Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Feynman
Ttllmun l'ruJ~u(Jr oj Ph,·.lc.
California It.. ti, .. te oj Tf:ldHlOlolIY
A. R. Hibbs
Jet l'ropul8ion l.abof'Ofory
CuliJurnin III"tirUh~ oj Tf!f.'hnolo,y
,
R. P. Feynman
.d. R. llibb.
I
Contents
'''-,.•• I ,
1·1 1·, 1...1.1111)· 111 (JUlIoIIUUII ., t"... IIAni~ :I
1·2 ·n l'h'''''.'III) 1"111'-11". 0
1-3 'ul..tI...tllll .' It " " 'IIoII\ _ t:1
IA """IIII1.",, ) ..r ",,,I.. ,.,Iot) t"un...·.,l. '0
I .A "'"'11'" 1I.......hli". 'l1.tI" "I,l. :tJ
IA:) on... I'u,•.-", ,,, "1M. I~...~ ;tJ
••
xii
The Quanlum·nl",IIan Ic. I La w of Mollon 2.
2-1 The CJl\..;:<Kal Ad ton 20 .
T he Qlltlhtum-mcr htlniclli Amphtud e 28
2-2
2.J Tbe Cl:L~irn l U mit. 2!)
2-1 TIl(' Sum over Paths 31
2-S Events Occurring in Succession 36
2-6 Some Remarks 38
.
D eve Ioping ,I.
• Coneo.ls 'W ith Specipi E.8mplctl 41
3--1 Th e Free Pe rt ielc 42
3--2 Diffract ion through a Slit 47
3-3 Results for a Sharp-edged Slit 05
3-l The WS\"e Function 57
3-5 Gaussian In tegrals 58
3-6 Motion in a Potential Field 62
3--7 Systems with )Isn)' Variables 65
3-8 Separable Systems 66
3--9 Th e Path Integral as 11 Functional 68
3-10 Interaction of a Part icle and a Ha rmonic Oscilla tor 69
3-11 Evaluation of Pa th Integrals by Fourier Series 71
loci... 3S9
1
The F'uridarnerrtal
Concepts of
QuantuID Mechanics
.-1 PROItlDlI.rn· IN QUANTl'M MECIIAN ICS'
•
I·rom nIIOU t u' 0 Iuo.:,..
· ...... innln.... of the twentieth century experimental
eo I '
.
physics Rl lUl...
, __ I
....~'U U
n impressive army or st range p rcuomcnn which
~ • • '1'1
d(';no ll~tmtt'tl the iIlUU{'CjllUCY of closs icul physics. 1(' att.emp ts to
I, J--
)---------i------- 2
A e c
F ig. 1_1 T he expcrtmcn tol nrrn ngement . El ect rons emitted a t A make t heir W fI) · to th e
detector at screen C, hut n screen IJ wit h two holes is int er posed , TIle .Ietrct or r('~ ist {'J'~ fI.
count for e ach electron ,,·hid l llrrive.!l ; ti le fract ion \\ hich a rrives when t he det ect or is
placed at u detence e from t he center of the screen hi measured nnd plot ted agnintlt x, nil in
F ig. ]-2.
B has two holes, 1 and 2, through which the elect rons may pass.
Fi nally , behin d the screen n at u plane C we have 3 det ector of elec-
trons wh ich ma y be placed at various d istances x from the center of
the scree n.
If the detector is extremely sensitive (as a Geiger count er is) it will
be discovered t hat t he current arriving a t s: is not continuous, but
corresponds to a mi n of particles. II t he intensity of the source S is
very low, the d etector will record pulses represent ing the arriva l of
individual part iclcs, separntcd by gaps in tim e during which nothing
arrives. T his is the reason we say electrons an' part icles. If we had
detect ors simulta neously a ll over the SCfl'CU C, wit h a very weak source
S . only one detector would respond, th en a ft er a little t ime, a nother
would record th e arrival of nil elect ron, etc. Th ere would never be a
half respo nse of th e d etector; eith er an entire elect ron would a rrive
or nothing would happen . Ami two detec tors would never respond
simultaneously (except for t he coincidence that t he source em it ted
two electrons within th e resolv ing I ituc of t he dct cct ors-r-a coincidence
whose p roba hility cun be decreased by furth er dcc n-nsr- o f th e' source
int ensity). In other words, t he detector of Fig. 1-1 records t he> 1':1$-
sage of 0. single corpuscular ent ity travel ing from S through u hole in
SCl'l'C11 B to the point r ,
T his particular experiment has never 1>cp11 done in just t his way .
In t he following des cript ion we nre stnt illJ!: wbn t t he res ult s would I"K.'
uecord ing to tile lnws wh ich fit every r-xpcrinu-ut of t hi,; IYIk.' which
bus ever been perfor med . ~O lll (' (' xI Jt' r i lll ('nl ~ whi ch d irectlv illus-
trute 11 m co ur-l us lo us w e' lin ' n'nchill~ lu-re ha ve lxx-n 11Il III.' , hut ~ III' h
ex periments nrc usua lly 1II0rc cemplicutcd . "'C'
p n.'ft·" Icr pl... I;IJo:1 1~i l'a l
11'n~11.~. t 0 ~(' I.- ... I "xlV'nUl(,llts
. I"
which nrc
..
t he simplest in principle a nd
di...:n'~llfll ti ll' clillicull ics of nctually dOIl1~ t bem, ' .
. I I II if 0 11(' prefers nne could Just n~ well usc hp;ht IJ1Htcnd
I llC'u t'u u )', " , •
of electro ns in this expC'ri lUC'ut . 'file sumo POlllt~ w~u ld IK~ Illustrated.
TI ';'OUI1:'C S could be a source of monochromatic light aud th e Ii(mlji~
ti ~; ~1l'h'('lor 8 photoelectric cd l or, bett er, a ph otomultiplier which
would record p ulses , each being t h(' arrival of a s ingle photon.
What we shall 1UC'nJ:\u n ' for vari ous positions x of th e detector is tile
mcnn number of pulses per second . In other words, we s hall deter;
mine experimenta lly the [rvlnti vc] probability P that the electron
passes fro m S to x, as n fu nction of x.
The graph of this probnhility as n fu nction of x is t he co mplicntcd.
curve illustrated qualitatively in Fig. 1-2n. It has several maxima
and minima, Bud t here arc locations near the center of the screen at
which electrons hard ly ever arrive. It is the problem of physics to
discover the laws govc ruiug t he st ructure of this cu rv e.
We m ight at first suppose (since t he electrons behave as particles)
IhAI
1 p ---- ~ ---
---
f'J feJ fdJ
. 'ig. 1·2 n~ull8 of IIltl CXp4"rimcl I' _.
nguinsl the Jlflsit iofl x of II .• I It . ffJ l.t1l llht y of IIrth ',,1 of elect roue lit 'Z IllotlCtI
Ilr·r(: nl (0). If 11111,)' one h ICi Ul' ector. 1'1,(, n- -ult r II " ." . I
' " 0 If' Cli.lll'rlmcll t or I' IJC . I_I l ~ ploW't
, II) To· • 0 c lB cpcu IlO t )1I' vleetr I ' II
III :. ror J1Il1l l,(,le 2 0 IK"A, it itI (el. Jr WI" O~18 r Ail gu t Irou R,h .JUtll Iloll' I , ti ll' rc.~u
11011. or t he otl lf'r we (:xp' I II - 1I1J1I~IIJC C'1 U" h vlre t rou ' ll",l goes tll rou~h one
·.1 ' "C
('oru.luf·naLly d i(Jcr~nl (r om ' 1 t .
, (: f"Un'(' (II) _ (b ) + I
e) WI ICII both IIOll"tl li n! op en. 1111s LS
..
'" III we lI t lUtlll,r get, (0 ) .
5
1M
Fig_ 1-3 An enntogoue experiment. in wnve int erference, 'The com plicat ed curve P (z ) in
Fig. 1-2a is the sn mc 8 5 the intensity l (x ) of wnvea whieh would arrive at z tit nrti ng from S
and corni ng t hrough the holes. At some point s x the wnvc lcta from 1101('8 1 nnd 2 in t erfere
destructively (e.g., tl. cres t (tom hole I nr rivc-'i nt t he enme ti me as n trough from bole 2) ; fit
others, constructively. Th is produ ces tbe complicated minima and maxima of the curve
1(,).
2 and measure the chance of arrival at x with only hole 1 open. This
gives the chance P 1 of arrival at x for electrons coming through 1.
The result is given in F ig. 1-2b. Similarly, by closing 1 we find the
chance 1>2 of arrival t hrough hole 2 (Fig. 1-2c).
The sum of these (Fig. 1-2d) clearly docs not agree with the cu rve (0).
Hence, experiment tells us definitely that P ~ P, + P t , or that II
is false.
pnd
P. = I ~ I' (l-J)
1', = I~I I'
.
I t('I"S \\ C shall discuss in detail the actual calculation of
111 later CIPP , I
d A. JlrJ'C we AA""' ouly thnt ~h for example, may >c en culolcd
. 1 8 1) ~ =. J • . 1' f
~n ~U I t iIO JtO f a wave. equat
~,
ion representing waves ep rcnuiug . rom
t he source to 1 and from 1 to r . T his reflects the wave propertlcs or
electrons (or in the case of light, pJ~otons~ . .
To SUIiUllOri Z(": We comp ute th e lI1tcIlSltY .(I .C,., the absolute square
of the amplit ude) of waves which would nrrlv~ . m the nppnm~us nt.x
and then interpret th is inte nsity us the probability that n particle will
arrive at e.
----
Sl,'nllr red h~' t he <'I('..- ( roll :< Il :l~""
init I hrouitl1 11(111' I or h,)I.' :!.
\Yit h II !' t r l1ll jt l i ~h t t'\' IIrt'(' ('\"t'ry
clee t ron i ~ illtlt·(·,1 ftllllu l 10 I\~'-""t
by 0 111' or t lU' lit her h n ll· . BUI
L now t he l) rO""" i l i l ~' of lu ri\ :11
~l
III r j " no IOIl t:.·r it i \ -I' II ,,~. I ll,'
eu rve of Fi l{. (.:!,I. hill i.11I1:oIl·I\,1
A gin 'lI Ly Fi ~ _ 1 ·~.I.
B
(.funll'''''' rr1('("lum i f'1I fUllI I HI". ill'('8"·I,,
8
II
II
I
I'
W u- II r I
t l"t
' 1 Ihu' u('t' cl r o:- l1I ~ ....
su '1 ,II' I-tIro ns "'l,.'nltr n• ng Ii~' hL nt I)...., we get (very ncar.1y
. " I ( np .....-ars at ei t her 1 or 2 n
I ) P
uu w c ca n
,
2 of 1-ilt. 1-2c
H l'pu.rn te
•
We shall state the uncertainty principle IlS Icllowe: Any determi nnticn
of the nlternative taken by a process capable of following more than
one a lternat ive d estroys the interference between ulter nntivcs.
Heisen berg's ori ginal stat ement of the un certai nty prin ciple wa s not
given in the form we have used here. We shnll interrupt our argu-
m ent for a few paragraphs to d iscuss Heisenberg's original st a t ement .
In classical physics a particle ca n be described as movi ng along ll.
defini te traj ectory a nd having, for example, a precise po sition and
velocity at any particular t ime. Such n picture would not lend to t he
odd results that we have seen nrc cha rac terist ic of quantum mecha nics.
Heisen berg 's uncertainty principl e gives t he li mits of a ccuracy of su ch
classical id eas. For example, t he iden that a particle has both a d cfi-
nite posi tion and a d efinite mome nt um has its limit at ions. A real
system {i.c., one obeying quantum mechan ics) looked u pon from a
elnssicul v iew appea rs to be one in which the posit ion or moment um
is not definite, but is un certain . The uncertai nty in position ca n be
reduced by careful mc usuremcut, and other mcnsurcmeuts may ma ke
the momentu m d efi nite. nut, us H eisen berg stated in his principle,
both can not be nccurutcly known simultaneously; th e product of the
uncertainties of 1ll 01 il L' Ut 1l111 an d posit ion involved in nny oxpcriuwn t
ca nn o t be small er Until u n umber with the ord er of mngni t udo of h. t
Thut Much U res ult is req uired by physica l cousist cucy i ll the situntiou
10
. . be l'hown by consider ing s t ill another way
we have M il d ISC:lISSlll g call
. .1 . _ ' c throu...h which holc th e elect ro n passes.
of t ry ing to UCI.·t: rnun eo
~
r..s u m p
IC.Notice
J\
th at if nn electron is d eflected in passing
•
through
I
~
our 0 I IItc J10 J...~, its vertica l componcnt of momentum IS C Ulng c u .
•
Further more, an electron arriving nt t he detector x after passlII g
t lI rOI I"~ h hole 1 is dcncctcd by a d ifferent amount, and. thus •
suffers u
ditT(,r<'llt change in momentum , than an electron ernvmg at x via
hole 2. Suppose thnt the screen Ilt B is not rigidly su~portcd, but is
free to move up nnd down (Fig. 1-.5). Any change 111 the vertical
eomponcut of thc momentum of an elect ron upon passing th rough 8
hole will be accompan ied by an equal a nd opposite change in the
momentu m of th e screen. This change in momentum can be Incas-
urcd by measuring the velocity of the scre en before a nd after the pus-
sage of all electron. Call 6p the d iffere nce in momentum change
between electrons passing through hole 1 or hole 2. Then an unam-
biguous determinat ion of the hole used by a particular elect ron requires
a moment um det en ninnt ion of the screen to an accuracy of better
thnn ~p .
If th e experiment is set up in such a way that the momentum of
screen B can be measured to thc required accuracy, then, since we can
~
:J ---__.. . ~ ~~ --__
---
F1g. 1-5 .An<ltL('r rnod ifir lll io" or tilt' t' . ' .
move \"<'rttrlllJy . H l be ~ltttrOI . , Xpt'rIlJ1('nt o r FJK. I-I . The s creen n is (crt free lo
:~mplC!J,. i~ . b ddl~lt'd uP","af~ 1:~f~~OI:,;. ~1I11 nrri \"('lil at tll~ d ct erlof {li t z _ 0, fUf
l"OUJ;11 \l: JII( h t he elee trcn pa.",-~ 'S C"n1 1.e .1 . t( .n /J ,.llouloi ft'tod duwnward , 'The huh'
6Crttn a l resr m d T ' I l.ll:l t-rlll lllt"11 fOf C I " I
II.eesen
" I I IIcw;u rml; ~· IJl· III (.r it ". .,. a c I !mlWoII KC I,>· l'l1n rt lllK Willi I Ie
x:rg'. unc t ' . Vi f'·(·UJ UrI( up I
tirrcen /J'" I.> I ~~ aUaly p,mt·i,,!f', l,oW~\·I'r 11" I
.. <1(1 u ...: InC OI ",. '. . . , r-
0.' ( (l ~' 11 " rtt'rwnrJ
I Ilrtt.·l'iC II' I
. Af"t o rt lillg 10
nol l. ~ I \.,Lo; e n t \'o"It ll Iln·ur I " k I Olllt'll UUl Illf'lI.:m n ·llIf"n t lil un
l)("b Urc t Ill tth e ~ t r . II t hoW''' 'Keor ·ts . I
gel lhjIJ srueercd I~ n l:r lIle or t !l(' II(j)l '... is r OITl'ttl • l V('rl IClll POSit ion, so we coub
II JlUe III the v~rtklLl If .' ~ L'let. I n lll t'lItJ or / ' (z ) o r Fig . 1_20, we
eecucn, 60 It looks like F ig , 1-2d.
ThejlUldomentol concepts oj quantum mecllonica I
.~
II
determine the hole passed through, we must find that the result ing
distribution of electrons is thnt of curve (d) of Fig, 1-2. T he inter-
Icrcncc pattern of cu rve (a) 1HU!<lt be lost . How can this hnppcn?
~
o
To understand, note that the construction 01 a distribution curve in •!
th e pla ne C requires nn accurat e kn owledge of the vertical position •
of the two holes ill scree n B . T hus we must rnensurc not only t he
mom entum of screen JJ but also its position, 11 the interference pat-
tern of curve (0) is to be esta blished, th e vertical position 01 R must
I,,
be known to nn uecu racy of better than d/ 2, ,... here d is the spacing ,,l
between maxima of the C1U"·C (a). For suppose the vertical posit ion
of R is not known to t his uccumcy; then thc vertical positio n of every 11
point in F ig. 1-2a can not be specified with nn ac curacy greater than
d/2 , since th e zero point of the vert ical scale must be lined up with
some nominal zero point on B. Then t he vnlue of P at any particular
height x must he obtained by averaging over all values within a d is- 1
tance d/2 of x. Clearly, the interference pattern will be smeared out i
by this averaging process. The result ing curvc w ill look like F ig. 1·2d.
The interference pattern in t he original experiment is the sign of a.
wave-like behavior of the electrons. The pattern is the same for an y
wave motion, 8() we may usc the well-known result from the theory of
ligh t d iffraction that the relation bet ween the sepa ration a of the holes,
the distance 1 between screen B and the plane C, the wavelength X of
the light, and d is
(1-4)
(1-5)
~p a (1-0)
11 =::: 1
as shown in Fig. ]-7. It Icllowa that
d=~ (1-7)
~p
• tin d ,nnth illl~6rob
QUOII,,,m n1rrl'lIIl1t:l ...
12
-e
CD
o
...L
®
r: d
1-
I
B C
Fi,:. 1-6 Two beams of light, start ing in pbese at IlOJCS 1 nod 2, will interfere conslnlc.
lin"I)' when tbey reaeh the screen C if the}" tnk e the same t ime t o travel from B to C. This
means t hat a maximum in the dilTr:u:tion pattern for light be am s passing t hrough two hoJes
will ceeur at t he cent er of the screen. As we move down t bc sc rcen , t he next maximum will
occur nt a detanee d, U'~ k h is (ar enough from t he center t hat, in t rnveling to th is point,
the beam from 1I0ie 1 will beve traveled exactly one wavelengt b x Iarther thon the beam
from hole 2.
11-- - - - 1 -- - --1
p
.j
'r~6PI
--
p
B
C
Fi s · 1..7 Th c dcRcct ion afan .-!cel roo ir .
a cbenge in ,?oment um 6p. T hill d um'; P Ill' I'oln /; tl lNltJp;h II hole in the s creen IJ is nd unl!)"
morhl'n t UJll Ul a dirl.'C:tion lIPPrtlJl:imllh .f , lthlount~ to tl, e a d d it io n of n s m al l eompoe cn t of
n Ie c"ll!mpe in ' ,) ,* r'K'Jld lCu la r I II "
.. cnCrr;y III ("om ple'tt ly " '" 0 IC OnK lIltIllIIOhlcn t u lIl ,'('d or.
mObl('n l um vect k net!; IIl:I • c . }"or II I '
,
III rI'IJret;Cllh .d t '
Inlt,,"
or l'('Jlf! t he 1';1Ime I (
II til e 't,. proxi
"'m n ( 'lIet tioh IUJ ~I t'S the lottll
t I ) .. •
frloln hole I ,. ; 1I " ('r)' goo d ll' llJf<lAiltlhl iun b,)' l ~ 1; 1'1l e )' . J hen til e deflec tion 1t1l~lc
rC:l.ea ILl' wu I ~1I0lu cn tu m III lIrK.! Ihe otl .('r ( ~ , 1·1. U h t"o (.. I("(,t ro ut'. one tll:tr ling
IOU!>t difTe;'~"~ point ~D U.e sereen C. tllch til(" '" II lIIR (r(>' ll Ilole 2 wil h moment um p~.
has come . , : aplJrtJxl~'lIItel)' all, [o;illl't. We " ,'n·n" ',n1 tI 'fUugh ",Ilid . till')' were lldlcdf'd
• .", tlll("('rlulIll y ' I 110 t;lly II I '
rocth'~ (on ~. « , ' I III ( .e \'t'rl iea! Cf ' I lI'"(1Ut; I willch hole UII elcetrcn
1'......<lUlg ( lroo I II . . l II lOrl('h t tI(
angl e. Thill.,', I ,RI lel;('",,-,ll lI m ufi t ll(" _ ' I
cs "' nllon I /
PI - P ~ I IPl _ /' II I I
II .' -
I' '
"'llLlv /I ('Ill to
alt.
,
Ili omelitum whid a the (']cd roh
'
II !'; Unccr lft lil ly in dellcetlon
T/lcfllndamrntal conupt. o} quantum mee/lOnic.
13
which agrees (in order of magnitude) with the usual statement of the
uncertainty principl e.
A simila r unnlys is CDn he applied to the previous measuring dev ice
where the scattering of ligh t was used to determ ine through which
hole t he elec t ron passed . Sud } an analysis produces the same lowcr
limit for thc uncertain! ies of measurement.
The uncertainty principle is not " proved " by considering a few such
exper iments. It is only illustrated. The ev idence [or it is of t wo
kinds. Fi rst, no One has yet found any experimental way to defeat
the limitat ions in measurem ents which it implies. Second, t he taws
of quantum mechanics seem to require it if their consistency is to be
maintained, and the predict ion of these laws has been confi rmed again
and again with great precision.
I'
the result}, but this iuform nt.ion is 1I0t used. N eve rt heless in this
case a Bum of pro In )1"'"I' I If'S (iIII the ordi nary se nse) must he carried • out
o'h·~r l exclusi!'c nlle rnntivcs. These exclus ive a lternut ivos nrc theM
w tc 1 could hove been sepa ra I eIy .I( Ieutjficd "
by the informa t ion .
" SOl ~_~ IlIluhtratio~lI. When a lterna t ives cannot possibly IJc
'50 \ h J Jy any cxpcrlll c t tI I " "
tru tioll f II · z.. - 1 n .. ley a ways IIl lt·rfl·rf'. A s triki ng illus-
of- ~ ·'1 . .
o 118 1::0 t he 8ca tlrr llw
. ""
or
I " I '
\\ a lIue CI at 000 ~a y i ll the c('n u 'r-
gravI y "'~tcm, as Illustrilt.t'tl · F. . " .
an a pa rt icle a nti n I III Ig . 1-8. Sup pose ..1 rl·prt·:.lo(·nho
A 18 •
Hentt.erctl to SOlli. e. ot Ie' r Ilucl(·u~. . AS k rUI· t Irc pruhabi"I"It )" IIIII I
P<>OIIOIl I nnd /1 to 2 1'1 " I' ,
tliJl ll(t ,2). TIIf! prnba bilit f I . _. . ro nt upl ituc C I~, SlI) ,
not dist inguish ""hat k:I:Jo~ t us IS P = . 1~A. II(I ,2>1:t. ~UPpO~(' we (~()
nucleus urrlVCt:! at 1 that is wh ether It
• •
Tile fundamental concept. oj quontllm. mecllanin
IS
F iR. 1-8 Scnltcring of one
rfl' nud clltl II)' n n llt )tr f in th e renter-
of-~ mv it y 8ytllcm . T he sc ee-
lcrill" of t wo iden tienl nu clei
A - - - --;- - - -o . • B
~ I IO W H li t rik inJ;in (rr(crcnc(' effects.
T here nrc t "" 0 i n ( c rl(' ri n~ nlte-m n.
t iH"l( hcee. The pnrt icle which
IIrr i \' ~ at I, fillY. can hav e teee
3
~
"r
I
-, or from n. If t he or iginal nuelei
rn~
were not «Ientienl, 1(.'81s of
ident ity at 1 could deter mine
whh-h nltcrnntive had act ually
bee n ta ken, Ml t he}' ere ('ltdusi" c
el tem utives a nd Ille s pn:ia l inlcr_ ~
Icrenee effects do not BJ'iae in -1
UIU case . 1
i
is B or A . If it is B , the amplitude is tb.t1J (2. J) [which eq ua ls t/I.4D (I ,2),
because we have to ken a gOo a ngle}. The cha nce that some nucleus
arrives at 1 and the ot her nt 2 is
(1,9)
•
I
16
is t hen
(I-II )
I~ (I .~) I' + 1 ~(2. 1) 1 · - 2p
ActUD II)'
thls i
-rcng and remarkably, electrons obey n different rule.
IS IS w I • . ' 1 " I I
I
lit Ie fo, nn event in winch t he identity 0 n pmr 0 c cct ron,
T I1(' nmp I ut
"IS rcvcrscc• con tributes lSOIl out of phase. •I' IJut "IS, t h C cese 0 1 be th
spin up gives
I ~A "( 1 .2) - ~A .(2.1) 1· (1-13)
In our C8:iC of 90° scattering 4>AD(1,2) = 4>Ao(2,1), BO this is zero. I
Fermions and Bcsons. This rule of the 1800 phase shift for alter-
natives involving exchange in identity oC electro ns is very odd, nnd
its ultimate reason in nat ure is st ill only im perfectly understood.
Otb cr pa rt icles besides electrons obey lt, Sueh particles a re called
Iennions, and are said to obey Fer mi, or ant isy mmctric, statis tics.
Electrons, protons, neutrons, neut rinos, a nd p. m esons nrc Iermions.
So are compounds of an odd number of these such as 0. nitrogen atom,
which contains seven electrons, seven protons, a nd seve n neutrons.
This 180" rule was firs t stated by Pauli am} is the full quan tum-
r ncehanicnl basis of his exclusion principl e, which cont rols the chnr-
acter of the chemists' periodic table.
Particles for which iutcrch uuge docs not niter the phase nrc called
bosous and arc said to obey Bose, or ay nuuct ricul, sta t ist ics. E xam-
pk-s of boscns are photons, 11' meso ns, uud syst e ms containing an even
number of Fermi particles such as an a part icle, wh ich is two protons
and tw~ ncutrous. All pa rticles ere eit her vile or the other boSOIlS
or rcrl~lons. These interference prop(~rti("S can have profound uud
my1:ilcrlous
. effects . Fa r eXJ.lI UP le, lreI"unu 1-I<IUIO " Illude of o.to ms 0 1
ntcmie mess 4 (boso ) •• _ • .
. ns au U:lll pcrnturL~ of o ne or two deg rees h.c1nn
COn flow wit hou t • I ••
Ii U1id any reststance t hrough sma ll tubes wherea s the
14 made of a toms r .. 3 (Ieru llOus)
. .. ,
T1 . 0 Inns.., docs not have t his property.
re concept of iUl'ntit.y of p lrf I . 1 I fi
nile ' . : re C':I 1.~ ur more com plete a mi ue •
callymh(~~npn ll lull.lI l lllecJII~lUc~ thnl~ it is ill cln."iSicul mechan ics, C lu.o.;si-
, r rc es w rich 6('('111 l(!<·utieul 1_' 1 . ' I r
id('llt ieul for nil ti 1 co u u >c ucnrly ItlcutlCll . 0
pruc ICtl purposes i l tI I be SO
clO!i(,·ly equa l that res . . , I U' ~' Ilsc t lnt they may .
P ent CXpcrllllclltal tec hniques cannot detect any
r',e f,"ltlamenrnl concepta of quantum m ecllaniCIf
I
17
difference. However, the door is left open for some futu re technique
to esta blish t he differr-uce. In quantum mecha nics, however, the
sit ua tion is differen t. \Vl" en n give a di rect test to dete rmine wheth er
or not pa rticles arc co mple tely indistinguisha ble.
If the particles in t he experiment diagrnmed in Fig. 1-8, starting
at A und B, were only approxima te ly identical, then improve mcuta
in experimen tal te chniques would enable us to dcter miue by elOHC
scruti ny of the particle a rriving a t 1, for exa mple, whether it came
from A or B . In this situa tion the ultcmnt ivee of the two init ial
positions mus t be exclusive, end t here must be no interference bet ween
the a mplitudes describing; these alternat ives. Xow t he important
point is that this act or scruti ny would take place af ter the sca ttering
had taken plncc. This menus that t he observa tion could not possibly
affect the scnt tc rlng process, and this ill turn implies tha t we would
expect no interference between th e amplitudes describing the al tcruu-
tivcs (that it is either the pa rticle fro m A or the parti cle Crom B which
a rriv es at 1). 111 this case we must conclude from the un certainty
principle that there is no way, even in principle, to ever d istinguish
between th ese possibilit ies. That is, when a particle arr ives at I, it
is complete ly impossible by a ny test whatsoever, now or in the future,
to det ermine whether the particle started Irom ...l or B. In this more
rigo rous sense of iden ti ty, all electrons a rc ident ical, Il9 nrc all pro-
tons, etc,
As a second exam ple we conside r the scattering of neutrons from 0.
crystal . 'Vhell neutrons of wavelengt h somewhat shorter tha n the
atom ic spacing a rc scattered from the atoms in a crys tal, we get very
st rong interference effects, The neu trons em erge only in ecrtalu d is-
crete directions de ter mined by t he Bragg law of reflection, just as for
X rnys. T hc interfering nltcr uativcs which enter this exampl e nrc
the alternative possibilit ies tha t it is this, or thnt , atom which docs the
acnt te riug or a particular neutron. (The amplitude to sca tter neu-
trons Ircm any atom is so s mnll that we nerd not consider alter natives
in which a neutron is sca t tered by more thnn cue utour.) T he waves
or a mplitude describing th e Illation or n nout rou which stu rt from these
atoms interfere const ruct ively only in cortu iu definite di rect ions.
Now t here is all intcrcs t iug cour plieution whi ch enters t his appur-
ent ly s im ple p iclure. Neu trons , like elect rons, cnrry u sp in, wh ich
call he uunlyaed in t wo s tat es , spi u u p nud :-p in do wn. ~uPJ>()~ t he
scnt tt 'ri uj!; ruaterin l i:ol COlllptl'U-U or uu a t omi c :o>1)t'{' irs wh ich has n s im i-
lar sp in proper ty , I'lilch a s carbou-Ll. In t his C:l Sf' UII f'x !)(' r ilul' u t will
reveal two uppurvu tly difTt'f('ut typ~ Il f scu t toriug. Jt is Ioun d t hut
ht.o- sidL'K t he Hcu tu'rilli; ill discrete tlin' rt iults, ItS described in the pre-
I•
•
I
•
I
. d patl. in'f!Sral~
Quontun1 n1edlfJlIl('$ fJll
18
h . d "lIused sealtering in nil directions. Why
ceding parngroph, t ere 1S a I
should th is be? f tl esc two types of scattering is provided by I
\. clue to t he source 0 1 ihi I
, " b ' (OJ Suppose nil the neutrons W lie 1 enter tll(~
the followmg 0 serve I I . I I ' . di "
. afro in such n manner t hat t rcrr spm U'Cctlon
cxpr n lll(>llt nrc prep . ' 1 •
.}s up. II t I,e ~P "Il1 d"I-~"' 1"L"\.:10
ion of the erncrgmg neutrons IS una y zcd, It
_ • •
will be fou nd t hnt there are some up and .some dow 11, those which
&1"II havc s in p up atel scattered
· only at the d iscrete Bragg angles, while .
those whose spin has been changed to down come out scat tered dil,
{usely in all directions!
1\0 ,. in order thnt a neutron flip its spin from up to down, the law
1 \ • I '
of conservation of angular momentum requnes that t ie SPin of the
scattering nucleus be changed fro m down to up, Therefore, in prin-
eiple the part icular nucleus which was responsible for scattering that
particular neutron could be determined. \ \'e could, in principle, note
down before the experiment the spin state of all the scattering nuclei
in the crystal. Then, after the neutron is scattered, we could rein-
vest igate t he crystal and sec which nucleus had changed its spin
from down to up. If no crystal nucleus underwent such a change in
spin, then neither did the neutron, and we cannot tell from which
nucleus the neutron was actually scattered . In this case the alterna-
tives interfcre and the Bragg lew of scattering re sults.
H, on the other hand, c ue crystal nucleus is found to have changed
spin, t hee we know that th is nucleus did th e scattering. There nee
no interfering alternatives. The spherical waves of a mplitude which
emerge (rom this particular nucleus describe the motion of the scat-
~red neutron, and only t he waves emerging Irom this nucleus enter
Into that description. In this case there is equal likelihood to find
the scatte red neutron comi ng out in filly direction,
. ~he concept of searching through all t he nuclei in a crystal to find
: hich one I~~ cha nged its spin state is surely a ncedle-in-t he-heystack
~ of ac tJ" lt~, but l1?lurc is not concern ed with the practical dim.
~: tres..O hl, ex~rnllclltatICJn . Thc importa nt fact is that in principle it
poSSI e Without producir diIS I urbunee of the scattered ncu-
trc I d O . I Ig any
n 0 etcrntina (iu this Jet te 1 )
which eryst I I . r case w rere the sp in states ChlUlgl'
this POSSihil7t nl~\::~ OJCluully d~d the scattering. The existence or
deten llinatiOl; we ~~: t 10.t even If we do not actually curry out th is
nonil1 terff'rillfl') alter 11.lc\Ocrthcless dealing with exclusive (uud thus
- ... lA: na Ives
i~
for example, the overa ll amplit ud e for a rrival at z is given in Eq .
