordon second-order equation. or

(1.15)

T

Relo "" •• de quan I.u". n.ecllfln'c.

Iterating Eq. (1.13), we find

{J~ -hi -- - ai'

We may resurrect (1.15) if tho fODr Illatrices a~, ~ obey thc algebra: a",. + a.a, - 23,.

aJJ + ~a~ - 0 a: -.~, - J

(1.16)

What other properties do we require of these four matrices a~, ~, and can we expliciUy construct thcm? The a~ and ~ must be hermitian matrices in order that the hamiltonian H., in (1.14) be a hermitian operator as desired according to the postulates of Sec. 1.1. Since, by (1.16), a: - ~, - I, the eigenvalues of ai and ~ ere ± 1. Also, it follows from thcir anticommutation propcrtica t.hat the trace, that is, the sum of the diagonal clemente, of each ai and ~ is zero. For example,

a~- -~alJ and by the cyclic property of the trace TrAB - TrBA

one has

Tr a, - + Tr ~'a, - + Tr ~alJ - - Tr a, - 0

Since the trace i.8 just the sum of eigenvaIuCl, the number of positive and negative eigenvalues ± 1 must be equal, and the a, and ~ must therefore be even-dimenslonal matrices. The smallest even dimension, N - 2, is ruled out, since it can accommodate only the three mutually anUcommuting Pauli matric68 0', plus a unit matrix. The smallest dimension in which the a~ and ~ can be realized is N - 4, and that i.8 the cue we shall study. In a particular explicit representation the matriccs are

a. - [~. ~I] (J - [~ _~] (1.17)

where the O'~ are the familiar 2 X 2 Pauli matric08 and the unit entries in ~ stand for 2 X 2 unit matrices.

To construct the differential law of current conservation, we fint introduce the hermitian conjulatc wave functions t/lt - (~~ • • • t/I:>

DIrac equation

and left-multiply (1.13) by ~t:

Next we form thc hermitian conjugate of (1.13) and right-multiply by",:

(1.19)

where al - aC, ~: -~. Subtracting; (1.19) from (1.18), wc find

a 8 he a

iii at ",t~ - L 7 ax' (",ta"~) II-I

or

!. p + div j - 0 at

(1.20)

where we make thc identification of probability density

..

p - ~t~ - L ~:"'.

.-1

(1.21)

and of a probability current with three components j' - ell a"'P

(1.22)

Integrating (1.20) over all space and using Green'e theorem, we find

(1.23)

which encourages the tentative interpretation of p - ",t~ as a positive definite probabili ty density.

The notation (1.20) anticipates that the probability current j forms a vector if (1.22) is to be invariant under three-dimensional space rotations. We must actually show much more than thiB. The density and current in (1.20) must form a four-vector under Lorentz transformations in order to ensure the covariance of the continuity equation and of the probability interpretation. Also, the Dirac equation (1.13) must be shown to be Lorentz covariant before we may rea:ard it 88 satisfactory.

9

4 NonrelotlvlAtlc Corrc.pondenco

Before delving into the problem or establishing: Lorents iuverlance of tho Dirac theory, it is perhaps more urgent to see first that the equation makes sense physically.

'Ve Inay start shnply by considering a free electron and counting the number of solutions corresponding to an elcctron at rest. Equation (1.13) then reduces to

i"~ - 8~ at

slnee thc de Broglie wavclcngth is infinitely large and the wave function is uniform over all space. In the specific rcpresentation of Eq. (1,17) for 11, we can write down by inspection rour solutions:

thc first two or which correspond to positive energy, and the second two too negative energy. The extrancolls negative-energy solutions which result from thc quadratic Iorm or 112 - p2Cl + mlc' are a major difficulty, but one for which the reeolution lcads to an inlportant triumph in the form of antiparticlcs. 'Ve come to this point in Chap. 5, Here we confine ourselves to the "acceptable" positive-energ:y solutlone, In particular, we wish to show that they have a sensible nonrclativistic reduction to the two-component Paull spin theory. To this cud we introduce an interaction with an cxternal electromsgnotie field descrlbed by a four-potential

A": (cJ',A)

The coupling: is moat simply introduced by Incans or thc gaugeinvariant substitution

(1.25)

DIrac quadon

made in classical relativistic mechaniee to doseribe the interact.ion of a point charge e with an applied ficld. In thc present case

pIA -+ iA a/ az,. • p"

according to (1.5), and (1.25) takc8 thc Dirac equation (1.13) to

ill:t - (~a. (p - cA) + timet + ct» ~ (1.26)

Equation (1.26) expresse8 tile "minimal" interaction or a Dirac particle, eonsidered to be a point charge, with an applied electronlagnctic ficld. To emphaalae ita clauical parallel, we write in (1.26) H - 11 G + 11', with 11' - - eo · A +~. The rnatrix co appears here a8 the operator transcription or the vclocity operator in thc classical expres8ion ror the interaction energy or a point charge:

H'I I~_' - - ! v • A + ec)

a ••• _ C

Thi8 operator eorrespondenee Vop - Co i8 again evident in Eq. (1.22) for thc probability current. It also Iollows ir we make the relativi8tio extension or the Ehrcnfcet rclation8:-

d i

dl r - A [11,rJ - Co • Vop

and

die a

iTt (,,) - A [II,,,) - c at A

~ (tI) - n [ E + ~ VOD X B ]

(1.27)

with ox. p - (e/c)A the operator corresponding to the kinetio momentum and

1 aA

E - ----vc)

c Ot

the field strengths. Equation (1.27) is the operator equation of motion for a point charge e. More general couplings in (1.26) would lead to specific dipole and higher multipole terms in analogy with the clauical development,

III takillg the uonrelatlvietle lirnit of Eq, (1.26), it i8 convenient to work in the specific representation of Eq. (1.17) and to express th'

and

B - curl A

'Pauli, Schiff, and Dirac, op. cit.

11

wave function in terms of two-component column matrices IP and ~!

(1.28)

We then obtain for (1.26)

iA ! [:] - Cd • • [~] + .~ [:] + mel [ !~]

In the nonrelativi.8tic limit the reet energy mel is the largest energy in the problem and we write

[:] - ~-'I'I' [=]

(1.29)

where now fP and x are relatively slowly varying functions of time which are solutlons of the coupled equations

iA ~ [:] - Cd'. [;] + ~ [:] - 2mcl [~] (1.30)

The second of Eqs. (1.30) may be approximated, for kinetic energie8 and field interaction energies small in comparison with mel, to

dOfi

X --t' 2mc

(1.31)

Equation (1.31) reveale X 88 the 44small" components of the wave function ~ in comparison with the "large" components fP. Relative to t', x is reduced by ~ "Ie «1 in the nonrclativistie approximation. Inacrting (1.31) into the first of Fqs, (1.30), we obtain a two-component spinor eq uatioll

... ~ _ (d. fI d 0 fI + Jt..)

'n at 2m e'Y fP

This is rurthor reduced by the identity ror Pauli spin matrices d 0 ado b - a· b + id • • )( b

(1.32)

or, here,

dO" d • fI - "I + id • fI )( " eft,

- fli - - d. B c

(1.33)

Then we have

itt '!!e _ [(P - (e/e)A)1 _ eft, do B + e+]

at 2m 2me fP

(1.34)

Dirac equation

13

which is recognized I B8 the Pauli e(luation. Equation (1.34) gives us confidence that we are on the right track in accepting Eqs. (1.13) and (1.26) as a starting point in constructing a relativistic electron theory. The two component. of f/J sufllee to accommodete the two spin degroos of froodom of a spin one-half electron; and tho correct magnetic moment of the electron, corresponding to the gyromagnetic ratio g - 2, automatically emerges. To BOO this explicitly, we reduce (1.34) further, keeping only firet-order WI'1ll8 in the iuteraction with a weak uniform magnetic field B - curl A; A - J1B )( r:

• af/J [ pi e ]

,A - - - - - (L + 2S) • B f/J

at 2,,, 2mc

(1.35)

Here L - r )( p is the orbital angular momentum, S - )'lad is the electron spin, with eigenvaluCl ± 1./2, and the coefficient or the interaction of the spin with B field givee tho correct nlagnetic moment or the electron corresponding to a g value or 2. '_

Fortified by this succeasful nonrelatlvietie reduction or the Dirac equation, we go on and establish tho l~orentz covariance of the Dirac theory, as required by special relativity. Next we must investigate further physical consequences or this theory; especially we must interpret those "negative-energy" eolutlons,

Problem.

1. \Vrite the Maxwell equation. in nirac torm (l.13) In term_ ot alix-aomponen& field amplitude. "'hat. are the mat.ricH corl'Oll)Ondlnl to • and fJ? [See II. E. MOIO_, Pltl/'. llftt., 118. 1670 (lU59).)

2. Verity that t.he matricOi (1.17) eatiaty the alpbra ot (1.10).

3. Verity (1.33).

4. Verity (1.27). 'Ibid.

..

\

2

Lorentz

•

Covariance

of the Dirac Equation

1 Covariant Form of the J)lrac I··quatloll

It iI neceesary that the Dirac equation and the cont.inuit.y equation upon which its physical interpretation reats be covariant under Lorentz transformations. Let us first review what ill meant by a Lorentz transformation. I Two observers 0 and 0' who are in different inertial reference frames will dcscribe the same physical event with the different. space-time coordinates. The rule which rclates the coordinates ~ with whlch observer 0 describes the event to the coordinates (%'I)' used by observer 0' to describe the same event is given by the Lorent.z transformat.ion between the two sets of coordinates:

•

(%')' - I o',,;e- • o',,;e- 1'-0

(2.1)

It is a linear homogeneous transformation, and the coefficients 0'" depend only upon the relativo velocitles and spat.ial orient.ations of t.he two rcfcrence frames of 0 and 0'. The basie invariant of the Lorentz transformation is tho proper time interval

d,' - 0., d%'l dz' - dzI' dx,.

,

(2.2)

This is derived from thc physical observation t.hat the velocity of light in vacuo is the same in all Lorent.z framee. Equations (2.1) and (2.2) lead to the relat.ion on the transformation coefficients

(2.3)

Equations (2.1) and (2.3) serve as defining relations for both proper and improper Lorcntz transformations. In tho lorlner case the determinant of the transformat.ion coefficients satisfies the relation

det 101 ~+1

Proper J orents transformat.ions can be built up by an infinite succession of infinitesimal traneformetlone, They include transformations to coordinates in relative motion along any spat.ial direct.ion as well as ordinary three-dimensional rotations. The improper Lorentz transformations are the discrete transformations of space Inversion and of time inversion. They cannot be built up from a sueceeeion of infinitesimal ones. Their transformation coefficients eatisry the

I W. Pauli, "Theory 01 Belativity," Pergamon P ..... NI"1I York, 1068. uThe Principle 01 Belativity." collocted pal,el'l 01 U. A. Lorentl. A. Eimtein, fl. Minko,,·akJ. and H. Weyl. Dover Publicationa. Ine., Now York, Un3 rellsue.

16

I'. covarIance oj the DIrac eqlladon

17

relation

det lal - -1

in both cases.

Our task is to construct a correspondence relating a given set of observations of a Dirac particle made by observers 0 and 0' in their respective roferenee fraln68. In other words, we seek a transformation law relating the .wave funct.ions ~(x) and ~'(%') used by observers 0 and 0', rospectlvcly. This t.ransrormation law is a rule which allows 0' to compute ""C%') if given I/I(z). According to the requirement of Lorents covariance, this transformation law must lead to wave functions which are solutions of Dirac equations of the same form in the primed 88 well as unprimed reference frame. 'l'his form invariance of the Dirac equation eXpl'C8868 the Lorentz invariance of the underlying energy ... momentum connection

pjjpJ' - mlcl

upon which the considerations of Chap. 1 were hued.

In discussing covariance it is desirable to exprese tho Dirac equation in a Iour-dimensional notation which preserves the symmet.ry between ct and z'. To this end we rnult.iply (1.la) hy ~/c and intrad uee the notation

'YO - fj

i-I, 2, 3

This gives

. (0 0 0 0)

fh ')'1 - + 'YI - + 'YI -:-.. + 'Y' -- ~ - 'I1IC/I - 0

ox' OZI oz· OZI

(2.4)

'rhe new matrices 'Yjj provide an elegant rest.atement of the commutation relations (1.16)

(2.5)

where 1 is the 4 X 4 unit matrix and hea-caft.er will not be explicitly indicated. It is clear Irom their definition that the 'Y' are antihermitian, with ('Y')I - -1, and that 'YI is hormitisn, In the representat.ion (1.17) they have the fornl

[Oat]

'Y' - -tT' 0

I _ [1 0]

'Y 0-1

(2.6)

It is convenient to introduce the Fcynman dagger, or slaeh, notation:

A - 'YJli1. - O".-'YjjA· - 'Y'A 0 - "'( • A

Relol'lI',.dc QIIOrlt.Um nIfH!hon'c,

and in particular

-

Equation (2.4) then abbreviates to

(iA' - mc)~ - 0

..

(2.7)

· h ... iJ

or, Wit P" - 11' -8 ' x,.

(p - mc)~ - 0

(2.8)

Addition 01 the electromagnetic interaction according to the "minimal" subetitution (1.25) live8

(,,-4-me)ft-O

\ This in no way influences con8iderations 01 covariance, because both p" and All, and hence their difference, are lour-vectors. J

2 I'roor or Covarlanco

In order to eetablleh Lorents covariance 01 the Dirac equation, we must satisly two requlrements, The first is that there must be an explicit proscription which allows observer 0', given the "'(x) 01 observer 0, to compute tho ""(x') which describes to 0' the saIne physical etate, Second, according to the rclat,ivity principle, ""(x') will be a solution 01 an equation which takes the form 01 (2.7) in the primed system

"

(ift'Y" at;., - me) rJ/ (z') - 0

The of- satL~fy the anticommutation rclations (2.5); therelore .yot _ .yo and in - -i· a8 required ror a hermitian hamiltonian. A8 Inay be shown by a lengthy algebraic prool,1 all such 4 X 4 matricca .p are equivalent up to a unitary translormation· U:

o.

. ... ....

... .. II

. .

18eo R. II. Good. Jr .• RffJ. J.Vod. PAl/'.' 11, 187 {I 066) , eepeelally Sec. III, p, 100.

o,...ral.. co.,orionce oJ 'he Dirac eqllotlora

(t and 80 wo drop the distinction between oj" and ')'~ and write

•

with

(p' - mc)""(z') - 0

, ." a

p -, Y' 8z-'

(2.9)

\Ye ask that the transrormation between", and ~' be linear, since both the Dirac equation and tho Lorentz transformation (2.1) of tho coordinates are themselvoe linear. 'Ve introduce it in the form

•

"" (.%') - .p (ox) - 8(0)"'(z) - 8(0)~(a-lz') (2.10)

where 8(0) is a 4 X 4 matrix which operates upon the Iour-eomponent column vector ~(x). It depends upon the relative velocities and spatial orientations or 0 and 0'. 8 must have an inverse, 80 that ir 0 knows ""(x') which 0' UBe8 to describe his obaervations of a given physical state, he may construct his own wave function "'(x)

",(x) - 8-I(a)""(z') - 8-1(0)",'(cu) (2.11)

\Ye could equally well write, UBing (2.10),

"'(z) - S(O-I)",'(a.x)

which provides the identification

S(O-I) - 8-1(0)

Tile main problem is to find 8. It must satisfy (2.10) and (2.11).

If 8 exist!, observer 0', given 1/1(x) by 0, may construct 1/1'(x') using (2.10).

By recxpre88ing the Dirac equation (2.7) of 0 ill terms of ~' (x\) with the aid of (2.11), 0' could then check whether 1/1'(x') satisfies his own equatlon (2.9). ,,1ft "'ould find after left-multiplieation by 8(0)

[iflS(a)'Y"B-'(o) a~ - me] tI'(r) - 0

Using (2.1) to write

a ax" a a

az;. - 'a:rl' "iJiT"r - 0'. az"

•

the primed equation is found to be

[iflS(O)'Y"S-'(~)o." a! .. - me] ~(r) - 0

This is form.invariant, that la, identical with (2.9), provided an 8 can

19

R.'od., •• dc ql&oncum rrulcl.on'c,

be found which has the property 8(0).,..8-1(0)o'j6 - ")"

or equivalently ,..

(2.12)

Equation (2.12) is the fundamental relation determining 8. In seeking ,.'; we are BOOking a solution to (2J2). Once we show that (2.12) has a solution and find it, the covariance of the Dirac equation is eatablished, By way of terminology, a wave function transforming according to (2.10) and (2.12) is a four-component. Lorentz spinor. We anticipate that 8 will present novel features not. found in tensor calculus, since bilinear forms in", such as the probabilit.y current (1.20) are expected to form four .. vectors.