(1-2) as
( I- H )
'0
A
roO.I.IO More and mo re holes nrc cut in t he 1ll' t(.ocn 8 nl "f) nnd •• ~~~ tu II II
--"-'5 tire romp Icte Iy r
'.'
lu<
"
II C( I " n.t II I10 I('S. nud tile electron Ims n
"nuo
.....nu
s. "., n -n n rY. Ie
'
""''-'-'' "u u. rn nge 0 ~I_
tjons;, up nnd .down alo ng me-It scree n , ,nt ,,:h ie), it ea n ptUi8 th roulI:h th e p'~i(ion or t he
f (l'ff'fl. In t his (:91'(" t he 6u n~ ~r nllcrnn tlves becomes n doulJlc in ll'gr nl over t he con t inu ous
p.'1 r:lmclrrs r o nnd r~ dcsrnblnJ!: the nlt crn nt ivc IICigiata nt which tile elect ron passes the
~it ioD of the screens at lID and JiE .
23
•
•
•
•
1
•
•
•
•
2
The Quantum-mechanical
Law of Motion
I s this ehnpter we' intend to complete our s~('Ci fi.CD.lioll of ~onJ'('lnli.
vistie qunut um nn-ehanics which we 1 J('J!;~U1 III .L hup. 1. f hcre WI!
noted t he existence of nn a mplit ude (or eneh tr:1JI~clory; here we! bhnll
~i\'(' tlu- form of t he a mpl itude for cec h trajectory, For ~ while', !or
simplicity , we shall rc-tr lct ourselves to t he CU;--;C of n partic le !1~OVlIlg
in one dimension. T hus the position nt nny ti me cen be specified hy
a coordinate x, a (un ct ion of t. By the pnth, t hen, we mean a Iunc-
t iou r (/) .
If a particle at an initial time t. starts (rom the point x. e nd goes
to n fi na l point x. a t t ime l ro , we sha ll roy simp ly t hnt t he particle goes
(rom a to b nnd our funct ion .r(t) will ha ve the property that x(/ = z, G)
S - t L(x,z,t) dt (2-1) I
wher:c L. is t he lagrangian for th e system. For 11 part icle of mass In II
movi ng HI a potential l'(ZIt), which iH a fun ction of posi tion a nd time
the lagrangian is ' I
L = ;' x' - V(z ,t) (2-2)
T he form of t he ext rem um path ret) is determined wit h the usual pro- I
cedures of th e calculus of variatious. Thus, suppose th e path is varied I
}6
I
Tile quanrum~mecllonicol la ", oj mOliun
27
uwny from % hy an Amoun t ~%(t) ; the condition that the end points
of r. nrc fixed requires
6>:(1.) = 6>:(4) = 0 (2-3)
The condit ion thnt x he un ext remum of S means
~S = SIx + 6>:] - S(x] = 0 (2-4)
S( 1 + "(aI,+ b 8 1,)
-= x I. I. ~:to:t
ax
til (2-5)
!!:dl (8L)
8:t
_aL
ax
= 0 (2-7)
28
I
r ro II,('1n 2- 2
With T equal to I. -
For a hnnnouic oscillator IJ = (,"12)(zl - Wl,zl .
. . .
I. show that the elusslcnl action IS I
S = m.. 1(,.. + z,') c08 .. T - 2z"%,) (2-D)
tI 2 sll1wT~
Ptob Irnl 2-oJ• F IIId S<f for n particle under 0. constant force F, tha~
is, L = m,i:1f2 - Fe. \
Problem 2-4 Classi cally, the momentum is defined na
•
I
aL (2-10)
I' ~ at
Show that the momentum at an cnd point is
(2-11)
H inl: Consider the effect OIl Eq. (U) of a change in end points.
Problem 2-5 Classically, th e energy is defined as
. E ='L + tp (2-12)
(2-13)
Now we can give the quantu m-mechanical rule. \Vc must say how ,'
much each t rajectory contributes to the tota l amplitude to go From
a to "" It is 1I0t tllat just the particula r path of extreme nr tion COl'· '
tributes ; ruther, it is tho.t all the paths contribute. They ccut ributc
I
equal amounts to til" tolul amplitude, hut contribute nt different .
phases . The phase of the eont ributin u Iro m 0. given path is the ac tion 1
S for thut path in units of the CI UllUlu lII of action ft . Th at i~ , tu 8 \1111- I
rnuriec: The probabilit y }1(",O) to go from 0. poin t r. at th,· time I. to
the point z. at 4 is the absolute sq uare }'(b,a) = IK(b,a) 11 or nil o.mpli:
29
tude K(b,a) to go from a to b, This amplitude is the Bum of eentrl-
butlon 4'Ix(t)J from each path .
K (b,a) = I
ov ..r ,,11 1' 1111,"
<t>[x(l)] (2-14)
f ....m Q 106
The action is thnt for the corresponding clessieel sys tem [sec ECI.
(2-1» . T he const nnt will be chosen to normalize K conveniently, and
it will be taken up later when we d iscuss more mathematically just
what we moan in Eq . (2-14) by n. sum over paths.
30
,
.I
X
.' iF:.2-1 The clnl'tojeRI pnth I, 1 {t), it! t bnt Ior which n ccrlnin intcgrnl, t he ncl ion S , is
minimum. U the puth is vuricd br u (l). to pnl lt 2, the integral 8UfTC1'8 no firsl-ordcr
change. This det ermines til t' equa t ion oC motion.
In qu nntum nL('('l mn n , tilt'! Amplituu e to gc from a to b is t he Bum of amplitudes Icr each
intc rfcrinll; eltem euvc pnlb . The empht udc Ior a giv en pnt h, t ·BI A, has n pbesc proper-
l K>nnl to t he art ieh.
If t ilt' act ion is YCI)· Jargc com pa red to Ii. ne ighboring paths such as 3 nnd 4 have eJighUy
d ifferent actions. Su('h pe t he win Ibeceesc of the smellnesa of M hnve very dirrerent
pheees. Th dr contributions will cancel out. Only in th e vi cinity of tile d nssKtll petb
l {t ). where t he act ion ch n n ~ lillIe when the pnt h varies, will ueigh boring petbe, eceb as
1 and 2, ('()flt rihu tc in t he M inI.' phase eed const ru ctively interfere " Thot. is why the
Dpproximu t ion of d Sb5it"ol pll}"t;i('t;- llml onl)' t he roth 1 (1} need be considered-c-ie vnlid
..ben tb c a clion is very large compared to Ii.
(2-17)
wh ere the summation is carr ied out over th e finite set of points Xi, as
shown in Fig. 2-2.
The next step is to define A as thc limit of this sum as the subset
of points an d thus the subset of ordinate'S, becomes more compl ete
o r-e-beca use 0. finit e set is never any mensurab le part of thc infinite
conti nuum-we may better say us the subse t be comes more represent-
ative of the complete set . ' Ve can pass to t he limit in all orderly
mnu ncr by taking conti nuully smaller nnd sruu llcr values of 11 . 11\ so
doing, we would obtain a d ifferent ~UIU (or (' Ildl value of h. No I h~, i t
exists. I II order to oh tu in n limit t o t his process, we IU U~t specily
HOl lie normnlising Inetor which should depeud o n h. Of cou rse, for
the Hicmnnn iutcgrnl, t hiAfacto r is j U!4l h i L~dC. Now the limit exis ts
QunnWIII n1fO('1lRtlIC'lf ond pRII. i'.'f'Il Nlb
32
/
h
} (
I
.lo x. Xz I ,
"Y ',-t-' K
Fi,.2.2 In the definition of t he ordinary Riemnnn int('~rl~ l , n set of ord inntee ~ drown
(rom the a bscissa to th e curve . The ord innt es lite spaeed n d ts t n n~t> h upnr t. The IhtCKrnI
(artll between the curve nnd tile nbscissn) is AI)proximAtn l by h IIIII CS th o s um or the ordi-
nat es. Th is npproximntion nppr onclJ<'S t ilt> C'O~t('d value ll~ 11 " PIJrollChcs ~ero •
•.o\n analogous definit ion can be used ror pnth IhtCJ;.tllls, 11 1 ~ III C'lI 1;Ure wh ich goes to zero
in tbc limit precess is t he time inleTVnl t bet ween discrete pomte on the petbe.
(2-18)
A = (2m~,)" (2-21)
'"
We sha ll sec la ter (e.g., Sec. 4-1) how this result is obtained. With
t his factor the limit exists Bud we may write
where
Slb,a] = f.... L(x,%,t) dt (2-23)
is e line integral taken ove r the traj ectory passing through t he points
Xi wit h straight sect ions ill between, as in Fig. 2-3.
'.
I trs · possnue '11 t 0(,-loflnc 111C path in n somewhnt • .. more elcgnnt . manner,.
'1 '
I Ill ' l i lt I of st rni..htIi ucs 1)('1wI'('11 the 1)()lIll~ t und , -I- I, wo could use
l' ... I 0 ' II ' .
sections or rill' rlll ~... i('lli or bit . Then we could My l ra t .J 18 U~ 1l1l.1iI~
mum \"11 Ilit' ef II.c. - " III'IP
l"'
rul. of the Ingmnginu .
over all the paths which
. , •
The Path Integrul, There nrc many ways to define n subset of all
the paths bet ween a RlH] b, The p"rticllhl~ definition we have used ,I
here 1119 ,)' not be t he best for 5-0IH ~ ma t homnt ica l PUfJ)OSC13: For cxnrn- I
pic, suppose the lngruugjnn depends upon the nc~clc~tl~n of:, In
the way we h3YC constructed t he path, the velocity IS discontinuous
at the various points (ro,l;) ; thnt is, the ucceleratiou is infinite at these
points. It is possible that t his sit uat ion ,,"DuM lend to t rouble. Ilow-
ever , in the few such examples with which we have had experience thc
substitution
i =
•~ (X,H - 2x j + Xi_I) (2-24)
has been adequate. There may be other cases where no such substl-
tut ion is available or adequate, and thc present definition of a sum
over nil paths is just too awkward to usc. Such n situation arises in
ord inary integra t ion in which the Rieman n definition, as in Eq. (2~18),
is not adequate 811d recourse must be had to some other definition,
such ns thnt of t he Lebesgue .
T he necessity to red efine the method of integration docs not destroy
the COIlC<'pt of iutcgratiou. So we feel that t he possible awkwardness
of the special d efinit ion of the sum over all paths las given in Eq.
(2-22)] 1UUY eventually requ ire new definitions to be formulated.
Nevertheles s, the COIlC('pt of the sum over nil paths, like the concept
of nil crdinury inbogrnl, is indepcnc.J £'l lt of n special d efinition and valid
ill spite of the failure of such dcfillitioJls. Thug we sha ll write the sum
Over nil pat hs in a Jess rest rict ive notation us
wh ich ~ve s!lnll call ~ pull, irllc(Jrul. The identifying notntion in this
C.~Pres:'IOIl Is the scnpt ar, Only rarely shall we return to the Iorm
given III Eq , (2-22),
P,fJ',lena 2-6 The class of functiollals for which path integ rals can
be d efined is surprisingly varied, 1:50 far we have considered Iunc-
The quontum-mechanic-al laUl oJ morion
35
'1".
~
•
,~ Fill. 24 The path of a relativistic pnrticle
lrnvl'lin$C ill two dim cnJolion!'l i!'l a t igzap; of
" f
8t rail(bt flf'J!:rnclllJl. TI le s tope of t ile ,.t-'Jl:~
mente is ec nstnnt in rungnit ud e Anll di ffcn
only in Ili p;n h um xiI{ 1(1 Zflp::. TIle Am plitude
for n pnrticull1r pAth, ns well M Ill!:! kernel to
I,
go from a to 1,. depends on t he nu mber of
"t corners Jl Klonp; n path. as shown by Eq..
(2-20) end (2-27 ).
I
•
•
•t t ionals such as that given in Eq. (2-15). Here we shall consider (Illite
n different type. This latter type DC functional ar ises in u one-d imcn-
sionnl relat ivistic pro blem. Suppose a particle moving in onc dim en-
sion can go only forward or backward at the velocity of light. For
convenience, lye shall define the u nits such that the velocity of light,
the mass of the particle, nnd Planck's constant nrc nil unity. Then
in the xt plane 0.11 trajectories shuttle ba ck and forth with slopes of
+ 45°, as in Fig. 2-4. The amplitude for such a path can be defined
as follows : Suppose t ime is divided into small equal ste ps of length e.
Suppose reversal s of pnth direction can occur only ut the boundaries
of these steps, i.e., at t = t.. + ae, where n is an integer. For this
relativistic problem the amplitude to go along such a path is d ifferent
from the amplitude defined in Eq. (2-15). The correct definition for
the present case is
X (b a) =
,
f'.'4
f X (b,c)l\(c ,rI)l\(d,a) dr, dr, (2-32)
K(i + I,') = ~ exp [~ L ("'+1,- x,, xO+'/ x,, 1,+1/ I.)] (2~)
which is correct to first order in E. Then by the rules for multiplying
a mplitude of events which occur successively in time, we have
N -I
4>[r(l)) = lim n K(i + I,,)
_0 i - O
(~5)
for the amplitude of a complete path. Then, using thc rule that nmpli- j
tudes for altemafive paths add, we arrive at a d efinition for K(b,a ).
It can 00 seen that the resu lting expression is actually the same 85
Eq . (2-22).
I
I
2-6 SO~IE RE~IAnKS I
I
In the rdntivihtic tJ ~('ory of the elcctron we shall not find it possi ble
t~ expres s the alll phluu(~ for a path as CI~/", or ill a lly other simple
way . JIowc\'cr, the Inws for eombi uiug a mpli1udes still work (wah
The quantum.mechanlcal'alD oj motion
39
some small modifications). The amplitude for a trajectory still exists.
As a matter of Iect, it is still given by E£I. (2-3;1) . The only differ-
ence is that K(i + J, 1.') is not so easily expressed ill a relativistic theory
as it is in Eq. (2·34). The complications ar ise from the necessity to
consider spin and the possibility of the production of pairs of electrons
and positrons.
In nonrelntivistic systems with a larger number of variables, and
even in the qunntum theory of electromagnetic field, not only do the
laws for combining amplitudes still hold but the amplitude itself Iol-
lows the rules set down in this chapter. That is, each motion of a
variable has nn muplitude whose phase is I /A times the action associ..
nted with the motion.
'Vc shall take up these more complicated examples in Inter eheptera,
\
I
l
3
Developing the Concepts
with Special Elxarnpfes
I
I
IN this chapter we shnll develop the keruela governing some Elpccinl
types of mot ion. W(' t'lmlJ explore the pbysicn l m~nni l~K of. t,he muthe..
matirtll results in ord er to de velop some physicn l intuition nhout
motion under qu uutum-nwchun icni laws. The wave function will be
introduced nnrl its relat ion to the kernel will be described. 'This rep,
resents the first step in conn ect ing the present approach to (Iuantum
mechanics with the more tmditiounl nppreaehcs.
We shall al so introduce some specinl mathematical methods for com.
put ing the sum over nil puths. The ideo. of a SUlII over all paths WDS
described in Chop. 2 with the help of a purticular computational
met hod. Although that method may clarify the concept, it is an awk,
ward tool wit h which to work . T he simpler methods to be int roduced
in this chapter will be of great usc in our futuro work.
Thus the present chapter hes three purposes : deepening our under-
standing of quautum-mcchnnical principles, beginning the connection
bet ween our present approach a nd other approaches, and introducing
some useful mathcmntical rnethods,
~hus. with the help of Eqs. (2-21) to (2-23) the kernel for n free par'
ticle 15
I
. N
I
K(h,e) = ~~ If . . • Jexp [~;". ,-II (XI - x",)']
I
dXI • • • dXN ,1
(?' ) '"
-::" ' . (3-2)
I,
I
This rep resents a set of gnussi • tc I · •
f[cxp ( _ 2») dx J[ Ian2 In gru a, r.e., iutegrnls of the fonn
gaussjnn ~ s . . or e~p (-cu + b.l)1 dx. Since t he integral of"
ga m a ga ussian we may -,.,
r, . .
. II , . Plo o.u I I" re U1t e~ru.tlons
.
011 one
varia) 0 after the other Aft the ] .
limit m b tnk '. or . re mtcgratlOlls nrc completed the
uy e en. 1 he result 18 '
K(h,e) = l2.-it.(t. -
III
I.)]"" im(x. - x.),
exp 2 h(1. _ I ) (3-3)
42 •
De l'eloping tile concep ts wid.. lfpecial e xa lll/lle.
(
2riA• 3')- "
m
[ m
exp 2ifl. ' 3E (X1 - XO)1
]
I~ this way a recursion process is established which, after n - 1 steps,
gives
(
2riAn,) - "
m
[ m
exp 2i fl.. nE (z,. - XCI)!
]
Since ne = I. - to, it is easy to see that the result after N - 1 steps
is ident ical with Eq . (3-3) .
T hcre is an al t ernative p rocedure. Equation (3-1 ) can be used to
integra te over nil the variables x. for which i is odd (assuming N is
even) . The result is an expression Iorrnally like Eq. (3-2), but with
half M many varia bles of integratio n. T he remain ing variables arc
defined a t poi nts in ti me spaced Do distance 2E apart . Hence, at least
in t he case t hat N is of the form 2· , Eq. (3-3) results from k steps of
this kind .
•
I
\I
R/(K)
Fil:_ ),1 TIle ft'fll p:lrl of amp lit ude 10 arrive at vnrious d isl tlnC~ z from the origin after II
ti me I . TI le i nm ~i n :t. r)· pnrt (nol l'JlOwn) is an d nllloJ:Oll~ wave ~ out o f plUllK", 110 thnt l h4!:
absolute sqU:lo rt' of t he amplitude is eonst en t. . TIle WR\·c!rn.-:t." IS 8h or~ nt larg e x , nnmely.
"'bcre . dtL.~ ic.:r. 1 JlllI"t icic rou k! nnl" C only if It moved with h l~1t \'('IO<'lIy . Generally, t he
wan k ngtb en d etessieal momentum are im'crscly rf'lalctl lEq . (3-10».
Fro m u clnssieal point of view a particle which moves from the origin
to x in t he t ime interval I has n velocity z l l and a momentum mxl L
From the qu nntu m-meehauical point of view, when t he motion can
be ad equately descr ibed hy assign ing u c1n...sical moment um to the
purt.icle of 'P = mIl l, then the amplitude varies in space with the
wavelength
x- -ph (3-10)
'Vc may show this relation still more generally. Suppose we have
some large piece of apparatus, such as B magneti c analyzer, which is
supposed to bring particl es of a given momentum p to a given point.
' Ve shall show that, whenever tile apparatus is la rge enough that clas-
sical physics offers a good approximation, th en the amplitude for a
particle to arrive at the prescribed point varies rapidly in space wit h
a wavelength equal to hlp. For as we have seen, in such a sit uation,
the kernel is approximated by
(3-12)
•
I
Qunnllu" mrrhonic'. and lH1th inlelml.
46
1(KJ
m
'" ~ 2fl
(x)'
I
(3-14)
Since m(x/ t)' / 2 is the classical energy of a free particle, this equation
says
Energy = '"" (3-15)
This relation, like the one relating momentum and wavelength,
holds for any apparatus wh ich cen be adequately described hy classical
physics ; and, like the previous relation, it can be obtained from a more
general argument.
Referring to Eel. (3.11), any variation of the time 1&of an end point
will cause a mpid oscillation or tho kernel. The resulting frequency is
1 as"
'" = " iit (3-16)
(3-17)
Developing tile concepts tdtll special C%anrplf'B
47
In this way th~ concepts of momentum and energy arc extended to
qua ntum mcchnnice hy the following rules:
'Ve have just, shown that this ru le will agree with the definition of
energy and moment um in the classical limit.
Problem .1-2 Show by substitution that the free-particle kernel
K(b,a) satisfies the ditTercntial equation
h oK(b,a) _ h' a'K(b ,a)
- i a 4 - - 2m a",'- (3-18)
,
i
QUlIn""U rurr/lRnirs arul,m'h in'f'INIlll
III
TT
t
I
I
I 2b
I
T ...j r
1 I
'. ,
nF:. 3-3 A pnrtidt' sla rtinf1; nt r - 0 when f 0 is determined ~ pese bet~-cen x, - II
co
and z~ + b nt f - T. WI' ,,·k;11to rnleuletc th e probahility of find mg t he pa rtlrlc at some
reint x at a l ime r Inler, i.c., when f ... 7' +
8('(' According t o elcss lr nl tnwe, th e pnrticle
T.
would hevc 10 be betw een z J.TIT ) + 6( 1 + TIT) a nd r rJ.,TIT) - /,(1 + TIT ), that is.
between th e rectilinear t'). ( ('~ion" of th e original ..lit. However, qunn tu ln-lIlcchn nical
hun; she w Ihal such particles have a nonzero probab ility 01 ap pearing outside tb ese
l'1s......o;iC'al limits.
We cannot lloppronc:)1 th le problem by a single application of tile Irce-part ld c law 01
motion, since the particl e ili nrlually const rained by the slit. So we b reak th e problem up
into two successive Irec-perticle mol ions. The fi rst takes th e particle Irom x = 0 at l ",. 0
10 z - z~ + Y at I - T, where 111'1 < b. The second takes the pnrticle from Xo + 11 at
• - T to:r at' "" T +.... The ovcrarl amplit ude is an integral of th e product of tb eee two
frec-p4rt id c kerne ls, as ehc ...·n by Eq . (~H ).
49
event is t ha t t he particle proceeds from the slit to the point x further on.
'I'hc slit hns u fi nite width, and pussngc tilrough eneh elcmcntnl inte rval
of the slit represent s nil ultcruufive wny of proceed ing along t he com-
plcte path. T hus we must integrate ovr -r t he width of th e slit. All
pnrtk h's which mise t hc slit arc captured and removed from th e
f'xpc'rimcnt, Ampl itudes for such pnt ticles nrc IIOt included . All the
particles which get t hrough the slit I HOVC ns free particles with kernels
given by E q. (3-3). Thus the a mpl itude is
where
This fu nction has the sha pe shown ill Fig. 3-4: T h.c rffccl.;l-C u·idth of
such a curve is related to th e pnnn uetcr b. 1-or this part ~cubr func-
tion, approxima tely t wo-thi rds of t he nrcn under the curve 11('s be t ween
-b . I<d +b. . . .
" '(' do not know how to dvxiJ!:1I mctnl purts for ou r nungmnry ('x.llt·n .
ruent which . will
, product' sue II n gauss' "I u •~ ti t. Il owvvor, t here I ~ uo
•I
i
so
GI,I
(3-23) I
0t Iexp [im
J
II-(x) = t:
_ . 2>rih m 2f, (x,
-;: + x')
;.
+ imh
(_ r:l: + x7'.) Y + (im im - 2b'I) u:-J)rly
2"~ + 2h7' (3-24)
t: [exp (ax' + /lx» ) rlx = ~ "a exp - :~ for ne (a) <0 (3-25)
Sl
The classical velocity to get from the origin to the center of the slit
is t o - xol T . Whr ll we usc t his as u substit ution a nd rearrange some
or t he termss, th e expression for tile amplitude becomes
'..,.
.
'I
!/-(x) = ~2:fj (T+ + ~~2)-}i T TT
m
P(x) dx = 2w11T h
Ax exp
[-(X - V'')'] dx
(t.x)' (3·28)
2',
Xo x
Flg.3-S TIle rw l 1l8 01 pnrlid~ movinK th rough a JtIl.ussl~n . !in~ . ]( T lb
+eY' part " • ohf")"ro,
icles vc t he
. ,- . -, . ( 1it'll'2l Al l llnl! woo" ne
d W1P!lf'" al la \\-II tll llIol~lfI . t llt'n t h e d llltn m llon U pur l' 1 I I,·, ou t
. . 7' ' 1 n r n:hrr woo u )(' Oil )' II ~ lln::U I';
I!llrne form "J'I the t1i"lri1JUhOl~ II I Illite • 1(' t I. \ . ,,;jllt h uf the d b tr ihulillli Wtlllit l he
" lI)pm l iu fl ral lo lIw t i llll: o( Ih ~ h t . ' 110 d lltrnr h"l"l:oh I . I •. .1
Illr +
. n 'lUfl'd (ro lll 2-'. to :!b l, IV1Il'r e h , - u' (T .)IT , . Fur q tl:t n lu lll-m ~' I K nlCtl 1110 toll, IU
IIcl uuJ Illlrt'mlilig ill greu h~r li l/Ul tliUl.
52
. -~lrtl the tli:4 rihution is B gnussinn centered nhont the
As wc expec , I r I .
IlOlIl ' - ' • However, till! Hpn~.tll o t rc d l1'\ tril..,.
. I I I = "t,T of 1'...-'1. ('1·')'1) •..
tiou ,u, i~ l:uW'c limn t he expected vuluc b1 of g ' l· (:J ~2:J) . T hi!i can
he interpret ed i ll th l' full()w i ll~ IlUU UI ('C. SI1P I'~e U I and al n~c ~""o
indt'pl.ntlent rnudom vn riublcs whose ~t-lII en.II-:S(l uurc tl~~UltIOIIS
ab out t heir n\"t'rn~c va lues nrc res pectively 0'1 and 0'2· 1 hen if
a = a + at th o ruts d eviation of al ubout its uv crngc val ue is
0 : = (~11 + ~,~) h. ~ O\\', th e rms deviation in u purt.icular d istri.
burien is u measure of t he spread , or width, of the distribution. As
1f1b1
a muller of fact, for t he gaussian d istribut ion e- · t ile rms vol ue is b.
Th us in t he present case we find t hat the q uuutum-meeha nienl sys-
tem nets as if it hnd un extra random variable X l whose rms d eviation is 1
~T
lui = - (3.:JO)
mb
It is th is ext ra deviation Ax l , or spread ing; rather than the appa rent
ext ra variable I I , which has p hysical sig nifica nce . That t his k rill is
qu auturu-mcchani cal in nature is clea r Irom th e inclusion of the con-
stant fl. Such a term is important for particles of s mall mass and for
narrow slits.
T hus quantu m mechan ics tells us that for small p articles, passage
through a narrow slit makes the future posi tion uncertain . T his
uncertai nty .0.%'1 is proportional to t he t ime interval T bet ween passage
throug h th e slit null the next observation of posit ion. If we int rod uce
the classical notion of velocity I we enn say t hat passage through a slit
C3 tL')CS a velocity uncertainty whose size is
~
60 = - (3-31)
mb
53
Th is m ust be the proba bility that th e particle gets through the slit,
since th e integration includes th ose particl es and only those particles
which did get t hrough. Hut we have another way of obtaining this
result . Suppose we take th e ab solute square of the kernel K (x o + y,
T ; 0, 0) , wh ich com prises the second half of t he integrand in Eq, (3-20).
Th is is just the probability per unit distance t hat the part icle arrives
at the point %0 + Y in til e slit. Thnt is,
P(r , '" dy
+ y) dy = 2rAT (3-34)
-
'"
fI This probability is independent of the position along the slit. Thus,
if we were to multiply it by the width of this slit, we would obtain the
" total probability for the particle to arrive at the slit . This implies
that the effective widt h of our gaussian slit is V; b. Had we used
,
" the origi nal sharp-edged slit, we would find the effective wid th to be 26.
•0: Problem 3-3 Dy squari ng the amplitude given in Eq. (3-20) and
then integra li ng over a, show that. the pro bability of passage through
the original slit is
will appear. This is the integral representalion of the Dime 6fu nction
of a. t
Thus the qunntum-mcehunicul results agree with the idea thnt the
probability t hat a part icle goes through a slit is cqunJ to the p robe-
~
," biJity that the pnrt icle arrives at the slit .
~ Momentum UI)(.I En ergy. Next we shall verily again that, when
"
~'
the moment um is definite, th e am plit ude vu rics Il S e'''' . 'Vc return to
t Sec tho tII,Me vC in l4'p;rllll4 in till' AI'pl'mlix a nd I.. J. ~hilT. " QUlUIlu m
'I I . " •• W ."' · J McGraw.lHU lloo k Company, N~' Yurk o 1t156.
, - -, PII. UV-oI-
j l (.0(: 1n.IIlOl, ~ l....,... .
51
. stml • of the ampl itud e ~ivr.n in ~CJ ' (3-20). ThiH time. lYe
n dctmlN.l . ) 1'1 ' . 'III our l'x l)('r nncn l so thu,t the pnrllf:le
I lit· to urrungo coni 111m.
S IU • f) . II 0 1l1.11 the slit is known nstLccumb'ly M pM"Ii lJlc.
,,1."10('11\' ark r p·l..... III~ rr l'" ' .1 I' II
, ' 'f •
U
• unnlUm-ll11'chanicnl COI IS lu (' ru lOllS, tore i,
QUltC' npnrt rom n ) (I • F . _
n elassicnl Ullc('ltninLy of biT in the ,'(·I~c~ty. I ' 0' , nn~ glvTcn Rilt
, - k thi unccrtninty Il('ghglhl c )y C roosmg very
width we cun run c IlS U t U
large. \\' c can also make Xo cx.tr<'mcly hlr~c ~ III io averoge
l IT - .10CS not go to O. COllslllcrlllg Vo and T us con,
vc octty ze - 1' 0 U • I' 'I fl T-
. - 0"011 for the nn1plitudc III t he Hili 0 urge Is
stant ~ I tb c express
const [ imZ! mt(x - TVo)' ] (3-37)
~(x) ce -VI + ."Iimb' cxp 2". + 4"'.'(imI2". 1/2b')
Next we must arrange that the quantum-mechanical uncertainty in
momentum Illb is very small. That is, we take b v ery large, so we can
neglect l ib!. Thcn the amplitude can be written as
We not ice that this is a wave of definite wave number k = pIA. Fur-
thermore, it hDS a definite Iroqu oncy (at = pl/2mli . This means we
can say that a free part icle of momentum p has n definite q uantum·
meeheuical encrgy (dcfiucd es A times Irequcncy) which is p!/2m just
as in classical mec ha nics.
The pro bability of arriving at any particular x, which is propor-
t ional to the square of the ampl itude, is in this ease independ ent of L .
Thus exact knowledge of velocity means no know ledge of position.
III arruugiug the experiment to give un uecurutc ly kn own velocity we
have lost our chances for an accumte prediction of position. ,Yc heve
a lready 6 C'Ctl that th e reverse i H true. T he existence of the qunntue'"
mechanical spread, inversely proportionnl to the slit with 21,. i1upliCS
that .an exact knowledge of position prec lud es uny knowl edge ~
velocity. So, if you know where it is, you ca nnot say how Iust it IS
~?i~ lg ; n~d ".if you know how fa st it is J;oiuJ!:. you cannot suy where it ls.
lilts again illustrates the uncertainty principle.
55
k
,,
3 06
r-,
,
"\
2 PM 04
r-,
PM
I o2
r-,
\: I"
0
: '- 0
'<,
o 1
,2 3 4 o ,I 2
(rJJ {bJ
,
0.045
•I
l-,
Q04 3
P(xJ
004I
"
'\
0.039
1\
1'\
7 I 2
x
Quont''''. lUrc/ldni,.. and ,mil' i.. rcItNJ "
56
This dist rihut.ioll is expressed hy
where
r - T1', - + ,IT)
b(t x - Ttl, +
/>(1 + ,IT) (3-41)
II, - (T",l mHI + , IT) II, = (T",/JIl)(1 + ' IT)
a nd C(ll ) and Sell) a re the real and imnginnry parts of the F resnel
integral. The fi rst factor in this probability d istribution is identical
to th e probnbility distribution of u Ircc particle given in E q .: (;t-{j),
T he remaining fact or contains a combination of real Bud imaginary
Fresnel Int egrals.' It is th is factor which is respon sibl e for the veri-
atious shown in th e curves of Fig , 3-0.
Thus for both slits th e geucm l result is the same. The most prob-
able place to find th e particle is within the classical projection of the
slit. Beyond th is t here is UJ(!' qua nt um-mech a nical spreading.
We have treat ed this problem as if it were a combination of two
ecpcrute motions. First the particle goes to thc slit , and then it goes
Irom the slit to the point of observation. T he motion seems al most
disjoint ed at th e slit . It migM be asked th en, how does a particle
with such a disjointed motion "remember" its velocity nnd bend in
the gcneml direction predicted by classical physies f Or, to put it
another way, how docs making the slit na rrower cause n ' 11055 of
memory" until, in th e limit, all velocities are eq ually likely for the
particle?
To unders tand t his, let us investigate the am plitude t o arrive ot the
slit. This is just th e free-pa rticle amplitude given by Eq. (3-:J), with
Xv = to = 0. x" = X II + .'/ . and t" = 1'. As we move across the slit (vary
y), both real and inmginnry pa rts of the amplitud e vury slnusoidully.
As WC' have seen, the wuveh-ngth of this variat ion is connected to the
momentum Irdt·r to Eel, (a- lO)J. The subsequent motio n is n ",soH
of opt ical-like int erfcrenec muong these wan'S. T he interfere nce is
~Ilstru ct ivc in t he ~C'III'rul direct ion pred icted by elass icnl phys ics and.
III genera l, destruc tiv e i ll other dirvet iona.
.If there are J~U\Il~ wn\'('II' Il ~UI!; ucross t he slit (i.e., the slit is "Cf)'
wide) t1le. f('stJ l tlll ~ IIl l~' rrf~f('IlCe pnttr-rn is quite »hurp a nd lIll' motion
IS approximately classieul. But suppose the elit ili made so narrow
. ' n f'(l'r to ~, :I-I of Eo J rllankf' and F, EnJde. " TuLICti uf Funcficne," Dever l'ultlica-
IlonM. ' nc.• New York , 1tl43.
Deteloping the concept. with ..pedal cxomplcil
57
thnt 1l0~ f'V~" one wh.ole wSl.v.clml1;th is included. There a rc 110 longer
a ny osciilntlons t o J!:lVC all intr-rfereuce, uud velocity information is
lost. Thus in th e limit n.~ the slit widt h goes to zero all velocities arc
~ CtlUally likely for t he particle.
58
at. n p.'l.rti('ulnr time, th en we enn ctllculnt.c everything that can happeu
to th nl pnrticle a lter t hnt Iir uo . All of 11I~torY 'M .(·lTcc ~ upo n the fut:u rc
of the universe could be obtained from 11 sm gll! gigantic wave fun cbon .
" ' e a rc finished wit h the physical portion or this chapter, a nd we now
proceed to mnthcmnt ieal eonsidcra tlons. \Vc sha ll introduce some
additional mat hema ti cal techniques which will help us to compute
t he sum over paths ill certain sit ua t ions.
Thc simplest path integrals n rc t hose in which nil of t he variables
appea r up to t he second degree in e n expo nent. \Vc shall call tho n
gaussian integra ls. In quantum mechanics t his corresponds to n cesc
in which the a ction S invo lves the path x(i) up t o a nd incl udi ng t he
second power.
T o illust ra te how t he met hod work s in such a case, consider n pur-
ticle whose lugrangian bas the Ionu
L = a(t),;' + b(I),;z + « t)x' + " (I)'; + c(l)x + f (l ) (3-44)
The action is t he integral of t his function wit h respect to time be twc<'11
t wo fixed end poi nts . Actually , in th is form th c lagra ngian is a Iitl le
mor e general t huu necessa ry . The fa ctor .t could be removed from
th~ Icr.llls i~1 wb ieh it is linea r t hroug h a n integra t ion hy putts, but
t his Iaet IS unu nportuut for our p resent purpotic. We' wish to d ctcnnlse
t
b
j
~
x
I . I t h f (t) nnd some ~ih k! 1Io !t('mnt il"e
I Fig. 3-1 The d ifference hcLween the d afl.!l't'n, I'" .t I II, , ....ch the Rune Nl<1 l-li nts,
• !o: ' t he ..... t UI mus XI ~ ..
JWlth d t) ill t he fn nf"t 1Ofl y( f ). • !nrc' t ·· · ) I ' c env ' (lr lll ~i nN' t he rL\.~irn.l
I If (l .) - 1/(4) .... II. III hct wecn t IIl'toC 1"11 IX} ·
d illt !'l N(' ('li n IlJ<
ti
• •
t h r (/) r.s ,..... u iv al cn t (0 t h l" esso-
• 1 ' t" ·n t he altl." mll l\'Cptt ' I .
pnth III co DlJllc tcly fiu~, • nllY vnrln 111111 . th ·n l". , u 1 t1\4' ,>n th lIil1l'n'llh:tlll.r(l)
"(, . .
"lilted v nr iu tioll ill Ih (' , III1I'rt'II(' (' V '
(I)
"' ) +
T I IlI!l III II I Il I I
(0
. •
III (I,is ro rlll i'{l) is n ,"tllll,(llIlL
ea n lie rl."llIl"m l IIy 1Jy (l J, nlll i tl w l"Ilil
. . ovr-r 1m IIIS. {'
;reo hy 1(,' ,II. I~ll h Vllr iuh k fltt) i8 rl'lIlril'l l"l1 In
"'urllJ{'rIllO"~ " I It't iIII "I 1, t" 1\ nnth "Illh'j!;rnlllll . I" 1)('11\ I('II'
fI,r tile mll'j( f!. llfln
. '11 " I IlI1Itu 1011 e ll l. ,.
b kl' t h C" VIIIIIC n lit hll th c u d ))()U1 111. 1111 till
lIf I'lkl-poillL Jl0f<itiolllJ.
At ('nrh I th e vn rinbles z nnd ll d ifT('f hy the cOlu~tn~l t.f. (O f COUNl c,
th is is n d ilTl·rt·nt constant Ior each vnluc of I.) Therefore, c)('lu ly,
tI.ri := ,fYi for each specific point I , in the subdivisio n of time. hi gen•
eral , W. lIlay sny :Dx(/) - :Dy(I). .
Thc integral for t he action can be written
srx(t) ] = sr.i (I) + y (llI = I.:' (a(I)(;!' + 21ft + Ii' ) + . ..J dt (3-48)
I£ nil the terms which do not iuvolvc 11 arc collected, thc TCsultirlg
int egral is just S{i (t) ] := Sci . H nil t he terms which contai n y as 0.
linear fact or nrc collected, the resulting integral vanishes. T his could
be proved by ac tually carrying out the integration (some integrution
by parts would he involved} ; however, such a caleulet ton is un ncces,
sary, since \\'0 already know the result is true. T he fun ction i (l ) is
determined by t his very req uirement. That is, .f is so chosen that
there is 110 change in 8, to fi rst order, for variations of the path around i .
All that rema in are t he second-o rder ter ms in y. Th ese can be easily
picked out, 50 that we can write
The integral over paths do cs not depcnd upon the classical path, so
the kernel can be written
Ult~grnl over paths ca n be a fun ction only of t imes at the end points,
'I'his means that thc kernel can be written as
(3-51)
~ K is determined except for a. fun ction of I.. and I". In part icular,
I ts d ependen ce UpOIl tile spatial vnriables r " and Xli iJi cmuph-t ely worked
out. ~t should he ,noll'd t hat th e dcpcnd t'llee of til e kernel Up Ol1 till'
C?CffiC I£> l1t.S of the lincnr t erms rI(l ) a ut! c(t) and the remnining eoctli-
cicnt J(I) IS also compl etely worked out.
• T his ~I11S to be churacterist ie of various met hods of doi ng path
mtegral!!; a grea t deal ca n be worked out by som e general met hods-
Del'e loping LI,e concept, with special e~tJlnlJ'e,.
61
bu t often n multiplying Inetor is not Iully determined. It must be
determined hy SO Ul f! other known property of t he solut ion, ns, for
exnmpk-, hy Eq . (2-31).
It is i l1 t('rC'~tillg to note that the approximate expression K"" e's.J'"
is exact (or th e case that S is n qundtutic form.
The result is t rue in situa tions other tha n the classlc nl limlting case.
For exumple, suppose 1" is n quadratic Iuu etion of x. T hen th t'! solu-
t ion is exact, since t he expansion of V ~ in Efl. (3-57) contains no
n. poWCN higher tlrun the second. Some exa mp les of this type arc
given i ll the problems. As a nother exa mple, !lUp p OSC I" is a slowly
vary ing Iuuction. I n pa rticular, if the thi rd and higher deriva t ives
~
nrc extremely small, the result given a bove is 11 very accura te
approximation. T his par ticular case is calico the WJ{D a pproxima-
•• tion in quantum meehn nies.
There arc other situa tions in wh ich the approximation is good.
Suppose the total time in terval for the motion is very sh ort. If a
parti cle moves along a path d irT"riug greatly from th e classical pa th,
it must have a v ery la rge extra velocity (t o go out from th e init ial
poin t a nd the n return to th e finnl point in t he allotted t ime interval) .
The extra kinetic en ergy is proportional to t he sq ua re of this large
veloc ity, and t he act ion contains a term roughly propo rt ional to the
kin etic energy m ult iplied by t he time interval (thus, the square of the
v elocity mu lti plied by the ti me interval). The action for such path s
will be very large, a nd t he phase of th e a mplitude will vary greatly
for closely neighboring paths. I n t his case aga in it is reaso na ble to
, drop tbe higher-ord er terms in the expansion of V.
I
I Problem 3-8 F or a harmonic oscillator the lagrangian is
I
m """,
L =-,;'-
2 --
2 x'
(3-58)
K = F(T) exp 12h:':",T [(x. ' + ",') cos ",T - 2r.r,l) (3-59)
,. where T = lJ. - 1... Note that the mult iplica t ive funct ion F (T } has
not been exp licitl y worked out . It can be obtai ned by other mea ns,
and for t he harmonic oscilla to r it. is (cf. Sec. 3-1)
F(T) ~ ( 2.-ih"""
. ill ",T
)" (3-W)
Problem 3-9 Find the kernel for a part icle in a constant external
field J w here the lagrangian is
(3-61)
(3-62)
where T = t. - 1...
Problem 3-10 The lagrangian for n particle of charge c nnd massm
ill a constant external magnetic field H, in the % direction, is
K = ( m )" (
2rihT
.. T/2 ) ex
sin .. T/ 2 p
(im.
2h
;(%, -T z..)'
+ 2.. cot "2
..T [(x, - :r.)' + (y, - y.)'! + .. (x.y, - x,y. ) )) (3-64)
where
+ 2z, I." .
;;;;; t, f(/) SID ..(I - I.) dl
27. I."
+ ;;;;;, ,. f(/ ) sin ..,(/, - I) dl
- m'..,'
2 ,. 1."1.',. f(/)f(o) sin ..,(1, - I) Sill " (0 - I.) do dl ] (3-00)
with T = t. - I•.
De~elopinB the concept3 kith 3pecial e~Qmple3
6S
then, using Eq. (3-42) and the results of Prob. 3-8, show that
II-(x.TJ = OX p
I-
i.,T
- 2- -
""" .
211 Ix' - 2ax......T
m
2 ••
J."(t (I)' + 0(1)' + : (I)'J dt
T he kernel to go Irorn some initial point (ro,yo,z..) at time 10 to a final
point ( XlJ IYb,Z,,) alto is
K(:r.. y. ...,1. ;X••y•••••t.) = J.' loxp [U~' ; (t ' + y' + :') dt]j
'!>x(I) ~g (1) X..(I) (3-69)
I H.. P . Fey nllmn, SPIlI't'-T illie Al' llrollr h to Non-nolBt ivistic QUlln tunl
Mcd ll\Uir.Ho J(rfI . M url. /"1. 118., \ '01. 211, no. 2 , I'· 371, ItHS.
Quan ru rrt ml'C'hnniC''' 111111 parh i"'''ll m b
One might understand t his genera liza tion of Eq. (2-25) matbe-
mati caUy. Thus one m igh t consider t he motion of n point in some
abstract two-d imen siona l space of coordinates z, X. However , it is
much eas ier to think of it phy sically as rep resent ing th e mo tion of two
separa te particles whose coord inates arc respectively % a nd X . Then
K is t he a mpl itude t hat th e particle of IIUl.';S m goes from the po int in
space-time (x. ,I..) to (r ",t.) nnd t he particle of mass M goes from
(X ••,.) to (X . ,t. ), T he kern el is t he n t he su m of an a mplitude taken
over all possible pat hs of both particles be t ween their respect ive end
points. T he a mp litude for a lly particular combina t ion of paths [i.c.,
both x(1) a nd X(I) nr c specified] is eiS Il , where S is the action defined
in £':1. (3-70). :\(l'lthcmoticnlly, the amplitude is a functi onal of h 'O
independent fun ctious x(1) end X(t), and the integral is oyer both of
the variable (u nctions.
Supposewe have u sit ua tion in wh ich two particles nrc present, both
JJlO\'lIIg III one or perhaps more d iuWIl!SiollS, Let t he vec to r x rcprc-
sen t the eoord inntea of one part icle and t he vector X rep res ent t he
DerelopillS die t on cep h IClith apeciol examplea
67
coordi nates of t~e other, as in the paragraph above, except th at now
we extend the .plctu ~ to n three-dimensional space. It may happen
that the resulting ucuou con be separated into two parts, as
S[x,XI = S.lx] + SxIX] (~-7 2)
where Bz ~m:olvcs ~nly : he paths x(1) and Sx involv es only th e paths
X (/) . ~Ius IS the situation when th e t wo particles do not interact.
In this case the kernel becomes t h(' product of one factor depend ing
on x and Another depending on X alon e. Thus
K(b,a) = /.'
40
i I." ':' x· dt + 11.'.2
/.'/ex [_A,.2
p sI." AI s» dt
+ tI.:' V(x,X,t) dt]l Dx(t) DX(t) (3-75)
First, suppose. we carry out the integral over the paths X(l), The
result can be written formally as
69
functional is a nu mber whose val ue depends on specify ing n complete
funct ion. For exam ple, t he nrea under u curve is u functional of the
curve A = fl (lI) dll. To find it, a fu nct ion (the curve) must he
specified . We write n Iunct iouul as A (J(II)] to i nu icat.c~ t hat A depends
on the fu nction I (y) · We do 1I0t write A (f(y» , for that migbt be
interpreted as a function of n fun ction, i.e., that A just depends on
wha t value I tnkcs at some specified point y . T his is not t he case.
A (J(y)J depends on the enti re shape of the fu nction J(I1) . It docs not
depend on y in any way.
The fun ctional d efined in Eq . (3-77) is the amplitude that th e X par-
t icle alone goes between its end points X a and X,. under the influe nce
of u po tent ial 1' . T his potential, which d epend s upon both x and X,
is computed assum ing x is held to be a fi xed path as X changes. T hus
it is the poten tial for t he X particle when t he x particle is mov ing along
a specific trajecto ry. Clearly , this am pl itude T depends upon th e tra-
jectory chosen for x (l ) , so we write it as a functional of x(l) . T hen
the total amplitude is obtained by summing over all pat hs a func-
tional consisting of the product of T and th e free-particle kernel
for x (t).
T hus the amplitude X, like all others, is a sum over the amplit udes
of all possible alternat ives. Each of these amplitudes is a product of
t wo Jesser amplitudes. T he first of these is t he amplitude T that th e
X particle goes between its given end points when x has a specified
trajectory. The second is t he amplit ude that x has that specified
t rajectory. T he fi nal sum over al tcruutivcs becomes t he sum over
all possible trajectories of e . It is important to understand this con-
cept clearl y, Cor it includes one of the Iundamcntul pri nciples of
quantum electrodyna mics, a subject which will he taken up in a lat er
chapt er.
Of course it is not practicnl to usc th is method unless th e integral
T can actually be work ed ou t, either exactly or a pproximately , for
th e possible val ues o! the tr~jeet~ry x~t? As we h~\'-c _S~C~1 (d . l:r:ol!:
3-11) onc exact case IS t hat II I which ~\ IS a ha rmonic oscilla tor. 1.111:;
is a very importa nt practical case. For example, when a particle
interacts with a quant ized field, the field is an oscillator.
We Hlml! oousid cr ill more detnil the interact ion of II pnrf.iele ami 0.
ha rmonic oscillator. Let t he coord ina tes of the pnrt id e be x und
.0
those of the oscillntor be X. T he octio» can be written o.H
.~lx,X ) = S , lx)
I. "
+ ,. Olx (I),I )X(I) ,It + I.",. "2
AI (V'.
A
'X')
.,
dl (3-78)
I
+ olx(t),I)X(t) dl)] :oX(I) (3-80)
71
Therefore, the kern el (or the present situa t ion can be written
Consider the path integral for the harmonic oscillator problem (Prob.
s-s). This is
Using the methods of Sec. 3-5 this path integral can be .rcd u('cd to n
product of two (u nctions, as in Prob. 3-8. The more Important of
th ese two fun ctions depends upon the classical orbit for a harmonic
oscillator and is given in Eq. (3--59) . The re maining function depends
UPOIl t1 w time interval only uud is written down in Eq . (3..t.iO). This
latter funet ion call he written as
(3-84)
('11m,'''"1 mt'f'lulfIlf'lt and IHI". in '''IfNl/II
-:2
W(' sha ll solve this, nt l(,fiSt 1-0 within n Inct or indc'.)('lIdcnt o.f w,lly a
I\lrl hol! whirh illustmt cs stil l nnot her wuy of IUlIUlhllJ!; path 1I1lf'l;mlll,
~i ll(,(' nil pill hs !J(t ) It.0 fro m 0 nt I = 0 to 0 at I = T, such pntllH COIl
1M' writ ten I\S a Fourier sine series with a fuudnmental period of T.
Thus
(3-85)
•
It is possible then to consider the paths as functions of the cocm.
clouts of a.. inst ead of fun ct io ns of !/ at nny particu lar value of t. This
is a Iinear trausform nti ou whose jacobian J is a constant, obviously
ind epe ndent of "", 1tI , a nd fi.
Of course, it is possible to eva lua te this jacobian d irectly. However,
hero we sha ll avoid the e valuation of J by collect ing all factors which
arc independent of w (including J) into n single constant factor. We
can always recover the correct factor at the cud, si nce we know the
value for'" = 0, F(T) ~ "';",/2";1,7' (a free particle).
The integral for the action ca n be written in terms of the Fourier
series of £<1· (3..s5). Thus the kinetic-energy term becomes
1.o
T •• d
u: t = ~~ If
~ ~ 12.
r
mr a a
.. ..
1.
0
T
(·ctJ
nrl
T
m7f't
('O" T dt
• m
(3-86)
integration is
The first product does not depend on wand comb ines with the
jacobian and other Inctors we have collected into a single constant.
The second foctor has the limit (tlin wT)jw7T B as N - 00 J that ts,
as E - O. Thus
N N 1
JVN;- n ii-+
(f) .-1 1 (3-!J.l)
4
The Schrodinger
Description of
Quantum Mechanics
Tm; pnlh int l'~ml:o which we hnvc discussed so Cor ~with the ~xc('Pti.on
of Eel. (:I.....'\:!») hnve intcgrnnds which nrc cxpollellt.l6l~ of uctioua Wllh
the property
S I2,1) = S(2,3] + S[3,IJ (4-1)
Such path Iutcgrnls can be nnalyzro in terms of thc properties of inte-
gml cquntious which can be deduc ed Iro r» them. w e hn~c a!rca.dy
eecu this ill Chnp. 2 [e.g., Eq . (2-31») nnd Chnp. 3 [e.g., I'.ll· (.1-42)).
A still morc convenient method is to reduce the path intcgrnls to
differenti al equa t ions if possible. This possihility ex ists in q ua ntum
mechanics eud is, in (act, the most convenient way to present t hat
thoorv. It is in almost eve ry case easier to solve t he differ ent ial equa-
tion than it is to evaluate the path integral directly. The conve e- ,
tional presentation of quantum mechanics is based on this differential
equ at ion. called the Schredinger equa t ion. Here we shall derive this
equation from our formulation. 'V
c shall not solve t his equation (or
a large nu mber of exa mples. because such solut ions arc presented in a
deta iled and sat isfactory fnshio n in other books on quantum mecha nics.'
I II th is chapter our purpose is twofold : (1) For the rea der primarily
inte rested in qu a nt u m mecha nics our aim is to connect t he path inte-
gral formul ation with other formulations which nrc found in the sta nd-
ard lit erature and textLooks so that he can continue his study in thos e
books a nd can learn to translate back and forth between the two dif-
ferent lan guages. (2) For the reader primarily interested in path
integrals th is chapter will show a technique which is available (or a
certain class of path integrals to reduce these path integrals to differ-
ential equations. This te chnique is best shown by the particular
example of quantum mechanics which we shall develop here.
I Fur e XIlIIlI,h!, Ik'\.' I.. I. <ilT, " Qunn t u lII !'llti:llILn irp ," 2d cd., McGraw-Hili
flock Coml,uny, ~cw York, HIM .
16
Ti'e Sc'lre;dinger description oj quantum medtaniCB
77
giv~ the wave function at a time I, ill terms of the wave function at
a tune 11 . J,n ol'dc~ 1.0 ~h~in the differential equat ion that we seck ,
we apply thi s relationship III thc special ease t hat th e ti me I diffe
onIYIl)' a u 111 uit esimal iinterval E from II. Thc kernel K (2 1)2 is ro-
infiiuitesima rs
portional to the exponential of i l li times th e action for th~ intc':val
II to I f. " . For a sh~r~ interval E the action is approximately Eti mes the
lagrangtan for this interval. That is, using the same approximation
l1S that of Eq. (2-34), we have
'Ve shall now apply this to the special case of a particle moving in
a potential l'(x,t) in onc dimension, i.e., that for which L = m.:i;' /2 -
V(x,I) . In this case Eq. (4-3) becomes
78
.I'((r +(,)/2, I} II1n)' he rrpln ""d hy ,V(r,l) be~aU8e th e error i~ of
higher order t hnn I . Expand ing the lert-hand side to fi rat o~er 10 I
RUtI the right-hand side to first order in I and second order In 1J, we
obtaiu
If we take the lending term on the right-hand side, we have the quan-
tity of(r,l) multiplied by the integral
On the left-hand side we hnve just of(x,t), In order that both sides
agree in the limit f approaches 0, it is necessary that A be so chosen
that the expression of Eq. (4-7) equals 1. That is,
A - (2
- .-ih
')"
m
(~)
and
J--. A1 . '/.,
- e'''''' .... 1]'dJ} = iftl
-
m
(4.10)'
Probl~tn ,1· 1 Show thut for a single pnrticie moving III three
d imensions in a potential I' th e Schrodingcr eq uation is
AO'f A'
- i at ~ - 2m '17"1- + V,I- (4- 13)
This equation was discovered hy Bchrcdi nger in 1925 and formed the
central featu re of the development of quantum mechan ics thereafter.
(4-14)
Here H docs not represent a nu mber but ind icates a n operation on !/t"
It is called th e hamiltonian operator. For example, in Eq. (4-12)
d ~~ t his operation is
H=- -
Ira a'" + V
- (4-15)
2m' ax'"
't j Such a n equation with operators on both sides means this : If any
function / is written afte r each opemtor on cnch side, the equa t ion
will be true. T hat is, Eq. (4-15) sy mbolizes the state ment: The
relat ion
HI = - -
A' a'i + VI
- - (4-16)
2m 2 iJx'
holds for a ny fun ction [ ,
110
_ ~i a",
at = _I (~'" _~ A) .(~ '" _~ A) '" + c#-
2m J C J C
(4-18)
11 = _1 (~'" _ ~ A) . (~ V- ~ A) + <<I>
2m! et C
(4-19).
A af' (4-20)
i8/ = (IIW
The nota ti on for opera tors ca n be described by giving 8 number 01'
exam ples . Fo r example, the opera tor x menus multipl ication by %,
t he opera tor .:t ~ menus multiplicat ion by x 2 , t he operator Vex) (some
fu ncti on of z) means multip licnt ion by Vex), the operator a/ax means
partial diffcrent int icu wit h respect to I , iJr/t/iJx l etc.
lf ..t and B arc operators, t hen the opera tor A B mea ns thnt we first
ap ply B a nd t hen A . that is, An", means A (B Ij-) . Thus, for example,
the operator zeal aI ) means x t imes aljt/ ax. On the other hand, the
opera tor (a/ a:r).r menns t he partial derivative with respect to x of~.
or (a/ ar )(xf ) - r (a",/a r ) + "'.
We sec t hat in gene ral the operator AB and the operator BA nrc
flCJt iden tical. We further define th e opera tor A + B by the rule that
.tl + B opera ting on ~ is A Ij- +
n",. For example, the previous equa-
t ion can be written us all equation a mong opera t ors as follows :
a
- z - z - +1
a
ax ax (4-21)
Thi!l equation plus lh(' boundary eou dition of ECI _ (4-28) ~olll~ t:«~rv(!
to cll'lilll' 1\(2,1) if nile' \\"('1"(' to huv e s turted out Irmu til(! HC,'II"tHI11IK!'r
equnt iou ns t he fu udnmentnl defin it ion i ll quantum lIH'ehu~ lICH. It ill
CII'Sf that the qu antity K (2,l) is n kind of Groen 's Iunc ti ou for the
ScllI'i idingcr equat ion.
- -A' J"
2m - . dz'
-d'U·
- f dx + J"
-.
I'U·f dx = - -A' (d- u' f - U' -d~
2111. dz dr
1+-
-.
- -A'
2m -.
J"
U· _d'f-dx
dx J
+ J- - .
VU·fdx (<h'll)
(<h'l2)
and if 1Jt satisfies the wove equation (4.14), this becomes
mJ-. D~
d
Jat '" dx + J",. at <Ix = <II (J ",.", <Ix) = 0 (4-33)
83
somewhere, which is certainty (or 1) and is constant . Of course as
fa r as the W~\'C ('{Iuntioll. is concer ned tit ca n he mult iplied by a ny con-
'I stant and still he a sol~ tlOn. T hen "'-'" is multiplied by t he square of
th e eousta nt , an d the integral is t his eonatant sq uared.
It is (~J1llnmcllt.al.to . OUf d efinition of '" as prohability a mpl itude
t hat t he I II tcgm l of '" tit IS constant . 111 te rms of the kernel t his mea ns
that if f is t he wave function a t t ime 'I, thcu at time /, i t has the same
SQuare iutogrnl. T hat is, if
where t. > It > It. Compare this to JK(I ,2)K(2,3) dx. = K(I ,3) for
'. > II > tao 'Ve may d escribe the S('COIU] relation this W8)' : Stur ttng
t at ,~, K (2 ,3) gives us the nmpfitudc at ti le lntcr ti me l~. If we wish
to go to n st ill Int er ti me I" we C8 1l do SO h)' u~i lll!: t he kernel 1\(1 ,2).
0 11 t he other ha nd if hav ing t he a mplit ud e at I, We' wun t to work llock
to find it at all cn rfier t ime I I < I., we can do tl li~ II)' using til l' Iunc-
Quan.um 1IIf'C'/lRrlir" OIul IM"1 i'ltellrrl 11t
81
t iou 1\·(2,1) acconlillp; to FA}. (4-38). 'rhat is, we ca n sny K· (2,1)
undoes t he work of K(1,2).
or
(4-10)
T ile I.eft-hand si~c of this equa tio n docs not depend upon x, whereas
th~ rlgh~hand Side is independ ent of t. If they nrc alw ays equal,
neither Side can d epend UpOIi eit her vnriuhle I or x, 'fha t is, each side
must be a constant , Let us cull this constant J.,:. T hen ! = _ (i/ Mt1,
O.f f . C-· ~'I. with un a rbitrary constant factor. Thus t he special solu~
tton 18 of the form
1/-(%,t) "'" e- C./ A)I:'41(x) (4..11)
where 9 satisfies
(4-12)
TI~n~ is, for this sp ecial solution the fun ction oscillates with
WBV C
a defi•nite frequency
, •
\ Vc "".. " 1 •• , f .
U \\ " IU" ... ie requeuey with winch n wa\
-e
function oscillates corrcsp . L · I · . .
OI IUlS, III C usaicul physics, to the cllCllP '
The Sch,Qdinse,. description of quantum mecllOniCIJ
85
Therefore, we say that whe n the wave fun ction is of this special form
the state has a definite energy H. For cnch value of E a different
particular funct ion t/J In soluti on of Eq . (4-42» must be sought.
T he probability that a particle is at :r is the absolute square of the
wave fu nction 1ft or 11f11• In view of Eq. (4-41) this is equal to ItiJI2
end docs not depend upon the time. That is, the probability of find-
ing the particle in ally locat ion is independent of the time. We say
under these clreu mstancos that the system is in a stationary statt-
,
<
stationary in the sense that there is no variation in the probabilities
as a functio n of ti me.
-, Th is sit uation is somewhat related to the uncertainty principle; for
I,
in u situa t ion in which we kn ow that the energy is exactly E we must
be compl etely uncertain of the time. This is consonant with th e idea
that the properties of an atom in a specific state are absolute ly inde-
pendent of the time, so that at any time we would obtain the same
result.
Suppose that E 1 is a possibl e energy for which Eq. (4-42) has B solu-
tion ~1 end that E" is another value for energy for which this equat ion
has some other solution 4»2. Then we know two special solutions of
the SchrOdinger equation, namely,
"'1 = e- li/AlE,I4>l ( X) and "'2 = e- W A) E,I4>2( X) (4-43)
Since the Sehredinger equation is linear, it is clear that if tI is a solu-
tion, SO is CJ/!. Furthermore, if .pI is n. solution and tI" is a. solution,
then the sum of and "'1 "'2
is also a solution. Evidently, then, the
function
(441)
t: 4J:'" dr ~ t: 4J1~: ch = 0
WhrTC two fuuctionsJ, g satisfy fr odx "'" 0, we say they nrc orthogOflaJ 'I
Thus Eq. (-i-Hi) ~lYS that tw o statr-s of d ifferent cll~~rgy a rc ortllOgfJftal.·
Later we shal l lea rn 811 intcrpretntio n for expressio ns s~lcb ux II', tIt, '
and we sha ll fi nd that Eq. (4-Ui) records the fnet that If D. particle b
known to 1111\'c all energy f.'a (a nd hence 0. wave ~ullction Ifl = e;e,lJ.~J,
t hen t he a mpl itude that it is found to have u d ifferent energy E. [I.e., '
\\'6"C fun ction e'''',I'''I/>.) must be O. 1
f __" 4>.(x)
. 4>.(x) dx ~ 6._ (~7)
(-1-491
T hat is,
a. - t: 4>:(x)!(x) dx
TIle ScllrooiRger delJcription o} quantum mechoniclI
87
,
f:, [L
.·1
~.(x)~:(y)] f(y) dy
= (4-51)
•
!' An interesting ulternatlve way of expressing this result makes use of
'l the definition of the delta function. We write
"
!(x - y) = L
.-. ~.(x)~:(y) (4-52)
,._1
(4-54)
Th is Iinnl expression determines the wave Iun ctlo u .at the tin lC 'I
I
eomptetely ill tt'rlUl'l of f (r) , th e wav e function ~t the time '.. l'rcvj. \
ously WI ' represcutrd Lhis rclntiou by th e equation
Problem 4-11 Show that for free particles in three dimensions the
solutions
4>, = e(i /Al,.., (4-60) I
go with th e energy E. = p'J2m. Consider the vector p as an iadcs
n and note the orthogonality. T hat is, as long as p ;! p', I
/' 4>;4>,. d'r = 0 even if E, = E.. (4-51)
Therefore the free-particle kernel must be
K c(r"I, jfJ,tl ) = r e-hIAI' "lr, - r,l e - il" U, - tl)J ' '' ''' (4-62)
•
Since the p's arc distributed over a conti nuum the sum over the
"hrdicca" p i9 rea lly equivalent to an iutogenl ever t he values of p, \
na mely,
I
l( )-/'( (4-63) I
•
T hen-fore, we find t hat the free-particle kern el is given by I
K o(r" l, ;r"ll) = f I' e- WAI,..(,,.. ,,Jc."'(Ir-I.JJ:.' .s»:
(2.11)'
(4-61)
,J
,
The Sc/lrudill8f!r dC14cripl.;on oj quantum mechanic.
89
:rhe d('ri.vnti on of t he ke rnel for n Iroc particl e, 8.'1 given in Prob. 4-11,
IS unsnti sfuetory for tw o related reasons. First, th e idea of a. sum
over d ist inct s ta tes n used in U I_(.1-G2) is not satisfactory if t he sta tes
lie in n conti nuu m, as t hey do for n free porti clc where a ny p is allowed.
Second, th e plane-wave fun ctions for free pa rticles, a lt hough orthog-
ona l, cannot be normalized, that is, 1--.
4J*t/1 dx = 1--..
l ox = co , so
th e condit ion of Eq . (4-47 ) used in deriving Eq. (4-(;2) is not satisf ied.
Both of these poin ts can be remed ied to get her in n perfectly st ra ight-
forward mat hemati cal way. Starting all the way ba ck when we
expressed an arbitrary function n.s a sum of eigenfunctions,
fU ll ct ions~ ? The answer depends 011 the bc uudnry cond itions odin.
jug ¢ ut x = -LI Z a nd r = + 1../ 2. The easies t conditions to under.
stand physically arc those for walls which offer very high rep ulsive
potentials to t he part icle, thus confining it (i.e., perfect rcficctol'5).
They correspond to .p(x) ~ 0 at x = -L/ 2 and x = +L/ 2. The
solutions of the wevc equat ion
(4-66)
or a ny linear combination . N eit her e ;!z nor e-oJ:z Cal l satisfy the bound.
ary ~Ildit ions, but with k == nr / L (n an intege r) sa t isfactory solut ions
arc grven by half t he sum (which is cos kx) for n odd e nd >i!i l inn'S
th e difference (which is sin kx) foe n even, us dingrumed in Fig. 4-1.
T hus th e states nrc sines and cosines and the enecgy levels arc separated
{i.e., not in a con tinuum},
If th e solutions arc written as
J LI2 (
- L/2 'Vf2I. cos kx)' dx = 1 (4-67)
, 91
•
l - - - - - - p!' 4
Fi g. 4-1 The for m of the
one-di me-oionn! wave Iunc-
t iona wh ich hnv e been Dor -
mnl iecd in n bo x. T he fi rll1
- - - --- - 17 =3 Iunr eee shown. T hc corrc-
"pomlinp; cncrJ;Y lev els Arc
£ . "'" 11.1"' / 2,,. /.', B, ... olE..
E, - 9E I , nnll E , - 16E •.