\Ve first construct 8 for an infinitcehnal proper J ... orentz trans-formation

with

(2.130) (2.1:lb)

. .

according to Eq, (2.;l) for an invariant proper time interval. Each of the six independent nonvanishing ~, gcneratoa an infiniteshnal Lorentz tl'ansformation,

~WOI - flfJ

for a tram~forrnation to a coordinate system moving with a velocity c ~13 along the x direction,

for a rotation through an angle ~",. about the. axis, and so forth. ~~xpanding S in powers of ~w'j6 and keeping only thc linear term in the infinitcshnal generatore, we write

and

.

8-1 - 1 + i flj6' ~-, (2.14)'

with

fI". - -fl,,,

.

by (2.13b). Each of t.he six cool~ients fI". is a 4 X 4 ry.trix. as are

the transformation 8 and the unit matrix 1. Inacrtin.a(2.IJ) and (2.14) into (2.12) and kccping first-order terms in Il.w"'. we find

R" couar'o'ice oj ,he lJ'rot' eq"o"on

From the antisymmetry of the generators ~CI>'" there follows 2f.W'.')'I - 0',')'.) - [')",er.,) ()

(2.15)

The problem of eetablishing proper Lorentz covariance of the Dirac equation is now reduced to that. of finding six Inatrices er., which satiBfy (2.15). The simpleet guees to make is an antisYlnlnetric product of two matricee, and directly we find, uBing (2.5). that

(2.16)

is the desired matrix. According to (2.14), S for an infinite8imal Lorentz transforlnation is given by

(2.17)

W c now complete our tMk by constructing the finite proper transfornlat.ions by a eucceeslon of infinitesimal ones. First. to build up (2.1) from (2.13). we write

(2.18)

where ~w iB thc infinitesimal paramcter. or "angle of rotation" about an axiB in the direction labeled '1&, and] PI is the 4 X 4 (in space-time) matrix of coefficicnts for a unit Lorentz rotation about. this axis.. P and IA label row and column rcspcctivcly. Thus for a transformat.ioll to a primed SystClll in motion along the ~ axis with an inflnitesimal velo('ity c ~w - C l1{j

0 -I 0 0
I'Jj - -I 0 '0 0 (2.IQ)
0 0 0 0
0 0 () 0
BO that
111 - ]10 _ -111 _ + 110 - -I Using the algebraic property of 1'", that

1 0 0 0
I' - 0 1 0 0 and 11 - +1 A
0 0 0 0
0 0 0 0
we can write thc finite transformation for uniform relative ~-axia 21

motion as

x"" - lim (0 + ~ I)· (0 + ~ 1)01 •

. N-- JV al Jl' a,

- (ee'I) • .a"

- (cosh wi + sinh w/)·.a"

- (1 - II + II cosh e + I sinh w)·.a"

For the individual components this gi'\'01

•

or

xOl cosh w - sinh w 0 0 Xl
x·' - sinh w cosh w 0 0 Xl
Xl' - 0 1 Zl
0 0
x" 0 0 0 1 Zl
xOl - (cosh w)(XO - tanh w x·)
x·' - (eosh w)(x· - tanh w XO)
Xl' - Xl
x" - x· (2.21)

(2.20)

where

1

I - )

tanh w - /J

and

cosh w -

relate the Lorentz rotation angle w with the relative velocity r/J.

ThiB reBult. can be generallaod to include motion along any direction or Bpatial rotation about any axis. The six lIlatriee8 I·,. generating the six independent. Lorentz rotat.ions are the Iour-dimeneional generalizations of the three-dimensional Bpace rotations familiar in the nonrelativiBtie theory.

Turning now to the construction ot a finite splnor transformation S, we have from (2.14) and (2.18)

,,'(:t') - S,,(x) - ~~r;. (I - i ; .,.~.1=· r "'(x)

- exp ( .; i - .. 10::) ,,(x) (2.22)

Specializing again to the transformation (2.19) we have

"" (x') - e-U/I)1I4.1"'(x) where x' and x are related by (2.21).

(2.23)

I. col'orlonc. oJ tI.a 1)'NlC a'la,otlon

23

Similarly, for a rotation through an angle 'IJ about the a axis, /11 __ III - -1 and

(2.24)

where

in the reprcaentation (1.17), with

u, - [~ _~]

the Pauli 2 X 2 matrix. 'Ve recognize the similarity of (2.24) with the form of rotation of a two-component Pauli spillor

(2.25)

The covariant "angle" variables w,., in (2.18) are aesoeiated with the Lorentz transformation in the same sense that the rotation angle and dircction in w arc for the three-dimensional rotation. Thc appearanoo of half-angles in (2.24), u in (2.25), is an expression of the doublevaluednese of the spinor law of rotation; it takee a rotation of 4r radians to rcturn ,p(x) to its original value. BecaUBC of this, physical cbservables in thc Dirac theory must be bilinear, or an even powcr in ,p(x).

For apatiall'otatioll8. S - S. is unitary, since the U'J are hermitian, and

B~ - etU/&,.tU.,., - ,-U/6,,, • .,,. - 8;,1 I

This is not true for transformations to a moving coordinate system

S - B£. For instance, for the transformation (2.23)

S£ - .-U/I) •• ,. - 8-(.,/1'-1 - 81 " 8l,1

However, S£ does have the property

B"il - ")'a8L")'a

found by expanding B£ in a power BCriCS. Since [")'o,u·JJ - 0, this can be gencralizcd to include rotations

B-1 - ")'oS''')'a (2.26)

The continuity equation is also covariant. The probability current (1.21) and (1.22), in the notation of (2.4), is

n» - c.;t(z)")'o..y"l"(z)

R.'oth".t.'c Qllont"". "t cl&on'c.

and undcr (2.1) transform8 to

3'" (01:') - Ct/t't (o1:')-yo~' (01:') - rYtt(z)st')'G-y-S~(:t)

- CI/It(:t)')'oS-l-y"S~(:t)

- c~,~t (x)-ya"Y'¢' (x)

- tJII,3'(01:) (2.27)

Evidcntly rex) is a Lorent.z four-vector and the continuity equation

~ .. ~ -0

fJzIA

is invariant. AI8o, the probability density J-O(o1:) - cp(x) tro,nsf'orma as the time. component of a conserved four-vector. Thie is the desired result noted in Sec. 1.3 for an invariant probability.

Beeauee the combination ¢'t-yo in (2.27) occurs 80 of ton , it i8 dignified by a new notation

(2.28)

where ~(z) i8 known a8 the adjoint spinor. It.B Lorentz tran8formation propcrty is given by

•

~'(01:') - ~(:t)S-l

(2.2Q)

3 Space lIefiection

""'

Vle now expand our outlook t.o take into account the cxistonce of

the improper Lorents tran8fOl'mation of space reflection

x' - -x t' - t

Again covariance requires a solution of (2.12), but in t,hi8 case we cannot, build it up from the infiniw8iaual transformation8. However, it. i8 easy enough to solve (2.12) directly. The transformation matrix i8

- 0"

(2.30)

o o -I o

o o o

-1

1 () o -1 o o

0'. -

o

o

Denot,ing S - P (or the coordinate reDoot.ion, (2.12) becomes p-1-y"P - 0",),'

(2.31)

which is sati8fied by

(2.32)

The phase factor is of 110 physical interest here and nlay be narrowed down to the four choices ± I, ± i if we require that four reflcctions return the apinor to itsclf in analogy with a rotation through 4", radians. P in (2.32) evidcntly is unitary, p-l - pi, and satisfies (2.26) as well, Equation (2.32) tells us that

fez') - wY( -x,t) - e~""'G"'(x,t)

(2.33)

In the nonrelativistic limit. '" approaches an eigenstate of P, and by (1.24) and (2.6) the positive- and negat.ive-cnergy states at. rest have oppositc eigenvalues, or intrinsic poritl~S.

The discussion of thc otber improper transformations, such as t.ime reversal, is more involved; it. is given in Chap. 5 .

• 4 Bilinear Covariant.

By forming products of the ")' Inatrices it is p088i hlc to construct 16 linearly independent 4 X 4 matrices r:, which appoar olten in applicat.ions or tho I>irac thcory. These are

rB _ 1 rV - "'"

Jj '.

r.p - i'Y°'Y1'Y''Y. - 'Y •• 'Yllo

rr _

,." flJj"

r! - 'Y1Io'Y,.

(2.:i4)

By using the antlcommutntlon relations (2.5) the r" are readily established to be linearly independent by the following argument:

I. For each r-, (rll)' - ± 1.

2. For each rll except rB, there exists a rllt such that

From this it. follows that. the trace or r- vaniehes:

± 1'r r- - Tr r"(r"')' - - ')'1- r"'r"r'" - - Tr I""(r"')' -_ 0 3. Givcn r- and r', a " b, there exists a r" " r" such t.hat r-r- - r"

This follows by direct inspection of the r'S. 4. Suppose t.here exist numbers 0" such that.

L0,.r" - 0 "

Then multiply by r"'" r" and take the trace; using (3), we find a. - O. If r- - rB, wc find 0, - 0, and all cocfficicntft vanish.

Thi8 eetablishee the linear independence or the rR. It follows that any 4 X 4 matrix can be written in terms of the r".

'Ve may now write down the Lorcntz transformation propcrties of the bilinear Iorms ~(x)r"~(x) constructed from the 16 r". We need only the observation that

'Y"'YI + 'YI'Y· - 0

(2.35)

and therefore

['Y6.CT ,.,J - 0

or

[8.'YIJ - 0

(2.36)

for all proper Lorentz transformations. AA a special case of (2.35)

P'YI - -'Y.P (2.37)

Carrying out calculation8 8imilar to (2.27) we find:

JI(x')~'(x') - ~(x)~(x)

a scalar

y/(x')'Ya~'(x') - ~(X)S-l'YaS~(X) - det 101~(x)'Y~(x)

a pseudoscalar

JI (x')'Y'VI (x') - 0' .~(x).",..p(x)

a vector

JI(X')'Ya'Y'VI(x') - det 1010' .~(X)'YD'Y»,f(X) a pseudoveetor

y,'(x')"-"'V(x') - ~AO'dt(X)gO~(x)

a seeon d-rank tonsor (2.38)

ProblemA

I. Verlfy (2.26).

2. Vorify tho transformation Jawe liven In (2.38).

S. 'Givon a free"plU'tlole IJpinor u(p), coru41.ruct u(p + ,) for ,,. ~ 0, "Uh po, ~ 0, in termlJ of u(p) by makhag a I~rontz tranllforrnation •

•• Show t.hat. there exlBt four 4 X 4 matricea r.luch that Re Pel - 0 Ir,.,r,) - 2g,.,

[ ir. a~ - m ] ~(z) - 0

that la, tho Dirac equation is real.

3

Solutions to the Dirac Equation for a Free Particle

lr{-- • I c; c· ·
• , • I • / 0/
.. f • #\ J r-' .., . . .. ,
\ . , ,. . .
• _. J ••
1 Plane-wave Solution. 'I ( ( . )
A We have soon that the Dirac theory meets the requirements of Lorentz covariance and that the p08itivc-encriI_JOlutions __ to the Dirac equation have a sensible nonrelativistie eorrespondenee, CJ

Furtllcr insigbt into the nature and interprCtation of eolutlons of the Dirac equation may be gained by eoneidering the free-particle equation. The four eolutions corresponding to a froo particle at rest were given in (1.24) and are written in the combined form

with

",r(z) _ W"(O)r".,IItCI'/AU C!r_!+1 -1

r - I, 2, 3, 4 r - 1,2

r - 3,4

(3.1)

Tile spinore are
1 0 0 0
WI(O) - 0 101(0) - 1 101(0) - 0 10'(0) - 0
() 0 1 0
0 0 0 1
(3.2) in this representation, Eq. (1.17), of the Dirac matricea. The fim two solutions describe the two ~p~n degreeB of freedom of a SehredingerPaul! electron. The "negative-energy" solutlons, r - 3 and 4, remain to be interprcted. They are all eigenfunctionB of tI. - tin with eigenvalues + 1 and -I. Tho Lorentz tramformation (2.10) may be used to build tile free-partiele solutions for an arbitrary velocity. By tranBforming to a coordinate sYBtenl moving with velocity - v relative to that of the eolutione at rest, we construct free-particle wave functions for an electron with the observed velocity +v.

In order to exhibit the gencral Bpace .. time coordinate variation, we need only exprCle the exponent in (3.1) in invariant form:

~ exp ( • u, ~ t) exp ( - it. Po ~'Z') - cxp ( - u, ~.') (3.3)

where :ttl - ~..x' and 1'jj - 0",1'·(0) - ~omc; our notation throughout iJ such that pO • E/ c - + Vp2 + mlcl > o. The positive- and neg&ti vc-cnergy solutiona tranBf'orm among themselves separately and do not mix with each other under proper Lorentz transformat.ionB, as well as under Bpatial inversiona ThiB iB Been to follow from (3.3), since the four-momentum of a free particle is time-like, P"'P,. - mlcl > O. Therefore, PI' is within the light cones in p space, U ndcr the trana-

C\ 28

don. to dlG DIMC equation/or a/Nfl particle

formations mentioned above, tho future and past light cones, and hence the positive- and negative-energy solutions, remain distinct.

'Ve tramd'onn the spinors with

(3.4)

according to (2.23), where for simplicity we have specified the velocity to lie along the x axis. The Lorentz angle '" in (3.4) is given by '" - tanh-I (-vIc) - - tanh-I (vIc) and differs by a minus sign from (2.21), since we are transforlning to a system moving in the x direction with velocity -v.

Applying the transformation (3.4) to the splnors (3.2), we find

W'(p) - rl""I)'''W'(O) - (COllb i - 1111 sinh ~) 111'(0)
1 '"
0 0 - tanh-
2
0 1 w 0
- tanh-
'" 2 W"(O)
- cosh-
2 '"
0 - tanh 0- 1 0
2
'"
- tanh - 0 0 1
2
(3.5) From the form (3.2) (or 1D'(O), it is clear that the rth column of this transfnrlnat.ion lnatrix is identically the column spinor corresponding to Uf(p). We nlay reexprees it in terms of tile energy and momentum of the particle by utring the trigonometric identities,

_ tanh ~ _ - tanh", _ vIe _ pc

2 I + Vi - tanh' '" 1 + VI - (vIJel) E + mct

and

h.. ~R + mc·

COB 2 - 2mcl

(3.6)

Also, we may generalize (3.5) to the case of arbitrary dlrectlon of the velocity v. In th~ case the lnatrix I ill (2.10) is replaced by

- cos fJ o

o o

o - COSer

- cos fJ

- COB"),

- cos")' o

o

o

- C08er o

o

o

where cos er, cos 8, and COB")' are the direction cosines of the velocity

29

v, and in the tra.llBformation iuatrix

a·v fI#jp/:p - 2(flol cos a + flOI COB ~ + flOI COB -y) - -2i lVf

This gives, with the aid of (3.6),
" "( u.v)
8 - cxp - 2lVf
1 0 p.c p~-
E + mel E + mel
0 1 p+c -paC
_ ~E+me. E + mel E + mel
(3.7)
P.c p~
2mel 1 0
E + mel E + mel
- p+c -p.c 0 1
E + mel E + mcl where P •• P. ± i~. The general form of a free-particle solution is

(3.8)

where the rth column of (3.7) gives the corresponding spinor wr(p) in the representation of the -y matrleee given by Eq. (1.17).

The wr(p) satisfy the following useful relations:

(, - 'rmc)wr(p) - 0 tDr(p)(, - .,me) - 0

tDr(p)W"·(p) - 3;',tr

..

r 'r1O~(p)tD;(p) - 3",

r-I

(3.00) (3.0b)

(3.0e)

Equation (3.00), obtained by applying the Dirac operator ("V - m) to (3.8), states the Dirac equation for a free particle in momentum space. For r - 1 or 2, 'r - + 1 and (, - mc)w"(p) - O. This is the equation for the two positive-energy solutions given by the firBt two columns of (3.7). In this representation their third and fourth eomponenta are the "small components" in a nonrelativistic approximation, and they reduce to Eqs. (1.20) and (1.31) in the absence of external fields. For the negative-energy solutions the "large" and "small" components are interchanged in (3.7). \Ve also introduce the adjoint spinor according to the definition in (2.28)·: tDr(p) .• to"t(p)-yOo It eatisfica the adjoint wave equation

(3.10)

Ion. to t.lao Dirac quadon/or a/roo particle

31

which i8 obtained by tAking the hermitian conjugate 01 (3.90) and multiplying from the right by ")'0 with the aid of the identities (")'0) I _ + 1 and ")'O-yJ't")'o - "),J'.