TI le nll9tllu tc mngn lt ude of
- - - - - - - - - -- 17=2 l hc cnef"IU",wh Khd l" pl'nfbJon
IIle size of nu r fictitious bo x,
is not import a nt for more
real istic p roblems. RAther,
it is the relnt jon bet ween the
-------- --- 17=' onergy Ievets of t he vnricue
st at es which h8!l6ignw r nDcc.
x =-L/2 x=o
F2[ . (n + 1)%
'\}I., S in 2 7
. % ] F2 2n +1%. %
L - sm 2m L = 2 VI. cos 2... 2 L sm 2... 2L
(4-69)
and remember to add the result for the two kinds of wave functions,
namely, vilL cos kx and V2L Bin ks.
92
It iK oncn incon,"('ni('lIt to 11:'«! Bill ks: nud c.os k~ for the wav e fune.
tions as we would like to usc t.he linear comhll1ntlons
cot. "'" CO$ kx + i sin kx
Bud c- il.r "" cos k.r - i sin kr, We were forced by our box ~ Use sinl'8
and cosines nud not the lincnr corubiuntio u, lxcnuse (or a grveu tz(JII(\
but 1I0t hoth of the Iuueti one is n soluti on. n ut if we enu d i!'l J'( 'Knrd
smnll errors arising from these small dillere uccs in k, we ruighl lttill
expect to be able to get the correct results from th ese new lillear
combinat ions. Normalized, they arc
s- nnd
Since t he W8"C e-"· can be th ought of as eLl. but for negative values 1
I
I
These arc called peri odic bou,ulary comlitiona, because the same onCS
,·..o~ld result by th~ rl'Cluir~IU('ll t ,t ho.t ", is period ic ill r in all space ,,"jlh
pcrlO~ s: = T~. I,t IS readily verified that the (unct ions V l / T. ",.. nrc
sol~t~ons. Ilo rl ll,nh zed to runge J~. provided k = 2..,,,,/ 1. wit h 1I 1l1l illh·gt'f.
posi tive, 1l(.' ~tlVe" or zero. F rom this OUr rule follows directly ,
I II thr~(! dimensions we CD.II I'l'C wlcu hnppc.·lls hy using a rcct nngultlf
Lox of tildes L., L lI1 L. ill the three directions. Let us use periodic
77le ScllrVdinger (Ieacriprion oj quantum mechaniclJ
93
boundary condi tions. That is, the nmgui t ude an d first derivative of
a WU\~C' (un cti on. a t u ~i llt. on our. fuce nrn respectively equal to the
1~1Rgll ltUlh' ~Ild hNt dcrl\'nt~\"c a t t he correspondi ng point Oil t he c ppo-
S i te fnce. f he wav e Iuu et iou for a Ircc part icle normal ized is
TImt is, USC plane waves normal ized to volume V : $ exp (ik • r) .
The number of states in ra nge d' k (differential volume of k space) is
V d'k/ (2,,) ' _
Let us apply this to Prob. 4-11 and recall th e connection bet ween
momentum a nd wave nu mber 7' = Ak brought out in Sec. 3-1. In E q.
(4-64) we must make two changes. F irst , since the wave func tions
used were exp (ip · rllt), whereas we should ha ve used
,I
Alea/lllrelllentll ond operator.
i 97
2hT J'
1m exp imx' _. (exp A
- imx y
,,(z, n ~ V2ri1i'l' T- )
(cxp ;my
2AT') j(y) dy (5-2)
-
I of large T , there results
I
I P(p) dp _ ;:"1 J:. [ex p e;~ i~~)] j (y) dy I' (5-1)
!
I , ••
v c assumed ca cnrl
r tor tl16t . ,'" itian),, the 1'- oert tcle wouldI be' "restricted
I
, ithi
to a region W I 1111 - 0 I + b [ t l C ori mn
~. .
'rhis menus •
that t
I
ic iuitie wave
it d
' j( ) d to 0 [0 val ues or y lnrgcr III nbso ute nmgm 1I o
run cti on ,y ro p~7' I r Inrrr" 1I 1t~ (IU" llhty . . '." ..,/ T I . '
U'W-/ _
t ita n b. )\ ow us rccomcs ~- . , . j
• I r I'
K COIIU.'S
, 'I I II
negI Ig i ) Y 8 11111 . Q C
.... iucc t bero is uo colll n but ioll to the integrn 0 ..q.
98
____________-9-
I
,.T I
,. I
I
• I
I
E
,:
, \
, yo-
•
I
•• I
v I
Fjg. 5--1 Thc emplitudc (or par ticles traveling frcd y to arrive at the poeition % in the time \
inlen'a) T is determ ined Il)' th e produ ct. of two Iunr-tiona, The first is the nmplit ude f b )
•
(or t he p4rtirl e to st art at ro me posit ion y, as sh own by the ebadt'l1 cu rv e in th e liguno.
TIle second, t he Amplitude to get (rom 11 to r , is the Ieee-pa rt icle kernel K (x.T ;y,O), as
shown by the lIinc wa ve of alowl}' cbnnging wa velength in ti le figure. (He re we t reet the
fins) position z W! t he origin of t hiM (u nct ion, s ince 11 is our v a rieble. I If the point z is far
frow t he oriJ;in, compa red to t he distance - b to + b over whje h f ey) is nonzero, t he WAft
has no npl,roxinILLt d )' consten t ""O\·(·II'np;th . Its a pproxim Ate lorm is exp I( - l Ilt) (m.r/ T)~I·
The t wo (uncl itlns are rnu lfiplied t ogcl lu'f and t hen integr a t ed over y t o get th e final
am plitude for arrival. Since ull peructcs heve t ra veled lIpprox imately tbe same d istance s
(again ll&-w n inp; x » b) in tile sa me t illle T, t his finnl amplitude is equivalent to tbe
amplit ude t hat th e particles have t he momentum p ... mx/ T.
(5-5)
99
,
I PositiOn. ¥
'b
,•
..
I
,
I
On tile ot he r ha nd , suppose the
wevclcngt he dilTcr (or some Dew fun c-
I t ion 1'(1/) as shown in (b). ThCD,
, I when t he p rod uct is taken, t he con -
I trib ut ions t o t he integral fr om dif-
, I
I
I
fcltlot v al ues of 71 will eeacel eech
ot her, :-';0.' t he probab ility t h nt the
• I
-".,...~ momentum is rIlzlT is sma ll.
"
,, Position. ¥ r-
I If a Dew posit ion r is chosen DoS a
final point, as "hOWD in (c), then a new
+b
region of th e keene! curve overlies the
spac e - b to + b. For a corr ect
choice of Z ' t the wavelength of the
kern el in t his new region is t he sa me as
lbl t he wavelength or F (y ) and II. la rge
pecbnb i hty res ults . Thllt hi, t here ts
R large probnbilit:r t hat such pnrt ictea
,. -----------1
~
have the new m omentum value
p ... nu'/T.
or
~
,
P
"b ' \
/ '
leI
I
100
nif.ion to tho thrce-dimcuslennl ClUl C where Lho o.mplitur}o for the \
momentu m is
We shnll call this the amplitude that the particle hns momentum p at
the time I .
l It is often useful to analyze pro blems in this momentum rop rescota-
t lon rather t han in the coord inate representation, or, us it is often
, atntod , in momentu m space rather than in coordinate space. Acta-
, ally, t he tra nsformation Irom one representa tion to t he other is just
\ a Fourier tmnsform. T hus if we have th e momentu m representation
'J and wish to find the coord inat e representation, we use the inverse
transfor m given by
;
," (5-8)
'Ve can describe this last formula in the same physical terms we
have used to describe th e structure of other ampl itudes. The empli-
tude that t he particle is at th e position R is given by the sum over
alternatives. In this case the alternatives correspond to the produ cts
of two terms. One of these is the amplitude that the momentum of
the particle is P, given by ~ (p) . The other term, cxp (ip ' Rift) , is
the amplitude that if the momentum is p, t1len the particle is at the posi-
tion R. This second fac tor is not new to us, for we have discussed
such all expression in Prob. 3-4.
Note that in the transform of Eq. (5-7) the exponent has a minus
sign. Su ch a term cnn be described in a manncr parallel to that used
in the preceding paragraph. Thus we can say that (':'1."1> ( - ip · RIft)
is the am.plitude that if a partid e is at posil-ion R, it has the momentum p.
S IIS tit
._u ti , E' 'In ' ("• -u
1 II mg III
n) for ·'·(R
\1" 1, I, } t he.' expression of Eq,• (t....
•
~)
and Laki ng t he Fourier truneform of ~(R 1,11) to gt -l !Jl(p 1,11l}. us III Eq.
(}lllmtulII IIIrr/lOrti.." tlUf' pnt'. intt'J(rll'"
I
102
\
t OI ;S
' :ifl:' 5-3 Thr kerne-l fllr Il. Irce \
1'1Irt iri c ill reo mentusn flP"r:ll ill
\
(5-13)
The energy E is not equnl to p' /2m , hut is instead an extra independent
variable (the coefficient of time) needed to defi ne the kernel. Only if
the syste m exists in th e same energy state for an infi nite time ca n an
exact measurement of E be made to establish the relation between
energy and momentum.
M an example, we shall work out the kernel for a free particle. For
this case th e integra ls over R 1 and R, have already been work ed out.
with the resu lts given in Eq. (5-12). T hus we arc left with the inte-
+
gmls over II a nd I,. Make the substitution I, = 11 7 . Then th e
double integral can be written as
(5-14)
(5-15)
uu
' I ' thnt till' inverse lrnll~rorm of tile kernel would
l\ rcsu II wounII IlUIl ) t I' k I will J •
not J.::in ' bnek the originnl coord inat e rt'Jlrt':o\f'1I -R ,1011 er ne I WlIf;h
we sta rted . Such l\ tmusforur would d iffer. from th e> cor~'ect kCrt\(~1
.m t Il:lt I' t wou III' 110 t I 1(,' Z"".... I"O for values or tnm-
. h~ss than
' zero. O'le
way to o bt uin t he correct result from such iutegruls IS ~o ,Plnce the
pole 011 infinitosima ! dis ta nce nh?ve the real axts, This 18 nCCOtn_
pIL...IH'd hy leuving e in the exp~8s lOn.
If we raticual izc the expression ns
t . d I'lm +
o e"" T= _ow .
• te
J
K
Fill'-'. s.1 The rcnl pnrt of the free-
p nrt.i ele ke rnel K o plott ed V!'Io t ime.
The funct ion il'l zen> fur negutiv e
T t imes , then IItnMB wit h 8 ehnr p
jump nt t "" 0 nnd cont lnc ce /1..'1 n
eosine wnve of consta nt cm ptuude
end frequency .
,
,
i
Ell _ Pal/2m + iE =
( i ) + (E.' - 2mp')
P.P. E l' _ p'l/2m 11"& (5-19)
Here the first term on the right accounts for the transient effects that
result from the sudden start at t = O. The second term gin'S th e
steady-state behavior end shows that, if 'We wait long enough, the only
energy found is the usual p'/2m j but near I = 0 the energy is not given
by this classical formula.
Problem 5.2 If we tronsform only the time and not the spatial
variables, defining
k(Xt, EtiX.,E.) = f f eU/A)£,I.K(Xt,lt ;:t.,I.)e- <ifAl&,t, dt t dl. (5-20)
where ~'" and E", nrc the eigenfu nctions and eigenvalues of If.
106
eral ti mes it will be possible to cons truct the probability function P (o)
which gives t he probahility tha t in any particular experiment the
numer ical values of A will be found to be equal to a.
In general , it is possible to assoc ia te a probability amplitude with
such a probnbility fu uction. T his amplitude would be d efined in
terms of th e measured varia ble, together with ot her variables nCCl'S-
sary to complete the epecifl ea t lon. Let us 6CC whnt is Invol ved by
generalizing our exa mple of a moment um measure me nt , F irst we
sha ll take just one dimension, but the extension to seve ral d im ensions
will he obvious. \Yc ask : Does t he sys tem ha ve the property G? For
exa mple, r; w ight stand for the stntc mcnt : The value of the quan-
tity A is equal to a. \Ve must have SOUle wuy to unswer t his expcri-
~~ntnlly. So Itt Us imagi ne some equipment can be designed so that,
if It ha~ the propcrt~ G, Lhe pa rticle will pnss through the equipmcnt
and nrrt vc at a cer tuiu locati on on some SCfCCn or rueter,
T he probability of t his Illay be written
P(Gl = II K ... (r ,x)/ (x) <lxi' (;'-22)
if J(x) is the wav~ function of tile system to he measured, KUI,(f,x) is
the k~rn ('1 for ~~lIlg thro~gh t ile particular oxperimeutul Ilppnwh1:',
and r I~ ~Iw posuiou of nrrl\·?1 for particle,s with the properly G. ThlS
pro bability 11M the al tc ruutivn mathematical form
P(G) - 110' (x)/ (x) dxl' (;,..23)
.
I
I
JUco~urcmenl. and operators
107
It~ where we have put
\, (5-24)
L" (Dotining t his ua the complex conjugate of a function is just for con-
I, venience, as we shall sec lntcr.} So we ca n say
~'l.
\ ~(G) = Iv'(x)/(x) dx (.,-25)
'. is the amplitude that th e system has the property G. This concept
is further described in Fig. 5-5.
The property is defi ned hy th e function g-(x} for the following
reaso n. Suppose some other experiment with different equip ment, and
hence [l, different kern el Ku p{q ,x ), should he built to measure the same
property. In this second experiment the particle arrives at 'J. Then
the probability of fi nding that the system has the property G is
l
t or II v" (x)/(x) dxl' (5-26)
I Since the property measured is the same, we must obtain the same
result in every cese for peG) as we did with the p revious experiment.
That is to say, we must have
II v" (x)/ (x) dxl' - IIv'(x)/(x) dxl' (5-27)
for any arbitrary func tion fez) . This means O·(z) """ O'·(z) within
at least an unimportant constant phase factor ell. That is, all methods
y
I
I
t
I
I
I
I
Fi,:. S-S A d evice design ed to
measure the property G is placed
I be tween the init i:,r po int or t he
I incoming pa rt icle (with wave rune.
l ion / (z) ) llnd l hc f.nnl poinl z - t .
T he equipmen t nH)o.l if~ the ke m el
rOf t he motion (('Ofll p:1 fC Figs . .5- 1
ami 5-:.!), IlIRkin..-: it eq ual to u( x ) .
TI,e prod uct !(x )U(z), inll'l!:f:lh,,1
over z, i.. the 11lllJll it ll<lt, to /lr rivc
nt { IIrter pn."'.lIing th rough t he
equipmcot.
I
I
1011
.• I ' II sau te property cnrr"po"d [within u phas e) to tlt e \
t o ue I'TmUlC 1(' ' SOil we call O· (,r) t Irc cIlaraCtcri ' I uncllQ1l
c rUjI ec . \
salli e g. (J') . I',or tl us
t .,
reas
01 tin' "mp,'rty G, I II I tc I( ) be
We may usk another question. What n~us te s a z . 80
, , I ' tl c property (l"f (1' or exampl e, wh at IS Ule
th st It IS sure to lin C I . fi ' 'I) T
, 1ulle I'1011 1or n p nrt lcle whose mom entum IS de. Hute .,
wave hat i., t
,I fi ud
w(' W I~ I to I ll u n n
on"
f (:r ) , """ . t F( r) • so thnt t he part icle gomg
•
t, hrouglt
•
the appnrnt us will certainly arrive at r eud ~t no other point r · 1?c
amplitude to arrive at r' should be proportlounl to li(f - r) (that Ill, I
zero unless r = n. Helice •
I
JKu,("r)F(r) dr ~ 1(, - r) (5-28) I
This we can solve by the relation of the complex conjugate of a kernel
to its inverse, discussed in Sec. 4~1. \Vc have from Eq. (4-37)
JK",(r',r)K:., (r, r) dr = I(r - r) (5-29) \,
60 that I
F(r) - K:',(" r) = girl (5-30 ) I
T hat is, O(L) is th e wav e func tion of n particle having the property G
with ecrtn inty. 'Ye con say either (1) the particle ita."! the property
G or (2) the particle is in the state o(x) . So we find : If a pnrticle is
in a state /(x) , the amplitude it will be found in a s tate O(x) is I
(~I)
I
"'(G) = Jg'(r)/(r) do: ~ '!>[g(r»
I
For more dimensions, x becomes a space of several variables.
We might say loosely : T he probability that the particle is in the
state oCr) is IJg' (r)/(r ) dr l', This is nil righl il we kn ow what we \,
mean. The system is in stetc j'(z) , 50 it is not in O(x ) ; but if 0. mens-
urcment is made to as k if it is al se in U(x), the answer will be unirl1ls- I
ti ve with the probability I
-I J:. g'(r )/(r) do: I' - <Vlg(r»
P(G1
A measurement which asks : Is the state o(x) ? will always have the
(~2)
II
answer yes jf the WAve fu nction ac t ua lly is y(.r). For all ot her wuve
fu nctions, repetition of the experi ment will res ult in y es some fmetion I
l' (JJctW('C1l 0 a nd 1) of the tries . T his is a central result for the
proba bilist ic interpretation of th e theory of qu nntuur mechunics.
For all of this we ded uce an interesting inverse relat ionship between \,
a wave fu nction end its complex conjuga te. In neeordunce with the
iuterpretation of Eq, (5-25), O·(x ) is th e am plitude that if a sys tcJl1 \
,,
•
1IIea~uremenr. and op e ra t o r.
109
is in position x, then it has t he property G. (Such u sta te ment is put
mathem a t ica lly b)' su bsti t uti ng u 15 Iunc t io u for f(L ) in Eq . (.h.'H ).)
: On th e other han d, geL) is the amplitude that if the sys tem has the
"
~, propc r ty C, it is ill position x . (T hi!'! is just a wa y of gi ving t he defi-
I, nit ion of a wave fu nct ion.) On e function g in 'S t ile a mpl itude for:
•
If ..:I , then n. The othf'rfunct.ion gives t he umpl it ud e for : I( fl. t hen A•
T he Inversio n is a ccompl ished simply hy tnk ing t he comp lex conjugate.
E quatio n (:N I) can he intC'rprclro as follows : T he am plitude that
a syste m has property G is (1) t he Dmplitude J(x) t hat it is at x t imes
(2) t he a mpl it ud e o· (x ) that if it is a t x. it has property G, with this
product summed over the alt ernatives x,
Prob/cm 5-3 Assume I_.. /·(x )/(%) dx, which is the probability
that a particle of wave fun ction/(x) is somewhere , has been normalized
t o t he va lue 1. Unde r t his const ra int, show t hat t he sta t e J(x ) \\ hich
has t he highest probability of having the property G is f ez) = o(z ).
J-". X:., •.... (zlx.·.~ (z) dz = ~(a - a')~(b - b') 6(c - c') ($-35)
The fun~tion XCl.6.~ • •• • (x ) is t he amplitudc that if the system is in the
state described by a, b, c, . . . • thou it will he found at x. T he fune-
t ion X:'6.~ ....(x), which we have called the characteristic funct ion is
the ~mplitude tl lat. if t he system is at z, it will be found in the state
specified by a, b, C, • • • •
If the sys tem is k nown to be in the state ! (x) , then
F = J•
_"x
• (z)f (z) dz (5-36)
is the am plitude to fi nd the system in the state specified by A having
the value a, B having t he value b, etc.
T he qu ant it ies FCI.6.~ ..•. are lust as good a representation of the state
as the fun ction ! (x, Y,z, . . .). In fact, if we know the function
Fa .6.~ ..... we can reproduce the function j (x ,y,z• • . .) by means of an
inverse transformation.
T he fu nction F..6.c• •• • is called the A, B . C, • • • representat ion of the
state. (In t he preceding section we had an example of t his in the
moment um rcprcscntat ion.) The function!(x,Y,z, . . .) is thecustom..
ary coordi nate representation, or x, Y I Z, • • • represeut ut icn, of theslate.
Transformations between t he two nrc carried out with t he help of the
functions X and x". In part icular, the function x:.•.~....<..r,y,z, .. .)
is t he trunsfcrnmt icn function necessary to go from th e x , y, z• . ..
representation to t he A , B. C, . . . representat ion, while t he func-
tio n Xcr.lo (z .y,z, ) is the transformnt ic n funct ion necessary to
go the other d irection. T hus the inverse of t he transformation given
by E q. (5-36) ia
! (x .V,z• • . .) = LLL . . . F.. 6.C•• •• XCl.• ,C••• • (x.Y,z, • . .) (5-3;)
• ••
This says t ha t the amplitude to be found at x is the amplitu de F• .•.c •• • •
to be found with A = 0 . B = b, .. . t imes t he amplit ude X d ,c-.. • • (x)
to he at x if A = a, B = bl ctc., summed over alternat ives a, b, c• . . . •
112
B}' subst itu t ing this rclnt jcu into Eq . (5-36), nnd usi,ng the orthogollal
properties of x 8 5 defi ned by EC). (;,....15), show Ulnt. F•.•,~ ... . = F•.• oC. '" •
Problem 5-6 Suppose A, Dr nud C nrc the three ca rtesian com.
I
poncuts of moment um I'., PII' 1'.· \Vhnl is ~hc form of .the rune,lion I
x.....t ( X . y:)1
,
Usiurr the results of Sec. 5-2, verify the relations obt:uncd
1)0
in Sec. •fi- L
Problem 5-7 Suppose thnt the A, B, 0, . .. representation does
not correspond to either coord inate rep resentat ion or 1lI01llC'nt u m
representation, but instead is some th ird way of representing the state
of the system. SupJ>Cl8C we kno w the funct ion X• •• •I • • • • (x ,y,z, . • .)
which permits us to transform back and forth be tween coordinate
representat ion und .II , B, C r . . . representation. S uppose furtbcr'
t hat we know the transforma tion fu nction necessary to transform back
a nd forth bet ween coord inate representation a nd momentum repro.
scnta non. Whut then is t he fu nction necessary for the transformation
bet ween momentum representation and A, B, CJ • • • representation?
5-3 OPERATORS
Pea) = ~~ IF j. (5-39)
In th is cq ua tion S Ut 1" •
in tI c ti 11I~18. Ions nrc enrriod out over all possible values
I COli IlI UOUS or dlscrcte!:lcts of b 'C J • • • •
ftloolJlfremenr& nnd op e m ror&
113
IFo.!l...... I'-
- F'..,6.1, • •• F ..... r..•• (5-41)
= t: f'(x)R(x) dx (~)
In the second line of this equation we have made use o( the subst itut ion
R = <If (5-1; )
where a sta nds for e linear operntor which .o pcr~tes on th e rUnCli~ll/.
1 the present cnso a stands for t he opcmticu displayed on th e fight..
h:nd side of Eq. (r.....l:n. l hnl is, lllultiplicntio~1 by t he fUII.d ion o, and
iutcgmticn. T he nperutor a is nssoolnted with the phys ical quantity
A . Using th is notation, we cnn write ~
Any operator, such as ClJ for which Eq. (5-47) holds is call ed hermitian
(ef. Eq. (~O» .
Problem 5-9 The transformation fun ction between space repre-
senta tion and momentu m represcntat ion is
(5-18)
d
where 6'(z) = dx 6(x) . \Vith this result determine thc op erator eor-
responding to the x component of momentu m a nd show t hat tile
expected value of tlltos component of momentum ca n be written OS
Problen ~ ~.10 SUPJX>sc the qua ntity A corres ponds to the x coonli·
na te of posit ion. Show that the correct formulas for thc mea n value .
of z result when the fun ction GA(L',x') is taken to be I
G.(x,z') = zI (z - r W<U - y') ~'(z - z') (5-0 1)
lllpfllrIl remenf" nnd operntor"
lIS
Eigcn f1J n~lion8 a n d Elgcn vnfues, The wove functi ons Xa .b.r• • • • '
as discussed III Sec. ;)-2, show a particularly simple behavior when sub--
jected to the operation a. T hus
hax
- ~ - = bx (';";0) ,
1, iJy
t. ~: (XI) , which is the amplit udc that if we are at XII thcn we arc
in the state n.
2. e- f./" s, l fl -f' ) , which is thc amplitu de to be in state n at the time
It if we nrc ill the state n at t he t ime 11. t
3. 9~ (:r!)J which is the amplitude to bc found at Xt when we know
we arc in t he state n.
121
In gcol1('rnl, we have found t bn t if t he po tentia l is quadrat ic in z, the
hell e! CRII be ddl'rlllill('(1 exact ly, whereas if it is sufficient ly slowly
vary ing, tll(' ~l ll idH!-...ienl npproximntam ~'llu h·qunt {' . TIIC'TC nrc sonic
otlwr types of potcutin ls which ea u h(· succee-Iully t rea t ed with the
help of SchriKli nJ!:('r's cq unti cn. XOW W (~ nre st udying a technique
which is ort (,11 useful if til e effect of the potent ia! is s mall.
SupJlO8C tl1l' poh'llt ial is small , or more pr('("il"(·ly . sup pose the time
iut ('j:rnl of th e potenti al along n po.th is amnll compared to It Then
t he part of t he expon ential of Eq. (6- 1) which depends upon V(:c,l) can
be: expanded as
cxp [ u:
- Ii s, V(x,1) d!] u:
= 1- Ii •• Vex,!) d!
-1 (i)'
+ 21 -h [1." f..
vex I) d!
'
]' .. . (6-3)
where
K U' (b,a) = - i J.' [expG 1.:' ~' dl)] J.:' V[x(s),. j d. :ox(l) (6-6)
K''' (b,a) - - 2~' J.' [cx (i. J.:' n~' dl) ] 1.:' I1x (s» ) ds
p
12:.!
,
. ·i/l' 6.1 A JlIIrli"lr Mt nrlM r.m~" tJ ''1111 mOVr'll l1ll
I.
II fn'l' luul il'le to c. I b-re It 1M Ilf'lf.'d UI)liQ,
ltf'nll f'rt'\I, hy the !'oll'IIliZlI 1'(.1 ("),111 ... ~
'n u'll':lltt'r it IJI OVCf! IU' I' Irf'C I'ut ti r.lc to "
1'1tr muplilUllc ror .. udl n mol ino ill J!:iVl'h i~
s c loA] . (H. Ill). 1£ thiM ItIllJllih adc is intl'grlltf'«
e ver nil Jlfl~ ib l(! JXlSi t in l1l~ 01 ti le IJo int t , the I
tl.'tIult L" the firrlt -otfh·t term in t he lx:rlurl)8tirJQ I
III - - 0
t' XllOfVjion. I
Here the limits on the integral over x hnve boon written as ± ISJ .
In n. p~ct ical problem the limits will he estublishrd by the potrll1 hl.I
(wh_lch III 1110St cases d rops to 0 when r becomes very large) or by the
equrpmcut , which restricts t he m nge of x.
b b
• (1)
• (21
b b
• (31
•i lZ:. &-2 In () II pnrticlc m o VC8 from " to b t hroul:h t he ~t~tial I " witho ut lKoinJ: lIrn.~w
, cree . '1IC a mp I'Il u l Ic ror " IlK
. III" nI.' . (I'." l . III r-1) liar , l'urt " K'1r ' lK ti(' D IlIl' h,"l"l o nce' nt (' "I~, II. ,
" hroue h
mov('II t ITonp; t he Jlou'Q III • ' \' '11 1(' lilli I' l il ue l.·
·
for t1 L111 1ll ,
I\ III
'
(b,. ). n (.n u« l':art 1 K l ts
, - " I ,', r r ill " , e}• llm l i ll (-I) •It I.i f('lI tt l'R't1 ,.,"
HCfi tere I I wiee . "It I l lc "11111 (1 lUI e n
t llJll':", I 1(' I,U'I.
" . ..
___, , ' k ' In ,
...... I' TIIl II; b , IQP; (I CC II C'.
11 I
e tota l . h1ll1 it Ulw flit motion Iro iu a to b ': ll h rtll) RUlli r
ol 1ol:Att.,ringa i.It K . + K ill + K II) + .. . + K'· ) + . . . .
J21
O IlCO I
tl lC RCIt n )ntlvrs• which COlll prisr this kernel cOI1f;ist~ of II I~
, tlou The l>n rt i('lc stnrts from a, moves (l~ n fn 'l! Pllrl~I.·
folInwmg 010 I • I ' • ""It
to t I1(' pom int Xli I~ -- c, is• there scat tered by• urc potentI ' ial I, (e) I 8rt(-r
w hiICII 't ,- as a 1"'0 pnrt icle . from the po in t e to t ro final po:silio..
I 1I10' ~ "
I.
.. u.
The nmplitude for such a path IS
where dT = dz dl. Ren ding from right to left t his for mula means:
T he particle moves as a free particle from a to d. I At d the particle gets
scat tered hy th e potential I' (d) at that point. It then mons (ISA
free particle from d to c, where it is scat tered by t he potent ial r(t).
After t ha t it Ill (}Y<'S from C to hi again as a free particle. \" C sum over
all t he alternati ves , na mely , all places a nd times that t he scatt rrillg
may take place.
1I('r~ w~ han tacitly assUlIlC'd t hat t, > ld. I n order to avoid the
cOJllpll('nt lon of having to introduce th i~ as snnrptio u explicitly in each
such example, we ehall tuakc USc of t he convent ion a dopted in Chap-4
(c!, Eq. (4-28)] . lld a.sUIIlC
Tlle~1 Eq. (f,..13) is COrrect with out any rest rictions on the range of inte-
gmtion of If uud I• .
The ~'8dl:r wonder wha t Imppc'Il,'{1 to thl' fuctor H whicb
I URy
alP Pc~rs III h , . (U..7) but is uPI}an 'lI t1y oniiucd i ll 1-'(1 (fi-l:J)- ~ ott>
t iat 111 E'I. (6-1 3) t lre runge of integra tion for ld is sti
" ,ll Irom I. to I.,'
Tll cpcrturbarion m crhod in quantum m ccllanic.
125
•
• however, t~IC ra nge of I.. has been restricted (hy the definitio n of EtJ.
(..-0 » to lie between I.. 8 11d I". This restrir lion cuts t he value of t he
~Io\lhle ill tC'~!nl ill ha!£. To see this more clearly, suppose the dou ble
int egral of E(I· (6-7) IS rewritten 8.!J
If the variables 8 and 8' nre interchanged in the last expression, the
va lue of the double integral rema ins the sa me. T his means tha t the
first and second te rms on the right-hand side of Eq . (H-l ii) nrc equa l,
so eac h one is half the value of t he original double int egral . This sam e
sort of argument will account for a factor l / n ! in a n expression for K l-).
126
127
b b
"
~
'~
III 121
Fig. 6-03 In (1) the pnrt ide moves from a to b through t he pot ent ial V ns e free pllrt id r..
desenbcd by the emphtudc /\ ..(/1,11 ). In (2) th e pArticle is eentt ored one or morc t imcs hy
V, with the Inst aent tcring Ink ing plnee nt c. Th e mol ion from a to r is dC!locrih('d I.y
K r(c,o ). nnd HlDl from c t o b IJy /\..(I.>,c). A cCIIllh inntiofl of the two situnne ns, when nil
posit ions of C are a ccou nted for, eoYCTS nil possible ClUICS nn d gives K ,·(b,a) in the form of
Eq , (6- 19),
I
point b. as represe nt ed by the kernel K o. This interpretation is d ia-
gramed in F ig. fi..3.
Since the lust scattering could take place at any point in space
and t ime between a and b, t he am plitude for t his compos ite motion,
represen ted by the integrand of t be IIlSI term of Eq. (&-19), mus t be
Int egrated oyer all possi ble posi t ions of thc poi nt c.
-
f,
-r -
a K .(b a) a'
+ -2h'max"
- , K ,(b,a) - i M (4 - 1.)I(x, - x.
) (&-20)
tat" •
Show, from this result and Eq. (11-19), that the kernel K v satisfies thc
differential equation
- ~~ ~ Ky(b a) ~ dx,,!
+ 2m ~ K,,(b,a) + V(b)Ky(b ,a)
al~ t
128
seribiutf" t he. Illat ion of the system between the two tim""
I «I '.,...
t I1(' k·('n l(' • • '-0 ,
For our pr('SC'lIt pnrllo~l'S thirs equation ca n be written lLS
Using this def init ion, the series of Eq. «()"23) can be rewritten as
- ~ df ~ . V"" + 1'", _
+ 2m 0 (6-27)
t at
w orking in cue dim ension only , show how the SchrOdinger equation
may be deduced from t he integral equation.
$ s' o
Fig. 6-4 E te- t rons hoil orr n hot fil11m cn t lit.. 0 , Ilre tl('f'C('nNI in to II be am h)' C'O lIim nti ng
holes ill II a nd II' . lind t hen "l ri kr II IlIin-foil ( lt ~rt "I 0 , 1\10:-1 of Ilit' r'1('C·I ~mi r~" t:t nt lF:ht
on wi thout "I'ill" tlr'll Hptl'C l (if th r-ir "II r' rjt\' i!'l j1; ~1 "llCI\lj1; h nil' ! lIu' tllrK~1 IS Ilu n I'II0u.-:I.).
"'II
but JIOIIIC' nre ,11'1I'~I·t c',1 I I)' ill Ic' nll' l iO ll!l wi Ih 11h .mlt III t hi' ! ll r Kl ' 1 IIll1l ~l'~ t t crcrl , ~or III p Ie.
t hrn U II'; ,I 11 11 n u ll'; (. s t U b . A!'i 1•' 1., ' uuter ,., I, i!'i. 11111\'1,. 1 \II' 11 11,1. 111"" " . Ihl' rd llh nn hph .·l...· n
I
UIC rdl1tive 1I1l1ll1)pr o f ~nltl'rillgs " a nd Ibe llClIHt'rill~ lingle 0 ca n he 1IIt"IlllUn....1,
(lrum fII"' ",('('/,nllk,. ancl Jln.', ill,,,/(rn'"
130
c
o r:
• o .- -'
b
Fig. £..$ TIll' gOOlllctry or ti le U - I
• r~ Jlllrt id e to ~ " "I M" "t - ·IWII f't mJl: p ro ,INn . n il' d r d ro h st af ts ftt 0 and rno\·('!l -U
it JU(}\'I"!I to lilt' f:~UII;t./' It ~ t1t:n , t"hl>r~od IJ)· 111(, 0 ( 0 111 it' po le'll t 1111 I' (r) , AI tor t 11(' "r llt It·rio~,
. II ', ,, I ll" I l.!4 10 t:ll h 'i l llt II . , I r I · I ,h.
l:lCalt'·rmlt ("("nh', 0 In , , · Ie • III U t u· fntl 'U>l \ ·I·... tor 11 & rom
.
Ifl ClUill fl·,j f ruul lIll' ,I" n '.I '
.,
r,
l lll l " lIl ' (·.....'1 the 1·It't tror I
I '
, I I; IO U 0 I II ' n UIl'l l' lIt ten- I I '
I
"j', '
d •
I lIIH u'r - n Ht' ll ll l ,rt'l l l h ftl Ul/:h t lll' ltn.- t •
I t till'
urHt-ord.'r Ikln . ll" ' Ih,xin , . If I . ' ' Illl lt , us l 'ftlt't'~'1 t' tJrn,,,pOll t :i I' •
" III 10 11. t Ie 1I111/lht ll t , I , . Id •
mdutll·d then the I ClltIl t " II I , t UI " '0 ecnt tenugs, };IIY . lit C d Pl ,
, " III Ie H('( "Ollt -on l l 'f 1IJ1llroxinull ioo. e tc.
The perturbation method in quantum mechal.iC8
131
where r. z:< IR. - r ] and '. = !Rr. - r]. Using these definitio ns, we
write
2R.· r (&-30)
r, = R. ( 1 - Rd'l + Rr o")" ~ R. + L.' r
r = R, ( 1 - 2Ro·r
I I,' + R.' T')" "" n. - 4· r (6-31)
132
Icr terms in only in the expone ntial factor 6in.....
T
kN'P till' first -crc I ' I . I ' ~
. f t ' . U ·I· ~ sensitive to small rc nu ve C l u nges III p U1SC. II""",
tills nc or 1:5 q "" ~ ~ -c
",cnero
(r. + r.)' ce (B. + B.)' + 2(B. + B.)(i.· r - ;'. r) (&-32)
Using these approximations, t he kernel can be written as
m(~+~' ("')
w = 2h T' "'"
With u defiued llS above this means that the energy is mu' j2 [el. Eq.
(3-15)).
Differentiating the phase with respect to R" yields the wave nunlber
at the point b as
k = ~I1·+R.
I
h T ((;.."l5) I
which means that the magnitude o( the momentum is mu (ef. Eq .
I
(3-12)J.
133
ti me at which the elect ron would a rrive a t th e center of the a tom of °t
1ll0Vcd IIna CI DSiS 'ICD1 manner. I I
p. ee mul.. (IMlG)
a nd the outgoing vector momentum p, as
131
~- ' .....
~--
. --<,
Fli~·,6-16 IltillrilJIt~ of 1'01Ii'IiRtio n
-. ..
to t'ii lllilllll C I III' wro-ordee t eem at b Dill}' d « ttotl'
'" 11(' I 1I.\.~
·
Il("en tlC1I11' I " f' .
t tl.. :t eas t (JIII'e r un get (rom 10 II wi lh nil)' rt0 ' ltiU llllhlt- I'tOI I:I )1 -.
11 IWI t II~ Z'OfU-f)t dl'r 1<'f1 ' ' I I ' .
I lit\,·
. "
I I. " III Ie !wor(ur JIl II UII e X I)ll Il~ i" n II( 1\ 1, ( 11 II) will t"on ttihu 1e 1\ 11\
KI I e llmOU li l lind CUll he Ill...1• 'I ..1 '11 " . ' , .
-.. ce (.'\1, ie .tnt u-nn of nllpott ,ance 18 I\ UJ (b,o ),
Th e perturbation m edlod in quantum m et:llon if"1J
135
b
FiB· ':7 It d nntl II nrc the Mme lotn l dis tnnce from 0 , nnmcly, n", t hen th e dilTerence
(or.mt lo) of num bers of [-k~ t ~o ml nrrivilll{ nt t be t wo point./! enn df'pend only nn t he N'Ilt _
t (-ntl~ p lien_o~lcnon . If d 18 10 . t ~c d irect line of Jlo nSf>nHf'n'f1 f'I<Ttm ns, t he rul in of t he
n~ ~bM' n rrJ\"l n~ nt b to t hat ern vmg at d if no sc::at lcring source were prctlf'nt is t he proba-
bility o f scatler lng tv b.
60 that
We shall int erp ret the last factor geometri cally in t he next sect ion,
where we shall also give more detailed attention to the function v(q}.
region d, where t hey arc sprrud out over fi ll nrcn « N.. + R6 )/lt P 11(1.
Instead th ey are Ke it t out ill n solid Angle flU toward b uud nrc therefore
, Rp rend out o ver an urt-n Ub2 tlU t here, us shown in Fig. H-S. Hence
t he ratio of thc probability of find ing Ull~1U at b to that of fi nding t hem
,
I
Quant..", mrdla"icl!I and pat'l inl,.grob
136
FiJl. 6-8 Pnrt id l" !'t rild nJ;an ftrt'l\ cI. or t he t arget ar c clc n('("tcd t hrough An l\nRIc II into u
measured l.~· t hr- w li\1 angle Iln. If no t nrgel hed been present, the pert icles Vo"OIlL!
ft rc::t.
have proceeded 10 t he point d. IIl....tend, they p roceed to t he po in t b, lSprcadinK OUl ltllo
th e nrCt Il,: eu. TIle proh.. . bil it)· of finding II port icle 1\1. (I is inve rsel y Ilroporbonul to the
An'.:l 0 "<"1' which th (' bea m would bnve sprcnd in ar riving lit d. Simila rly , t he proLnbilit)' 01
finlhng the p:lf(K-!" at b Is i n\"('~ly proportions ) to t he a fCR R ht tin over whic h t he he3llld
SCo"\ttl'rro pnrticlcs spreads in tr:n-d inp; from t he t arget t o b. If we take t he ra tio of thlS!
areas, we have the inverse rnt io of t he n..."SOCit~t C'( 1 prcbubilit lee. From t his point or vitw
we f.lI,)' that all or the pnrt iclcs ,,·hich I.it the ta rget area dtT arc eent te rcd, and through it.
p:u1.iru b r angle O. Of course, nd \l:lllr only n Icw particles wh ich hjt the t ar get nrc seal.
tcred at a ll and cnly 0 fraction of t hese through the e ngle 8. Thus, t he eree r lemenl d.
,,"bieh we I.:a,'c used in this ce teulet lc n is th e rf/rdite eross-ecctio na! o ren for 8cattttir:g
tl,rougla the engle 0 measured in terrae of the clement. of solid angle dn into which the
parliclel lU"C 6CUtt crOO.
On comparing Eqs. (lJ-42) and (lJ-43) we see that the cross section per
uni t solid angle is '
da =
dll (2rl'rn)' I,(q)!' (6-14)
137
Ix'~l ind the ~rp;('t ili not t here .in th e classical SCIlBC (wit h sha rp hound.
nm~) : for smce we nrc dealing with a wove phenomenon th .
di.IT metlon
. .mto t Inc S1muow,
.1 ' ere IS
Sup pose 1'(1') is t he Coulomb potential Ze"t/ r. I n t his case the integra l
for I'(q) is oscillatory nt the up per limit . Bu t co nvergence of t he
integral enn be a rtificially forced by int roducing t he fa ctor e- » and
then taking t he limit of t he result as e - O. Following t hrough t his
ca lculation , show t ha t t he cross section corresponds to t hc Ru t herford
cross sec tion
~.
(6-46)
- 16(mu' / 2l1sin (8/ 2)J'
where e = cha rge on an electron
q ~ 2p sin (8/2) ~ 2mu sin (8/2) (6-4;)
8 = angle between the vecto rs io and it.
T he result of Prob. 6-6 is, accidentally, exact , That is, the f irst-
order Darn approximation gives the exact value for th e probability
of scat teri ng in a Coulomb potential. T his docs not m ean t he higher-
ord er t erm s a rc zero ; it Ulmus, ruther, that t hey cnnt rihut c only
to the phase of the scat tering amplitude. Since t he proba bility is the
a bsolu te squa re oC t he am plitude, i t is independent of t he phase. Th us
a first-o rd er Born npproxima tion, which gives t he correct value for t he
proba bilit y, is not exact for t he a mplit ude. This case of a Coul omb
sca t tering is amusing , for t here is also a nother nec ideut. A complete ly
elcs..ieal treatment of such a scattering problem, i.e., t rcut tug the
electrons as charged point mas..scs, gives t he sa me result .
138
By assu ming that p(r) goes to 0 as Ir l -> ... multiplying E'I . (f>-l8)
by cxp (iq ' rill) , nud integrating tw ice over r, show that v(q) call b,
expressed ill terms of p as
The quantity in the brackets is coiled the form faclor for elec tron
scattering. (Incidentally, u similar form factor appears in X·ray
scattering. The theory of X-roy sca t tering shows that only the atomic
electrons, and not the nu cleus, contribute to t he scattering. Thus the
form factor for X-ray sca t tering is the same but with the Z umitted.)
In a u atom the potential follows the Coulomb Jaw only for \"cry
small radii. As the radius is increased t he atom ic elect rons gradually
sh ield , or cancel out, the nuclear charge until, for sufficiently lnrgr
values of r, the pot ential is zero. Th e shielding effect of atomic
electrons con be accounted for in a very rough approximate manner
with the formu la
ze
VCr) = -' - e- b / o ) (6-51)
T
I
In this expression a is ca lled the radius of the atom . It is 1I0t t he
same as the outer radius of t he atom as used by chemists but instead I
is given by ao/Z~\ where aD = fI,!fmc= = 0.528 A. ' I
Problem 6-8 Show thnt ill such a potential
-h-Ze:h z
.(q) = q' (fo/a)'+ (6-52) .
and hence
- z"lmu'
rT -
2 2
A']1-'
sm .8)' + __
e - - [.... (.
(pa)'
«(j..\~) I
The total (,TOSS sect ion rTf' is defined 8S the integral of (T over tile uuit I
,
sphere; thus
•.
rTf' = f.
0 rT dn
The perturbotioJ1 method in quantum mechanic.
139
In the present example show that tTT is given hy
Z'c ' / (2 11~)'
~T - .a' I + h'/ (2pa)' (6-,';5)
(6-57)
where fA. and In arc the amplitudes for scattering by t he two a~oms
individually when eac h atom is located a t the center of a. co~rdll.lnte
system . The atomic biudl ng docs lIot change the ~1U\rgc d,s~rlblltlons
around the nuc lei very much (except for very hght nuclei such as
hydrogen) because t he binding forces uffcet only u few of the outermost
electrons.
t
1\0
· )''.( 1· ("II....'.-, ) , ~JIO\\"
{TSill':: .
t ha t the prohnhility of scattering at n pnr-
t iculnr vuluc of q is proport iona l to /A2 + IlJ 2 + 2fAiv cos (q • d), where
d is a - b.
Proble m 6-12 Assume tha t I' (r) is independent of time and show
~t the t ime integral of t he second-order scat te ring t erm K C
%J(b,a)
gIves
141
------- -- ~-
b
Fill' 6--9 T o incrc:'se t hf' neeurney of "eRHc-ring clllr.ulnt ions, w e cnn t ake neeeunt of
SN'Oru l~ rdcr 1,I'n ns In t he perturbntion f"xpnn"ion, I kre, lllI in Fij(, G·;! (3), we pir ture t he
clcd ron ns IIf"lnJ: ~n ll("r('f II\ 1 two M"P:IIIIIC points in t hl' nto mie rw,lf>nli nl. Thus Ihe etce-
tron sta rts A I 0 : IlfOl'C'f"'tls n.~ A Free p., rt ir le t o r, where it is SNl llr l1~ l ; tl u' n moves ns 8 free
p-, rt id e t o J, wbeee it is S(':IltCrM n ~ain; nn.1finnII}' moves ns a free p.u tid c t o b, where It i!;
rollcd cd h~' the count er. 11u;' poiu te C a nd d eun lie III a ny posit ion in epece. Th l' n(omil'
polNlt inl n.t t hese positio ns dep ends u pon the md iU!l veeto ra r, and r... men.suroo from the
center of the et om O.
- R... The magnit ude of the momentum is P, nnd it is approximately
unchanged by a u clastic sca t tering of the elect ron from t he (relatively
massive) atom.
One might expect t hat in a situat ion in which the Born approxima-
ti on is not adequa te it wou ld be worth while to compute the second-
order term as a correction. But in practice it seems that in this
application Eq. (<h19) is a kind of asym ptot ic series. If th e second
term ma kes an appreciable correction (say 10 per cent or more) thc
higher terms arc not much smaller and the true correction can not be
gotten easily by t his method . Of course, if it is a problem in which
th e errors of the Born approximation nrc small (sny less than I per
cent), the second term will be adeq uat e to fi nd t he correct ions.
Thc Wm.-c Function Treatment of Scatteri ng. I n t he sca t ter-
ing experiment whi ch we ha ve described we ha n assumed that t he
initial st at e of t he incoming electron wus th at of n free particle with
moment um P<I. w e hav e assumed that the va lue of t he momentum is
det ermined by a t ime-of-Ilight technique {i.e., the total t ime required
t o travel the distance fl.. + HI. is '1') .
It is not IlCCf 'S.... ury to usc such a techniqu e. Any device which
ena bles us to det ermine the momen t um is equally sntis faetory . So
su ppose we gcm-ruliac our pict ure of scattering phenomena with the
help of the wa ve fun ction l1lC'thod.
SUPI>O~ the iucomi ng c11'CtroIIS a rc known to . have mOm~llt ulll. p..
and clIC'rJO' H.. = p:t/ 2m. T hus t he wave [unct ion for the mconu ug
electrons is
I (6-00)
i
QUlin'" PI m('{".oll ;e-S (.lId fUI' 'I ;n ,cgrnltf
112
Then usr-ug tlte f rst two
" terms or Eq, (6-2;'), t he wave fun ction 1_
Ul
first order for the outgoing clef.f.ro ns 18
",(R.,I.) = e h /AJ,•.• "C- ( i /A llih
The first term in this expression reprc~:nts t h.e wn~e of free part icles
which have passed through the poteut.inl region without being sent-
tc rcd. The S<'COnd term is th e amplitude of the scattered electrons.
II 4-. represents this term, then "'. is t he scattered wave.
m . f" Tile
+ 2".h- .!. e,·/A,,,·V(r,)e""" '-" -l (&-62)
wbcre n, is the dista nce from the final point b to the variable point of
integration c and p is the magnit.ude of the momentum of the elect ron.
Once more, suppose that th e pote ntial drops to 0 for distances
which arc short compared to either R. or R.. Show t hat Eq. (6-li2)
can be written as
.'-In
'1'\6'6,'") = e (" AI.l'aJ1e (i /Al p. ·R1 + f eWt.J
fl. ~ R.
(6-63)
where the scatte ring amplitude J is defined ill t erms of v(q) [sec Eq.
((;.:l9)J us
m
f = 27h' v(q) ((;.IH)
1Th~1 last I~rrn 01 Eq. (lHl3), UI R.) exp (ip R./ h) can be thought
Of as IC ~pntllll part of the scattered wave function . ' It has t he fonn
0, a ~PII<'1'~IJ CBI Wll\'~ rudinti Tlg outwnrd f rom t lte center- of tht. srnUcring
a om. Ie :llilphlud 1 II · ' . I r
. e (J Wi spherical wave nt 80 III e P:lrt ICU It
scatteon g nllglc t}('Pl'fI(Js tl . r
whiel L t.:t ( .upon mt allglc thruugh the funct lOlI 11
tl I, Y, I· Q..(H ) , vnrlC'S with the 11I0 1l1('1I1U111 t runsfcr fl. 1'hus
IC eomp etc ,,"U "C Iuu t ' f l J"
th ou ht of C roua or t ie r lceh olls ufter sca ttering ('nn
g US the sum of two ter ms. T he first ter m is the plane ,,"pvc
TI,e perlurb,"ion med,ad in quonlum mecllOui~1I
H3
Fljt'. ()..IO A benm of elect rons. represented by its equivalen t wa ve, moves toward the
at omic nucleus at O. TI, e sl ro n ~~ t fm etion of the bea m 1Il0 H 'S on und ist urb ed lL'! n plane
weve with mom entum p... A ~Rl1\ 11 fmc t ion of t he beam is sca ttered from the nto mic
nu cleus an.l mo ves n"'ny from 0 116 a l'phcri l"nl WrlVC pnttl"rn. The I'C9tllting st l"l"ngt h of
th e' ..'nYC of clcet rons (t he number of elect rons) a t some point l.o loca ted at R... measured
from t he a t om ic nu cleus 0 , is t hen m ade up of t ...c porta. The fi.nl t is th e ncnscnttercd
bea m given by n ,e plane wave exp (rp.. • & /Ic). T o t his is a dded t he sca t tered wave \\'ith
t he aphcrice t ton» giv('D by ( 1/Ji~) exp (ipR,JA) and multiplied by t he funct ion J. ",,·h il"h
det erm ines the a ngular dependence. The com bina tion of th ese two waves gives t he
spatial part of (he sca tt ered wave fun ction .
of nonscattered electrons, exp (ipCl · ~/h), and the second term is the
spherical wave of scat tered electrons, as indicated in Fig. 6-10. Use
this point of view t o derive t hc form ula for t he cross sect ion IT.
PrQblenl 6.14 Usc the wave function approach to d iscuss t he
scettcrlng of an electron from n. sinusoida lly oscillating field whose
po tential is give n by
V(r,O = U(r) coo wt «(Hl;;)
IJ.l
ed kernel can be expanded 118 (now in one dimen..·
t hat the unper t urb ~l()n
for couvcuictlre)
Kv(2.0 = L ¢.(r,)¢=(rJ'-(;··"'c.~',) for I, > II (fHoG)
•
. t erms 0 r tbc c,'gcufunct iol1s t1J.. and
10 •
eigenvalucs
•
E.. of til e unpertur
•
L •
U\.'Q
pro II) cm. Let us look a t ou r senos for K 1'(2,1) after •su.bst it uting t1.:-
IICj
('xpl"C§ion for l\u. 'Vriti ng ou t t he first t wo t enus, It 15
K,·(2,1) = L¢.(r.)¢=(r,).-(;··""'~'"
•
- i LLf ¢.(r,)¢: (%.) V(r ,,'.)e-C·· ·,."'....,'¢.(%.)¢:(x,)
• •
X e-C" ~''''''''~ d:r. dl , +." (6-6i)
It is clear that in each term the Xl will appear in some wave function,
Iike 4': (z. ), and the x2 likcwise, so we can always write K v in the form
(6-68)
• •
where th e A's arc coefficients depending on t f , ll. We shall call these
coefficients tranBilion amplitud es, To zero order in V, this must reduce
to K u, 60 to this order A.... = 6.... e- h l:'..IlJ(Is- " J. If we expand Ain a
series in increasing orders of 1', we have
... (6-69)
(&-72)
= L~,.4>,(xJ (ll-73)
,
That is, the wnve function at 12 is in the form LC... r>... (x z) •
•
This expan sion in t erms of cigenfunetious was first int roduced in
Eq. (4-48). New we can assign n deeper menning to the constant s e.
'Ve can interp ret e...
as the amplitude t hat t he system is in state 4J .
In this particular case, e.. = "-..... is th e ampl itude to be in th e state
¢m at time t, if at II t he state is ¢ n'
\Vith no potential acting, a system once in state n is always In
state n with un amplitude varying in ti me. So, to zero order,
'" _ ., ...-(lE,J A)(t r' J
1'0 ..... - "' .....c.
We can interpret the fi rst-order term by the rule (Fig, ()"11 ) : The
amplitu de to be scattered/ rom slale n to m in a lime dl is - (i/ h) I"mn (t) dt.
Problem 6- 1 -t In terprct
....
Eq (6-i2) by explai
.. . . •
ning t he meaning a
c: la in and verify the equation for the sccontl-<ln!<:f
ench tr-rm. TI jcn exp
cocfiicicut
).. ..
en = - .! J." [J.loh L'\' e-(·'~)B'..(I'- 'J l' Ml (t.)
11: II
•
X e - (i /",EI(t.- I. )V.... (lJ)rWA
1B. (',-,,) dl a] dtt (6-74)
•
'
L• .
dl. >". (1,) = - ;; e',""E.-E.I',V.... (t,)~ .. (t,) - T. Em~m.(t,) (G-iG)
First-order Trnnsitions. First, let US take the case that the final
state m is different from t he initial state n and let us consider only
the first Born app roximation, i.e., the second term in Eq . (~69) .
The result will be applicable for small values of V. The amplitude
that we make the transit ion from m to n is
(&-77)
1
,
c' , IAU /(.- K..l T -
f ""V.... /.'.... _ n.
1.'
(&-78)
L·
Quant"". n1f"f"n"ruc. d nnlh in If-lrals
an ,--
148
t VI
o T
r
F'i 6-12 The potential aacet inl/.: the lrnD8it ion from m to n is turned on a nd ofT sloltty
As lh~ ti~c factor Lccom~. smoother (e.~.•48
\\":5; the lime ytlri:lt ion f (O, t-Ilown here.
discont inuitil'S llppc.U in successively higher dorivetivcs) t he proba bl ht.y of n trn.tlSllion
becomes 6runllcr.
If it should happen that E... and E.. nrc exact ly the sa me energy, we
find P (lI - m) = IV..... 12T 2JIl2. T his RroW 6 IlS t he squa re of t ile t ime.
It mea ns thnt Do concept of the transition proba bility per u uit time is
not mea ningful in t his casco T his formula hok.l s Dilly for T short
enough t hat V..... T « Ii . I t tu rns out t hat. if only t wo states of exactly
the sa me pertu rbed cnergy nrc invo lved , th e probability of being
found in t he first goes as cos t C1 V'",.. IT/ h) and of being found in the
second as sin 2 (IV..... ITI ii ), while our formula is only a first approxima-
tion to this.
, . vT + . v)T · .. vT
1\11 = - "-,; 1. BAa _ ••• = - I SI" T (6-82)
, Such systems swing ba ck and forth from one sta te to the other. A
, further conclusion can be dro wn from this result . Suppose the pe r-
t urbation nets for an ext remely long ti me so that I".....T/I. » 1. Th en
if t he syst em is invcstigutcd at uu arbitrary t ime T , which is somewhat
indefi nite t he proba bilities of being in eit her t he first or second state
arc, on U rI C average, eq unl. That is, a small indefi nite perturbation
acting for a very long time between t wo sta tes at t he sam e energy
makes t hese states have equal pro ba bility. Th is will be useful when
we discuss the theory of stnt lst .ical mce lumics ill Chap. 10.
T he case of great importnucc is thut ill which t he values a~lowcd
for H.., t he energy or the fi lial sta te, nrc 1I0 t scpnmte nnd discrete
but lie in u continuum or at lcnst nrc extremely closely spuccd . Let
, us bay that pCB) dE iafh c nu mber of levels or slates in t ile range of
f
Quan ,.,nI m H'honicI' and pa'h in'Plrnll'
ISO
energy }; to E + (IE. Then we ca n ask for the proba bility to go to
some sta te in th is continuum. F irst we sec th~t to go. to an y fltnh!
Icr which E.. - H.. is la rge is very un likely. It IS most likely t hat lhe
fi nal state will be one of ncnr jy t he sa me energy as t he originnl £..
{within an error ± 1-~....). The totnJ chalice to go into any state is
Since t:
[(sin' z )/ z'1 d. = w. t he integra. of Eq . (G-84) has t he value . .
rT12A and we obtain t he result that t he probabiJity Ior a t ra nsition to
r
I
t
I
•
~
l
o 3. t
Fig. 6.13 10 t1IW figure t be cncrgy di ITcrencc J: _ • • .
When t hese two ~nerl;i('!l lire npp roxirulltel '" E . IJ! ecnluced by the vnrmblc. r.
(Kin t 7.)/7.' IlJIJlrood lNi it t! m n xim UIn v al ue ~, C(lt'll (tllU.", :z IH very fll n ull) t h e lund ~1I \
IJCI:Omc.!l Very Kma ll TI LUK In exp . • ' . or fl IRt· va luf'>! 01 tile d ilTerence the fun d IOn
. • ~ rl'tillIOnll Ulvolvil U · I
oon trihu tion ll eome froUl ti le ('t.'u tru.l .....,.;". I I
' .
. lit I ~ un enc n, t he Ill0000t Im portll l1
I ~
.
n p proxlIlIll. t ely t'fl uni .
-.0-_"' I " t L't . U..
I... r 1'8)00
' wItere th e two cnerg J{'
. ::l ...-
•• • I.
I.
ri,e pertu rbation method in quantll.n& mechoniclt
151
"
I some state in the cent inuum is
':
l Pen --+ m) = 2rIV•• 1' p(~.)T (f~85)
\
and that the energy in the final state is th e same ee the CIICrgy in the
origina l state.
From these results we can write the probability of a transition per
unit time in the form
(6-86)
Then when we sum over a group of states m, only those with E.. - E..
survive. Since L f dE. p(E.), we get as • resu lt Eq. (6-86) .
m
--+
Ip,' ~ Ip,l ~ p
Quonlll ni 1IIl'ChaniC'1' and pndl integrala
152
ny our usual eonvention for writing c.! ilTc~ntinl elemel~ L~ of 010nJcn. "/'
"r of states which ha ve t heir momenta 111 t he volume .-I '
t urn , IIIC 11\ I I1,1...- • • ~ )a _ '! d tlf! ;'
clement of momentu m space "1>, IS d~J (2"'1 .- P P / (2rl)', -/
whore tin is the (') (' III (,l1t of solid a ngle wh ich COtl~tIIS th e m Olt lCl\tu lll ;!
vector p:. An cl(,l1l('ut dR of the energy range IS couucctcd to the
clement of momentum space by ~
• d
dE - d p. - p p
2m m
Thus the density of momentum states for particles traveling into the 'f.
(ll-89)
,
solid a ngle em is
(6-00) rJ
..
"
Subst ituting these relations into Eq . (6-86), we find the probability of
t ran sition per second into the element of solid angle em is given by
*'ri·
(ll-91)
•
j::
\Ve define en effective target area or cross sect ion for scat tering
into d!2 8 S da (eL Sec. 6-4) . T he number of particles that win hit
this nrea per second is th e cross-sectional area times the velocity
of t he particles coming in, U l = pJ m. This is because we have started
-
,.
with wave functi ons 4>" which were norma lized. on a u nit volu me; or ill
other words, the relative probability is unity to find one particle in
any un it volum e. Thus
dP P, ,'1.
dl an = u 1 d4 = m da (ll-92)
du (m)'
2.';; I·(vl'
dlJ = (6-93)
'.
which is just what we obtained in Eq . (6-44). "
't
Tllr prrlurlmrion rnellwd in quunlum mee::hun ira
153
dP (n - m)
- dt-
2" 1 I .
- -" M .-. '[1(1' . - E. - hw) + I(E. - E. + hw» )
(6-94)
ProlJlelll 6·25 It has been argued that th e equations of th e clec-
trodynntnica must, like those of mecha nics, he converted to a quantized
for m On the basis of ti le photoelect ric effect. Here nn elect ron of
energy liw is occasiona lly emitted from a thin layer of meta l under
t he influence of light of frequency w. Is this impossible if matter
obeys t he quantum laws but light is st ill represented. as a conti nuous
wave ? What a rguments call you adduce Ior the necessity of giving
up n cla ssica l description of electrody namics, in view of the results of
Prob. 6-241
Problem 6-26 Suppose we have two discrete energy levels E 1
and E 2 , neit her of which is in the continuum. Let a transition be
induced by a pote nt ial of the form V(%.t) ~ V(%)/(t). Show that th e
probability of transition is
P(I- 2) - IV" I'I"("',>I' (6-95)
if f (i) is representable by the Fourier t ransform
. . d",-
I (t) =
J- . " (,,,).~.
d.
(6-96)
the exposure t ime nnd (2) the intensi ty of light nt the frequency
(E, - E ,)/ I, of absorption,
e+(·J~l(E.,.lr-E.I,I'\
~.....
(I) -
- - h! L
1~ V IT
""" '"
J.
0
T dt
•
J." dt
0 '
• X e UlA)( F...- E ll ' oe(l/ A)( E, - Jr.)I,
e WA HB. -Ea)T _
E. _ E.
1) (&-98)
V...... V...
M._= V•• + LE E• . (6-100)
• • -
•
- 1E
The first integral is that of an odd function and vanishes. The second
approaches a finite lim it as T -:-+ co (and therefore as 6T/ A_ ao).
'That is,
2 ,·!..• . sin
- y-Ydy-- 2'"•
.'
QmUltuni mcchanics and pada intcgra"
156
so no large transit ion proba bility occurs. A large effect ca n arise only
in rose H.. a nd b'.. nrc essentially equal, (or th en t he dou ble coincidence
of t he two pok-s from (E" - H.. )- l nnd (R... - 1':,,) - 1 call mnke t he
second term important. T herefore, we continue t he a nalysis, D..<tsum_
ing E.. a nd 1:.. nrc nearly equal.
T he SUIlI of k in Eq. (~!)S) can he divided into two regions by
choosing n very small energy 6 and brea king the SU IIl up into n part A
for which IE" - E"I > .11 and a part n
Ior which lEt - E..I < 6.
We choose Ii. to be small enough that th e fact or V...t V"" docs not vary
appreciably when 1.'" va ries ar ound E.. over t his energy m ngc 26.
This is some finite energy, and we shall take T so long that hIT « 6
~...hich menus that IBn - B... I « Ii..
•
F irst for part ...I, IE, - E... l > Ii.. Then t he second term cannot
become large; for its poles arc avoided. Only the first contributes
nnd t he cont ribut ion is J
,,", - I T
a x Ii (&-102)
The sum extends over all E, except for within + Ii. of E ThO .
I'd d - ... 1S sum LS
valu y.m cpcn cnt of ~J ~nd as 6 - t 0 it is t he dc6 nit ion of n principal-
va uc integral, That 15, In the limit 6 - t 0 we can write
where Pep is th o. I
to .. . . c prmerpn part and we have reinstated t he first-order
rm, III case I t docs not va nish.
For the region B we take V V t b
..." ,,, 0 e constant at its value for
E - _ . CI!J
A: E.. - O. That IS, we repla ce tV..." l' ""F(E,) by
b= f V••V••~(E. - E.)
(6-10., )
Tile perlu,.bolion mell,ad in qllonlilm mecl,n n lCIt
157
"
and
,.
\
I = f E. + 4 dEl
} g. - . E. - E.
( Cf i/ ll CIl'. - Rol T -
E ... H...
1-
e CIIAj(Il'.. - Eo)T -
',' J'~ .
1) (6-106)
L ... -
Now we put (E. - E .)(TI t.) = % and (E. - E .)(T I t.) ~ y so that
(E. - E. )(T I t.) ~ % - Y. to get •
I _ T
h
f 4 TIA
-'TIl Y
dll (e 1_e"f
U
-
%
r
%
- . ) -
Y
1) (6-107)
The lost brocket CDn he written (EI< - E", - if)- l in the limit as
II! --+ 0. as we have writte n in Eq. (6-100).
From E q. (6-100) we lenrn then that even if 110 direct transition
is possible from n -+ m, nevertheless the transition can occur, as we
MY, t hrough u vi rlual state. That is, we can imugiue that the syste m
goes from n to k, then k t o m. The amplitude for nn indirect transi-
tion prec ess is given by Eq . (G- 99). 'Ye note that it is not right to say
t hat it actually goes t hrough one or nnoth r-r intcrn u-diutc state k,
hut rather that in charnetcrist .ic quallt ulll-lIl('c!ulIlit·nl Inshion there
is u certain nmplitudc to go via the va rious intcrmcdiutc states k , an d
the cou t ributjous interfere.
T he intermediate "U1U~ nrc IIOt of the snme energy ns the init ial
.1 and final states. T he conservation of energy is 1I0 t violated , for tho
•
158
Sometimes 8 t ran sition cannot take pla ce except by the use of two
or more intermedia te virtual states. Analysis of such t ransitions
requires the calculatio n of third- a nd highcr~rdcr terms in the pertur-
bation expansion.