Equation (3.0b) i8 a covariant normalization statement, The bilinear form tD'(p)w"(p) i8 a Lorentz scalar a8 diseusaed in the preceding chapter [see Eq. (2.38) J, and 80 we veriry (3.9b) simply by returning to the rest solutions (3.2). The probability den8ity w,t(p)Uf(p) will not be an invariant but tran8formB as the fourth component of a vector according to (2.27). Calculating from the columns of (3.7) we find

(3.11)

Thi8 shows that the probability density acquires the correct factor E/mcl to compensate the Lorentz contraction or the volume clement along the direction of motion and to preserve thcreby the normalization of the invariant probability. Notice that (3.9b) i8anort.hogonality statement between a 8pinor and it8 adjoint of the same momentum P, whereaa ill (3.11) the positive-energy epinor i8 orthogonal to itB hermitian conjugate spinor of negative energy and reversed momentum. Thua two plane-wave solutione of the same 8patial momentum p but of opposite energy are orthogonal in the sense that ""t(z)~"(z) - 0 if r - I, 2 and r' - 3, 4, or vice versa.

Equation (3.ge) is a eompleteneee 8tatement applying to the four Dirac 8pinorB for a given momentum. It i8 clearly true for a free particle at rest, To prove it for an arbitrary momentum, we can make an appropriate Lorentz transformation to the rest 8ystem and then use (3.2) to find

4 4

L Mo~(p)1I)_(p) - L ~8 •• ( - ~) lO~(O)t»t(O)S;:J ( - ~)

,-1 r-l

- 8 • .,3., .. s~ - 3.,

That t» and not tot appeare in the completeness relation i8 due to the relation st - "),OS-I"),O derived in (2.26) and again reflect8 the fact that the Lorentz tran8rormation i8 not unitary.

By using the rotation operators

8 - e"'I)" ..

•

upon the solutione (3.2) for the electron at re8t and polarized in the • direetion, it i8 possible to form Btate8 which are polarized in any arbitrary direction I. In particulart tho defining relation for such

8tatca i8

d· BW - W

if the spinor 10 eorreeponds to a particle polarized along direction of the unit vector B. The specific form of these solutions is similar to that of the two-component Pauli theory owing to the 8tructure of d in (2.24).

In thi8 description it iB convenient to introduce a different notation. Let U(1',8) denote the apinor which is a positive-encrgy solution or the Dirac equation with momentum 1'" and 8pin~. Thu8 U(p,B) satiafieB the equation

(JI - mC)G,u,(1',B) - 0

(3.12)

The spin vector Ii' iB defined in terms oC the polarization vector I ill the re8t frame by Ii' - ~.It, where I" - (0,1) and the ~tI are the transfonnation coefficients to the reBt frame, that. iB, 1>" - ~,,'PtI, where j)~ - (m,O). Notice that 8' - -1 and that 'P"'Jj - 0 and therefore 1"'8,. - O. In the rost frame u sati8fies

d • lu('P,J) - u('P,I) Similarly let v(p,.) denote a negative .. energy solution

(, + mc)v(1',8) - 0 with polarization -I in the reat Irame, that iB,

d • iv('P,1) - -v('P,1)

'fhe U(1',B} and 11(1',8) are related to the lO'(p) by 101(p) - u(p,u,)

101(p) - u(1',-u.), "" toI(p) - v(1',-u.) 1O.(p) - v(1',u.)

with u: a four-vector, which in the re8t frame takeB the form u: - (O,u.) - (0,0,0,1)

An arbitrary spinor is thuB specified by the momentum 1'~, the sign of the energy, and the polariza.tion in the reat frame ' •.

(3.13)

(:1.14)

(3.15)

(3.16)

2 Projection Operator8 for Energy and Spin

In practical calculationB, it Is often convenient to have operator! which projcet out a spinor of given Bign of energy and polarization.

'on. to ".ft I)'roc quotlon for I. free p,.rl'l"le

3:1

These projection operators are the Iour-dimensional analogues of the nonrcla ti vistie t wo-cmn ponent operators

P 1 ± tI.

:I: - 2

which project out or an arbitrary state the spin-up or spin-down amplitude,

For the Dirac equation, we search ror four opcl"awrs which project from a given plane-wave solution of momentum p the four independent solutions corresponding to positive and negative energy and to spin up and spin down along a given direction. \\re would like these operators in a covariant form 80 that. we may transrorlll with case among difTerent I~orcntz syslenls, as will prove useful in practical calculations.

The four projection operatore are denoted by P,(p) • P(p_,u.,') and are defined to 8Il.tiBfy the following properties:

P,(p)W"'(p) - 8",w"(p)

or equivalently

(3.17)

P,(p)I~,,(p) - 3",P,(p)

An operatol' which projects out positive- or negative-energy eigenstates fol' a given p Inay he found directly Irom (3.9a), already in covariant form. \Ve denote it by

A () e" + mc

, p - --2mc

or, alternatively,

± +mc

AJ:(p) - -

2mc

(3.18)

By direct calculation, using" - pi - 1nlcl, we verify that

A ( )A () mtcl(l + e,e,..) + me,.(" + e,..) (I + e,e,,) A ( )

, p "p - 4mlc· - 2 ' P

that is,

A!(p) - A+(p) A+(p)A_(p) - 0

A+(p) + A_(p) - 1

To exhibit the analogous operator Ior the spin, we go to the rest frame, where the spin is most easily described, and try to find a projection operator which may be cast into covariant form, The natural candidaw ror a spin-up particle is (I + tI.)/2. In the same

Also notice that

n.'nrlv'."c quonl •• nl mochon'c.

way as tho two-component nonrclativistio 8pin projection oporator iIJ liberated from explicit dependence upon the a direction by rewriting (I + (1,,)/2 as a scalar,

we t,ry to write the Dirac spin projection operator in scalar form by using the four-vector tl:, that is

1 + (1. 1 + 'Y6'Y.tl:'YO 1 + 'Y6tl.'YO

---'!_..; - --

2 2 2

This may now be cast into covariant form by oliminating the 'Yo. Because wo are in tho rest frame, 'YG acting upon tho Dirac spinor becomes ± I. \\lith the conventions established in (3.14) and (3.15), tho covariant Dirac spin projeetion operator is finally

1:(u,) _ I + 'Y6'" 2

or for a general spin vector 8", with rpjj - 0, 1:(8) - I +2 'YD'

(3.10)

Thus in the rest framo

1:(tl.)w1(0) - 1 + 2'Y6tl. Wi (0) _ I ~ (1. Wi (0) _ WI(O) (3.20)

and I( _tl.)WI(O) - w'(O)

Slmllerly, for tho negative-encrgy spinors

I( - tl.)wl(O) - I - 2'Y6tl. w'(O) _ I + ~DlI.'YG w'(O)

_ I + (1. w'(O) _ w'(O) 2

(3.21)

and

In terms of the definitions (3.16) of the 8pinors tt and v, these are I(u.)u(p,u.) - u(p,tt.)

I(u,)v(PJu.) - v(p,u.)

E( - tt.)u(p,u.) - I( -u.)v(p,u.) - 0

Because of tho covariant form or tho projection operator E, wo may

II. don. to tl •• Dirac "uatlorl lor a IrIl. particle

write for any polarization vector "'("'1'~ - 0) that

1:(8)'U(1',8) - U(1',8)

1:(8)V(1',8) - V (1', 8) 1:( - 8) U(1', 8) - :t( - a)V(1',8) - 0

'Vith the four projection operators AJ:(1') and 1:( ± 8) we can now completely 8pooify free-particle motion in term8 of four-momentum 1'J" sign of energy ., and polarization ". with r1'. - o. In partioular, we con8truct from (3.18) and (3.19) the four projection operators

(3.22)

Pa(p) - A+(1'):t(u.) p.(p) - A+(1')1:( -u.) p.(p) - A_(1'):t( -u.) Peep) - A_(1')1:('U.)

Notice that [:t(a), A±(1')] - 0 for all vectors 8lLtisfying 8"p,. - 0, since, antieommutes with bot.h -Y. and ,. From thi8 it follows thnt \ these P r(P) sati8fy the defining relation8 (:i.17).

"Ve shell rely upon theBe projection operators very frequently in developing mpid and eflicient calculational techniques. They permit U8 to usc closure methods, thus avoiding the necessity of writing out matrieee and spinor solution8 component by oomponent.

In order to achieve an invariant formulntion, we have introduced negativo-cnergy solutions of momentum p which, according to (3.8), are eigenfunctions of the momentum operator p with eigenvalue - p. Similnrly, acoording to (3.10) and (:t21), the negative-energy solutioue representing spin-up and spin-down st.atce reduce in their re8t frames to eigenfunct.ions of tI. with eigenvalues -I and + I, respectively. The physical motivation for thi8 apparently backward associat.ion of eigenvalue! for the negative-energy solutions will appear when we come to the hole theory in Chap. 5.

3.3 ('hyaleallntcrpretut.lon or Free-I,nrtlele Solu liona und Puekcta

"Ve may now superpoee the plane-wave solution8 at. our di8posaJ to eon8truet localized packets. These packets are still solutions of the free Dirac equation, a8 requircd by the auperpoeition principle, since the Dirac equation is linear. \Ve study them to gain further insight into the interpretation of the Iree-partiele solutions.

35

To begin, we form a packet. by superposing positive-energy solutions only:

To normalize the expansion coefficients b(p,a) to unit. probability, we call on the epiuor ort.bogonaIU,y relations (3.11) and find I

/ ~(+)t(x,t)~(+)(Xtt) d'x - / d'p ";1 l b·(p,,')b(p,.)ll'(p,a')ll(p,a) :1: .. :1:"

- / dip llb(p,a) II - 1 (3.24)

:t-

The average current for such a packet iI!I given by the expectation val ue of the velocity operator

(3.26)

In evaluating t.I1is we use the following huportallt relat.ion between the three four-vectors that, can be formed from free-pertielo eolutlons:

For ~I(X) and "'I(X) any two solutions to the Dirac equation, (p - mc)",(x) - 0,

(3.26)

'1"'0 prove (3.26), we observe t.hat, if o· and bJj are two arbitmry fourvectors

(3.27)

I We collect bore familiar proportiol of tho Dirac I runotion used in dorivinl (8.24):

f• dt ,'(a-'. - 2.1(. - a)

-.

! dlJ 1(. - a)/(.) -lea)

(IDlerv. ]

InC!ludinc _ - •

if /(.) baa no linlularitiee ill tile Interval or intolration;

Tho I 'unction II mathematirally f'08l*'table in tllo lienee or diltributlon thtlOl')'; !HIe, ror inltanco, M. J. Lilhthill. Ulntroc!uctlon to Fourier AnalYlil and Generalizod Functionl," Cambridge t;nivcnity Pre .. , London. 1968.

10". to the Dirac equatlo"Jor oJ,... IlfJrtl,.Io

37

Then with the Dirae equation we construct

- -

o - Jtl( -P - m.c)~\ltl + ~I~(p - mC)"'l

- - - -

- - 2rnc~I~\ltl + ~I[tV'PJI - ia"p·u "., - P"O. + ip"o·u ".]\ltl

-

("' - p.n·)

and (3.26) emerges &8 the coefficient. of an arbitrary vector 0·.

This identity is known as the Gordon decompoeition.' J t. expre88C8 the Dime current as thc sum of a convection current. similar to the nonrelativistic one, and a spin current.

With the help of (3.26) for the special case "'I - "'1 - \It and (3.23), we now find for the current (3.25)

d' dl , 1

JC+) - f dlz f P p_ mc ~ b·(p' tJ')b(p tJ)e·(P'-JlI)"~ .. I'

.. (2d) I I!.'E ~ , ,

~ :I:.t :1:1'

X 2k iJ(p',B')[ (p~ + PI) + ial(p~ - p.) ]u(p,tJ)

\

- f dip p~1 ~ Jb(p,tJ)II (3.28)

:h

According to the normalization (3.24), the current can be written

JI+I - (clI'h. - (~) + - (v •• )~. (3.29)

where ( >+ denotes expectaUon valuc with respect to a positive-energy packet, Thus the average current, for an arbitl'ary packet, formed of positive-energy solutions is just the claasical group velocity. The corresponding statement is familiar in the nonrclativistic Sehrl)dinger

,

t,heory.

Now wc come to an important. difference in the relativistic theory.

In the SchrOdillger theory the velocity operator appearing in tho cUl'rent iIJ just p/m and is a constant of the motion for free psrtlclee. The current, is not, however, proportional to the momentum in the Dirae theory, and whereas the I~hrenfest relation (1.27) has shown that.

~ p - 0 for free-particle motion, the velocity operator Ccr is not constant, since [0,1/1 "0. Indeed in construct.ing eigenfunct.ions of ccr we have to include both positive- and ncgativtH)ncrgy solutions, since t,he eigenvalues of c(i are ± c whereas I (co')+ I < e, according to (3.20).

I W. Gordon, Z. PAI/lik. 10, 630 (1928).

Let us now enlarge our considerationa to include the negative&S well 88 poaitivo-cnergy soitlt.iona in forming a packet from the complete set of frce-part,icle solu tions, 'Ve generalize (3.23) to

~(x,t) - f ~R.Ji: }; [b(p,a)u(p,B)e- ...... /A

*.

+ d·(p,8)V(p,8)""·P~."'''J (3.30)

again normalized to unit probability. A short calculation givca for the probability

f d'x ",t(x,t)~(x,t) -. f alp ~ IIb(p,8)11 + Id(p,8)11] - 1

*.

and for the current for such a packet!

J' - f dip {}; IIb(p,a) [I + [d(p,a) [I' ~

*.

+ i ~ b*( - p,a')d*(p,8)eIUfPI/At1( - p,8')cr"V(p,8) :1:1.*"

- i }; b( -p,a·)d(p,.).,-I .. ...t..,(p, .. ),,··u( -p,a)} (:1.:11)

:I: •• :t"

In addition to the time-independent group velocity there now appear eroaa terms between the poaitive- and negative-energy solutions which oseillete rapidly in time with fl'Cfluellciea

2poC 2mcl

T > T - 2 X IOlt 800-1

Thia rapid o!Cillation, or ,itterbewegung, I is proportional to the amplitude of the negative-energy solutione in the packet. We have &a yet no phyaical interpretation of these solutions, but we may ask when to expect them to be present in the packet with appreciable amplitude. The general form of a free-particle solution (3.:iO) shows ~xplicitly by the time independence of b(p,8) that a packet initially formed with poaitive-energy solutions only does not develop negativeenergy components in the absence of forcee. However, a packet formed to represent an electron somehow localized initially in a region

I Delpite a certain ineollli.uDcy, ",·e de not. her(larter U(Vpl + ml,-p,.) .1I(-p,l) with similar cODV8nt'onl lor expanl'oD eoeffiaientl b, d·, eta.

IE. Bohr6dinpr, 8ibbfr. Prtua. A~. lVi.,. PhJl.ur-A/alA.~ 2'. 418 (1030).

WI •• to the IJ'roc ."uollo,. lur a IrfHJ partld.

39

of finite extent generally includes eolutione of both 8ign8 of energy. Consider, for example, the. solution

(3.32)

which correeponde to a Gau88ian density distributlon of half-width "-Id about the origin at time t - o. At a later time t it can be expreseed &8 a packet (3.30) with the coefficients band d* fixed by the initial oonditions, vis., at. t - 0

Taking the Fourier transform and using

we find

~1 b [b(p,B')U(p,,') + d*( -p,")v( -1',.'») - (t:.)" e-' ...... /A·wl(O) The orthogonality relation (3.11) gives

(3.33)

Thu8 the amplitude d· of the negative-energy solutions in the packet (3.32) iIJ nonzero. Relative to the positive-energy components b it i8 reduced by the mtio of the upper, or small, eomponents of f,I to the upper, or large, components of ", that is, by "-Ipc/ (B + mel). This ahow8 that the negative-energy amplitudes are appreciable for momenta "-Imc. 'Ve also eoe in (3.33), however, that the packet is eomposed predominantly of momenta p S It/d. Therefore, thi8 packet mU8t be localized in a region of 8pace comparable with the electron Compton wavclcngth, that is, with d "-lit/me, before the negativeenergy solutions enter appreciably.'

• For a di.ecullion of the polition eeerd i nate of a poaitivo-enorlY Dira(4)}eatron ... T. D. Nawton and E. P. WilDar, n.,. Alod. PAil'., 21, 400 (1049).