(6-113)
AE. = ' ''_
I
This derivation of the first-ordcr energy sh ift is not satlsfnctory
if the system is degenerate, i.e. if there arc init ially very many states
of exactly the sa me energy. 1t turns out tlwt in such a case terms of
,.'
second order in V give equally large effects.
Addiug in the socoud-order t erm in the perturbation expansion for
j
the transitio n clement gives
160
rformcd easily to give thc result
terms k # m can a Iso IlC pc
Iii \'....
T - 2h!1 v.....2 1 2
e-(· /l IE.. f)..... = 1-
The first three terms On the righf..hnnd side of this,c,q ua t ion rcpre-
sent an expansion . tlIrO ugh second order of •exp ( -, I • • T/ Il). b . The
first of the summation tenus, t he onc proportional.to T, e~n C Inter.
preted as a second-order energy change. T hat is, t~c mcrcm~n,ta1
energy IS. no t Ji USt I' ......, but contains higher-order corrections.
•
' Vr
b ltlDg
.
out the energy correction through second order in t he pcrtur aticn
energy, we get
liE.. = 1 .... -
r
...
~ V..I' ..
L E.. _ £ "
This Inst equation gives the correct expression, through second order,
(6-116)
f of two levels each of which has a decay rot c .." the resonance is not
I sharp but has a defi nite sha pe. T he cente r of th e rcsona nc:c determines
the energy difference, and the width of the resona nce gin 'S the su m
of th e ..,'s of each level.
,
"(
I
I
I
,I
I
•
7
Transition
Ele:ments
I s the preceding chnptcr we developed the eoueep t of 11 ~crtUthnl ion
• 01 sta te in n qunntu lIl- lIIcchnlllcal syslc m
t n-at mont 1or c I1R1l ~'''' • • • ••
Wt' ca rried out nn in\"('St ignt ioll of this method us It I: n~phcd to
systems whose ullpprturhcd 11o.milton inns n rc cc nstnut III tllIlC. hi
this chapter we shn ll conti nue t he c.JC'\'C'lopmc nt of the perturbation
concept a nd genera lize the trea t ment ~ cover . 8Y5t C~llS ~'hcrc the
u npert ur bed state may have n h31lli1tomn~1 vary mg With time. We
sha ll int rodu ce n more general type of notataou a nd att empt to bronden
and deepen our underst and ing of the ways in wh ich changes of state
ta ke place in a qua ntum-mccha nical syste m. T he notat ion to be
introduced applies to a type of fun ction which will be d efined in th e
first portion of th is chapter. The function is caned a trcneitum
d ement,
The cha pter is divided into four parts. The first part, comprising
Sec. 7-1, gives e definit ion of tran sit ion a mplitudes nnd transition
elements, with the help of examples based upon the perturbation
t heory of Chap. G. The second part, comprising See s. 7·2 to 7-1,
gives so me inte resting general relations a mong tmnsition clemen ts.
T he third part, consisting of SC'C . 7..JJ. shows the con nection between
tra nsition elements defined with the help of path integra ls and the
t reatment of qunnt um-iuccha uical transit ions d efined in terms of the
more usual operator notat ion of qunntu m mcche uics. I n t he last
pnrt, comprising Sees. 7-6 and 7-7, the results learn ed in the preceding
sections are applied to two interesting problems of quantum mechanics.
Here F is any functional of xCt) which docs not involve x(t) at the
end points XI or X 2 or beyond the end points. In th e special case that
F = 1, the integral of Eq. (7~3) defi nes a transit ion amplitud e.
i It is difficult to und cratn ud transition clements on t he level of
"
intuitive physics. Onc approach to ward s~eh und~rstnll~i llg ill \'ol~'cs
I a classical analogy. Picture a small particle movmg WIth brownian
"
166
ever the va riables %1 Dud I 2 of th e initt ul and final states-a step I
,•
<xIV[x(t),tJlIf)s, ~ J.,JJ x · (x,)e"""V[x(t),tlHx.) dx. <lx, :!>x(1) (7-9)
u - I."I.',.
1 wT it
77kol Si n
0[x(/),tjO[X(8),S] sin ,,(I, - t) sin "(8 - I,) d8 dl
(7-17)
with g(x (t),tI characterizing the interaction of the particle nnd oscil-
!? lator, und T = i z - t••
'Ve have noted that path integrals involving such complicat ed
'~ actions are very hard to evaluate indeed; but if the effect of t he compli-
cated term v is expected to be small, we can obtai n useful result s with
l!
less effort wit h the help of t he perturbation expansion of Eq. (7-4) .
,.. To illust ra te, we find the first-ord er term in such an expansion {i.e.,
the first Born approximation) . Using Eq . (7-17) for v, we must evalu-
..
.' a te t he te rm (i/m <X I(l'I~}s.. This term can be writte n as
T his expression enn be su bst ituted into Eq . (7-18) to obta in the finul
resu lt for t he first Born approximat ion, (t"/ II)(x lo!1Jt)s•.
Qllonl,"n nted,oni~1J and pall. inle#rnlrt
lin
Tra nsition elements will come up more frcquel~tly in sue~ing
chap ters. In each example they can be ('vn luntc;I III t he st rnlghtfor•
ward manner " hich we have Illust rated here. Fer t.h.n.t reason , Ycry
HII IC' of th e material in the remai nder of this chapter IS really ~s(,ll linl
to th e work that follows. !\eY('rt hd eES, t here nrc two reasons for the
inclusion of th is material in this book. F irst , it is possible to ohtain
a very general relat ion bet ween transition clements. Th is rela tion
might well serve as all alternat ive starting point for the d efinit ion of
quantum mechanics. Second, for man)" people already fa miliar with
the more conventional opera tor notat ion of qu antum mechanics, it is
helpful t o 11:1\'e exa mples of the tran slation fro m t he more custcmnry
representation into that which is used in this book, such as expressions
of t he Ion» of Eq . (i-3 ).
Wit h the rules for tra nslation ava ilable, the subject ma tte r of tbe
later chapters, developed as it is from t he pat h integral a pproach, can
be a pprecia ted in ter ms of more fa miliar sy mbolic concep ts,
Th e relations d iscussed in the remainder of this chapt er are inde-
pendent of the form of t he wave functions which describe either the
init ial or final sta t e of the system, and they nrc used in defining the
integral of th e transition element. F or th is rea son we shall abbreviate
our notation by omitting a ny specific reference to these wave fun ctions.
Thus a transit ion clement will be written as {F)s instead of (xIFI'f').s.
This 8F/b($) depends on the functi on x(t), of course, and also on the
value of 8. Thus it is n functional of xC ') and n function of t ime 8.
'Ve may look at it another way. Suppose time is divided into
~,
." very many steps of small inte rval e, the values o f the t ime being
to (1~ +1 = e + I.) . The function x (i) can now be specified approx i-
mutely by giving the value e . that it takes on o.t each of the t imes I..
The func tional F{x(I )1 now is a number depending on all the Xl; that is,
,
"
it becom es an ordina ry fun ction o f the va riables Xi,
< F lx(t)) ~ F(• . • ,X.,X'H, . ..) (7-21)
Now we can consider its derivative with respect to one of these vari-
ables Zi, dF/az;. Our functional derivative is just this partial, divided
by e and taken at th e point t. = 8. That is,
IF I aF (7-22)
-- ~- -
Ix(.) • ax,
" This we can sec as follows. I£ we alter the path from x(1) to x(l) +
~(t), we ehange all the x, from x, to x, + ~. [where ~. = ~(~)J, so that the
first-order change in our function is
F(. . • z, + 1]i , Xi+! + 111"+1 •• •) - F(. . . ,:1;:,x'+I, •• .) _ '" aaF "'
L x.
.-
,4
,.
'.
•
(7-23)
,., ,
aj ~ ~ aj
~aXi
ax.
•
60 we can write for th e first variation of any functional
W ) (7-26)
6.r(. ) = ~(. - •
is given by
.W
~J (r••)
= [ _ 1R(r - r', ' - l,)j(r',I') dr' dl' ] F (7-27)
= 1F(x(I)le'"" ' " '''' + 1[I .s: ~(.) dS] eW" SI.'Ol
fu (t)
~x (,)
fu(t)
+ i1 [I ~:(.) ~(.) -l
Flx(tJl + . .. e"''''\' '''' :Ox(/)
(7-29)
expanding the exponential a nd d i~playing only to first order. The
z('n::ort!('r term ~ exactly (F)b' aga in, so the remaining te rms must all
vanish. In purt iculur, the lin.l-onler ten" must va nish for any 'J(')
so that we conclude the relation '
i(F ~(.)
- Ii
h \
/. (7-30)
T hia general relation hD.J:I many importnn t
consequencCB.
Tranllition e le naen t.
173
,. by parts to get
,. i JF ar,
as
J iJF
aXI:
e hlA}SI:ltOI :Dz(t) = -
h
el iJAIB l:leJ)I :Dx(t) (7-32)
"
.'
dropping the intcgmtcd part.
I• for t hen the specific varia bles on which F and S depen d need Dot be
:I ind icated .
,
" "roblf'nl r.s Argue that Eq. (7-3-1 ) may be misl('~ding. for :AJ.
(7':l:t) up pliea only to rcctnll~lnr eoonJ inntC's. ~o t his by studying
t he corres pond ing rclnt iou where spherical coordinat es, for exa mple,
nrc used und we wish to fi nd (iW j arl:)s'
Qllan!um m echanic. and pad. in!c,ral.
IH
S - l""lmt'
2"" - Vlx(t}] fit ) (7-35)
Upon application of the small variation 6%(t) to each pat h there results
(to first order)
&S - -
1""Inti: + V' (x)] 6x(I} fit (7-36)
Using Eq. (7-34), " 'C havc
S= .f. [m 2~ xJ' -
N- I
I~ ~\'e select a parti cular time II; and, as before let z be th . ted
position of a pa th • tl ten J a e 8SSOClB
as
dx.
= m ( XI;+1 -
•
XI;
-
XI; -
•
xa_l) + V'(x. }. (7-39)
Upon applicatio n of Eq. (7-33) there results
Since this result must hold for any arbitmry choice of h(t), it must be
that
(mx) ~ -(V'(x» (7-42)
,
at nil values of time. This is the quantum-mechanical analogue of
\ Newton's law. Making usc of the classical a nalogue for a tmnsition
clemen t, described in Sec. 7-1, this result says that th e weighted
"average" of t he mass times acceleration at any time " ave raged" over
, 0.11 paths with t he weight cI S / " is equal to th e weighted " average"
"
j ', momentum variables mi. In the fi rst terru the momentum IS evnlu-
,, I
QuantuhI m echanics and pori, in tegra'"
li6
and
(7-47)
These two term~ differ from each other only in order I, since they are
the same qU8 ~ t1t~" cal~ulat ed at two ti mes differing by t he interval f .
~h~.l\~~~~.J~~~e~:~hstituting Eq. (7-47) for the second term
/ r... . - Xl ) h
\m e (X' -X"') = i ( l ) (7-48)
AltemativcJy, we can write t his as
If some average velocity is dcfined for B short time interval Al, as,
for exam plc, tr (I + ~ ) - x (l » / .6.1, th e " mean" square value of this is
- Il / {im At). T hat is, t hc "mcnn" square value of a velocity averaged
over a short lim e interval is finite, but its value becomes larger as the
interval becomes shorter.
It appears t hat quautum-rncebenlcal paths are very irregular.
However, t hese irregularities average out over a reasonable length of
time to produce u reasonable drift, or " average" velocity, although
for short inte rvals of time the "average" value of the velocity is very
high.
,
t
.
"
z
. I rt id e are hir.=:h1y i rr('~1 1U' on a
. 1..::. 7.1 TY]liral Jlll t ~ or n. (I U/t ll hllll~lIl f r IItl lllm 1)11, . - • If I
. h n . lr bn u gh 110 11It'/t 1l \"l'll'lt'lly C/ln .(' ( co Ih('<l • n o
fine IICOk-, as Ilfl O W Il III tl l l' t;k rl r . \U", K I • . " , • • • _ n tJlld i rrl'n'nllul"'~
. '
mco.n-i!(i U6 rc vcloetC.)· CXI!lhl ut tin y IlOm • " In oUwr "'-t.rlf'l Ie pal I . ...
' •
.,
118
Problem 7-6 Show, for a particle moving in thrce..<Jimcnsional
space x, V, e,
flO
«r•• , - ..)') - {(y•• , - y. )') - «2••• - 2. )') = - im (7-.1O)
I I
2\m
(r'H•- X')'\I (7-52)
for this quant ity becomes infi nite as e approaches zero . H ow shall
we find an appropriate expression to represen t the kinetic energy?
\Ve might make t he heuristi c guess that only those functionals F
which might appear in some kind of a physical perturbation problem
rna}' be of importa nce. How ca n we get the k inetic energy by a
perturbation? I{ th e mass of the particle were perturbed by a.
factor 1 + 'PJ (with 'PJ very small) Ior some short interval of time At,
th e action would be perturbed by 'J 6 t(m/ 2)j :, wh ich is proportional
to t he kinetic energy. We are led to ask : 'Vhat would be the form
of th e first-order perturbation (o)s. if m were changed to m(l + ,,)
for a short time?
F or simplicit y we ca n take thc short time to be just E, the step
used to define th e time spacing; so t hat the first-ord er term divided
by ('PJ is the kiuctie en ergy- T he perturbat ion in S of Eq. (7-38) (it
thc m in the i = k tenu is changed to m +
'Jm) is clearly f7J (m/ 2)(ra,+ 1
- Xa,)2/f. 2, But this is not th e ouly cha nge in the path integral if m
changes. TIle normalization Iactora A for cnch ?It vary us m+h , so a
Inctor (1 + %,,) is introduced from thi s. lienee the entire first~rder
change i ll the path integral when m is 80 changed becomes after
d iv idillg~~1 '
(7-53)
whieh
.. •
should be satisfactory for ti ft timca., the ki 1(' t IC
'
energy,
. t SI.ng Eq. (7-49), one might expect this to van ish; but Eq . (7-49)
.1B valid Iionly
. as ( - - t 0 to the order 1/. . TI IC quan 1'1 ' I' (7-.53) is I
I y III sq .
10 teet , lIut: as ( - O. T he cxpn."sioll ce u be rewritten b)' cxpnndiug
the quadratic term. In Eq , (7-10) let P I 1£ of
)C X"+I - z". terms
Tronsilion element.
Thus we can define the left-ha nd side of Eq. (7-.'>4) us the transition
clement of the kinetic energy.
We see from this result that the easiest way to produce satisfacto ry
transition clements involvi ng powell! of the velocities is to replace
these powers by n product of velocities, each factor of which is taken
at a slightly di fferent time.
In simple pro blems the transition clements can someti mes be evalu-
atcd directly. For such proble ms the same results can also he obtained
'. by using the relations nmo ng tran sition clements which we derived in
Sec. 7·2. These relations may supply us with solubl e differential
I
equations for the transition clements. We shall give a few illustra-
,• tion s, but it will be readily seen thnt the examples for which the
·'. method works must be so simple that a direct evaluation would not
really be much more d ifficult.
For our first example, consider the case of a Iree particle going
I
, fro m X l to X 2 in the total time interval T . Let us find the transit ion
element of the position at the ti me l , that is, x Ci) . Of course, this is
·• some funct ion of l and it is clear that
(x(O» = x,(I) (x(7') - x,(I) (7-55)
"
Since any potentials act ing on the particle arc constant in space (i.c.,
no forces act), the second derivative of the transit ion element of posi-
-
.. tion is zero in accordance with Eq. (7-42). Thus an integration gives
I
./
Quant.. n. 1llf'c1lanicle and path intesrab
Ill()
Thus the ~rst derivat ive of the function I with respect to t jumps
by the quanti ty A (B) - O(B) lIS I goes from just below to just above B,
end In accorda nce with Eq . (7-59), A(B) - a(B) = A/mi.
Thc boundary condit ions sta te that
a'f
~.
= (!!...)
~
~(/ _ B) ~)
+ m~T I(T -
(7..()4)
= [ X(/).t(B) .) ] (I) for / <B
Transition clemenrs
181
182
g(1 ,
.) = . ~ wT sin ws sin w(T - l)
1n1"" Sin
8 <t
S m
="2 f :i;
2 m",'
dt - 2 f xtdl
then
m;;; i;«
X [ 2 X' . w(1 -
" t) ern I.) dl 2% , f," f(l) sin ",(I, -
+ m;;; h I) dl
- m~w' f,:' f,: f (l)f(o) sin w(I, - I) sin "'(/ - I.) ds dt]l)
In this and the following sections we shall sec how transition clements
look in the conventional notation of wave functions and operators.
Th is will help the reader who is familiar with that form of expression
to relate the res ults of path integral calculations to other results that
he already knows.
]f F is n function only of x at a single time, say, the function V(Xt)
at time tt, we know from Eq . (7-10) how to evaluate its transition
clement. Sinrilarljr, if F depends on the value of xCt) at two different
times, Eq. (i- I5) tells us what to do.
Let us consider next the case that F represents the momentum at
time tl; and make usc of t he approx imation that the time axis is cut up
into slices of length f . Thus
m
F = - (Xl+J - %1)
(7-72)
e
Then we have
Jx'(x, I Jx'(x,I)2\I-(x,l) dx
+ .)2\I-(x, I + .) dx =
- ~. {f x '(x,t)x(ll\l-(x,t)] dx - J(ll'x'(x,tlJxl1-(x,tl dX} (7-77)
Dy Eq. (4-30) this lust integral can be writ ten 118 Jx' (x,tl lll2\l-(x,l») dx,
or more simply we have
186
. this. new. vana
• hie m ther tha n the old Xl will not alter the tmnsi~
Using
tiou a mplitude of Ell . (7-82). It becomes
J.
N- I
<>:11 111-) ~ J:. J:. J:: X(XN,/H) exp Ki) S[X••,.I••• ;X.,t.J
+ (i) Six,. I,; x; - 1>. I) III-(X; - 1>. I) :ox(l) dX; dx, (7-84)
where the path integral for th e kernd bas been written out explicitly,
using t he methods of Eq. (2.22). , . .
X ext, we ex pand SlI !, I! ; x~ - A, I] and ti'(x• •- A, t) in Taylor ser~cs
and keep only thc first-order terms. In pnrtieular, the exponent ial
(unct ion becomes
exp (I (i) S[x. ...I...;X••t.] 1(1 - (;.) I> a~ S[x,.t, ;X;.I,] I (7-85)
(x 1- a~, S[X,. 'I + '; x,. t,] III-) ~ (xiii ~ all-~:~tl» (7-87)
(7-88)
which is the same result as that obt ained in Eqs. (7-78) a nd (7-79).
Tramrition element.
187
\
Sometimes working with a complicated S that results perhaps
from the partial elimination of interacting parts, we would like to
identify the functional p(/) whic h corresponds to thc momentum at
t ime I . The work of the preceding paragraph suggests n. general
definition. T he first-order change in the transition amplitude (xl111J'),
if all coordinates corresponding to times previous to t arc shifted by
-A, is this A times (xlp(t) hf.). From this principle the momentum
\ functional may be found for nn arbitrarily complicated S . In a like
mBIU1('r, the hnm iltoninn or energy Iun cr lcnul can be defined by shifting
• the time variables, as we shall describe in Sec. 7-7.
, (7-89)
Consider the case that V is also a function of the time. Show that
t.he transition clement of dV/ dt is equivalent to the transit ion element
of the operator (il h)(HV - VII) + avl at.
Problem 7-13 Show that
and argue for any quant ity A, given in terms of an operator or other-
wise, that dA ldt is equivalent to M ia, +
(ilh)(H.l - .4l/).
N ext we consider an expression F involving two quant ities evaluated
in rapid succession, such as
(7-91)
•';
'~
<X\FI~) = ~ t: J:. x'(x; t + .)mxK(x, t + .; y, t)yf(y,t) dy dx
'J·
~.
,
- ~ Jx'(x,t)mx"f(x,t) dx (;-92)
!n (1
we find that this integral is
=
,
i
! f-". x"(x,l)mx""(x,l) dx + J--' x"(x,l)m(x/l - Hx)%,/(x,CJ <k
~~
Thus finally
I
( x m x,• • ,- x, -l"')~ if x"(x,l)m(x/l - Hx)%,/(x,CJ ax
= Jx"(x,I)1'Z'/(x,l) ax (7-96)
the last step follow ing from Eq, (7-78). This is on exnmplc of the
general rule: In writing the integral defin ition of the transition clement
for u set of qunutitics corresponding to a succession of times, the
corresponding operators are writt en in order from right to left, accord-
ing to the order in t ime of th e original transition clement. If there
is a finite time interval !:;J. between them, a K, or alternatively the
operator c;(./l ISA I, must be inserted. (For an example, sec Prob.
7-1 6.) As the time interval e between two successive quantities
approaches zero the K approaches a 6 function and the rule results.
288
with indisti nguishable part icles, one alterna t ive way for acco~plishing
011)' event nlwnys cxi~t s ; na mely , UU~ i nt~rchnllgc of l.wo pnrt ielos. III
such a case t he amplitu des for t he POftl clf's (1) as intcrchuugcd and
(2) as not. intercha nged must be added. (Th is add itio n rule applies to
Bose particles. For Fermi particles the cont ributions Ior amplitud<'5
which urise from odd permuta tions of particles will subtract fro m each
ot hcr.] Ordin ary helium ntoms arc of isotopic mass 4 and conta in
Ii parti cles: 2 protons, 2 neutrons, and 2 cl<.'C t fOIlS. Th is mean s that
helium atoms nre Dose part icles and t he a mplitudes for interchange
of part icles add. (For insta nce, we say that Bose pa rticles follow
symmetrical statistics, whereas Fe rmi pa rticles follow antisy mmetricnl
statistics.)
To sec how th is addition of am plitudes comes about, at least for
helium atom s, we can follow this line of argu ment: In the final state
the atoms cannot be d istinguished from cnch other. T hus, although
the uppcara nec of the configuration of atoms may be the same finally
as it was init ially , the identi ty of some of the atoms may have been
cxcbe ugod.
For example, an atom which we shall designate as I starts 0. posit ion
:r1(0). \\·c h3\'Cassu med that some atom at least will be in th is same
posit ion at th e d ose. Thus, Ior some atom :r(fj) is equa l to Xl(O)o
However, it way not be atom 1 which ends up in t his particula r place.
Ins.tl'3.d , ato m 1 lIl ~y go t o the init ial position of atom 2, say 2'2(0),
while at ti l t' same ti me at om 2 has moved into the init ial pos it ion of
atom 1. That is, it is possible that atoms 1 and 2 exchange places in
the fina l configuration.
To describe t his sit · ua t ion in the most general t erm s • let Px • s tand
f or some p(·r lllutQh~n umoug the atoms which nrc init inlly at r•.
Thus, for example III the sit uuticu in which atoms 1 end 2 were
excha nged and all others remained where they were, we would havc
• • , P x .• - XN (10-75)
In general, the f in al state cnn be any perm utation of thc initial state:
x,(ll) = Px.(O)
(1(l.7 6)
where L'
r
menns a sum Over nil permutations P.
I.r we were d enling with Fermi particles, e.g., the isotope of helium
wh ich h~s. three nucleons, we would have to includ e an ext ra fac to r of
+ I , positive for even permutations and nega t ive for odd per mutations.
There would a lso be so me extra features wh ich depend upon the spin
of t h e atom in our resu lt.
It is possible to give a mo rc d etailed d erivation of Eq. (10. 77) in
the followi ng manner. For helium-4 atoms thc quantu m-m echanica l
a mplitude for two atoms which start at pos it ions a and b to get to
positions c and d is
K(c,a;d,b) + K(d,a ;c,b) (10-78)
(Amplitudes for al ternative final condit ions add, since these co n-
ditions cann ot be d istinguish ed from eaeh other.I In this expressio n
K(c ,o jd,b) is the co mplex amplitude to go from a to c while the ot her
particle is going from b to d.
Since the particles are indistinguishable, their sy mmet ry properties
imply t hat the a mplitude to find t he two pa rticles eve ntually at thc
poi nts c a nd tl must be a sy mmet ric funct ion of c a nd d. T hat is, the
wave function "'(c,d) must be a symmetric fun ct ion of the variables
r e, r". That is,
"'(ell) ~ "'(d,c) (10-79)
If the par t icles were Fermi, t hc wave fu nction would have to be an
a utisynunctrie fun ction of these positions.. . .
If many particles arc involved, the rule IS Simply extended, that IS,
"'(1.2,3, .•• ,N) ~ "'( 1,3,2, . . . 'N)
= "'(1 ,2,4• • . . ,N)
= etc. (I ll-SO)
The simplest statement of the general rule is ~hnt t~e wave funct ion
must he sy unuetrjc (Ullti:oynmll"tric (or Fc~nll p:~rtldl'S) . .\ lt hou":,h
other solut ions of S('llriKli ll~' r'~ WllX C' equation l"XI~t, 0 11 1,- :<Ylllllll'~rlc
unt!"I)'1l1iIl Ct r ic on es
an d 6ntL ~'"
nppenr
r -: i ll nnt ure " lienee III th e sum d efining
Quantum mnhanit:' and path intesror.
t he partition function in Eq. (10-2), "'·C do ~ot wish the SUl~ over aU
energy eigenvalues of t he ham iltonia n H whic h ca n be ~bt.nlUcd from
ti f IrA ~ E.I. but onlv over those for which the wave
so Iu la n 0 ..... .." ..... , ~ dcnsitv rnatri
funct ion t/l.. is a sym met ric function . F OTex amp le, the CU!Hty m.a t :ue
p(x',x) is defined b)' Eq. (10-28) with a disrcg~rd for t~1C s tat istiCS
of the ..Y atoms involved. How can we reduce thia s u m to Includ e only
sym met ric wnve fu nctions ? . . .
To accomp lish t his red uction , we use the following trick. F irst we
notice that fro m any fu nction a sym metric fu nction cnn be produced
simply by permuti ng all variables a nd a dding toge~heT. the resulting
fu nctions. Thus, for any function !(XI ,X2) the co mbinat ion !(%1,%2) +
j(XI,Xt) is a symmet ric fun ction . It follows t hat for any wave func-
tion 4>(Xl,2:'l, • • • ,%N) the function
~'(x.) = L ~(Px,) (10-81)
r
is symmetrical. N ow if ¢,,(x,) is a solut ion of the Schrodinger equa-
rion, then ~: (.rJ as defined by Eq. (1Q-81) is a lso a solu tion, since the
hamilton ian II is symmetric for an in terchange of coordinates. T here-
fore, each interchanged Iorm ¢ ..(Px) is a solu t ion, as is the sum.
Some of t he energy eigenv alues E.. have eige nf un ct ions t/J.. which are
symmetric, a nd some d o not. Sup pose Eit is a n ene rgy eigenvalue for
whi ch the Schrodi nger equation docs not ha ve a symmetric solut ion .
Then the s u m L
¢ k(P X) mus t vanish , since if it existed it would be a
I"
symmetric solution for E. . This res ult implies that the ope rat ion
d efined by Eq. (10-8 1) selects just those sol utions t o the wave equations
wh ic? are sY~lD~etrjc. All other solut ions vanish . If ~ .. (I) is sym-
met ric,. then It IS equal to ¢ ..(Px) ; a nd eince there are Nt ways of
permuting the N atom, we have
L ~.(Px.) ~ NI~.(x,)
p
if ~. is symmetric
= 0 if ¢ .. is of any other symmetry (10-82)
These res ults give us an answer t o OUT questi on . \Vc can now sel ect
out. of the SU Ul defining .tlle density mat rix those particular elements
};'(Pr,x) -
.LL
which apply to sym mct n c states. Thus
~.(Px·)~:(x)",,·.
r .. r
.~
= Nt L ~.(r)~.(x)e-'"
•
- Nrp.~... (:r',z)
(UJ-83)
291
This is t he reason why in Eq. (10-77) d fi . . .
for sym metr ic statist ics we pcrmut 1I t ~ uung .the pnrtlt~o~ function
The resulting partition function eoe . edJXlrllclcs end divid e by NL
rrcspon s to
292
t ion terms should begin to pb y an impor ta nt. role in t he description of
liqu id heliu m.
At first sight, one would Ilot expect vcry. ela borate . cxchn~lgCS of
ato ms to eve r be important. An exponen tial fa cto r . lIlVO~ vlIlg the
spacing must be included each time nn a tom move s to Its IlCighbori ng
locat ion. If we roll this factor y, then for T ato ms to move to neigh,
bori ng spots the factor y" must be included , and since U is certainly
less than 1 at any temperature, y" could become q uite small for large r.
' Ye ecrtninly would t hink that as r ap proaches euy reaso nable fraction
of t he approximately 10:: a to ms in a cubic ecutimcte r of liq uid helium,
contributions from facto rs like y" meet be infi nitesimal. H owever,
th is first sight docs not take into a ccount t he fnct t hat with T atoms
permu ting, t here a re e n enormous number of possible permutat ions, r !.
Thus t he small weight of one particula r permuta tion is offset by th e
large nu mber involved .
Ano th er quest ion which arises in t he description of liquid helium
concerns th e t)'PC of permutations which a rc involved . Any permu-
tat ion can be descri bed by cycles : t h us 1-4, 4-7, 7-6, 6-1 is a cycle.
Arc t he importa nt cycles long or sho rt ? A careful es ti m ate shows
th at a t modera te t empera tu res, only simple excha nges of two a to ms
arc importa nt. Then as th e temperature falls, cy cles of three ato ms
become important , th en four, a nd so on. But then sudde nly, :1t a
certain critical tempera t ure, cyc les of much greater lcngt h 1. offset
by their great number the sma ll value of y L. At t his te m pera ture
cycles of importa nce become very long, in volving nearly nll of t he
a to ms inside a conta iner. At t his point t ile cu rve of speci fic hent vs.
tcm pC'~ture Eho\~'S .8 d iscout luuity. Below t his t cmpvruturc t he
beha\' l0~ of t he h(IU1~ Is very st ra nge . It flows t hrough very th in
tubes Wit hout a ny rcststunec for low velocit ies. It st imul ntes infinite
heat conductivity
. in bulk ' elc . '1'1 lese odd uu CInuuctcrtstics
' , nrc n UU H~ ,
((:stUtIOIl 8 of (!U3I1tum mecha nics , pnrtieuln rfy tl IC consI rne t I\"C .III I Cr -
fcrcnce. between . a mp lit udes for repla cing one a t 0111 W 'I 1"I a no,IIl'-r •
Qua
t l lltltnh . .\'dy, the det ails of th e hcl ul\'illr t lr tlH" spee if r I C Ile8 I JU
' S,t at
rc t.rnll~l tlQIl temperat ure lire (lot On n very fi rm fou ndat ion . But th e
qualitative ~son Ior t he trunsificn is cleur.!
TIle expression all u!oJ!;()UJoi tu F' I (1 0-77) r I'
H S ' I.. .'1 ' . . or -cr nu. part ,ie1cs, sueII B;i
C , IS u ..0 r usr Ywrrt t on do wn. H owever i ll tI . r I' ' I IIc'
Ihc err~t ......
of U Ir-ut i I'
It" po I ' ll III IS very lin d t
, I e ca se 0
I rq u n . ,
a ccura te man lier . The f{1l."'i011 f r r . 0 .l · \, a ua te q uu nt itut.ivvly III fi ll
o ti ll.,; IS t hat ti le ec ntri butjon of 8
1 A more d ct.ail~·, 1 .ti..rll""ion 01 t il.. t "t" . .
point of view rna)' lie found in It p Pe v 1 11m (UIIl'l wn o( li(llI id hcliu Ul (ronl (hI.!
. . ~)n nllan, Pit" •. RI!I'., vet. m, p. I 2tJ I, 1953·
Stotbliall mecllonit:8
293
cycle to the su m over permutations is either iti "
depCIHI"rug on WItet her the cycle has an odd' POSI lye Or negative
" c_
" - 1 II or ('\CII IlUIU IJl.:r of atoms
III Its _ t'nr.: ~t 1 ~. At _low tempera tu
'
re , tile c011 t rtib 11 tiIons 0 f cycles such
us L r- .>1 and L = 012 nrc very nea rly equal but °PPOSIitcC iu si
III sign, and
t~crc ore they v<'ry nea rly cnnccl, It is necessary to compute the
d!ITcTl'llce between such terms, a nd this requi res very careful cnlcula-
tl0~1 of onch term sepa ru.lf" Y. It is very difficult to su m an alternati ng
1 .terms which nrc decreasing slowly in mngnitud c w-hen n
seri es- of large
precis e ann ytie for mula for ench term is 1I0t availa ble.
Progress co~ld be made in this problem jf it were possible to arrange
til e ma t hemat ics describing n Ferm i syste m in a way that corresponds
to n sum of positive tenus. Some such schemes have been tried but
the resulting terms appear to be much too hard to evaluate 'even
qualitatively.
For molecules which nrc separa ted by d istances in the neighborhood
of 1 A we have seen that the effect s of exchange (the nonidentical
permutations) are i mporta nt only when the temperature is down to a
few degrees a bsolute. In contrast to this, consider the behavior of
electrons in a solid meta l. The mass of the electron is so much smaller
than that of a molecule that the crit ical temperature is much higher.