Ittig. a-I Potential barrior confininl oleotron of onergy E in rOlion I to tho loft.

This result can be equally well inferred on dlmenslonal groups using dp 6.r I"'OV A without reference to the particular gaussian shape. In discu88ing problems and interactions in which the electron is "spread out" over di8tancea large compared with its Conlpton wavelength, we nlay sirllply ignore the exlstenee of the uniuterproted negative-energy solutions and hope to obtain physically sensible and accurate results, This will not work, however, in 8ituations which find I electrons localized to di8t.ances eomparable with Almc. The negativefrequency amplitudes will then be appreciable, the zit.terbewegung terme in the current important, and indeed we shall find ourselves beeet by paradoxes and dilemmas which defy interpretation within the framework 80 far developed by the Dirac theory of an electron. A celebrated example or t.hC8e difficultie8 i8 the Klein persdox,' illustrated by the following example,

In order to localize eleetrons, we must. introduce strong external forces confining them to the desired region. Suppose, for example, we want. to confine a free electron of energy E to region 1 to the left. of the origin , - 0 in the one-dimensional potential diagram of It'ig. 3.1. I f the electron is not to be found more than a distance d to the right of , - 0, in region II, then V muet rise 8harply within an interval , S d to a height V 0 > E so that. the solution in II falls off with a characteristic width Sd. Thi8 is 88 in the SchrOdinger theory, until the confining length d shrinks to ~hlmc and V 0 - E lnereaeee beyond mel. To see what. happens, let us consider an electr08tatic potential with a sharp boundary as in ~'ig. 3.2 and calculate the reflected and t.ran8mitted current for an electron of wave number k incident from the left. with spin up along the, direction. The posltlve-energy solutions for the incident. and reflected waves in region I may be

10. Klein, Z. PAU.ik. N, 167 (1029).

,

'on. to the Dirac equation lor a IrtHJ particle

41

written

I o

'" la.. - oe"" ek Ih

E + mel o

(3.34)

o I

+ b'e-("" 0

ek,h

E + mel

For the tranemitted wave we need the eolutlons of the Dirac equstion in the presence of a constant external potential eel» - VOl Tbeee differ from the free-particle solutions only by the subst.itution po - (I/e)(]-] - Vo), 80 that in region II

~Ikllcl - (8 - Vo)! - mlet - (E - mel - Vo)(/l + mel - Vo) We therefore write the tranamitted wave of positive energy E > 0 88

I ()
0 1
~'r.DI - d~"" cAkl + d' ( .... 0 (3.35)
E - Vo + mel -ellkl
0 ]-] - Vo + mel The amplitudee d and d' are fixed by continuity of the solution at

V(.)

Vo ..... -------

\AA B

I n

--------------------~---------------- ..

Fig. !J.' 1~loC!trOltatio potential id('alizoo with a .harp bound. ary, witb an Incldont free olect-ron wal'O of onorgy E mOl'ina l4:I tho rip;ht in region I. For V. > 8 + Inc' the reflocl4:ld ourrent from tho potential OXeoedl tho incident one; thill ill a .. olal1l))I ... of tho Kloin paradox.

Relath.'.rlc qllantllna mechanic.

the potential boundary a8 required by current conservation: a+b-d

b kl E + mel d d

0- -r'IJ .,

It 1 ~ - V 0 + mel

b' - d' - 0 (there i8 no spin flip)

If Vo > 0 and IE - Vol < mel, the wave number i8 imaginary, kl - +ilkll, and the solution in region 11 i8 a decaying exponential eorreepondlng to damping in a dietanee d > ft/me. However, 88 we increase the height. of the barrier beyond Vo - E + mel in ordcr to furthcr confine the electron, thc transmitted wave beeomes oscillatory. The tran8mittcd and reflected currente may be computed, and we find

(3.36)

~rer _ (1 - ,)1 _ 1 _ ir.ran,

'I.. (1 + ,)1 ;1 ••

(3.37)

'VhereaB the form of theee resulta reminds U8 of the analogou8 predietlons of the Schrtklinger theory, we must now observe that, by (3.36) and the above condition Vo > E + mel, , < o. So we find in (3.37) a result cont.radicting our ordinary reasoning by indicating a negolwe t.ran8mit.ted current and a reflected current exceedi1l{J the incident one. \Vhat. i8 the source of a current in region II moving left in Ii"ig. 3.2 into region 1 in thi8 casc of Vo > E + mel? \Ve inereased the potential hcight. Vo beyond E + mel in attempting to localize the solution within one Compton wavclcngth ft/me, but ended up with undamped oscillatory eolutions Instead, Ilow do we underetand t.hi8? Only by understandlng and interpreting the negative-energy solutione, It is clear from tho packet discussion that they enter promincntly in solutlone localized within ft/me. It i8 equally clear from the above calculation of the currents that our physical pieture of what i8 going on also fail8 at these dlstancee.

\\r e shell taekle and resolve these question8 8tarting in Chap. 6.

Before doing thi81ct. u8100k in the vast, if limited, domain of phY8icai problems where the applied forces are weak and smoothly varying on a eeale whose energy unit i8 mel and whose distance unit is ftfmc. Here we may expeet to find fcrtile flelde for application of the Dirao equation and theory for positive-energy clcctrons.

ProblemA

1. Dorive (3.11) in a reprtMntatkm-(ree way directly (rom the Dirac equatioo.

don. to the Dirac eqr,ation lor a lree part.ie ..

2. Prove that (3.Oc) is independent of the lpooifio roproeontation of the Dirao Ipinorl.

3. nerive (3.31) lor the eurront in a lonoral packet (3.30). •. Verify (3.36) as the conditlonl for ourront coMervaUon.

6. Ji'ind the energy levola of a Dirac particle in a one-dlmenalonal box of depth V. and width o.

6. Verity tbe oompietoneu ralation

..

\' E

i.{ W.r(I"P )w""t (I'p) - m 6,.,

r-I

4

The

Foldy- Wouthuysen Transformation

1 Introduction

ABide from the negative-energy problem, the Dirae equation appears to provide a suitable description of the electron. I t has a sensibl nonrelntivistic limit, and it nutomntieelly yields the correct. magnetiG, moment. We now investigate the interaction or the Dirac electron with preeeribed external potentiaiB. In particular, we shall be pri. marily interested in low-energy properties, avoiding the diaieultil~s asaoeiated with the as yet uninterpretcd negative-energy sclutlons, which arc an e8scntially relativistic feature. ,\Ve anticipate from our diaeueaions of the packet in the preceding chapter that in practire they playa very minor role in a problem such as the hydrogen atom, which finds the electron localized in Bohr orbits of radius' 1/ am» I /m~

lVe shall see, in fact, that tho stationary energy levels deduced (rom the Dirae equation for the hydrogen atom arc in exceedingly elO8O agreement with the observed eigenvalues. However, before indicating the solution to the eigenvalue problem in the Coulomb potential, it is instructive to cast the Dirac theory in a form which displays the different interaction terms between the electron and an applied field in a nonrclativistie and easily interpretable form.

lVe consider, then a systernatic pro£..edurc d~~oE~ by.lc"oldy and Wouthuyscn,l nanleLv, a canonical tl'ansformation whWl.d,-:collpl the Dirac equatjon into two two-component equntions: one reduces to the Pauli description in the nonrclativistie limit; the othcr'(JcBCI'ibce the negative-energy states.

2 ."rcc-parU('le TranMforltlalioll

As a tint illustration of the Foldy .. \Vouthu)'8Cn transformation we consider the Dirae equation for a froo particle, most conveniently-for this purpose-writtcn in hamiltonian form and with the G' mntrieee in the representation introduced in Eq. (1.17). 'Ve search for a nnita transformation' U" which will remove from the equation all operator eueh 88 G' which couple the large to the small components. lVe call

II-Ienooforth we lOt A - e - 1. The Compton wavelength of the electron ill 11m - 8.86 X 10-11 om. and the rod. energy m - 0.611 MeV. The dimenllonl finCHtruoture oollltant iI a - .I/fr ~ 31.,-

1011 lOll

0.611 MeV - 8,86 om-I - f.29 MO-I - m

iD tbeeo unita.

I L. L. Foldy and 8. A. WoutbU)'lOn, PAI/" Btu., 78. 20 (1060).

47

any such operator "odd"; operatore which do not. couple Inrga and smell components nrc "even"; thus 0, 'Y, -"" etc., arc odd, and 1, 13, d, ete., are even.

'Vriting U" - e+" with S hermitian and not explicitly timedependent, the unitnry trnnBrormation is

and

1/1' - e+'B,p

i ~ - .-H8111/1 - e+lslle-fSI/I - 11'1/1' at

H' is to contain no odd operators by eonstruetlon.

Sincc II - 0 • p + 13m with I o,lJ I - 0, our problem i8 quite analogouB to that of attempting to find a unitary transformation which changes a two-eomponent spin hamiltoninn 3C - tI.n. + tI.n. into a form which eontaina only even opcratore (that i8, I and tI.). Such a transformation is 8imply a rotation about thc 1/ axle and the operator is e+Ul1).,._ - e+h., •••• , with tan 80 - B.I B». This suggeBtB

that a good operator to try in our case would be

wherc thc right-hand Bide is cstabliehed by expansion or the ~ential in powers or 8.

\Vith thi8 choice II' bceomes:

/I' - ( -lplll(P) + Pjp'IP Bin IpI6(P») (0 • p + pm) (COB Iplll

1J0· p . I I ) -,pl' BIn pO

- (0' p + pm) ( coelpl6 - Pjl Bin Iplll)'

- (0· p + 13m) exp (-2130 • pO)

- 0 • p(cos 21plO - ~ Bin 21p10) + lJ(m COB 21plO + Ipl Bin 21p10)

In order to eliminate the odd operator, we choose

tan 21plO - 1eJ m

Ite'allll"'tlc qr.an.lr. n. nI.chan'(~'

and the transformed hamiltonian is

II' - ~ vml + pi

(4.1)

as may be verified with the aid of the triangle construction of l·'ig. 4.1. The new hamiltonian is just the one rejected in Chap. 1, with the important change that now the negative energies are also accepted. The negative energies and four-component wave functions arc the price wc must pay in order to have a factorization or II' in (4.1) into a linear Dirac equation.

3 The General TranRforlnation

We turn now to the more general case or an electron in a prescribed external electromagnetic field and eearch for the corresponding transrormation S. The hamiltonian is

II - a· (p - tA) + {Jm + ~ -~m+f1+8

(4.2)

with e - o· (p - tA) and 8 - ~; as before, ~e - -e~ and

~& - +&.8

The fields appearing in (4.2) and hence the hamiltonian itBClf may be thue-dependent, In the general case the tranafonuation S is elso time-dependent and it is uot possible to construct an S which removes the odd operators Irom H' to all orders, 88 was achieved in (4.1). Therefore, we content ourselves with a nonrelativietie expansion or the transformed hamiltonian in a power series in 11m, keeping tcrms only through order (kinetic energylm)3 and (kinetic energy)(field cnergy)/ml•

Fig. 4 .. 1 Foldy-~'outhuYlen triangle coDitruction.

fioldy- Wor .. thuys8,1 frd,Ut!Ormat'O,1

Again we introduce the transformation by

finding

i :, rl8.f' - II{I - H '"", - ria (i ,!') + (i :, r'6 ) IV

Thu8

i a: - [6'8 ( H - i !) ria] IV - II'IV

Since S is expanded ill powers of lIm and is thererore "small" in the nonrelativistic limit, we expand the (IUantity in brackets in a series or multiple commutators, using the relationl

e+UJlle-'· - II + i[S,1I1 + ~>; [8,[8,11]1 + + (l)n IS (S fiT ' ,

• • •

... ,[S,II] • • .J + · . ·

Since S - O(1/m), to the desired order or accuracy we have II' - H + 11S,111 - ~ [8,[8,11)) - J [8,[8,[8,Il)))

+ ~4 [S,[S,[S,[S,P1nJ1J - S - ~ [S,S} + ~ [S,[S,Sll

To 8tart constructing S, \Ve consider jU8t the terms through order unity:

11' - Rm + S + 6 + 11S,,sJm

(4.3)

We require that the odd term in (4.3) vanleh; and taking our cue from the behavior in the Iree-particle case, we choose 8 - -i{Je/2m.

I Thil may bo verifiod by conaiderinR

•

l).n (BrtF)

F().) - ,().tlll,-i'l.1J _ _ --

nl ())." '1..0

"·0

(0)

It followl that

BF

- - ,..ui[8 111,,-,)..

II). ,

and thUi

B"F

~ - ,,'1. •• --[8,(8, • • • , [8,HI • • ·11'-""

from which the identity follow. upon letting). - 1 In (0).

We then find, to the order of aceuracy desired,

,18,IIJ - - e + 2~ [e,&J + ! ~el

m m

" ~el 1 1

I [S,[S,II]J - - 2m - 8ml [e,[e,&JJ - 2mi el

," el 1

al (S,(S,[S,II))) - 8ml - 8m' ~et

i' f3et

it [S,[S,[S,[S,II]1JJ - 24m;

~e -s- +- 2m

•

Collecting everything together,

( el et ) 1 i·

II' - ~ m + 2m - 8ml + s - 8nil r e,[e,&]) - 8ml [e,eJ

+ _!_ [e E) - el + ~ - R.wt + s' + f)' (4.4)

2m' 3mi 2m fJ"~

The odd terms now appear in (4.4) only in order 11m. To reduce them rurther, we apply a second Foldy-Wouthuysen transformation using the same proscription:

, -i{j -i~ ( ~ el i8fJ)

S - -- e' - -- ---leS) - - +-

2m 2m 2m' 3ml 2m

Under this transformation we find

/I" - e,a' (II' - i iJ) e-,a' _ ~m + &' + l_ le' &') + i~ft

iJl 2m' 2m

- {Jm + s' + e"

where 0" is now O(l/ml). Finally, by a third canonical transformation

S" _ -i{jo" - 2m

51

thiB term may also be eliminated in the same way, the end result being Hili - e'a" ( H" - i !j) Cia" - fJm + s'

( e' et ) 1 i·

- IJ m + 2m - Sml + & - Sm' [e,[e,&J] - Sm' [e,e]

Evaluating the operator prcduets to the desired order of accuracy, we find

e' (a • (p - eA»' (p - eA)' 6

- - 2 - 2 - -d·B

2m m m 2m

S~' ([ f),&) + ie) - S~' ( - ia • V~ - ia 0 A) - S:. a 0 E

[e u cr. E] - -~ [a • p a • E)

'Sm' Sma'

u ~ (.aB'\ 6 E

- 8ml L" Q,G; -f, -Ox') + 4m' d • X P

il

6. ie 6

- Smi dlv E + Sm' d • curl E + 4ml doE X p

and thus the reduced hamiltonian is to this order

H'" - IJ (m + (p - tA)' - pt) + 6ct - 6 • .!_ ~d 0 B

2m Sm' 2m

i6 6 6 •

- 8nii d • curl E - 4m' d • E X p - Sm' div E (4,5)

Tho individual terms in (4.6) have a direct physical interpretation. The tcrlllS in the first bracket give the expansion or

- --

V(p - tA)' + m'

to the desired order, showing the relativistie mass U!creue. The second and third torme are the clectrostatic and magnetic dipole energiC80 The next pair of terms which taken together are hcrmitian comprise the spin-orbit energy, and they have a vcry familiar Ionn in a spherically 8ymmetrle static potential. I n this case curl E - 0,

1 oV 1 OV

d·Exp- -, iJrdorxp- -, OrdoL

and this term reduees to

(4.6)

Rolat.'lI'.de quur.t.u, .. lIIoclrnr.'c.

Equation (4.6) is in agreement with the classical rcBult obtained by considering the magnetic field B' - - v X E experienced by the mov .. ing: electron. The interaction energy one would expect is thus

- __!_ d • B' - .~ d • (p X E)

2m 2m'

However, this is reduced by a factor or 2 owing to the 'I'hornas preceesion effect and indicates that the orbital moment of the electron has the standard gyromagnetic ratio or g. - 1.

The last tcnn-known as the Danvin tcrm-rnay be attributed to the zitterbewegullg. Because the electron coordinate fiuctuate8.over distaneea ar ~ 11m, it 8008 a somewhat smeared out Coulomb potential; the correction is

< oV 1 ~ o'V)

(aV) - (V(r + ar» - (V(r» - ar lJr + 2 ~ ar~ 3r, or~ iJrl

4J

Ot! art V.V Oc' ~ VIV (4.7)

-6 Om'

in qualitative accord with the sign, form, and magnitude of the Darwin term.