At room temperatures, electrons in n metal a rc described accurately
only by equations which include the exchange effects of these cyclic
permutations. From this point of v iew, room temperature is very
cold for electrons. The excha nge effects arc of do minant importance,
Of, to put it another way. the electron gas is degenerate. Of course,
the electrons internet by Coulomb's law, which is qu ite strong. But
since the effects of the Coulomb a ttraction arc of long range, they
tend to average out. To (J, fa ir approxima tion , the electrons act 3S if
they were independent, although, of course, each moves in t he sa me
periodicn lly varying potential produced by the arrangement or the
nucl ei a nd th e average or t he positions or neighboring elect rons. Fro m
the s tudy of t he ideal Fermi b'1lS neglectin g iuterncticns, we can learn
n lot about the behavior or elec trons in met als.
However, it is apparent that we cannot learn quite enough, for
the superconductivity of meta ls occurring bdo\~ a few degrees a bsolute
would remain n mystery. 'rlli~ phenomenon , 111 some metals at least,
involves all interaction in which the alow vibratory motion of th e atoms
is involved . 'Ve conclud e t his beCAUSC t he tra nsition te mpera tures
Ior two different isotopes of th e some metal depend on t~l<' utouuc ".~ss.
This value of th e isotopic mass would not be huportnnt If till' tra nsition
were simply n umtter or mutunl inh'nu~tioll between electrons, or
interaction of the electrons with nil idealized urruy of ji.u tl atoms. The
Qllnlllllm mechollics nnd pal" itl leB ro l.
294
idcaliantion t hat t he atoms arc fixed must be incorrect. But how docs
the mot ion of the atoms produce a sudde n j um p in speci fic heat in
some metals an d permit elect rical conductiv ity below this t emperature
wit hout rcsistanccf T his qu est ion was first an swered in n. convincing
way by Bardecn, Cooper, a nd SchricfTer. ' T he path integral approach
played 110 pa rt in their analysis, and in fact it has never proved useful
for degenerate Fermi systems.
kT 10 (2 sinh 2~';,)
T his gives the fr ee energy of a linear system as
F = kT ~ In ( 2 sinh 1;)
= kT L• In (1 - ......,,'1 + LIi.>, (10-85)
• • 2
Th~ last te rm in t his exp.rcssion is a grou nd-state energy of the system.
I·or an electromagn etic field in a box of volume V, the moore are
d(,~lIed by U~e \'e~tor wave number K, two for each K. T he zero-
point energy 15 omitted. T hus the free energy of the elect ro magnetic
field computed per un it volume is
d'K
F - kT / (2".), 2 In (I - e: 'K,"'1 (10-86)
~1~ ;lIt
w II C 1
('rllnl cnergy C! is thc partial deriva t ivc of {JF with respect to {J
lx comoe (puttlllg w = K c)
U = 2 / d 'K hw I
(2r)' e"w f"T - 1 (1G-81)
Thc volume cleme nt ill K space ce b .
, n e wrlttcn as
d'K = 4rK2dK = 4r ~ dw
c' (10-88)
I J . nard f'f'n. L. ~ . C"..onJ)("r a nd J n Sc.h .
aod vol. 108, p, 1175, 1957. ' " rK'lTer, Phy•. Rn ., vel . l OG, p. 1(;2, 1957,
SIoti8'iml mecllonir a
295
This menns tlInt the cnergy in the elcctrom . .
frequencies from (oJ to (oJ + dw is agnetie field In the range of
2 · 421' flw
(2~c) ' eI* /lT 1 ( 1Q-89)
Thls
,
is the fam ous blnckbody-mdiation
. •
la W dilSCOVCred Ily Planck. It
was the first real qu nlltltnbve (luuntul1l-mcehanicnl result disc ed
d . tl Ii t . I "'" rseover
an " US re nrs step III t io discovery of the new laws.
Another early tlunntum-m cchnuicnl triumph was t he explanation oE
thc te mperature d?I~Cnuellce of the specific heat of solids by Einstein
and b)' Dchye. This also COIUCS from Eq. (10-&; ), but the oscillators
a~ now the normal modes of the crystal, D.8 descrihed ill Chap . 8.
l~ or example, the ther:mal energy JK'r unit volume of such a crystal is,
like Eq . (1Q-87) [leaving out the zero-point energy) , just
u - '"
- 3. 6..~
f exp [hw(k)JkTJ
,",,(k)
-
d'k
I (2r),
( 10-90)
s = 21 J11IX' dl + 2k J%(/)%(1 + 0) dt
°
(10-91)
297
so that the equation of motion is
k
mi = 2 [x (1 + a) + x (! - a»)
(10-92)
In Kt·lIl·rnl.
,,. ••,,.,•
for I'll)' Arllnu funrlionnl, Ilk" ."1, UUIl d","", lIot Involvn
lin- t inu- t'xl,l ir iIJ)' (i.r., i" iU\'" riRlll f,lr thr IrllU",fUrlllnliull 1_'-}
nllllOl) Iln'n' i.. nil C'Xpn"NOion A'(T) fllr 1111' rfl.'rRY nl l illlr '/' whleh iJI
("unltl'r\'t"l.l. II cnu IH- Inuud iI)' ntlkillf( (ur till' fi",,'-orcJI'r d""IJ(" ill llu~
arliull S ,,111'11 nil Ilnllu~ nn- rllnnKI". Iroru .r(I) lo rfl + ,,(1)1, w"ere
,,(I) - -tfl:! (ur I > 7' nflel ,,(I) - - ./2 for' < T, with corUll.4nL •.
1.'\ be Iheu .1:.'(7/ (or iufill ih"fOilllnl f.
11
The Variational Method
IN us Chep to r we dlISCusa
t hi S
method based
. '
on no variational principle
...
for the approximate eva luation of ccrtuiu pa t h integra ls. 1< lrat,. we
shall illustra te the method by SO llie exa mples. Late r, we consider
those problems for which the method may be useful.
Z = J--. J.~' e Sf
' :OX(u) dx .
,
(11-2)
so long as we usc the "time" variable u as an imaginary quantity.
In Sec. 10-3 we developed 0. perturbation tech nique for the evulua-
tio n of the path integral d efining t he partition function for certain
special cases. we shall now descri be another tech nique, a pplica ble
in t hose cases where S is real. For ordinary cases without n. mag netic
field (and no sp in) S is real.
Throughout the remainder of this chapter, we assume t hat we
arc choosing units in such a way that the value of 11 is 1. Whenever
it is necessary to include It sy mbolically in order to visualize t he
qu antu m-mechanical characte r of 11 result, it can be so included by a
straightforward dimensional inspec tion.
Let us suppose that some other S ' CRn be foun d wh ich satisfies
tw o condit ions: First , :i' is simple enough that exp ressions such as
Je8 ' :o.z (1) or f Ces - 1)x(t), for s imple fUIl(;ti ollal~ G can be evaluated.
~ceolld, t he impo rtant pat hs ill t he integra l f e s :I.>.t(t) a mi t hose in the
J1I1 Ch'1'nl Jt s • :I.>.z (t) arc similar, that is, S ' and Sure simi lur when t hey
nrc I~th small Now suppose P' hi the frcc cuergy associated with S'.
T hat IS,
where {;r} = t he weighted average of:c. This follows because the curve
of ~ is concave upward, us shown in Fig. 11-1, 50 that if ~ number of
masses (weigh ts} lie along t his curve, the center of gravity of these
masses lies above thc curve. T hc vertical height of this center of
gravity is the average vertical posit ion (t"') of the points. It
-----------., ~
o.
0,
V>
itivo nnd look c e t hem as d ifferent
Fig. II-I we assu me t ho we iJ;ht ing r. ctors 0 , Il r e ~ of the "'ei~hted 1l" C",~ of r, tMl
ln lLS!l('ft ~itioocd alo np; A et rin" . T hen I h~ c;r;;:nen IA n Clit illl>J (, ' ) h l 't":tU >!C v f t he COOt"ftVO
il. l ·). m u...t l ie be low tile " ci~J.tcd a vc l"lI.rc, U " IU l~ t l.... c ur,",." but (c'). tile center of
nA ture o f t ile eu rvc c· , TI Ie va tee u of C" IIU b 1f' " I .. •
-
Rruvit y vf t he tlevcn.1 po lp hl, flI Uti re-t I' above the cu rve .
Qunn'um m edlflnits and po'" integrnl.
:102
') d i t f the eurve c- at the abscissa. position of the
exceeds cv , the or 1110. e 0 I ()
. hi 1 i the evcrege vu ue %.
center of g J1I.\"lt )· , w It.~ 1 \ E (11-5) we take t he avera ge value of
On the }rrt-hund $ 1 e 0 ~' . • I ts oS ' :Dx(t) where S' and S
er S' over pat hs with the positive we~g I . C cd the qu antity eS-s'
I II by Eq (11-6), tillS exec S •
arc rca . C~IC~, • of S _ S ' with t his same weighting
where (8 - S ) IS the Bverage TI "
scheme, namely, wit h t he werg c
" ht S ' :nr(1)
•
mt IS
,
(S _ S ') = ff( S - S' )e S ' :nr(t) d.J:. [Ife ' :ox(1) d.J:.I-·
S
(11-7)
where
303
value for the lowest energy state of th .~ E
C system . 0. Recall that
•
Z '= r,-6F = L e-/JE ..
• - 0 (1l~12)
"
Qllnntllm mechnn i t s a n d parh inregfflu
301
Th is path int egral is over paths which return to t he in itial !iO tortillg
poin ts ; and utter t he path integral h.RS I ~n e~ulunted, a furth er
integrat ion ove r nl! possible starting points IS carried out. .
I n See. I£).2 we consid ered t his same p~hlcm nr~d POI1~tcd out
bow the elassicu! a pproximation may be derived by inspection. In
the clnssical Iimit of high temperatures Of high val ues of k T com pared
with fl, t he value of (lfi is so small that pn.tbs which get very far away
from xo do not contribute. T hus, t he pote ntial can be replaced by the
constant val ue r(xo) , and the path integral contribu tes on ly a con-
stant, giving
m J.,-BVld ax
.-'E, (classical) ~ ~2~{J (11-17)
x = ~ J: x(t) dt (11-18)
Along any part icular path, th is su bsti t ute for the potential is a
constant, so th at t he new form of the action a long that p ath becomes
With th is more general form, it is possible to cal culate l oth 1'" nud
(S - 8'). )
P roceeding along this COurse we usc E q . (1l~I4). Substituting
into this exp ression, we have '
' 1 305
1t IS to )C reJnrmbrm l thnt Ih the
i h
o I E q. ( 11·20) nrc th ose whicl chpe to be- uscd In t C! path 'II1 tegrnls
and , ns in £ (1. (ll.W) a furth n~e the FUme initial and final points
. , er IUlegral io 11 ' •
to be carr'icd out " n over a end pom ta Zo is
No te that the nllnlcrntor of 6 is . "
introd uced in E (I (lO-f ')) ' f . qtnte ~lIl11lar t o the term l (z)
• • • Jo) ,I WI:' IY'st r.ct ou I-
specifi c avcm ne vn lue • a d ' rse \ cs to paths t hat have a
,.. , ~ -... n COu nt on Iut cemt i
values of ;f at n Inter staee of II I 1 . II egmtlllg over all possible
I') Ie en C'1I nticn n th
as were used in the d iscussion of J(,i we' Y c sa me argu ments
~ is indepeudont of t' \V I}' eoo that th e numerator of
" e can eva unte the path integrals in both
nu rncrntor and denomina tor hy the methods used in Ch 10 d
take the uuswcr from Eq • (10-"') \N ,
rc be ine
mem ring t at th ep . an
y = x,- i (ll-21)
t: f-'o /CXP[-6
(V(x,) - W(x) ] ; (x, - x)']l
~ = X ' c. PI ~Jr (i')JJ d.r, dx
(ll-22)
t: f_·./exp [-7 (x, - x)'J)
X I cxl' l- ~W (x)JJ dx,dx
(tl-23)
The form of l' (x) reveals the qunntum-mechan ienl effect we hnvc
introduced, T his Iunct ion is u weighted uvvrugv of I"(r c) with a
gaussian weightin g Iu uct iou just like- the funct ion U(r ,,} defined by
FAt. (l ~,s) , a nd t he- gnussinn spu'ntl is agniu (dA !; t:?",r~, For a
helium atom at a t r-mper utu rc of :!°l'-. t " ~ spn-ad 1I11l0UII l s to abo ut
0 .7 Jt At room tomperutu res, however, it is only ab out 2 per cent
of the 2.7-J\ dia meter of t he atom. The value of 6 eiu uow be written
Quantum mec/ln"ics and palh integrals
306
as
The next step, finding the optimum choice for lV(i), requires us to
determine the eITect of a small variation in the function IV(x) on the
value of F' - 6 and set this effect equal to O. Thus, imag ining W to
be repla ced by
W- W(x) + .(x) (11-27)
we find from Eq. (11-26) that the variation in F' is
dE' - f,(x)[exp (-PII'(x)JI dx
0 - Jicxp I-PW(i»)! d1; (11-28)
307
requires simply that
aE; - a~ = 0
(11-30)
which will be true if we take
IV(;;) = Vi;;)
(11-31)
This, r implies that 6 is 0 and that F'
• • inI tum, has th e same form as the
.ctassiea
h d firee. .energy defined
, ~ . (11 -17).
in Ell 1/on:ever,
· tl lC potential
.
1D t e c mtlon of F t has been replaced by V(i). That is, '
then, the conclusion holds.) To sec the relation between these two
equations. we shall consider the following exam ple: .
Suppose the action S is connected with a lagra ngmn su ch 88
'W here 1'(z) is independent of t. (Gt herwise, of cou rse, there arc no
fixed energy levels to scekl) w e shall limi t ou rselves to t he case of a
single variable x, but th e general case follows d irec tly. " 'c uote here
t hat if t he Iagru ngian co ntn ina the term xA - for exam ple, if the
lagrangian represents a particle in 8 magnetic field-then E q . (11-33)
is st ill corre ct . H owever, the action S is com plex. In this case we
suspect that Eq. (11-13) (or some simple mod ification of th is equation)
is st ill valid . However, this has not been proved. So, for the present
we shall limit our discussion to a cnse in whi ch no magnetic field is
present, Then in the limit for large values of {j we have
Now su ppose we usc for our trial action S' the form
which involves some other potent ial V' (x). This means that
E. + V'(z)
H' = 2m (11-45)
then
H'¢~ = E~~ (11-46)
so that
E; - 8 - !f/J~·H'f/J~dx + !¢~·V~d% - I¢~·V'thd% (11-47)
But the true hamiltonian can be written as
p' ,
H=-2m +v=L+r+v
2m -r-H'+v -r (11-48)
310
E (11-49) we couJd i)115 t cad pick the wave function itself and. then
q, E' (11 '9) -it l out ever bothcring about the potentin] to
evaluate q. - \\ 1 '1'1 ' . 11 f I' .
I . 1 II ' bitrary wevc functio n belongs. ic \ urrn ) o une IOn In
w IIC I l IS ar . , h II th t •
t hiIS process IS. II1('n the wave function . 410 rot 'or • IBnI o pch ential
. 1"()
f unction x . w e.. fi nd , theu , that this result IS simp y anot er way
of stating th e result of Eq. (11-33).
If the problems such as the one giv;n in this example were the only
ones In which the concept expressed In Eq. ~11-13? were useful, then
th -ould not be much point to this Jong discussion. But there arc
m~~~ : 10re complicated integrals for which Eq. (11-13) can be used in
a way that, at le8st as far as we can tell, is not so easily t~nsformnble
into Eq. (11-33). We shall describe such an example m the next
section.
(11-50)
P = k_ a.l:e,1r.-r (11-51)
k
then the charge density from the ions is
(11-52)
p = V. P = ka.l:e,k·r
a - -
V2'~
I
(.L !).,
, (11-55)
where € and e.. nrc the stat ic and h igh-frequ ency dielectric constants
respec tively. In n typ ical case, such as the crystal of N nCI, the vnlue
of Q mal ' be about 5. Thc values of energy which we shall ca lculate are
in units of fu.J.
No w we can study the quantum-meehunienl motion of the elect ron
solving t he mot ion of t hc harmo nic oscillat ors co mpletely. For exam -
~Ie, th e a mplitude that the electron statts at fl , ...· ith the osci llators
In. t~e ground state and ends at the position r t with the oscillators
still 111 th e ground stat e is
0 ,,(2.1) = f .'s :Dr (I) (11-56)
where we have used the results of Chap. 8, and
S= ~ JI~ r + J"';,..
ell a . ·· ··,,'.- ·....'. -.10-01dl d
8
d'k
(2,,>,
(11-57)
K(2,1) = f es :Dr(/)
(11-59)
where the varia ble t goes lrom 0 to fJ and
T his result is just that whi ch one m ight expect from the replacemen t
of t by the imaginary val ue it in Eq. (11-.58). Asymp tot ically. lor
large values of 13, t his kernel becomes proportional to e- EefJ.
\Ve no w have a relatively complicated path integral on which to try
our varia tional principle. Next, wo shall have to choose so me simple
ac tio n S', ,..·hieh roughly a pproximates t he true a ct ion S, and t hen find
E' and s,
W e note that in E q . (11-00) the particle considered at any particular
t ime ' " interacts" with its posit ion at 0. past t ime by 0. reaction whic h
is inversely p ropo rtional to thc distance traveled bet ween t hese t wo
t imes a nd wh ich dies out expone nt ially with t he time d ifference. The
reason for th is is that thc disturbance set up by t he elect ron in t he
crysta l lattice in the past takes some t ime to die out . T hat is, it
takes so me ti me for the ions to relax, and here ill this relaxation pe ri od
the elect ro n still "feels" the old dist urbance.
\Vc shall try a n action S ' which has t his same property, except
t ha t instead of involving the inverse distance as Do coupl ing law, the
attraction will have the geomet ric form of t he parabol ic well. This
would be n poo r approxima t ion if t he dista nce .Ir(s) - r (t)1 coul? ,~ery
often becom e exceedingly large" However, SIllCC there 18 8 limited
t ime available before t he exponential t ime fuetor forces t he interaction
to die out, la rge va lues of this ditTerellce will not make a ny importa nt
I Alt hougb I ie Eq. (11-61.1) iI:l not rt'all)" the t ime. Lut .a n. intl'gmt ion ' "ariaLIe
inst eed it ill useful to think !\bout it es J\ l ime, jwt as " ' C d id In Sec. 1()..2.
•
Q"an,"m mecl'Rn;~!t and pa". in . egrala
31~
d'X'(t) = 2C
dt'
J[X '(t) - X'(s)]c -·-·I ds - j(t) (11-68)
so that
(II-il)
d'~~t) _ w'[Z(t) - X'(I)]
= exp
[-
2Ck' • w' k'i T _
vito (1 - e- I-#Q - 2v 2 (f
I] (11-73)
(11-74)
r
on k which is a simple gaussian, so that substitution into A gives
To find B, we need (Ir(/) - r(s)I '). This can be obtained by expa nding
both sides of Eq. (11-73) with respect to k up t o order k'. 'Therefore
I
3 (Ir(,) - r(a) I')
4C
= v'w (1 - e~) + u'w' I, - al (11-76)
B = 3C
uw (11-77)
Finally ~·c need E'I the energy belonging to our action S'. This is
most easily obtained by dilTerentiating both sides of Eq. (11-6) with
respect to C. Onc finds immediately
CdE: _ B
de - (11-78)
ru.sn
and E~ = 3v/4. It corresponds to the usc of a fixed harmonic binding
potential in Eq, (11-37). For large v, C " can be neglected, so that
A = (rw) -·~av». For a less than ;). 8 a nd 10 = 0, E(I_(11-80) docs not
give a minimum unless v = 0, so that the 10 = 0 case docs not give a
single exp ress ion for all ranges of a . In spi te of this disadvantage, the
result with Eq. (11-81) is relatively simple and fairly aeeurnte. For
a > 6, only fairly large values of v arc important, and the asymptotic
formuln (good to I per cent for v > 4)
(lIoS2)
(lloS3)
The integral is
(lloS~)
2ur-'[{1 + tv)" - II = P
The problem of Eq. (ll.sO) then correspOnds, in this order, to mini-
mizing
(1l-85)
E = %w,' - a - a,{1 - P)
QlIflIII."" ml'chnn;cs ond po'" inl~grols
318
That is,
2ael - P) (11-86)
e= 310
which is valid for small a only, because E was assumed small. The
resulting energy is
a '(I - P)' (11-87)
E= - a - 3w
E - -a - ;; = . r-
(a)'
a - 1.23 10 (11-88)
It is not sensit ive to the choice of w. For example, for w = 1 the 1.23
fnlls only to 0.98. The method of Lee nnd Pines' gives exactly the
result of Eq . (11-88) to this order. The perturbation expansion has
been carried to second order by Hega," who shows that the exact
coefficient of the (a/IO)' term should be 1.26, so that our variational
method is remarkably accurate for small a.
The opposite extreme of a large a corresponds to large v and, as
we shall see, W near 1. Since v» w, the integral in Eq. (11-76)
reduces in the first approximation to Eq. (11-81), which we can use in
its usymptotie form. The next approximation in w call be obtai ned
by expanding the radical in Eq . (11-75), considering w/v «I. Fur-
thermore, eN is negligible. In th is way we get
E = ~ (v _ to)' _ a (~)l\ (I
4v ."
+ 21n 2 _ to')
V 2v
(11-89)
· 'T.
F re I I Lee.: and D. . Pines
. ' lnteracttcn of e Nonrelutiv leti P nruc
onre U IVU;tlC . ,e \Vlt
. 1I a ~.
o-alar
I wu I Applu'ohon to Sio . 1', . I 9"
p . 1)1):1. J ~ '':;:J. Yo ~ l '1:tro n lJ til Polar Ct}-st ubi, l'},YB. Rev., ,"0 . ..
s I~. Hugo Note on the BI n . h
H,I ~ II . p, 44!I, 1 !154~w ~ ectrcns m Q Pclnr Cr.Yl:lwl, Prugr. Th eorel- P lis.
( K JlQlo) ,
I S. I_Peka r ill Tlll'Or)' of PI . .. 190.
HI'W, hUH shown ti ll t E O uruns, ZII. Ek spenm. I Teor , Fiz., vel, 19. p-
I goc'fi WI - 0.108& 1 Icr thc CIUiC large a,
4 or
Til e "'"rintiOlw l trlel /lfJtI
319
TAIlI.E II-I
• 3 5 7 9
-• II
..E, 3 .44
2 . .'i.')
- 3 . 1333
4 .02
2 . 13
- 5 . 4401
5 .HI
I.(;()
- 8 .1 127
9 .1l'
1.211
- 1l .4R6 -
rs.s
1. 15
1.'; .; 10
Eo, - 3 . 10 -5 .30 - 7 .SS - 0 . 9S - 12 41
E. - 3 . 09 - 5 .24 - 7 .43 - 9 65 - II S8
E... - 6 fl.1 -1 0 .3J - 14 .7
323
On t he ot her hand, we might ask a .. .
kind. ' ''c miglll , for exam, I k : ~rohab.l h(y quest ion of a different
interval from one parlif')C il"' f'! , D.:t't " VllIa( IS the probah ility that the
• par. 0 t te next \ ·'11 1 '
lime 17 Actually there is no \J >C SOme perti culnr
hi '
t IS way. If we were to ask II
correct a nswer 10 th
b I':
. h
e question IJ rased
be equal to or great er thun ; e :;ro a ~J1Il Y that t!IC ti me int erval will
cxp (-pi)I. That is . ', 1<'11 we could give an a nswer [it is
, we ('an get 8 11 a nswer to a I ti bo "
with in n cert uin range Th "f . q IC'S IOn a III t failing
. I . u ~, I we arc int erested in a pnrtieular value
we mu st n low ourselves a n infin it esimal • d ask " •
WI t · I (' . . range an ask th e (IUcsf IOIl :
18 IS t io m fi mh$ lmnl) prcbahi fity t hat Ih time i I "
ithi I
WIt m t re range dt centered around I . T h
e rme mt crvn WI I fell
" .
. e answer IS written as
P(I ) dt = ~.-... dt
(12-2)
324
(12-4)
¢>ik(I)] -
f" '''''''')" 1'[1(1» :D/(I)
f/'I/«(») :J>!(I)- (12-12)
(J(a)) - - i 'k'~(
" a
) ¢>i1:(I)] I
~tfl_O
(1 2- 1 ~)
(12-15)
12-3 NOISE
327
associa ted with such a pulse would shn-· th 1 ·1 h h f
I ' ow a I as t C onn g(t) for 8
pu so occu rnng a t t =:: O. So, if the pu lse occu rred at ( the h f
th e voltage curve would be oCt _ to). 0, B ape 0
'Vc call the exp ression in parentheses A and mil e this resul t as A".
I! the number of events in the time interval is distribu ted in such a
way tha t th e Poisso n d istribut ion a pplies, l.e., the occurrence of each
even t is independent of the time of occurrence of an y other event an.d
there is a constan t ra te I.l. for the expected lI.umber ~f ev~n ls per u~t
time then the expected number of events III the time interval T lS
~T ~ ri. The cha racteristic functi on is
s-
eJ! =
L•A" -,
n . .....
(12-18)
The sum all the right-ha nd side of this equa tion is the ~x~nsion .of an
exponential fun ct ion, so that we ca n writ e the characteris tic function DB
eJI EI.
[
e- Cl- AJII = exp -pT 1 - ( f,0T e
i/ .t(l+_),(I) dl "s)]
_
7'
= cxp [ - "" }o
[T (1 _ eo/ i Cf+- ),,(l) . ) ds] (12.1 9)
Q,wntum mechonic. and padl integrals
328
Thus. we may determine thc ('hnradcri~t ic function for m~ny diffcrcnt.
situatio ns. "r
c next go on to d iscuss this result. u nder venous a pproxi_
mate circumstances. .
Suppose we imngiue t hat. the pulS('~ get very weak while the expected
number of pulses per unit li me, that IS, lA , be~11lcs large. I ~ that case
(I) is small, so we ca n expand e-1J,(H . h IUJ JJ III a power se nes a nd we
9 " iee f une tiIOn as
ea n a pproximat e the cham clenstl
exp [ ;p J:J: k(t + s)g(1) dt as] - exp [ipG f k (t) dt] (12-20)
where we have used the subst itu tion G - f get) dt, thc area of the pulse.
This means tha i lfl is in the form of Eq, (12-15) with F (l ) = lAG (n con-
stant ind epend ent of 0. That is equivalent to saying that J(C) is
certai nly p.G or, in other words, that t here is un it probability for
observing the functi on l(t) = p.G and zero probability for observing
any other I (t). That is to say , the p ile-u p of u la rge Dumber of small
puL"CS generates a nearly steady direct voltage of value equal to the
number of pulses per secon d p. times the average voltage G su pplied by
each . Xext, we go to one higher a pproximat ion and st udy the fluctua-
tions or irregularit ies of this nearly direct vo ltage.
Equation (12-20) is a 6rst-order a pproximation to the exponential
il
e J,(I+- I,,(l) dl in the description of the characteris tic functional of Eq.
(12-19). Su ppose now that we go on to the next-order approximation
and include the second-order term. This is
To simplify this exp ression, we define a (un ction whieh measures the
overlap bet ween two nearby pulses as
A(,) = fg(t)g(1 + ,) at (12-22)
By usc of this substitut ion, the second-order t erm is reduced to
p (T (T
- 2 Ja Jo /;(t)/;«()A(t - t') at dt' (12-23)
I nclud ing bot h first- and jsec d -ord cr terms, the characteristic
runcI 10" 11 ul 18
i ~. . . .on
(12-24)
The first rector in this ex " . . I
we might call tl c d I prc~lon 18 t he cons tant average level, WlliC I
I -e evel If we arc thinking about voltage pul~"
Other problenu in probability
.( . 329
W c can, we wish, nCJ!:lcct this) CVCI
1
varint ions around it hy shift in th .a~d conccntrntc only on the
g e O"gon
alw ays ta k.c o u t a fnel or ci f,t(I) FUI 4l b I " of J{/). °rhat .I S, we can
wri t ingJ{I) = F{I) +
nl) and stud y s nfl ong th e onlPu of J{I) [i.e., by
ret) and itscha rnctcristic fune tim lBJrn~:h.e prohahtl.lty distri bution of
we arc in n posit-ion to study th
d-e level.
n t w ~ make this cha nge of origin,
e uc uationa of voltage around the
We notc onc special a pproximat ion to E . .
adequate. Generally ~( ) . .q. (IZ-24) which IS often
. .' T' IS a narro w function of r, The ul h
get) rtscs nnd fal ls with a fi nite width 0 ,' ( I _ I purse s ape
dl . l oS " 0 pu eee nrc spaced a
very grout istauec upurt, their ovcriapiling area van ,' I 'Ihi
II L savi I S ICS. IS IS
Jano tor way 0 snylll g .t ant. >"(T) approach ca 0 raptidly as r becomes
urge. As a result or tins, .lr >"(T') is narro w enough, the second term in
Eq. (1Z-24) can be approximated by
(IZ-25)
where we have used the substitution q ,.". Po I:. >. dr; This is equiv-
alent to the probability distribution
If first we set aside the random nature of the events, we obtain a char-
Qunnlum medwnlclf nnd palla ;nlesra's
330
ecterietlc functional equivalent to that of Eq. (12--16) ee r
(12-29)
Of course, each of these probabilit y fu nctions for the values of OJ has .
associated with it a cbaracteristic function or moment-generat ing func-
tion. Suppose we call this function W[w] and definc it as
Now we can p roceed as we did (or the derivation o( Eq. (12-17) and
introduce the not ion that the exact time at whi ch a pulse occurs is
randomly distributed with B uniform distribution function over the
jn~er.val 0 < t :5 T. If we suppose that there are precisely n pulses in
this interval, the characteristic functional becomes
(12-32)
where
331
As a special example of this result we eseu
extremely narrow, In fat l we ru ' I me that the pulse shape is
• I 't'lUmc I tnt we Cal .
shape Iun rfion hy (L delta fuuet io Ih ' l upproxirnate the
charaeteristie functional is ", at lS, u(t) = &(t) . Then the
pea) da = V1 tr"',... da
(12-36)
2.. a
332
remarking that a nu mber of physical proc~ .actu ally seem SO dis-
•,
tribut cd by their nature. I n trad itional proba b ility tl.l cor:r the normnl,
"
or gaussian, d iIS t flu
ibutI ion fils \lh"si('n1
~
Ilhcuomcua wh ich nr c the result I
of the cc mb iuntion of n large number of iud cpcnd cnt .cv~n t s Occurring
l Tl is is the conclusion of t he central-Iuntt theorem of
mn d D ill ) . 1 l I d i "b '
pro bability theory. 1 The sa me cOllC'lu sio ll. e pp res 0 15t l'I uhon
functicnals and result s in t he rect t ha t ma ny important cases for study
of phy sical phenomena have ga ussian distributIOns.. For further .rc~~
ercnee , we writ e here the most general form of a gaussian ch nr acter18tlc
fu ncti onal as
l) = ei/ k (" F(I) fltr}i/ / k CI) k(f'IA e, ., ' ) <It 4l' (12-40)
333
wh ere now the function net i') is a kern I . I
is, the functions A lmd n a;c relnted bye recrproen to.4 (i,i'). That
fA (I,T)8(T,8) dT - ~(I - 8)
(12-42)
Problem 12-1 Prove this.
The funct ion B appearing in this Jest expression is the inverse of the
correlation Iuuetion A. That is, fB(1 - slAts) ds = 1(/), or, if
(12-45)
(f(a)} = - i Ik~)
(12-46)
~
Ik(a)
= [- f k(t)A(t - a) dt] 4'
(12-47)
~I ~
. b Thi . ealled the correlation function of the noise. (;(
ti mes a and . IS rs . f E (12- 12) twice]
Jt. is Iby difTerent iat ing bot h sides 0 q. ~
3.'15
whore "Y {w) is the Fourier tmnsform orour pul f ti ()
· I · . I I
SC une lon g t " re rnn
expI alii t us simp C result more uirt'ClIy lor
r bl ·
Our p ro em D.3 Iollowa
If the pu Isee occur at times t so /(I) = \' g(1 I) tb F · .
" L., - " e ouner t rare-
form of /(t) is <>(...) - L ~(...)c" ' . Tb~ the square of <>(...) h.., the
•
average
(1<>(...>1') = IL h(...)I'c'o·....,'!
,~
(12-54)
But, since the ~ ime~ I. ere at random and independent of I, forj #- i, all
the terms for t ;II! J average out, because thc average of e""'i.-I,I is zero.
Only the terms with i .:: i remain. Each is 1"Y{w) I!, and they are itT in
number; 50 the mean of 1¢{w)12 per seco nd is ph'{w) l'.
In the special case that the characteristic function ran be approxi-
mated by th e white-noise chamcteriatie of Eq , (12-.2i;), the function
A (t - t') = conet !(t - I') . This means Ihnt <p(...) is independent of
wand there is the snme "rower" per unit frequency runge {mean
1<>(...)1' per second] at all frequencies.
The distributions we arc describing can very convenienlly be
described b)" giving the probability distribution not for / (t) but for its
Fourier tmnsfonn ~",) directly and the characteristic functional not
in terms of k(t) but its Fourier transform K(...).
K(...) = Ik(t)e'"' dt (12-55)
Using these functions, the characteristic functional for tbe noise distri-
bution correspondi ng to Eq. (12-43) is
(12-06)
by d irect substitution of the inverse of Eq . (12'<;5) into Eq . (12-43).
The corresponding probabili ty functional of Eq. (12-26) is
P "= g-Hl1tt (...)II IlJ(J'I.. n fl,., I2w (12~5i)
,
I
so that Eq. (12-14) implies
(12-.'i9)
p = f~IK Mt {oo) ""/lr :DI\:(w)
If we now .n nc grue
. theI pcssiible v alues of
.
(oJ 10 be discret e and sepe-
. .
( rated by an infi nitesimal slml'i ng 6 , the Ul It'F:r:'ls III the CXPOllCl\ t 111
Ec:1S. (12....50) and (12-57) can be replaced by R ICIlUl IUl sums, and our
Quantuna mechanic6 and pad. inregral6
336
path integral becomes
p = n ! e-(.1 J21IK(",)llJ'cw)e';.b,K(.. I . M dK(w) (12-60)
•
The integral for each value of (0) can be done separately (by completing
the square), and we get
(12-61)
•
Putting the product together gives Eq. (12~S7). It is clear that what
happens at on e frequency is independent of.wh.at !lappens at another,
and the signal strength ¢«o) at Irequ eney (0) IS distributed as a gaussian
with a mean square proportional to a>«o).
It is usually true that the path integral method docs not really help
to get the solution to problems that cannot be solved in some other
manner. Nevert heless, someone who has followed us this far and who
is now familiar with path integrals will find its mode of expression and '-ol
logic very simple and direct when applied to probability problems. .
For exampic, in the theory of brownian motion we might have 8
linear sys tem-s-say , a damped harmonic oscilla tor being driven by 8
fluctuating force 1(/). Assume the mass of the oscillator equal to I,
and we must 60Ive
x. - 'Ix. + .,,'x = 1(/) (12-62)
whereful x is . 1 VUl.
r relat ed to . J'"'~q . (12"{)2). In general we must be very
care 0 thc reinlion of ~ (I) t " if( ) • ' fa
"jaeohianlJ b t ' . x 0 . ) I, there being an analogue 0
r I ted ( ~ wcell tl~e "volume" ~1~l1lcnts. Dut jf f und x nrcliucllrly
c th! as u >O\'c), this jacobian ls a (·onstnllt. SO if us is usual with
po Ultegral!:l 'we n t ' ) . ,
•, C n rust ourselves to be able to normalize our
Other' problema In probnbility
337
a nswer in the end, we have
P.(r(t) ] = const P,(x - "'rx + wo'.r) (12-6-1)
whi ch gives us a Iormal solution. If P, is gaussian, then P, is and
the problem may h~ wo~ked o~,t i~l muuy ways, the most evident being
by the method of Fourier scnce if wo' and "'r are independent of time.
At uuy rn!c, many problems can he set tip and solved or partly
solved by U~In~ 1-A1. (12-6-1) a"l a starting po i nt . We shall look at a
specific examp le, A fast pnrti ele goes through matt er in which it
receives small, sharp ult erat ioua in velocity es a result of passage by
nuclei. Aft er gciug through a thick ness T, wha t is the prcbnbility it
will eme rge n d ista nce D from th e origin (t he extension of its original
strai gh t-line path) and will be moving with deflection angle 8 as in
F ig. 12-1 ?
"-c assu me that the inte mc tions cause no mensurable loss in the
longitudinal velocity of th e particle a nd Ihnt the ma tte r through which
tbe part icle is passin g is homogeneous, Further, we assume that 8 is
always small and that th e mot ion is the result of 0. large number of
collisions ea ch of which has 0. small effect . We assume that the
expected number of collisions in th e infinitesimal th ickness dt is ,.,. and
that the deflection suffered in each collision is given by th e angle A,
which is governed by the pruhnbility d istribut ion p(~) dil. We further
assu me that this probability distribution results in a mean-square
value of l:J. given by
Wc shall confine our nttent ion to the motion as projected ooto a two..
dimensional plane contain ing the original pat h of t he particle. Motion
in a plane normal 10 this wiU follow similar rules. and the motion in
either plane cnn be co nside red independe ntly of the other. \Ve shal l
usc t to m easure th e dep th of penetrat ion inlo Ih e sinh , 8 to reprCSCnt
the instantaneous direction of motion in the plane we nr e consideri ng
and r to measure the posit ion of the pa rticle away from an cxtcnsio~
of its original path of motion , as shown in Fig. 12-1. These parameters
arc related by dx c= 8 dt, or x = 6.
We assume that the deflections of 8 occur suddenly. so th a t IJ = f (t)
where t he fu nctions J arc a set of randomly spaced d elta funct ion~
having random scale heights. This m eans that z = f(t) and PAJ(t)!
has the characteristic functional
(lUll)
where
\~7e note that the mean val ue of .d is assumed to be 0, and these deflec-
1I0llS themselves arc assumed small. Now if we expand G(w) as
(12-69)
This in tum im plies that
p Af(I)1 ee e-<1I2RJ/ U(llJ· eta
(12-iO)
Hence
P ./Z(I») = const ex
p
I I f r
- 2/1 Jo ['«/))' dt
J (12.7\)
\Ve wish to evalua te tl
gives the ,)robnljility U a t tile probability distribution P(D 8) which
!J.~ I'd a11fd(! of lIlotion ~ die parOtic "'iII exit with t he d is;)I~cell1cnt
lIutlal con " ' 115 .r(O) = 0u n or the
, d 1I10 n ~· · . hi e
. -,"Uillp t te ns t hat it ent era W i t t I
the exact pat h that the and ' x(O) - 0 . w e nrc concerned 1I0t \\ I"th
part icle tak
es .III th e mat eriel, but only t I)6 I
Oth er problem. in probability
, ~9
t he particle exits with t he di ,
0011 ' 100118>(1') - D d ' (1')
we express this p robnLiJity dist ib ti b - . an x ... 8. Thus,
• n u Ion y an Int egral over all paths as
I
I
P(D,O) = f exp. ( - 2".
-!- J. TZ'dI)3)%(1) (12-72)
6 (D -
P(D,O) = eonst exp [ - I/T' O!,\'
2'J - 21I0'T J (12-76)
In some pract ical cases we may really be concerned not with the
exa ct linear spa cing of the pa rtide away from our assumed origin
poin t but, rather, with the deflection angle nt which it lea ves thc stab.
Givcn the overall dist ribu tion fu nction of Eq. (12-ili), it is simple to
evalua te the distribut ion function in a ngle alone b,)' integrating ove r
all values of D. The result is c- (,tmm. This is a ll expected result,
because we have already assumed Ihnt t he mea n-square value of the
deflectio n angle which would be acquired in a unit thickness is H I 50
t his value in D. total t hickness T should be RT.
Suppose next we look only at particles which emerge t raveling in a
specifi c angle 8 a nd consi der th e distri buticu function of the emerging
.;
posit ions D of those particles. We fi nd t ha t th e probability dist ribu-
t ion has a maximum a t D = 8T/ 2. T his would be t he position we
would oxpcet if t he fiunl deflection angle 0 were acquired in u smooth
manner us a linear Iuncfion of th ickness starting from 0 and building
QlICm'"111 mCclUlllrC5 01111 parll i"r('gl'flt.
up to its fiunl value. In that ceso its av erage value during the passage
through th e slab would be 8/ 2.
Problem 12..2 Show that. the constant required to normalize the
probability funct ion P (D,O) d D dB is
("6 1 1 (12-77)
const .", V;RTi V~iT
3-11
qunntum·m('rhnnirnl system whose un " .
u nder t he illn Ucll r c of an external t ~~111~1('t)ro acnon I~ B(q). IS
is' po cn lt~ such that Its ecnon
1'!x(q) ;4>(q)j = IfJx ' (q, ) K (q,h;q" t ,) 4> (q,) ,Iq, dq.I'
= fJJIx•(q,lx (q~) K (q,lt.qu';)K' (q~,I, ;q:.I,) 4>(q;) 4>' (q:)
dq, dq: dq, dq; (12-79)
It is evident that all such problems can be solved if we ca n evaluate
(12-80)
Thc first fact or involves the path integral I t'Slq(l1! 3)q(I), whereas the
second complex conjugat e! one is f r 3 1, (11I :Dq(l) . Each integral is
over path s wit h appropriate end points. I n wri ting the product of
Eq . (1 2-80), we shall call the path variable in the second integral '1'(1)
and we can then express Eq. (12-80) as the double path 'integral
fJe;S1••,H SI".lI nq(l) :Dq'(t) (12-81)
The sum ming of such integrals over various end points gives th e
req uired probability.
If t he pot ent ial T" is acti ng, we should replace S in E q. (12-81) by Sv,
.1
a nd t he expression becomes
i
I [fe'l Shr lt)J-SI'l' (I)I+1 , (I) V(l ) dl-I'l'(ll v(I) .u1 :Dq(l) X>q'(t) (12-82)
,
i in a lar. 11 .
"" ,
Quanfuni nlechnnics and pall. infegrals
3U
Eq. (12-70) calculated for each 1'(1) and then averaged over 0.11 V(O
each with th e weight P ,-P"O)J ::o1·(t). This is then
Prob (~_ xl = ffffx'(q,lx (q;JJ (q"q; ;q;,q:) ~(q,)~'(q:) dq, dq: dq, dq;
(12-83)
with th e integrals taken bet ween appropriat e end points q(l.} = g"
q'(I,) _ q:, q(I,) = or. q'(I,) = q;. ActuallY, this choosing 01 end
point s and then int egrat ing over various values with wnvc-function
d istributions depending 0 11 the problem [as in Eq. (12-83») is simply a
sum of J 's for different end conditions, and we shall hereafter simply
forget this and speak as though with J we already have our probability
- it being left to the reader to remember that n bit more bas y et to be
do ne. T his is so that we can concent rate on thc main feature, the
cvnluntlcn of the double path int egrals needed to calculate J.
I n this form we call do the integral over " (t) explicitly and sec that,
to find th e probabilities after averaging, we must eval uate a double
path integral
•
E quat ion (1"." -8i) then 01 IS\\ ' erg our ch uIIenge to express the answer
In a ,form Y~hd aft er t he averag:ing. It invo lvC'S evaluntion of the
~.OU b~ path iutegrul. ~Iow to evalua te it is, of course, allot her ques-
lOll! ut thc meth~ s d lSC'uSSl'd in this book may be useful I n these
sections we arc d iSCUSS' I . •
formulated, lUg 0 11 Y how various p roblem s may be
As.:on example of the appli cntio l' of Eq. (12.-8;» sUPl>OSC V(l) is
gaussian norse With zero In • d I . . J •
M3
Because in th e new Iactor at 1 t h
rut ienlly, some of the met! d cas t .c q nnd. r/ appenr only quad-
forms may he useful. or CC)Ul~cg i r~v;o.u ~y UHicusscd .for quad ratic
ing to a harmonic osdllntor ih
fJ ~s ~ tsd( c(lIodrobc, correspond,
exactly by using the met hod' r"Se pal Integrals can be evaluated
S 0 ec, 3-5.
t Q slllndll for nny numh('r of eoordtnntcs-r-thie othc~ system meybc, nnd g C(I-:r-
n11y
'
l!l, very COIll P
Ic....
' \\'e t1hntl J'u:>1 (' art)' cue Q \'"ru,h le, but notlilng tss{'nlml
will be lost.
,
•
Qllnnfllm mechnnic8 nnd pafh integrnl.
The 6Um on I means we arc to sum over ell possible finnl states of Q.
Thi s is because no measu rement on Q is to he t aken, and all final states
of the environment arc possible. Therefore we must add together the
probabilities [i.e., th e J functions of Eq. (12-88») of all. In coordinate
rep resent ation, for exam ple, L
, just means that a t some final t ime t,
only rough: ~or F cont~ins the entire erred of the environment includin
t~~c chuuge 111 hcllB~'10r of the environment resulti ng from f('nctio~
with q: In t he cln.~,c nl analogue, F would corres pond to kn owing not
only "hat the ~orcc IS ~~ n function of ti me, hut also .....hat it would he
for ,every possible motion q(1) of the object, The (Of("C for n given
\ cnvlronn~clltnl tsY5( e~n depends in general on the motion of (}(O, of
course) smee the enviromncntal system is affected by interaction with
the system of interest q.
We arc therefore led to study the properties of influence Iunctionals.
We shell be cont ent Lo list a few such rul es and give some suggestions
on how they arc arrived at,
Rule I
F1q(I),q'(I))* = F1q'(I),q(t)) (12-91)
equivalent to
fK(Qfh;Q.,~)K"(Qlolf;Q:,~) :DQ, = ~(Q. - (/;) (12-96)
b)' Eq. (4-37). Thus, if the initial wave fu nction were ~(Q,). we would
multiply by ¢(Q,)¢"(Q.) as we did in Eq. (12.79) and integrate to get
1\0'" notice tb at if we put g'(1) = g(l) for a given fixed g(1) for aU time
in Eq. (12-90), ,~ c have an expression just like E q . (12-95) where the
effective (and definite) action is
S.[Q(I)J = S,IQ(I)] + S,..[g(I),Q(I)]
with
S. IQ'(t») = SolQ'(I)] + S••Jg(I),Q' (t))
as required, as !0I1g as g' (I) = g(I). Hence F [g(l),g(I») - 1. .
'I1IC same argument limited to the time range a < t < tl, usmg flo
relation like Eq. (12-00) but with t" Qi replaced by a, Q., shows that , if
g(1) = g'(I) for I > a, th e dependence of F on g(1) for I > a d rops
away, beeause the right side of Eq. (12-96) docs net depend on g(l)
fort>a.
Rule III is an evident result of the fact that probabilities arc deter-
mined by adding the value of J over va rious circumstances.
Rule IV is evident from Eq. (12-90) when it is rculiecd that the
conditions of the rule imply that the action that goes into Eq. (12.00)
is S,. [Q. (I)J + S ,., .[g(ll ,Q.(t)] + S, . IQ. (/)J + s... •lg(/),Q .(t)J nnd
that the exponential of the sum becomes n product as docs the integral
F, if the initial state is itself a product of wave Iunet ions.
Rule" is merely a statement of our results shown in Eqa (12-82) IlIld
(1 2-85). .
'l11CSC . arc so.me of the general properties of influence functionals.
Calculations With them involve the various meth ods for doing path
Other problem. in probability
341
integrals appli ed to Eq. (t~9) . We shall concIuc d this scctilon by
d tscussmg eertaiu Important Influe nce fUllctionalso
o 0 0 0 .
for arbitrary and complex I o , fJ, ")', and 6. The integrals on t', t nrc OVer
the entire interesting ra nge of time, but we always take t > t . This
is no loss of generality, of course, but it is convenient Ior Inter anntysis.
For t his to he a satisfactory influence functional, we must have from
Rule I
/1(1,1') = a'(I,t') (12-100)
and
,, (I,f) = 1'(t,I') (12-101)
Rule II gives us a .grent deal of infonnat ion, for putti ng q(l) = q'(I)
r
for I > a and, assum mg I > 0 , < 0 , the expression [which is part of
the integral in Eq. (12-99)]
exp {- Jf' [q(l) - q'(I)j[q(t')a(I,I') - g' (I' )a ' (I,n ] dl' dl} (12-1(}l)
In (he case tha t aCt I') is real sa '
to the cxponCJll ial ~f E: (12-8 y, A(t,t ) Our fUfl ct ional is equ ivalent
noisy classical Jlerturbn~'· ;), and we have (he equivalent of a
IOn. n general, in quanlum.mcchau icw
J These functions are dcfmoo only (or I > I'
I
I systems a is complex.
. Other problema in probability '
Ir
nothing because the states nrc orthogonal. The next term linear in a
has four pieces. One ia - a(/,t')?(/)q(n de d/. When this is
substituted into Eq. (12-89) in place of F and this evaluated as in Eq.
(12-83) with 40 = 40. and x = ¢., the int egral on :Dq(1) and :Dq' (l) is
seen to be the product of two factors. One, the integral on q, involves
The integral on 'Dr! is just [e'S(". 'IJtf and is the complex C?nju~tc. of
. . I t (I)
t h e transition c ernen ... ,..
Anah·zing the other terms In a similar
.. .
way, the total transition probabilJty IS
850
Problem 12-3 For m = n, verify P(m-tm) - 1 - ~P(m-n)
•
as required by conscn 'a tion of probabi lity .
[a (l) is 1I0 t defined for t < 0.] Then since Pen - m) in Eq, (12-108) is
proport ional to the time interval over which the integrals extend,
we can define a rate of transition per second and find the probability
of transition
P(n- m) per second ~ 2aR(E. - E.) !q••!' (12-110)
sY/'d em .~ gomg up 1TI energy V:. > Eft) nrc u nlikely. Hence lor such
CD -eu tn m meu l systems
Rate of transition n -+ m = 0
if E... > E" (12-115)
1
Other p roblerm in pro bability
I 351
I Si nce any a( , ) can be writ te n ee th e sum of one of the type shown in
Eq. (12-112) plus o~e ol. lhe ty pe shown in Eq . (12-11-1) , it is readily
\ apparent thet nny lImc-mdcpcndcnt gaussian fu nct ional is equivalent
t o n syst em in some cold environment acted on by n I luetuat ing elassi-
! eul pot en t ial described by e gaussian expression. T his conclusion Iol-
10"'5 from th e fact that th e produ ct of any two gaussian functions is
also n gaussian and from rule IV. H th e interaction of one environ-
ment on th e sys tem is represented by A ,(I ,t') in the manner of Eq.
(lz..87) and the interaction of th e other environment ua A , (l ,t'), th en
t he single in te raction t erm in t he single resulting gnussiun fu nctional
iBA, + A •.
Flq(t),q'UH ... J dQI K (Q" I/, Q" , ro h:lfIlloniC' osd llntor ..- ith / ln. .. Cq~t)
~. kernel or E q . (3-£16) ror a ror~ \" '.CfUllr l ioll or t he o!lt'lllalo r m
:~~r~~~9t~~~et with f (l ) ... Pl(I)'Q ¢~,Q)l\:l't: ~1~(,1l nrl~"r in ft f'llll~lt't ~a~"""'~l\n
the Rfou ml sl nl e. AU vn.f1nhlt'!J ; \Ye Ilha ll II&('n find It NI eesy to to e III c-
be directly Inh'grnl .
wayan d mllY
Quem',.," lIIer/lonic. and pnlll in'egrnb
3S2
with G OD given in Eq. (8- 138) and fJ* in Eq. (8·143) repl.acing 'Y by
Cq(/). Likew ise, the integral over Q' is the complex conjugate of Q,
similar exp ressio n but with .,..(1) replaced by Cq'(t) this li me. We
distinguish values with this subs titution with a. prime. Then the SUIll
on final states in Eq. (12-117) gives us
For example. tbe terms in qq' in Eq. (12-104) come directly from the
/l'tI in the exponential; for this product by Eq. (8-143) is
The quantity 0(') defined in Eq. (12-109) is therefore [cf, Eq. (5-17)
and th e Appendix]
e
0(') = 2" Jo
.« e-i"'e-'"
W
: dt = -2 C' [
W
-r
1
i P.P. w+JI + "~(,,, + .) ]
(12-122)
so that the real part of a( ,) is
xC'
aRC') = 2w ~(w + ,) (12-123)
I If many ind ependent osc illat ors of different frequencies nrc anacti ng,
353
then hy rul e ]\', their OR(l') funct ions add ; so any cold syste m (to
\ this gaussian a pproximation) is equivalent to n conlinuu m of oscillators
( in their ground stale. Thi s follows, since any function aR( II), for
negativ e .., cnn be buill up of Ii fu nctions of the form of E(I. (12-123).
Another intcreatiug exa mple is the int eract ion with nn oscilla tor
at finite temperature. If the te mperature is T, the initial slate is sl ate
n with relat ive probability e- S• fIrT • For our case, the absolute proba-
b ility is
tD. -= e-nAooI'u(l - C-,..,/l'T) (12-124)
l FII = 1.0....o:.. .
m
(12-125)
,
instead of the fonn in Eq, (12-119). Using rule Ill, we odd th ese
with probabilities w.,
60 our final F is
(12.126)
F -= L G"'.G'".·.. r"""J:I'(l - e-"",J:ry
m,_
The sum is difficult to work out directly from Eq. (S·H5), hut it is
F = G.,{;'.,e"·" cxp [ -
(fJ - fJ')(fJ* -
t!'",OT - 1
p'*)] (12.127)
The ••(_) that results from this in place of Eq. (12·123) i.,
. r man
oS("ilIntors constitu te the environ-
andsumsof such cxptCSSlonso ee-v (.. < 0) or up in energy.
Y.
. , go down Ul ene. e-r
ment. New tranSitions ca n . funct ion fails whereas if v < 0, the
\Vc note that if v > O. th e first ~ ,
second f nils, and that indeed
(12.129)
••(_ I_I) = ."""r••(+1_1) . •
. erturba tion theory, If E.. > EM,
This definite rc1al ion means that in P
if cr second up (m - 71) = e-{Ii"·-A"..lI l T
prohabili ty of a tronsl 1011 p ~nd down (It _ m)
probahility of 0. tra usitioll per sec (12-130)
11lUS , 'I
I
II f
IC sys cm q
occupies states n with relative probabilities
. . '11 1 I '"
C ".,fI.T the net nu mber of up and down tranSlhons WI in anee ~ut
an d th~ system will be in sl at i~l i('aI cquilibrium for weak pcrturbatlon
with the environment. This , 1S
' J, US I \\-hnt
1
we expect
" 1' - -
from the Inw6. 01
stnt istical cquilibriurn. Any environment at temperature : produclllg
a Quadratic influence functional will have: the p.rop erty of I'.q. (l2-1 29~ .
Fe r an atom as svstem q in iuternctiou With the electrom agnetac
T
field at temperature as the environment, a R(") is given by a n expres-
sion like Eq. (12-128) integra ted over all th e modes of .t he field of
various frequencies Ca1. H can be spl it int o the cold environment of
Eq. (12-123) plus e noisy external potential :
7fC!
aR(v) = -2 6(", + v) + e"'''T
1 "e' [6(", + v) + 6(", -
1 2",
v)J
'"
(12-131)
Th e first term produces only transit ions down in cncrgy and is called
spontaneous emissi on, The second produ ces transitions up and down
with equal case and is caned ind uced emission, or induced absorption.
" 'e say that the transition is ind uced by an external poten tial or noise
whose mean-sq uare strength at frequency .. varies with temperature
as l / (cl • l lT - 1). This is the way Einstein first disc ussed t he black-
body-radiat ion laws. As we see here, any en viro nmen t giving a
quad ratic influence functional at temperature T (we say it is an
environment respo ndi ng linearly) can be t rea te d in the same way,
Xleny people have extended Einstein's argument to o ther systems,
Jike th e voltage fluct uat ional noise in a resistor at temperature T,
Th e fi rst ~ e rm measures the rate at which energy is taken out of our
~·.st ~m q In a one-way mallt~cr, It measures the amount of "dissipa-
non produced by t he environment (e.g., electrical resistance of 8
metal or rad iation resistance of the clectromognet ic field) . At tom-
)J~r~t u r: T we eun then M )' that th ings hehave as if, in addition 10 the
dissipat ion, th ere is a noisy sib'lull gencra tod by the enviro nment whose
m ean squ are at car-h frequen cy is proportional to th e d issipation at
that frequency
• , .
and to (e". I.r - 1)-'• 'I'hi , - 'I d I d ' , / '
115 IS Cw e t ie lSS Zpa wn-
jl I
ur ua/ UrII theorem .
'Ve can not purs ue tbis subject further here.•
J:;
I111e l>IlbjCC'l or in nuC"n r
F. L. \ 'tt"nUtl Jr A70
ti I ' .
unc ,1fJ~a S ll> cIl:if:U&lfil in d et nil ily H.. P . Feynrnan and
PAy,. (N .y .): \·~i, 12 "~. I v;~~ . ) l've l. 2.1, P: lI R, 1!.l():J, nnd W . II . W('lIlt, .. h ll.
polaron i!iill H. "'C"}"n;nun 'n W · tJ I ~ ,1l 11,.hcoIlI KlO t o cnlcul ut ion or 1II0llilily of Ibl'
PI'1J'l. /{ctJ., vel. 127, p. 1004, '1002. e \\arth, C. K. Iddings, and P. M. l~nt zmann,
,
I Other problem.a in probability
I 355
12-10 CONCLUSIONS
I h
I
II
I
01· icr me l rode, 110t rcqUlnng path inlcgm'- arc'
G'J,
., bl
a so ave, n e to solve
l c p.ro) ern. One may rC'D...<:OIInhly clucslion the real utility of the
~at.h integrals. We ca n only say that if the problem is not gaussian,
Jt CBn a t Icas~ be formulat ed and studi ed by using path integrnls-c-
~nd that we ought hope that someday, when the techniq ues of a nalysis
Impr.ovc, ll1~rc can be done with it. The only example of a result
obtained With path integrals which cannot be ob tained in simple
manner by more conventional methods is the variational principle
discussed in Chap. II. We hope Ihat further study of these methods
may yield more such results.
In the meantime, however, it is worth pointi ng out that the path
integral method docs permit 0. rapid passage from one formulation
of a problem to another and often gives a clear or quick suggestion of a
relation which can t hen be more slowly derived in n more ordinary
fashion.
With regard to application to quantum mechanics, path integrals
suffer most grievously from a serious defect . They do not perm it a
discussion of spin operators or other such operators in 8 simple and
lucid way. T hey find their greatest use in systems for which coordi-
Dates and their conjugate momenta are adequate. Xevertheless, spin
is a simple and vital part of real qua nt um-mecha nical systems. It is
a serious limitation that the half-integral spin of the electron does not
find a simple and ready representa tion. It can be handled if the
amplitudes and quantit ies arc considered as qun ternions instead of
ordinary com plex numbers, but the lack of commutativity of such
numbers is a serious complication.
K evcrthcless, many of thc results a nd formulations of path integrals
can be reexp resscd by another mathematical system, a kind of ordc~ed
operator cal cul us. I In this form many or th e results of I he preceding
chapters find an analogous but morc general repr~l~tation (only for
the special pro blems of Chap. 11 is t hc gencrah~atlO1.1 not kIlO'''-!')
involving ncncom mu ti ng variables. For exa mple, III th is cha pter dis-
cussmg , ' IIUCI h...,
HI - fIl l Ict ·ionals it must have struck the reader that •nn
,
euvironmout e 110 I 10 the coonJinnlc q but to a ucncom muttng
coup led
• A 0 lltor Calrnlll!! 1I1n'i ng ApphcutioDe in QunntulD
I H. P • •' l',)"lI ll1 lln, II ' K"r ~ I
E lcetrOlly nnnlirll, /'l1l/s. Ret,., vol. 84, p. 1O~. 195 ,
t......,..... I'hN"H ..'n _"cI ,..'h ,,.,,,,,..,.
SAO
ClI»c'rn'nr, .tlrll lUI lit,. 1IIIIu, woull' IH'I All IlUlMlrlnnl nntl IfllI·" ... 'i'IC
1:,...llt'r.li,n'inn. :-;11 r-11 rll illW' mlillol I... N)m'f'lIil'ully ' ·Xllrf·M, ·d ;11 IIII~
pntll i111f'/otr.1 rurlllul",iull hut rAn I~ rllJlily f'Jlpfl'MM' ill II..• .·It....."
,,·I""'tl (ll~·r.lor 1".ll"lIhll',
An ...lTurt 10 .."Ie-ltd "II' I'l'llh Inlf'R:ml RppmlU'h I)("yon,' it,. """"'nt
Iil1lilll r'llluimll'" 10 he' A worthwhile JlUnOl;I: (or Iht'! Rrrnlf'flt vnhl" or t:!
thi,. Irrllllilill" rt'fI1"illl' in fIIpifr or ilA limi'nt iOlu" i.e., the flNIilllnfl~p' 'J
"bid. it Ioth't.,. 01""" intu ition in hringing together phy"i c41 in"ight And
IIlAlhl'nlQtiraJ Q(JAI.)·ala,
Appendix
Some Useful
\ Definite Integrals
I
\
I
t: . - -"
~ dx = -1
k + 1)
e-"~ :r,-
,1'1. - (H I)/"'r ( 1'h
o m
I
<
.
<
I
Index
Absorption, 153
A ~t'l c rnt ion , 174 CofTClntion (unction, 333-33t
Act iun, I iii, 26 - 28 Coulo.,n!' pot ent inl, n7, 239
AIllJJl ittltJe, .5- 9, 13-16, 19-21 CoUp11n1l; eoefJid ('nt, 70--1 1
Ht'lltter ing, 142 Cro!ill '!Celion, 135-1 39, 152
Annillilllt ion, 232 Cryllltll,212-2 17
App~x illl ntion o ( con t inuit)" 2 18-222 pol a r, &low c1Cl"trone in, 316-320
Ato on c po tcntinl, 13t-U I Curnnt den., it y, 237
Avc rnp;c, o ( Cxponcn t inl, 2&1 Cut-orr rul e, 245-246, 2S8
o( mcns u roo q un nt it)", 112-11 3
weighted, of po tctltinl. 305
d -e \'olt nge, 328-329
D~en l!rnte 8tn lf~9, 160
Dcthc, If " 2M Delttl (unction, 53, 87, 281
DJnck-body Rnd int ion La w, Planck Den ~i ty of lu 'cI'l, 151-1 52
294 - 295 • DenSoity matrix, 273-277
Boltzmann 's constant, 268 Dirrrnclion, eleet mn, 140
Bom approX~ll1l1tion, 128. 131, 169 o( Ji,::Jlt, 11- 13
Bom expne aio n , 128 of pn.rt icJ~, 47- 57
Bose particles, 16-17, 231,244, 288 D ipole npproximtltion, 250
Boson (see Bose particles) Dirae, Pau l A, M" 23, 2S5
Dou nd nry con d it ions, period ic, 92-94., Dime d cltn fun ction, 53, J03
213- 214 Di ra c cq untion . 36
Box, nonnatizing wave functions in, Dispersion , 225
89-92 DiSlIip3tioa-8uctWltion theorem, 354
Bro wn ian m otio n, 165-166, 336-340
I
t um , 100
P eriodir- boun dnry cond itio ns, 92-94.
M e90n , 231 , 254
213- 214
Minimum principle, 302-303 Pcriotlic pat h8, 284
Molecule, diatom ic, 139
I
P erm utat ion, 28S
polr etcmtc, 140, 203-208 P ertu rbatio n cxpnnMn, 120-127
:Mom cnt..ge n~ting tun ct ion, 325 firm:-ord cr. 2.'i2 - 2S.1
M omentum, 8- 1I, 28, 44-45, 184-189 time-d cpcndcnt , 1U-1 61
amplitude. 96-102 Perturbed po te ntial, 125
fun ctional, 187 Phase, cha nge of, 4S
measurement, 96 of pnt h, 2d
opemtor, 116, 185-186 Phono l1!, 2'22
or photon, 248 Pboto(!lectric effect , 153
probability, 9;-100 Photon, 231, 2-13
representation, 185 Pines, D., 318-319
epe ce, 101-1 05 Planck, M., 2-&3
Planck Block-body Radint ion Law,
294-295
Newton's law, 175, 251 Planck'! consta nt, 9
No ise, 326-336 P lessct , :\1. S., 23011.
classical, 350 Poisson d btribution . 322
correbuion funct ion of, 334 Polar crystnl, slow elect rons in, 310-
gnussten , 329, 33 1-336 320
NondeJ;cn crnte st at es, 160 Po la rization, 224
N onrclntivist ic npproximation, 237 oscillat ions of, 3 12
Norm al modes, 2U6-208 Polaron, 310
coordi nates, 208 - 212, 2 16-217 Position uieaaurements, 96
d istri h u tion. 50 Positron" 39
Normnlizatiou Iect cre, 178 Potcntinl, effecti ve, 2S5
of wa ve fun cl ionll, 89 -9-1. and moment um reprcsentat jon, 102
Nononlizing const nn t , 33, 01, 194, Z15 Potcnti,I1-cnt"r)O' term. 72
Fbtcnlill.l fun ct ion, ecriee expwl ilion,
6:Hl3
.• Operntor eefcutu e, 113
Opcmto r notation, 80-82, 170, 184-
Pewer, 153
mt"An, 33-l
spe ctrum, 334-336
189 .
Opcretcre (ete Opcmtor notatIOn) I'teisurc, 270
Quantum mecllCJnic.s and path integrals
3M
I ,......
165
Vt\ltap. ftut1Wllionll nf, m Wavftmcth and mom~lWlll.~. W
,'Numl' ... aonnahainK rnnd.ant.. WaY"'. tt'tardt'<I, 2-\1
03-G4 Nnninl. 2-42 -24.1
tr.lUITt'~. 22.\, Zl3
I,
,
I
,
I