4 Tilc Hydrogen Alolnl

I

\Ve turn to a discussion of the bound-state sol ut ions of the Dirac equation, considering in particular the encrgy levels of the electron in a COUIOlllb field. For this problem tile Dirac equation is

II", - fa· p + ~m + V(r»)'" - E",

(4.8)

with V - - Zalr. In order to separate variables, we take advantage of the fact that the angular momentum or a particle in a central field is conserved. Evidently J - L + S - r X p + ~d commutes with the hamiltonian (4.8) and thererore we rnay construct simultaneous eigenfunctions of II, J', and J.. To do this, we call on experience with the Pauli matrices, observing that in the representation of

I The eigeMOlutloll1 ill the Coulonlb potential were fint liven by C. Q.

Darwin, Proc. ROll. Soc. (London), Alii, 8M (1028). and W. Gordon, Z. PIt",it, 68, 11 (1028). !.o_r..! O:Qmplcte ~i,cUMiun and rcferoncee 01 the atomic applicatione of tta~ Dirac ~t~n lee U. A. Betho and E. f;. Sah,etor, "Quantum Mechanica

-or-on. anaTWo-eTeetrODXtoml," Aioadomie Prell Ine., New York, 10&7, and M. E. Ro., "Relativistio Electron Tbaory,'· John "'i1ey a: SoIllJ, Inc., New York,

1961. - ----

Foldy- Wouf'IUY •• " fran'4/orrrlodon

Cbap.3

d - [g ~]

is diagonal in terms of 2 X 2 Pauli spin matrices. Therefore, if we eonetruct t/t in terms of two-component spinors

the angular separation for the solutions of f/J and x is precisely tbat 01 the l'auli two-component theory. The two-component angular 8OIutions arc eigenfunctions 01 J', J., L', and S' and aloe 01 two typca:

Forj-l+~

(4.ga)

Fori - 1- J1

~'+ ~ - m y-Jt

21 + 1 I

_ ~,+ ~ +m y ......

21 + 1 I

(4.gb)

The spherical harmonics here are written with the convention

,

y~. - (_)111 Y 1.-111, and the solution "C-) exists only for I > O. The

two solutions above satisfy the eigenvalue equations

J'"c:.' - iU + 1)"/111

and

L • df/J}:l - CJ' - I~' - ~),,}!' - - (I + ")f/J:~

with

{-(I + 1) - - U + ~)

Ie-

+1 - +(j + J1)

i -1+ ~ i -1- ~

For a given j they are of opposite parity, since their 1 values differ by 1, and can be formed from each other by a scalar operator of cdd parity. This operator will be a linear combination 01 YiC6,,,) since it must change the I value by 1, and i8 therefore proportional to r. Dotting with d, the only pseudovector at our disposal, we form th ....

pseudoscalar d • r/r and find with the above sign convention d·r

cpC+1 __ cpl_'

1M r I-

(4.10)

The general solution to the eentral field problem for a given jm is iGj (+' + ."OJ c-,

r CPI- r "1".

"'1M -

We may finally break this down into two solutions each of definite parity .. Since V(,) is invariant under reflection of coordinates, we know that the energy eigenfunctions can be classified into parity eigenstates along with (j,m) j and therefore we form the even and odd solutions, which have the property under the transformation z' - -x

""(x') - 8';(z) - ± ';(x')

(4.11)

These arc given by

.,.1 _ "I".

(4.12)

where as a common notation we have introduced

to: j _.,+ M { r: j-l+3i
G'I - j -l- ~ FIJ - F~ j-l- ~
OJ
{ + j -1+ ~
, "s-
"1'" - - j -l- ~
"s- and have made use of (4.10). The parity of these solutions is (_)1 by the convention (4.11). 'Vith the aid of the following identities we can now find the radial equations following from (4.8):

d • p&1 "i' - ~ (d • rd· p) Ie,) "L

r .. r2 r pow

- d • r (~ r !. + i d • L) @ ~s-

,2 • Br r

_ [!.!®. _ i(l + «)&1] (d. r) CPI'

• dr r ,1 r ".

Fo.dy· WOll thuYHn tranlflornlodon

55

The radial equations are then

( Za) dFIJ(r)"

E - m + r G,I(r) - - dr + ,FIJ(r)

(E + m + ~)FIJ(r) - + dOt) + ~GIJ(r)

The bound-state solutions of these equations may be found by standard metheda;' we quote only some of the results.

The energy elgenvaluee are

E. - m [ 1 + (" _ (j + ~) + ~(j + )ill _ Zia.rrlt (U4)

(4.13)

where the quantum number n - 1, 2, ... , co is a positive integer and the angular-momentum eigenvalues range from 0 to j + ~ :S ft, with the restriction 0 SIS ft - I. Expanding (4.14) in powers of (Za)', we BOO that 11· eorreeponds to the principal quantum number of the nonrela t.ivistic theory

E .. m {I - ! Zlal [1 + (Za)~ (. 1 _ 1.)] + O({Za)') )

.. 2 nt n J + J.i 4"

(4.15)

The ground-etate energy is, with n - 1, j - ~,

E, - m VI - Z'al S! m - ~Zlalm - UZ·a·m + · · ·

The corresponding spin-up and spin-down normalized eigenfunctions are

iJ',,-l.J-M. T (r,8,c,o)

1 o

t(1 - .,,) 8

Za COl

t(1 - .,,) . 8.

-----.;;...;;.. sin "

Za

iJ' .. -I.I-M.' (r,8,tp)

o 1 .(1 - "')

_.__-~, Bin ere"

Za

- .(1 - .,,~ cos 8 Za

(2mZa)1t ~ 1 + or (2 Z )-1 "'s ...

--- - marwr

Vii srti +~)

I DarwlD, OOrdOD. Beth. and Salp.ter, and Ito .. , 0,. aC.

with 'Y - vi - Z'al• In the nonrclativistic limit 'Y _. 1 and (1 - 'Y)/Za _. 0, and thoy reduce to the SchrOdinger wave functions multiplied by two-component Pauli epinors, In the relativistic cue we Bee that as r _. 0, '" exhibits a mild singularity of order (2mZar)-C.r·atH2 which becomes important only at distanoos

For Za ~ I, 'Y is imaginary and the solutions exhibit an oacillatory 'Jehavior reminiscent of that found in tho Klein paradox, In this case there is no longer a gap between the positive- and negative-energy spectra, and again we lack a physical interpretation of the solution.

In classifying the energy levels (4.14) it is customary to denote them by their nonrelativistic labels, that is, by the orbital angular momentum l appearing in .;1", and by the total j. In the following table we lilt a few of the first terms:

n I j Eri/
m VI - ZIO'
is .. 1 0 H
28 .. 2 0 }' ~l + VI - ZOo'
m . 2
2Pw 2 I }' Jl + Vl-Z ... •
m - 2
~ V4 - Z.",
2/'" 2 1 "
2 The 2S~ and 21> .. states al"O degenerate, being the two eigenstates of opposite parity corresponding to the same nand j. Tho 2PM state is higher in energy than the 2P .. state; the energy diffcrence, [m(Za)·/32](1 + OCZa)' + · · .), iI the fine-structul"O splitting due to the spin-orbit interaction, (4.6), In general, the state of larger i. for a given n, lies higher in energy according to (4.15).

How do these predictions agree with observations for the II atonl?

Prior to 1047 the Bgl"Oement WBI completely Mtisfactory after tho above predictions were modified to take into account the hyperfine splittin~ of each level due to coupling between the electron and proton

57

B

381/2 =-=====

===3DIIZ

3P1/1===== __ ~t lIZ

3plll r 3D

L NPrty IquII (spilt by Lamb Ihift)

28Vf===

2PJII==::;

zplJI h r- f1nul,uc:tu,. (lpIn ·orbft coupl1na)

'Lamb Ihift

lS1/1 ==*L (triplet) ---,_

[_ (li"8let) __s- H~perflne spllttinl

Fi,. 4-1 Low-lying nellD' leve" oC atomic l1ydl'OlOll. 'I'he dIagram fa not drawn to Bcale.

8pin& In 1047 the Iamb-Retherford meaeurernente' of the l-I-atom fine structure confirmed an ea.·lier sU8picion of a 8hift of the 28" levele upward relative to the 2P" lines, Thie "Lamb ehif't," breaking the degeneracy of levels with the same nand j but differing I, arises from the interaction of the electrons with the fluctuations of the quantized radiation field. Both the hyperfine 8tructure 8plitting and the Lamb 8hilt have been measured and calculated to a very high precision with good agreement.:II

The hyporfine 8tructure l'68ul t8 Irom the intemction of the proton wit-h the electron magnetic moment.! This hae the effect of 8plitting all linea into doublets corresponding to the two poseible state8 01 total angular momentum compounded from the j of the electron

• W. E. Lamb, Jr., and n. C. UctherCurd, Plal/" Rev., '11. 241 (1987). For reforcnc81 to lJubtoqucnt work ICIO 13ctilo and 8alpotor, op. cit.; 100 allO 'V. E. Lamb, Jr .• R pll. ProUt. 1~1a1/'" 14 •• 19 (1061).

• For a reviow oC the Durrent Iltllntion, 100 R. P. Foynman. Proe. 1061 8cf&!tJl/ Con/., Intencionco. Now York. 1062.

IE. Fenni, Z. PAUlik, 80. 320 (1930); eo. al80 Bethe and BaII)Qter, 01'. ril., p. 183.

R Intlv,.tle q.,nntum m.c'wn'el

system and the half-integer spin of the proton. Let liS compute the magnitude of this effect for 8 states. For our purpose a nonrolativistie description of the electron suffices. The interaction is of the form

111-+~d.B

and B - ::. J d"r' p(r/)" X (l X ,,) 41rlr 1_ ?I

lIere 1 is the proton spin operator (I. - ±~) and per') is the mag .. netic moment density of the proton, owing to the fact it is not a point particle. Using the relations V )( (I )( V) - IV' - (I • ")l' and taking the angular average for spherically symmetric wave funetiolls 10 that

we find

B - ; u. fir. J d"r' p(r')I V' (41rlr ~ rll) - ~ u. 2M. I per)

The energy shift is then given, in nonrolativistic theory, by

2 0" J

AI: .. - (.p"H'.pft) - ! 4r:M, d • 1 dlr .p:(r)p(r).pft(r)

1 OpB2

... - - - d • 11.p .. (0)12 6mM,

with

I ( +~

d· -

-~

triplet states singlet state

The splitting 3" of the nth Htate level is thus

3. - ~ ma· n u. Z~~. (;;) ]

and is reduced by the masa ratio m/ Itt, relative to the fine structure.

Weltonl has given a simple qualitative description of the Lamb shift by considering the interaction of an electron, treated nonrelativistically, with the vacuum fluctuations of the electromagnetic field. Since the dynamics of a normal mode of the electromagnetic field is equivalent to that of a harmonic oscillator, each mode upon quantiz~ tion acquires a zero-point energy of w/2. As a reeult of this quantum

I T. A. Wiltoni Plav'. RIIV., 7'. 1167 (1948).

Foldy· WO""IIIY.ffn ,ranlflornaatlnn

S9

effect thero aro now fluctuating olectromagnetic fields even when no oxtornal fiolds are applied. Although the a vorage field 8trcngths arc zoro, their mean-square values are nonvanishing, and this leads to a mean-square fluctuation in tho electron's position eocrdinate due to its coupling with tho field. It is the amplitude of this jiggling of a bound oloctron in the hydrogon atom that we ostimato. It Implies, as we saw in our discussion of the origin of the Darwin term, (4.7), an additional interaction energy }.-6'«31')I}vtY from the emearing out of tho Coulomb potential VCr) seen by the olcctron. To lowest ordor, tho chango in tho energy level for tho electron due to this is then

~E-.(Lanlb) - }.-6'«3r)t)1 ~:Vtl'(r)~1t d.,.

- ~ Za«3r)2)I~It(O) It (4.16)

To estimate «3rp), wo treat tho electron claaically and nonrelativistically D.8 a charged particlo. Its equation of motion for

oseillatlon about ita oquilibrium coordinate in the atom is 31' - ~ E, m

whore E is the fluctuating electromagnetic field. For the wth Fourior amplitude we have

.I eE" ur., - - =::J t7k.J

and hence for ita mean-equaro amplitude

«3r. )t) _ ~E.,)I) ., mt;r

and «31')t) - :;;; f ~ «E.)2)

m w

(4.17)

To calculate the mean-square fiold strength, we eonsider the total vacuum field enorgy

1 I 1

2 f d1x (E2 + Bt) - L "2 W l-I i"

whoro the two values of). rofor to the two states of transverse polarization and the sum extends over all rnodes in a large box of volume

LI - f d1x

Since f d1z E2 - J d1x 82 and w - Ikl for free eleetromagneae waves,

Il"'atl.,,.tlc q.,ont~ m It.ecl.on'c

tho moan-square field 8trength in vacuo is

Inserting in (4.17), we find

«3rl» - et f ~

2""lml ""

(4.18)

whore tho frequency integral extends from 0 to GO. Because of the crudity of our approximate treatment of tho electron, tho integral diverges at both ends. This is not tho case for an aceurate relativistic troatlnont of tho olectron localized in a hydrogen atom. Wavelcngths larger than tho Bohr radius ",,(Zam)-l will not be effective, sinco thoro must be a minimum frequency for tho induced oscillation8 correspond .. ing to this typical atomic size; therefore, ""min"" m.za. There is also a high-frequency cutoff at distancc8 I"'tIo.1 the electron Compton wavelongth l/m coming from the relativistic structure of tho eloctron. This etruetuee corresponding to tho zittorbowcgung amplitudo suggests that frcqucncies highor than ""ma. I"'tIo.1 m will not be cffective in jiggling the electron. Hence we approximato J d""/",, I"'tIo.1 In (I/Za) and find for the mean-square amplitude of the oscillatione in the vacuum fiold, by (4.18),

{(fr)') - (~IO lor) (~)'

The resulting energy shift is by (4.16)

ilK. - 4~a' (~y III it. 1",.(0)1'

- [~ ~:~~ (III it.) ] (Maim) 31a

~ 1,000 mC/BeC for n - 2, Z - I, l - 0

This accounts for most of the m eaeured 8hift of tho 2SJt level in tho hydrogon atom; for tho p and highcr l states tho 8hifts are not precisely zero but are much 8maller because the wave functions at tho origin are zoro. By way of comparison with the ordinary tine structure wo ICe by looking back at the hamiltonian (4.5) that the ratio of the Lamb torm to the Darwin term is (Sa/3r)[In (I/Za)] corresponding to tho ratio of the mean-equsre fluctuation amplltude (4.10) to the sitterbowegung structure ~(I/",)t.

(4.10)

61

I. Derive (4.10).

2. The Dirac equation dllIcribinl the interaction of A proton or neutron "'ith An Applied electromagnetic field "'ill havc An AdditionAl magnetic moment intoraction reproeenting their observed AnomaloUl!llnngnetie momenta:

where

repreeente the fiold I!ItrOOlti18 I\I!I defined in Appendix I.

4. For the proton, i - p, '. - 1,1; for the neutron i - n, ,,- o. Verify thAt the clio ice of ICJI - 1.70 and If. - -1.01 eorrllIpondl!l to the obl!lerved magnetic momenta And chock that the AdditionAl interr.ction dOlllllVt. disturb tile IJOroratz covAriAnett of the equntion. Chock aiM thAt the llil'M! hAnliltonian l.a hem,itiall And that. Ilrobnbility il conlOrvm ill t.he l'fOIcnee of the a.dditionAI interaction.

b. Make A Jr·oldy .. \VouUmYICh tranlformAtion for the neutron, keeping terml!l to Ule aecuracy of (4.6), And Rive A pbYlical interpretatlon of tho individual torml. CalculAte tile eroae eeetlen for tile leatteriul of A elow neutron by Nt applied electroatAtic field. How mi&ht thil be meaaured? ISoe 1.1. 1.1. Foldy, RffJ. Mod. PA" •• , 80, 471 (106S).1

3. Solve for the eXlU't enerlY eigenvnlues and eiRenfunctionl of A DirlUl electron in a lIui(onn l!ltatic 1l1nJCftetie flekl. [800 I.. D. lluff, PAIl"~ Ii",., 8a, 601 (1931); M. II. JobnllOll And B. A. l.iPIJmnra, Ph",. ItffJ., 11, 702 (IO~).I

4. CaleulAto to lowclt order in 0'1 the flnt-ordcr 1cemall cffect for An electron in a hydrolen Atom. If the electron KyromAKnctic rAtlo difTen from Q - 2, hew Are the ZoemNtleveJI Altered (to firlt ordor in the difference g - 2)?

6. Diecu .. the preceeaion of the Ipin of A eiaal'lOO Dirac particle witl, an AnomAlou .. malnetie moment" in An Applied Itatic mfllnetie field. Show in pArticulAr til at tbe difference in the IJpin And orbital prect'll!lion frcquoneilll il!l proportional to Q - 2, or If. How dOIll it depend upon the mau of the pArticle? Boo:

II. A. Tolboek And S. R. de Groot, Phr';ca, 17, 17 (UlIH). K. M. Cuo, I'A",. liffJ., 108, 1731.1 (1987).

H. Mendlowiu And I\:. M. Cuo, Ph",. Ii",., 97, 33 (19M). M. Carralll, N'UO'Io Cirn.rno, 7, 624 (1968).

V. BAfIIIlAnn, 1.1. Miellel, And V. 1.1. ToIegdi, Pit" •. RffJ. lAU"., I, 486 (1969). Loulloll, Pidd, And CrAne, PAIl"~ Rev., 8"', 7 (1964).

Sohupp, Pidd, And CrNte, Pit",. RffJ., Ill, I (1961).

Charpu, FArley. aaf\\in, Muller, Senl, Telegdi, aDd Zichichi, PA" •. Rrv. Leur,." 8, 128 (1901.)

1("'atlll,,,tI,. qua" rum mechanic.

G. Conatruct an additional interaction term to roprolOnt a poIBibloanomaloua 01000! tric dipolo moment of a Dirac particlo. \Vhat !lapp.a to tho I,arity transformation 1 \\'!1at i. tho oRoct of sueh a term on tho bydrolcn-atom energy lovela, (Soo O. Feini>el'l. PAu •• nffJ .• 111, 1637 (1068); E. E. &lpoter. PiaU" R 'J Ill, 1642 (1960).1

7. Owinl to mMOn eRects (di.eeuued in Chap. 10). tho proton rhal'lo i. distributed. ovor a emlll ~gion 01 apatial extont --10-11 cm. Compute t,ho elTect on the Ilydrolcn.atom onerlY lovela of IUch a chaJ'KO cliatribuUon with mean aquare radiua r ;W 0.8 X 10-11 cm. Com)>A11I tho reault with tho Lamb abift •

...

5

Hole Theory

1 The Probleen of Negutive-enerllY Solution.

'rhc negative-energy solutions of the Dirac equation have been touched on in some of our earlier dieeuaaious, and their presence in tho construction of a localized packet, for example, has boon computed, However, we have so far managed to avoid coming to grips with tho problems of interpreting them and of understanding their hnplica .. tions.1 Let us now faco up to these questions.

By their very existence the negative-energy solutions require & massive reinterprct.ation of the Dirae theory in order to prevent atomio electrons from making radiativc tran8itions into ncgativo-onergy states and cascading down to oblivion. 'fhis is no problem if we completely neglect interaction of the electrons with the radiation field. We lnay then calculate stationary solut.ions 8.8 in the preccding chaptf .. and find energy eigenvalues and tranMition amplitudes whieh agree in general very well with experiment. However, the problem of keeping thc electron from tumbling into a negative-energy state exists in principle. as well as in practice. if we wish to ealeulete atomic properties to such great accuracy as requires inelusiou of the radiat.ion interaction. The tmneition rate for an electron in the ground state of a hydrogen atom to fnll into a negative-energy state may be calculated by applying scnlicla88ical radiation theory and using the wave functions found in Chap. 4. The rate for the electron to make a traneitlon into the energy interval - me:' to - 2mcl is

201 mea

,...., -;- T ~ 1011OC-1

and it blows up if all the negat.ive-cnergy stat.es are included. This i clearly nonsense r

We must find some troatmcnt of the negative .. energy states other than t~at suggested by the one-particle Sehrodinger theory if the Dirac equation is to survive, Dirac did this for us in I {l30. He Iormulated the "hole theory, U which reeolves the dilomma posed by the negativeenergy solutions silnply by filling up the negative-energy levels with electrons. in accord with the Pauli exclusion principle. The vacuum state iI then one with all negative-energy electron levels filled and all positive-energy levels empty, Thc stability of the hydrogen-atom ground state, for example, is now n.esUl'Cd, since no more electrons ean be accommodated in the ncgativo-cnergy sea by the Pauli principle.

There are In any eonsequenees of thil new n.esumption of a filled

1 P. A. )1. Dirac, Pr«. !lo'll. 8«. (/AnGon), Aile, 360 (1930). See alJo J. R.

Opponheimer, PAri'. R •• , aB. 939 (1930).

6·~

65

B

Unoccupied Itlles

Fig. 6-1 Pair production: a nelativeenerlY electron is excited to a poeitiveenergy atato by radlat.lon.

eca of negative-energy electrons, I t is poesiblo for a nogat.ive-enc.·gy electron to absorb radiation and be exelted into a positive-energy state, as shown schematically in Fig. 0.1. I f this occurs, we observe an electron of charge -Ier and energy +E and in addition a hole in thc negative-energy sea. The hole registers the absence of an electron of c hargo -lei and energy - E and would be interpreted by an observer relative to the vacuum as the presenee of a particle of charge + lei and energy +E; that is, the pottitron. This is the basis of the hole-theory interpretation of pair production. Correspondingly, a hole in the negative-energy sea, or a positron, is a tmp for a positive-energy electron and leads to electron-position pair annihilation wit.h emission of radiation, as shown in II'ig. 5.2.

\Ve recognize that with the hole theory we transit to a manyparticlo theory deacribing pfJrticks oj both signs oj charge. No longer does the wave function have the simple probability interpretation of the one-particle theory, since it must now also record t.he produetlou or annihilation of electron-positron pairs.

Recall, however, that the Klein-Gordon equation woe discarded and the development of the Dime equation was motivated by thc

Fit. 6-. Pa.ir annihilation: a pOIitiveGnorlY elootron falla Into a nOlative-onergy holo emittiDI radiation.

R,,'ad., •• dc quanl"rrl meCllan'C.

desire to establish a one-particle theory. Therefore, we may ask, why not abandon the Dirac equation too? We are reluctant to discard it for the simple reason that by now we have uncovered an impreeslve body of "truth" in the Dirac equation-it predict8 the correct hydrogen-atom encrgy spectrum and 0 value of the electron to high accumcy. Moreover, positrons as first predicted by the theory have been observed.

Thus the historical path of reasoning mapped out originally by Dirac haa led to the desired equetion for an electron, though we have now reinterpreted the theory and thereby renounced the motivation that started the development. The history of physics bas numerous othor examples of this pattern of progress. Therefore, we shall retain the Dirac equation and the hole-theory interpretation and reject instead the one-particle probability interpretation which we originally set out to achieve. \\' e note here that it should also be pOl8i blc to return to the second-order Klein-Gordon equation and rescue it by a suitable reinterpretation of the wave function there too.

The advantage of the Dime over the Klein-Gordon equation is that it correctly describes electrons of spin ~ with 0 - 2. The KleinGordon equation applics for spinlcN particles such as pions, as will be dlseueecd in Chap. O. For both equatione we have the invariant, quadratic energy-momentum relation for froe particles PJjP" - mi. III both casca we must rcinterpret the negativo-cnergy solutions in order to secure stable ground states, and this leads unavoidably to the existence of antiparticles as wcll 8S particles. The particles arc described by positive-energy soll1tiona-for the Dime equation, electrons of mMS m and charge -1,1: thc antiparticles are described by the reinterpreted negative-onergy solutions and, in the present instance, are positrons of Ina88 m and charge + 1,1.

2 Chnrge Conjugntlon

There thus elnerges Irorn the hole theory a fundamente! new symmetry in nature: to cach partlclo there is an antiparticle and, in particular, the existence of electrons itnplica the existence of positrons. 'Ve seck now a fonnal expreeeion of this synllnetry which we use to Ionn directly the wave function of a positron from that or the miuing ncgative-energy electron to which it eorrespcnds.

By our physical picture a hole in the negative-energy sea recording the aba nee of an energy -E(E > 0), and the absence of a charge e (for an electron e < 0), is equivalcnt to the prcacmce of a positron of

tlleory

67

positive energy +E and charge -e. We thus have a one-to-one correspondence between the negative-energy solutions of the Dirac equation

(·i' - e/t - m)tJt - 0

(5.1)

and the positron eigenfunctions. Since by their interpretation positrons appear as positively charged electrons, the poaitron wave function tJte will be a positive-energy solution of the equation

(2V + e/t - m)tJte - 0

(5.2)

Conversely and with historical hindsight we could equally well sta.rt with the Dirac equation (5.2) for positrons. Nowhere in our considerations haa the sign of the charge e played an eBBOntial role. Electrons will then emerge from the hole theory reinterpreta.tion as the absence of the negative-onorgy solutions or (5.2). We have thcreby a one-to-one correspondence betwoon solutions of (5.1) and (5.2) for both signs of charge and are led to construct an operator transforming the two equations into each other.

First we observe that it is necessary to change the rclative sign between the two terms IV and /t in transforming from (5.1) to (5.2). We aceomplish this moat readily simply by taking the complex conjugate: ia/a~ - - (ia/a~)· and A,. - +A:. Upon doing this, we find that (5.1) becomee

[(i £; + eA~) .... · + m] 11'* - 0 (5.3)

If we can now find a nonsingular matrix, denoted C .... o, with the algebra

(C .... O) .... ,,·(C .... O)-I _ -,.." we shall have the desired fonn

(5.4)

(if + e/t - m) (C .... '\lt*) - 0

with

(5.5)

the positron wave function. That there exists such a matrix C may be verified by explicit construction. Let us exhibit it in our representation of (2.6), according to which .... 01'~·,..o - .... ,," so that (5.4) becomes C,.."'''C-I - - ..... , or

C-ly-C _ - ,..,."

In this representatlon C must commute with "'1 and ,... and anti-

R.'ad,,'.dc quonturrl rrlGcllon'c.

commute with "0 and "" and a suituble choice is

C - i,.l.y° - -C-l - -ct - -C"

(5.6)

It suffices to be able to construct a matrix C in any given representation; the unitary t.ransformation to any other one when applied to this C will give a matrix appropriate to the new representation. We note also t,hat there is a phase arbitl'arineas in our dcfinition of C in (5.6); the similar circumstance for the parity transformation was dlaeuseed earlier. In the present considerations the phase of a wave function is of no physical intcrest and we do not pursue this question.

Letusexamineindotailwhatthetransformation'/t, - CJt" - i,.",,· does to a negative-energy free-particle eigenfunction. For a negativeenergy electron at reat with spin down we have the wave function

o

I 0

'/t" - (21f ).~ 0 e+i ....

1

The corresponding positron solution is then

0 0 0 • 0
-I
0 0 . () () 1
i,.'ti"· - i 1
0 • 0 0 -- ,-'.'
1 0 (21r)9i
• () 0 0 I
-I
I
I ~ 0 (5.7)
- (21r)H 0 ,-''''' - ti'l
0 That is, the absence of a spin-down negative-energy electron at rest is equivalent to the presence of a spin-up positive .. encrgy positron at reat. In the field-free case thcro is no difference between an cloctron and positron, and we sce by (5.7) t.hat the transformation (5.5) has Iormed just another electron solution.

I Applying the same tran!d'orluation to an arbitrary spin-momentum eigenst.ate, we find, using lC,,.tJ - () and ,.r - ,.. - ,.:,

!/t, - C~~ - C.,~· - C.,. (!l.2~ m)" C "'; ""Y !/t.

- C(q~2~ m)C _t·T).,~.

- C- f 2~ m) C "'; ., .. ) !/t, (5.8)

"."or,.

69

Again we 800 that the operation of (take the complex eoniugate) X (matrix multiplication with C,.o - i,.') has yielded from a negativeenergy solution described by Iour-momentum PIA and poln.rization alA a positlve-energy solution described by the aanuJ P" and a". In tenus of (ree-particle spinors, (5.5) reads

e'.C'··'v(p,a) - C{},'l' (p,a) e'fCJI'''u(p,a) - C[I'(p,a)

showing tbat v(1',a) and u(p,a) are eharge-coniugato spinors, within a phase fn.ctor ~(p.a).

Recall t,hat the solutions were constructcd such that

Po - + vp' + ml - E > 0

.

AII50 notice that a docs not change sign under charge eoulugatlon but

the spin docs reverse as we saw in (5.7). As discusscd in Sec. 3.2, this difference liOB in the (act that the spin-projection operator hae the Iorm (1 + "0 d • s/2) ill the reet SyStclll where a" - (O,s), and the sign change eomes from the "0 matrix,

The operator in (5.5) expli('itly constl"Uct.a the wave function o( a positron. 'Ve may develop from it an invariance operation (or the Dirac equation by defining the additional opemtor which changes the sign of the electromagnetic field. 'l'len the sequence of instructions (1) take complex conjugate, (2) mUltiply by C,.o, and (3) replace all A" by - AlA is a formal sYlnmetry operation of the Dirac thcory. It trans(orlus Eq. (5.1) [(5.2)] (or the electron [poaitron] into the MOle equation for the positron [electron] and is ealled the cbal"gC conjugation transforluation, denoted bye. The physical content of the tra.nsforlnation of charge conjugation is that for each physicn.lly realizable state containing an electron in a potential A,,(x) there corresponds a physically realizable state of a positron in the potential - A,,(x). Thus e changes spin-up eleetrona of positive energy to spin-up positrons of positive energy by tra.nsforlning u poeitive-energy solution of (5.1) to a negative-energy solution of the same equation, that is, to a positron according to the hole theory.

That the dynamics of a positron in a field - AlA is exactly the same a.s thn.t of an electron in a field + A" is not at all surprising to us from classical coneideratlone, The surprising and new result to which we have boon led by the hole theory is thn.t if there exist electrons of nlass m and charge e, there neee.arily mual also exist positrons of the same m&88 m but of opposite charge -e.

Rela tlv'lJdc quan fl. rra rrrocllClnic.

It is indeed one of the strongest votes of confidence in at least the partial validity of relativi8tic quantum theory that electrons or both signs of charge and of the saine nla88 are observed in nature.

3 Vacu trm )'olarizQ lion

The hole theory, while removing the negative-energy difficulty, leads to now fundamental barriers to be surmounted and new phY8ical predictions to be verified by experimcnt. For example, eonsider the influcnce of the vacuum on tho definition of the cllal'ge of the electron and upon the interaction between two cllargcs. A positivc-cnergy electron electrostatically repels the eleetrona in the negative-energy sea. It thereby polarizes the vacuum ill its vicinity, and the charge density of the electron, po(r), plus polarieed vacuum, pp(r), measured relative to tbe vacuum, is schematically shown in ],'ig. 5.3. The charge of the electron as soon by a macroscopic applied field, 01' by a teat charge at a large distance, is J dll,(po(r) + pp(r)] - " the "physical" charge. liowever, for a test charge probing at distances '0 < H, the apparent chaa-ge ia more negative until, as '0 _. 0, the charge becomes J dl, po(r) - '0, the "bare" charge, with leol > lei. This phenomenon ia observed in the hydrogen-atom spectrum. The electronic lJ levels are lowered relative to those with angular momentum I .. 0, since the l - 0 wave Iunctlons bring electrons close to the protons. This cfl' cct of vacuum polarization, calculated in Chap. 8, reduces the Lamb shift slightly. We also take up there the question of how to connect the "bare" charge of an isolated electron with its observed value at large distances.

r)

-----.r

R

FiU. 6-3 El1'eot 01 vacuum polarization on tho eleotron's chArlo dcnlity. PO II tho olaargo deDlity of the "baro" electron and P. tbat of tho induced polarizatlon .ac1oud" of virtual oleot.ron-pOIitron paira.

,'.eory

71

Another question ariscs from the hole thcory: \Vhat is the meaning of the infinite ncgative total charge in the vacuum as we have defined it? For thc present we sidestep this question, remarking that there is no prefcrred dircction in which an electric ficld from such a distribution could point. Only Inhomogeneltiee in this distribution due to vacuum polarization are observable.!

" Time Rever.aI and Other Symlnetriel

IlOt us turn now to the parity and the time-reversal transformatiolls. These are symmetl")' operations which are not included in the discussion of proper Lorcntz invo.riance of the theory. The additional symmetry of electromagnetic gauge lnverlanee is evident from the form of the coupling, p,. - eA", as remarked in Chap. I. It is verified in just the sallie way as in the Schrtidinger theory.

Recall that the parity, or space reflection, traneforrnatlon was found in Sec. 2.3 to be expressed by

PtJ(X,1) - tJ' (x' ,0 - ,'·,),oy,(X,l)

for x' - -x

(5.0)

tJ'(X',l) i8 readily interpreted as the space-reflected solution. For plane-wave solutions the parity transformation (5.0) invert. the momenta and leaves the spins unchangcd as we classically cxpect. This transformation on the wave function coupled with the familiar one for the vector potentials, expressillg their scalar and vector nature

P.z.(x,l) - .z.' (x' ,I) - .z.(x,l) PA(X,l) - A'(x',t) - -A(x,l)

for x' - -x

(5.10)

leaves the Dirac equation and all physical observables unchanged. The physical content of the parity invariance of the Dirac theory may be expressed simply ill terms of a sct of obeervatione on a state described by a wave function tJ(x,l). We record these observations on motion-picture film, aiming our camera at a plane mlrror fOl'ming an image of the cxpcrimental setup. \Ve say tho.t the dynamics underlying our observations is invariallt under parity if the movie we make of the mirror image describes a sequence of physically realizable observations, that i8, if we cannot tell from the sequence of eventa observed in the film whether we are looking at a mirror image or not. For thi8 purpose a mirror iallage is 0.11 that need be considered, although

I This qUOItion iI dieoulled alain In J, D. BJorken and S. D. DroU, ·'lle1ativiatlo Quantum Fiolcl8," MoGraw-Hill Book Company, Ine., in pre •••

H..I"tlv ... tlc '/111"1111 m m.CollCln'C'

it is not identical with spatial reflection. A mirror inverts only the eoordiaatee normal to ita plane; this must be followed by a rotation through e about the normal for the parity transformation. Such a rotation is already included, however, in our dlseueelon of proper Loreuts invariance.

Turning next to time-rcvcrsal invariance, its physical content may be illustrated again in terms of the motion-picture film which we U80 to record a ect of obaervations on a state described by -I'(x). Let us now rUII the movie backward. We say that the dynamics underIyillg the set of observations is invariant under time reversal if the backward-run movie describcs a set of physically realizablc observations. Tbis lnvarianee will be guaranteed if we may changc t to t' - - t and carry out a transformation which reproduces the form of the Dirac equation with the same rules for its interpretation. The transformcd wave function will describe thc original electron running backward in time and will be physically realizable, since it satisfics the Dirac equation.

To constl'uct the desired time-reversal transformation, we write the Dirac equation in hamiltonian form

i a-l'~~,t) _ 11-1' - lcr' (-1" - cA) + (jm + e¢»]-I'(z,t) (5.11)

and dcfine the transformation :I such that if t' - -t, -I"(t') - ~(t). Thcn (5.11) beeomes'

(5.12)

and time-reversal luvariance requires that either 'JH (t)'J-1 - - H (t'), or

'Jia-I - -I

'1'0 investigate the behavior of II under 'J, we must specify the behavior of the elcctromagnetic potentials A,. when we let t' - - t. Since A is generated by currents which reverse sign when the sense of time is reversed, we require

Sianilarly,

A'(t') - -A(t) ~ 0') - +.z.(t)

(5.l!i)

since it is generated by chargcs; also t" - + t'. since z' - +x. To restore (5.12) to the original form, it is clear then that tl"8.nsformations with 'J • • • 'J-1 must change i to - I; hence, 'J may be written in the

1 Th.lnMMDtlnl depondeDoc OD Jr "lJupprcllllOd here.

".eo".

73

form (take complex conjugate) X (multiply afterward by a 4 X 4 constant matrix T);

(5.14)

This gives a",'(t')

i at' - 1(- To:·T-l).( -1",,' - d/(t'»]

+ (Tp·T-l)tn + ~/(t') ItI-'(I')

In our usual representation (1.17), this meana T must commute with 01 and p and anticommute with 01 and 01; thus

(5.15)

is satisfactory; thc phase factor is again arbitrary.

To show that the transformation :I corresponds to what we mean classically by time reversal. we apply (5.14) and (5.15) to a plane-wave solution for a froo particle of positive energy:

:{, im m) C -+; ~ .. ) t(z,t)

- T (1· 2~ m) T-'T e + t'*) T-1t'(x,t')

- (t:.2~ til) C +2 ~ "') t' (z,t') (5.16)

where p' - (Po,-p) and " - ('0,-8) projcct a free-particle solution with reversed direction of space momentum p and spin 8. This operation. known as the Wigner time reversal. was first introduced in 1932.1

Since the space and time coordinate inversions P and :J are invarianee operations of the theory, we may just as well include them, if we wish. in constructing the positron wave function. Combining (5.9). (5.14), and (5.15) with (5.5), we find a simple corrcspondence between a positron wave function

"'"o".(x') • PC .... o(:I.;-(x»· - Pe:J",(x) - U''hy.tI-(x)

with x~ - - XII

(5.17)

and an electron wave function multiplied by U''hyl and moving backward in space-time. For a Iroe-partiele spin-momentum eigenstate tI-(z) characterised by (p",~) and. - -1, we 800 that

IE. J'. Wilner. G1Uling.r NG(M., 81. 646 (1032).

neladtt"t'c quantu rn " •• chanici

~"C1'(x') - .. '.,}" (-'~ m) e ~ ,. .. ) ~(x)

- (, im m) C -2"") t l'CI"(x')

(5.18)

E()uation (5.18) differs from (5.8) only in the direction of spin and tells us, therefore, that we may picture a positron wave function of positive energy as a negative-energy electron wave function multiplied by ie··'r~ and moving backward in space-time.

I~or an arbitrary solution in the presence of electromagnetic forces we may explicitly verify this interpretation by rcturning to the negative-euergy eigenvalue equation

[0 . (-it" - eA) + (jm + e~)wI' - - EwI' (5.ID)

and carrying out the transformation (5.17). Evidently, by (5.10) and (5.13), A; (x') - +Ajl(x) under space-time coordinate Inversion, x; - -XII: then (5.1Q) takcs the desired form

[0 · (-iv' + eA'(r» + (jm - e~'(x')lw1'pcr(%') - +EwI'pcr(x') The interpretation of positrons &8 negative-energy electrons running backward in space-time forms the basis of the StUckelberg-lt'eynman form of positron theory.' 'Ve shall use it often in the following chapters in developing soattcring theory, and we shall find that it offers great advantagce there.

In conclusion, we must notice that the structure of the interaction of electrons with the electromagnetic field was dictated by experience with both the classical and the nonrelativistic limit of the electrodynamics of electrons. The existences of the symmetries we have diseueeed are dependent upon the form of interaction. For instance, an anomalous moment interaction of the type diseueeed in the problem. for Chap. 4 for protons and neutrons adds a term of the form (1~~"'wI' to the Dirac equation. Its presence affects none of the above symmetries. In extending the Dirac theory to other particles of spin ~ such as Il meson or nucleons, and to other familiar interactions, it is very natural to 888ume that these symmetriea of :l, e, P are still preserved.

It was the great contribution of Lee and Yangl to realize that this is really an auumption to be verified by experiment and to SURgest that interactions such as {j decay violate the sYlnmetries of P and e.

IE. C. O. flUlckalborg, Hel&J. Ph" •• Acta, lt, 32L, 688 (1941); R. P. FayDman, PI." •• II ., 76, 749, 70ll (1949).

T. D. Lao and C. N. Vans, Ph" •• RCfJ., lOB. 1071 (l9l57).

7S

However, the symmetry under the oporation :lep is guaranteed by the much weaker 8Uumptions of proper Lorentz invariance and the usual connection between spin and stati8tica.

Prohlen18

1. Show that the rate lor an olootron in tho hy(~rolon-atom ground IJtate to radiate and fall into empty nelativHnorgy IJtatel (treated in Bore approximation) in the Inefl)' intorva! -,nc· to -2"1(· i. approximaLOly 2QIn,,·, .. fl .. 10. 1J00-1.

2. Reinterpret and roeolvo tho Klein paradox 01 Chap. 3 by UJing hole-tbeory idea.t.

3. Show that if "YJI and ..,; are two reprOlientat.ionl of tho "Y mat.ricH related by a unitary tran lormation U 10 that ..,,. - U..,;U-I. then C1 - (UIj-zCU, where C and C' aro th., corrNpondln~ matricee 01 tho charga-conjuRaUon tranelormation. Aro relation. (6.0) valid for C'f

In a .imilar waYt froo (6.16) lrom tho reprelentation (1.17) of the,. matrlcee.

4. In order that :t be a .ymmetry operation 01 the Dirat'l theory, tho ruiN of Interpretation of tho wave function ~'(t') nlu"t M tho aamo lUI thote of ~(t). Thie moanl that obearvablee com POled of forml bilinMr In rI' and ~'t mUlt have tho lamo interpretation (within a lJi~n, appropriate to tho oblClrvable) .. tholO of rI. eI. Vorlfy that thilJ il eo for tho eurront:

J~ (z') - jIJ(z)

and alIo

(r)' - (r)

(p)' - -(p)

h. Repeat th080 oaleulat.iona for tho charge-e.onJup.tion transformation C. In particular. show

~.(Z)"'JA"'.(z) - +~(z)..,,.!J.'(z) and interpret UJing the holo theory.

6

Propagator Theory

1 Introduction

We turn to a genernJ discussion 01 scattering proeeesee, Our aim i. to be able to calculate transition rates and erose soetione with the Dirac theory of electrons and positrons-in principle, exactly; in practice, to low orders 01 expansion in the interaction parameters. The p088ibility of altering the numbers 01 particles in such proceues aa electron-positron pair production or annihilation carrica us beyond the scope of the discussions in nonrelativistic theory. However, we shnJl delay aa long D.8 poselble the enormous task of developing the Ionnalism 01 quantized field theory in order to accommodate this production and annihilation of particles.

To this end we follow Feynman' in developing the propagator approach. The scattering process is dcacribcd ill terms 01 integral equations. The boundary conditions for their solutions incorporate the SUlckelbcrg-li"eynman physical interpretation 01 poeitrons as nep;ativc-energy electrons running backward in time. From this formulation a working theory with unambiguous rules 01 calculation for all phyBical processes emerges. t To begin, we review the propagator approach to the nonrclativistic Schrlklinger equation.

2 The Nonrclatlvl8tlc l~ropaRator

In IC&ttcring problems our attention is focused upon wave eolutlons which develop in time from initial conditiolls imposed in the remote past rather than on stationary energy eigenfunctions, that is, standing waves. CharacteristicnJly, given a wave packet which in the remote past represente a partiel9 approaching a potential, one aaks what the wave will look like in the remote luture.

We turn to lluygcns' principle lor a convenient way 01 viewing this process. If the wave function t/t(x,t) is known at one particular time I, it may be found at any later time I' by eoneidering at time teach point of space x 88 a source 01 spherical waves which propagate outward from z. The strength 01 the wave amplitude arriving at point x' at time f Irom the point x will be proportional to the original wave amplitude t/t(x,t). II we dcnote the constant of proportionnJity by iG(x' ,f ;J:,t) , the total wave arriving at the point x' at time t' will, by

1 R. P. Feynman, Pia",. nfJJ., 76, 741). 709 (1949).

I The quantum field theoretic baala ror theeo rul.1I provided In J. D. BJorken and S. D. Droll, URoiativlltic Quantum Fieldl," MeCira,,-lIill-Dook Company, Ino., in '(Jr"'.

78

a,at.or d •• ory

lIuygena' principle, be·

t/t(z',t') - if d1z O(x',t jX,t) t/t(x, °

t' > t

(0.1)

G(x',t'jX,t) is known as the Groen'a function or propagator, and it doscribea to us according to I1uygens' principle the influence upon t/t(x' ,t') of the magnitude of t/t at x at time to Knowledge of 0 enables us to construct the phyBical atate which develope in time from any given initial atate, and thus ia equivalent to a complete solutlon of the SchrOdingor equation.

'Ve must still give a complete formal definition of G. So far we have only claimed its existence on the basis of physical arguments. Let ua pursue theee arguments further in order to develop a better understanding of the propagator approach. Conaider first a free-wave solution. The motion of a free particle is completcly known, and it should not come aa a aurprise that the eorrespondlng free-particle Green's function 00 can be constructed explicitly. If we now introduce a potential, Go will be modlfled, Let V(X.,t.) represent an interaction potential which ia "turned on" for a very brief interval of time ~t. about t.o For times earlier than t., the wave Iunetion will be that of a free wave 'I, and the corresponding propagator will be 000 However, V(X.,ta) acta as a source of new waves according to the Schradinger equation

(i';'; - HO) oI-(x"t,) - V(x"',)l/t(x,,t,) (6.2)

The right-hand side differs from zero in the interval ~t.. It producea an additional ehange in t/t during ~t. above that taking place in the absence of V. This additional wave ~t/t(Xht.) is found by integrating (6.2) to first ordcr in ~t.o

~t/t{x .. t.) - -iV(z.,t.)rp(x.,t.) ~t. (0.3)

This added wave, by II uygena' principle and (6.1), leads at a future time t' to a new contribution to ';(x' ,t'), which is

~t/t(x' ,t') - J d'x. Oo(z' ,t' ;x.,t.) V(z.,t.)CP(x.,'.) ~t. (6.4)

'rhus the wave t/t developing from an arbitrary packet 'I in the remote past ia

t/t(x',t) - rp{z',t') + J d'x. Oo(x',t'jx.,t.)V{x.,h)rp{x.,t.) ~t. - if d1x [Oo(x',t';x,t)

+ f d'x. ~t. Go{x',t' ;X.,ta) V{X.,t.)Oo(x.,h :x,OJrp{x,t) (6.5) 1 The applicablllty of lIUYlena' principle without Kirchhoff'lJ modification II due to the faat that the SchrOdinler equat.ion iI firakrder in the time derivative.

79

R,,'at.,,'.dc qlUlt. f" m " .. ,("han'".

Comparing with (6.1), we see that tho Green's Iunctlon hore is given by O(z',t';z,t) - Go(z' ,t' ;z,t)

+ f d'x. ~t. Go(z',t' ;Xhh) V(Z.,t.)Go(z.,t. ;x,t) (6.6) It may be illustrated by the space-time diagram shown in Fig. 6.1. The first term (It'ig. 6.1a) represonts the propagation from (z,t) to (z',t') &.8 a free particle; Fig. 6.lb rcpreeents free propagation from (z,t) to (z"'.), a scattering at (x.,t.), and frco propagation from (Zht.) to (z' ,t').

If we turn on another potential V(ZI,tl) for an interval .1t:.J at time 'I > t., the additional contribution to tJ1(z',t) for t' > 't is, in analogy to (6.4),

.1'/t(z') - f d'zt Go(z';2) V(2)tJ1(2) .1" - if d'x dlxl.1tl Go(r ;2) V(2)

X [Go(2;x) + f d1x • .1'.00(2;1) V(I)Go(l ;z)JCP(x) (6.7)

in an abbreviated notation whose meaning should be clear. The first. term is illustrated in the diagram (Fig. 6.le) and represents single

,

(b)

(4)

t

(c)

(el)

Fig. 6-1 Space-timo eli.graml for propaKaUon from (z,t) to (1.',") u (a) a treo particlo, (b) with oDO Icatterinl by potential V(J.It'.) at (ZIt'l), (c) with linllo Ic.Uorinl at (ZI,"), .nd (el) with double lCattorml.

ecattcring at 'I; the seeond term is a double scattering eorrcction illustrated in Fig. 6.ld.

The total wave arriving at (x' ,e') is then built up by inserting (6.5) for tI'(2) in the right-hand side of (0.7) and adding the r08ulting

"'''' to (6.6):

tI'(x') - rp(x') + f d'z. "". Go(x';I)V(I)rp{l)

+ f dlzl ""t Go{x' ;2) V(2)rp(2)

+ f d'z. ",e. dlxl "", Go(x' ;2) V(2)Go{2;1) V{I)rp{l) (6.8)

Without further ado, if there arc n such time intervals when the potential V is turned Oil, the wave arriving at (x',t') will be

tI'(x') - rp(x') + L f daze' ""~ Go(r' iZ4) V(zdrp(z,)

,

+ L f dlx,"'~d'xj""iGo(x';X4)1'(z4)Go(X4iZJ)V(X/)rp{x/) -.J

('~>")

+ ~ f d1xe' /:at, d'zl "'tj d'z" """

fit

('~>">"')

X Go(x',z,) V(xdGo(x,;x/) V{X/)GO(XJ;x,,) V(x,,)rp{z,,) + . .. (6.Q)

By comparison with (6.5) and (6.6) the corresponding expression for the Groen's function G will be

G(x';x) - Go(x';X) + L f d1xe' ""~ Go(x' ;x"I1) V(Xi,l.-)GO(x"I1;X)

,

+ ~ f dl~. "'~ d1xJ ""j Go(x' ;Xi,l1) V(x"t,) II

(">")

X GO(Xi,l1",%j,tj) V{xJ,ej)Go(x",'liZ) + · .. (6.10)

We may lift the thne-ordering reetrictiona ~ > ej, ete., if wc dcfine Go(x' ,t' jX,') - 0 for t' < ,. With this boundary condition of propagating waV08 forward in time only, Go is known as the retarded propagator. Physically this just means that no II uygens wavelets ~tI' from the ith iteration (at time 11) appear until after 11.

In the limit of a continuous interaction the sums over time inter .. vals may be replaced by intcgrals over de with the result

G{x';X) - Go(x'j:t) + J dtx. Go(x';X.) V(x.)Go(x.;x)

+ f db. d4z, Go(x';X.) V(x.)Go(x.;Xt) V{x,)Go{x,;x) + . .. (6.11)

where

ne'atl,,'.t'e ql,antu m mechanIc.

This multiple 8cattcring series (6.11) is asaumcd to convergeL and may be summed formally to yield

G{z';Z) - Go(z';Z) + J d4z. Go(z'jx.) V(z.)G(z.;Z) (6.12)

'Ve notice that not only Go(x';X) but also G(x' ;x) vanishes for f < I, as demanded by our elementary concept of causality.

Equation (6.11) gives us an iterat,ion procedure for finding G in terms of V and Go and hence for constructing the wave function tI'(z',t') if it is known at an earlier thne. In particular, to solve the seattering problem, we must know the wave in the remote future, given a wave packet ~(x,t) rcpresenting a particle in the remote past approaching the interaction region. In order to define properly the scattering problem, there should bc no interaction at this initial time, 10 that ~ is a solution of the free-particle equation which ineorporates the required initial conditions.

A nulthematically convenient way of accomplishing this is to localize the interaction in timet by adiabatically turning ofT V (x,t) as t -t - 00; the exact solution r/I then approaches ~ in the remote past and thcre is no seattorod wavc. In the future the wave tI'(z',t') is given by (6.1)

tI'(z',t') - lim iJ dlx G(x',t';x,t)~(z,t)

'_-.

(6.13)

Expresaing G in terms of Go by (6.12), we BOO tI'{z',t') - lim if d1z [Go{z',t';x,t)

'_-.

+ J dfz. Go(x',t';I) V(I)G(I ;x,t)]~(z,t)

- ~(z',t') + J d'z. Go(z',t';X.,tl)I'(Xht.)tI'(Zht.)

(6.14)

'Ve have really not solved anything, since the unknown tI' appears under the integral on the right. However, we do have a formulation which includes the desired boundary conditions and which afTordi an immediate approximation procedure if the perturbing potential lis weak.

We are primarily interested in the form of the scattered wave 88 t' -t co. In this limit the particle emerges from the interaction region and again tI' becomes a solution of the froo-particle equation. A8 before, we adiabatically turn ofT the interaction as t' -t + 00 in order to ensure this condition. All information about the scattered wave

• We ignore hero the poeBibility of bound aut. in the potential V.

• \Ve might equally well build wa.ve packet! localilod in apace and initiaUy remote from the interaction "'lion.

opa,ator rhoory

nllly be obtained from the probability amplitudes for the particle to arrive in various final free states 'PI

'PI(Z' t') _ 1 e'~"l'-",,'

, (211")"

(6.15)

as f _. + 00 from a given incident wave 'P,; in particular. we may work with plane waves.! The probability amplitude for a given pair (f.t) is an clomen t of the S. or scattering. matrix and is given by

SI' - lim f 'P;(Z/.t')~~+)(z'.t') d1x'

,,_-

- lim f d1x' 'P; (z'.t') ('P,(x' .t)

t-_

+ J d·x Go(z'.t ;z.O V(x.O~~+)(x.t)]

- al(k, - k,) + lim f dlx' d"x 'P;(z'.t')Go(z'.t';x.t)

,,--

X V(x,t)~~+) (z,t) (6.10)

where ~~+)(z.t) is that solution of the wave equatlon (6.14) which reduces to a plano wave of momentum k, as t -+ - 00. By the ahorthand t -+ ± 00 we mean t _. any large finite time for which the particles are not in the interaction region (or alternatively when Y is turned off); in particular, t -+ ± 00 nlay mean the times when the particlo is produced and detected.

\Ve may expand ~(+) in a multiple scattering series by itcration of (6.14) and thus exprcss the 8 matrlx in a multiple scattering series. the terms of which correspond to the diagrams of Fig. 6.1.

6.3 Foemal ()efinltlon. and Propertle. of Green'. Function.

\Ve have unearthed the physical ingredients for solving a scattering problem. We now build the formal mathematical machinery to manufacture these solutions. Our goal is to investigate thc differential equation which definee 0. and in particular to solve for 00 explicitly,

I The plano-wa.ve IOlution. ar4' nonnalbed in the continuum ianlllage in (6.16). Altoma.tive1y, the box nonnalilation convention may be used, with

(2r)-M .... V-J'

where V ilJ the volume of Ule box in which the phy.ical intoraation ilJ confined. WiU, the box convention the Dlrac I function in (6.16) ill replnaed by a Kronecker Ifunctton

if k, - ~ Uk, ,. It,

80 that the expansion of G we have outlined can be explicitly carried out. \Ve start with Eq. (6.1), valid for t' > t, and rewrite it in a form valid for all times:

8et' - t)~(x') - if dlx O(x' ;x)~(.t) 9(1' - t) i8 the unit atop function defined by

9(t' - t) - {01 t: > t f < t

and has the following very useful illtegml representation:

(6.17)

(6.18)

. - If. dw e-'1If

9(,) - lim f,_. + .

..... o~, -. Col) 'e

It is evaluated 88 a contour illtep;ral in the complex e.) plane as shown in I·'ig. 6.2. For l' > 0 the contour may be closed along an infinite semicircle below the real axis in order to ensure exponential dalnping of the integrand, and the value of the integra! is I by Cauchy's theorem, For l' < 0, the contour is closed above and the integral vanishea because the polc at -it now lice outside the contour. Since 9(1') takes a unit [ump at l' - 0, its derivative is a 3 function:

d9(1') If.

-- - 3(1') - - de.) e1rtlr

d1' 21r -.

(6.10)

(6.20)

'Ve 1l0W attempt to find the equation and formal properties of O(x' jX) from (6.17). We know only that ~(x') satisfics the Sch~dingcr

equation; WI! are tberefore led to apply [i ir - H (z') ] to (6.17):

[ i ft, - H(z') ] lI(l' - t)I/-(z') - i aCt' - t)I/-(z')

- i f d'x [ i :" - II(z') ] O(z';z)I/-(z) (6.21)

Imw

Fig. B·' Contour in tho complox IW plano for intoll'atwi thu unit step runction 1(,,).

","tor d .. ,ory

85

Since (6.21) is valid for all solutions t/t we can cxtract from it the Green's function equation of the Schr&lingcr theory

[ i :r - H(z') ] G(x';z) - ,I(ll' - ll) a(I' - t) - a'(z' - 21) (6.22)

Togethcr with the boundary condition of a forward propagation in time, that is,

O(X'~) - 0

for t' < t

(6.23)

Eq. (6.22) defines the retarded Green's function or propagator appropriate for (6.17).

We can solve explicitly for the free-particle propagator when

110(x') - - ~ V:" I n this case Oo(x';x) can depend only upon the difference of the coordinates (x' ,t') and (z,O. This is because the wave at (z',t') emerging from a unit source at z which is turned on at t depends only on the interval (z' - z, t' - 0, and Oo(x'jx) is precisely the amplitude of this wave. We consider its Fourier transforln

Go(x'"jx) - Oo(x' - x)

(6.24)

In terms of Oo(P.CIJ), (6.22) is

(i 4, + _!_ VI') (;o(x'·z) - J dip ~(w - pi) Oo(p CIJ),--W-f)+.,.(a'-a)

at 2m ' (2T). 2m '

J d~pdw

_ crwu'-t)+.,o(Z'-Z)

(2T).

and hence for w ,. pi 12m

I Oo(p,w) - CIJ - pi/2m

(6.25)

A rule for handling the singularity in the denominator is neccuary to complete the expression in (6.25). This is determlncd by the retarded boundary condition (6.23). Recalling the discussion of the 6 function (6.10), we add a positive infinitesimal imaginary part to the denominator and carry out. the CIJ integration in (6.24) first. The singularity

1m til

--._--------~--------~------~~Retll

Fig. 8-3 Sinau1arlty in Oo(P."').

then lies below the real axis as indicated in Fig. 6.3, and we obtain

f dip f · dw ,-wU'-f)

Go(x' - x) - Tn::ti e,,·(a'-I) - - - •

\21-,' -. 2 .. w - pi/2m + u

- -i f dlp_ e,p,(Jr'-a)-'~(II-t)8(f - o (21-)1

- -i8(f - t) f dip cp,,(z',t')cp:(x,t) (6.26)

where the last form uses the notation of (6.15). It ~for the special ease of plane waves=-an example of a useful expression for the Green's function 88 a sum over a complcte act of cigenfunctions of the corresponding differential equation.' In general, if we can construct a complete set of normalized solutions to the SchrOdinger equation which satisfy a eompletenese statement of the form

L .s- .. (z',t).s-:(zJt) - al(z - x') (6.27)

"

where L is a generalized sum and integral over the spectrum of quan-

n

tum numbers nJ then, as is readily verified,

O(x';X) - -i8(t' - t) L .s-,,(x').s-:(x)

"

(6.28)

eatisfiea (6.22) with the deeired boundary condition. The special case (6.26) for 00 is establishcd by the connection 2 _. f dip when

..

integrating over the continuous momcntum spectrum.

I The free-particle Groen'a function in (0.20) may be expreeeed in cloeod fonn

O.{I'.I'iz.I) - -i (2"'('':'- ())" {axp [i~~~ : ~I~]} 0(1' - I)

Thia is rominlaoent of an expre.lon in the theory of brownian motion for the probabUity that a particle whioh waa at po.ition x at time I and which movee under the infiuence of random diatumancGI will arrive at r at time 1'. Indeed, the only ehal\le that needa to bo made is the roplacoment of (1.1') by (-il,-il'). Thil u.me chule traDlJfonnl the Sohr&Jinpr equation into the difl'Ulion equation.

ropo,tltor theory

From the form of (6.26) and (6.28) it follows that the same Green's function which propagates a solution of the Sehrodlnger equation forward in time propagates its complex conj ugate backward in time. Multiplying (6.28) by tJt ... (x). integrating over all x. and invoking orthogonality and normalization of the eigenfunctions. we reproduce (6.17)

i f d1x O(x' ;X)tJt ... (x) - 8«( - o I tJt,,(x') f d1x tJt:(x)tJt.(x)

"

- 8(l' - l)tJt .. (x')

Repeating this operation, only nlultiplying Instead by tJt!(x') and integrating over all r, we obtain the indicated result:

if d1x' tJt:(x')G(x';x) - 8(t' - t)tJt:(x) (6.2Q)

We now use thcac relations to construct various uecful forms for the 8 matrix.

From (6.17) and the defining equation (6.16) we can write a compact form for the S-matrix elements in terms of the cxact propagator:

8/i - i lim lim f d1x' d1x ,,;(x')O(x' ;X)cp.(x) (6.30)

f-. '_-.

This is not yet useful beeause in general we cannot solve directly for the exact propagator. As is evidcnt in (6.28) there is an enormous amount of infor.nation contained in G(x';X). All the solutions of the SchrfJdinger equation. including the bound states as required in the completenese relation (6.27). appcar with equal weight. It is no wonder that G ia difficult to compute,

We proceed, as in our earlier intuitive considerations leading to (6.11), by constructing an iteration procedure starting with the freeparticle Groen's function.

Regrouping terms in (6.22), we write, with 11 - 110 + V,

[i ~ - Ilo(z') ] G(z';Z) - 3'(z' - z) + V(z')G(z';Z)

- f d·x" 34(x' - x")[34(x" - x) + V(x")O(x";X)] (6.31)

where we have expressed the interaction term on thc right-hand Bide as a superposition of 3-function soureee. The int,egral of (0.31) with the desired boundary conditions is just the corresponding superposition

87

/fB'a ,,,,,., 'C quan II. rrl ".."chalt'.:',

01 free propagators:

G(r;x) - 1 d·x" 00(x';x")[3·(r' - x) + V(x")O(r';x)]

- Go(r;x) + 1 d4x" Go(x';x") V(x")G(x";x) (6.32)

which agrees with (6.12). Inserting (6.32) into (6.30) and Inaking U 01 (0.17) and (6.20) for free particlcs, we arrive at

8" - 1 dlx fP;(x)fP.(x) + lim 1 dfx1 d'x fP;(x~ V(XI)O(XI;X)~(X)

.... -.

- 3'4 - il dtxl '1';(1) V(I)fP4(I) - il dfxl dtxl '1';0) VO) X 00(1,2) V(2) '1'.(2) - il dtxl d·xl dtx. '1';(1) V(I)

X Go(I,2) V(2)00(2,3) V(3)fP4(3) + t •• (6.33)

This multiple-sen ttering series coincides torm by term wit h that devel .. oped from (6.16). It may also be finally summed up in terms of solution of the exact Schr~dillKer equation as in (6.16). To do this, wo note in the first line 01 (6.33) that we can write

lim 1 d'x 0(X";%)fP4(X) - lim 1 dlx G(x" ;%)-I',(x)

'_-. ,_.-.

- - i-l'4(:e")

if we refer to (6.17) and turn off V as we did earlier. Equation (6.33) becomes

SIt - 3'4 - il d·x" '1'; (x") V(x")-I'~+)(x") (6.34)

where the super8Cript (+) is now appended to -I' to indicate a solution which reduces to a free wave as t" -+ - 00 [see (6.14)]

-I'~+'(x") - fP4(x") + 100(x";%) V(x)-I'~+'(x) d·x

Equation (6.34) with expansion (6.14) and (6.30) with (6.32) are equivalent forme for the S matrix; both lead to the ruultiple-eeatterin series (6.33).

In practice we shall usually calculate only the first or first two nonvanishing contributioua to the S Inatrix for a given interaction in (6.33). The validity of this procedure depends on the weeknese of the interaction V and the rapid convergence of this scrics in powers 01 the interaction strength.

A general propert.y of the S matrix which results lrom the con- 8Crvation of probability is the property of unitarity. ,\,.re recall frolu the introductory remarke 01 Chap. 1 that hermiticity of the hamiltonian hnplies conservation of probability and thereby the result t.hat the

OIHJlln tor Illeory

inner product of two solutions is independent of time. We can write, therefore,

J dlx ~~+'·(x)~~+)(x) - lim J dlx ~~+)·(x)~;+)(x)

'_-.

- lim J dlx "t(X)"/(X) - 3/4 (6.36)

M-.

In the particular example of a plane-wave representation 3/i - 31(kJ - ~)

\\re Inay also projcct this inner product into the remote future in which case, by (0.16) and the colUpietene8s relation (6.27) for the ,,'s, we can expand the solutions ~~+) into plano-wave statos with the 8-matrix elements as the expansion coefficients:

Ibn ~:+)(x') - ~ ",,(x')S", (6.36)

"-+. "

(l- J dip for a plane-wave representation.)

"

Inserting (6.36) into the left-hand side of (0.36), we find

~ 8,.,8:1 - 31, (0.37)

"

or in Inatrix notation 8'8 - 1. If the ~l+), like the "" in (6.:JO), form a complete set, 8t - 8-1 and we conclude that 8 is a unitary matrix.!

6.4 The l}iropagalor in POAitron Theory

\Ve generalize our propagator development of the nonrelativistie theory and apply it to the relativistic electron thcory. Our startin,; point is provided by the picture of the nonrclativistic G(x':Z) as the probability amplitude for a particle wave originating at x to propagate to z', This amplitude, given in (6.11), is a sum of anlplitudcs, the nth such term being a product of Iaetora corresponding to the diagram of It''ig. 6.4, Each line in Fig. 6.4 represente the amplitude Oo(X.;%._I) that a particle wave originating at X,_I propagates freely to x,. At the point x, (represented by a 0) it is scattered with probability amplitude per unit epace-time volume V(x.) to 6 new wave propagating fonvard in time with amplitude OO(X'+1 jX,) to tho next interaction, Thi8 amplitude is then surmned over all space-time points in which the

'If bound ItntOl ooour, the completen .. sum in (8.27) mUlt &leo inolude the bound .. tate Ipoctrum. Thia dOli not altor the proof or unitarity.

89