Quantization of a General Dynamical System by Feynman's Path Integration Formulation*
K.S.Cheng
Institute for Theoretical Physics, State University of New York, Slon" Brook, New York 11790
(Received 7 July 1972)
The Schrodinger equation is obtained by Feynman's path integration method of quantization for a general dyna-
mical system. The meaning of the results is discussed.
I. INTRODUCTION + l3 (~~ i I + ~ iJ ~ a I) t,qjt,q mt,qnt,q
Bql/rnnl /all /rnn\
I + .. J.
J
We present a derivation of the "Schrodinger equation"
for a general dynamical system using Feynman's (6)
path integrall method of quantization. The result We also need the following expansions:
differs from the "usual Schrodinger equation" in that
there is an additional term proportional to the total -Jg(q(t» = "fg(q(t + E» _ t,qi a-Ji
aq'
curvature R of the coordinate space defined with a
geometry given by the kinetic energy~ This result + 2D.q'D.qJ
1 . .
-B2..fg
. - . + "', (7)
had been given before by DeWitt. 2 In a curved space Bq'aqJ
or in cases of constraints where R "" numerical con-
stant, the presence of this additional term would 1J;(q(t), t) = 1J;(q(t + E), t) - t,qi ~
change the energy spectrum of the whole system. In Bq'
Sec. III we discuss the meaning of this additional 1 '. a2 t/;
term. + 2 t,q't,J - - + . ". (8)
aqiBqj
II. DERIVATION OF THE SCHRODINGER EQUATION In these equations
We will give a detailed derivation of the Schrodinger t,q = q(t + E) - q(t) (9)
equation for a general mechanical system by using
the path-integral method of Feynman. For a given and \ i ( is the Christoffel symbol,
mechanical system described by a set of coordinates /rnnl
q (q 1, q2, ... qN), let the Lagrangian be
~ i I = gik[rnn ' k] , (10)
(1 ) /ulnl
Following Ref. 1, we can generalize Eqs. (1)-(18) to [rna k] =~ (Ogmk + og"k _ Ogmlz) (11)
, 2 oqn oqm aqk'
the above system, that is,
1J;(q(t + E), t + E) = (l/A) I exp [(ijn) S(q(t + E), q(t»] and (gi k) is the inverse matrix of (g i k)' Keeping the
zero-order term (1/2E)g ij t,qit,qi in the exponential
x 1J;(q(t), t)-Jg(q(t»dq(t), (2) and expanding higher-order terms into power series,
we get from Eq. (2)
where 1J;(q(t + E), t + E) and lJ;(q(t), t) are, respectively,
wavefunctions at time t + E and t, S(q(t + E), q(t» is
the classical action, that is,
t/;(q(t + E), t + E) = i I exp(2~E gijt,qit,qi)
S(q(t + E), q(t» = minimum of Itt+E L(q(t'), q(t'»dt' (3) x [1- 2~E gij ~r~nf t,qit,qmt,qn
with the boundary conditions
+ 8~Egii j;J3f t,qmt,qnt,qat,qB
In:n\
A is a normalization factor to be determined later
+ 6~E gii (B:l ~1f:n\ + ~~l\ l,:nDt,qjt,qmt,qnt,ql
andg is the determinant of (gij)' Taking the limit of _ gijgst j i (\ S I t,qit,qtt,qmt,qnt,q<xt,qB + .. J
Eq. (2) when E ~ 0, we can derive the Schrodinger
equation. Now as E ~ 0, the factor exp[(i/IZ)S(q(t + E),
81Z2 E2 Imnl ta/3\ J
2
q(t»] oscillates very rapidly. Only the vicinity of the x i";g(q(t + E» _ t,qi a-Ji + t t,qit,qi B -Ji .)
stationary point of S(q(t + E), q(t» contributes to the \' aq' Bq'BqJ
integral in Eq. (2). The stationary point is
x (t/;(q(t + E), t) _ t,qi at/;.
q(t) = q(t + E). (5) aq'
As we shall see the region which contributes to the
+ t t,qit,qi B2t/;. + ...) d(t,ql) ••• d(t,qN). (12)
qq'BqJ
integral in Eq. (2) is It,q 1= 1q(t) - q(t + E) I;S El/2.
Thus we can expand S(q(t + E), q(t)) as a power series The following are two useful identities:
of t,q. This is done in Appendix A and gives
S(q(t + E), q(t» = iE gij(q(t + E»
Ii: ... I exp(2~E giit,qit,qi)d(t,q) = (irrIiE)N/2 g -1/2
(13)
x [t,qit,qj - ~~n~ t,qjt,qmt,qn i 1: ... i exp (2~E giit,qit,qi)
x t,q<X 1 t, q <X 2 .•. t, q <X2m d(t,q)
+ ! j I j j I t,qmt,qnt,q<xt,qll
i
= (irrnE)N/2 g-1/2 (iliE)m{g<Xl<X2g<X3"4
4 Irnn\ Ia /3\ •.. g"2m-la2m
1723 J. Math. Phys., Vol. 13, No. 11, November 1972
Downloaded 06 Apr 2011 to 129.2.41.235. Redistribution subject to AIP license or copyright; see http://jmp.aip.org/about/rights_and_permissions
1724 K. S. C HE N G
+ terms with other possible permutations of space S of larger dimenSion. Does canonical quanti-
(14) zation in the larger space S lead to a unique Schro-
dinger equation in the subspace? The answer to this
There are altogether (2m-1)(2m- 3)···5'3'1 terms. question is no. To analyze this question, one would
have to investigate first the constraint to be applied
Using the two identities, we can easily find out the co- to the system in S so as to restrict the motion to the
efficients of l/I(q(t + E), t), al/l/aqm, a2l/1/aqmaqn. The subspace. This constraint is to confine motion to a
calculations are in Appendix B. thin layer of "thickness" t::.. (q 1 .•• qN) around the sub-
space and then to approach the limit t::.. ~ O. In classi-
Here we just write down the results. Equation (12)
cal mechanics any nondissipative constraint would
becomes
yield the same result in the limit t::.. --) O. The limit-
ing trajectories would satisfy the Lagrangian equa-
lj;(q(t + E), t) + E aa~ + ... tions for the q's, and one need not concern oneself
with the larger space S. In particular the thickness
:::::: (in; E)N/21l/l(q (t + E), t) t::.. can depend on ql .•• qN. E.g., one could have
+ tnE - --= -a-
.", [11
2 -Jg aqm
(1-
vg g mn -al/l) -
aqn
-R'I.]I
6
't" "
\
(15) t::.. :::::: A(ql .•. qN) E + O(E2), (21)
where and take the limit E ~ O.
R :::::: giiR ij , (16) In quantum mechanics, however, the constraint pro-
duces a zero point energy. The limit for the Schro-
Rij :::::: Rfja' (17) dinger equation would then depend on precisely how
the limit t::.. ~ 0 is taken. If one takes (21), and the
fact that A ,c const, the Schrodinger equation would
acquire an infinite term a(AEt2 which varies wildly
over the q's. Consequently, the Schrodinger equation
approaches no definite limit. If, on the other hand,
Compare the coefficient up to order E in Eq. (15). We
one takes A :::::: const, then everything depends on the
get higher order terms in O(E2) in (21).
A :::::: (innE )N/2 (19)
(f) To summarize, for a case R ,c 0, canonical quan-
and tization does not produce a unique Schrodinger equa-
tion, and embedding the system in a higher-dimen-
sional Euclidean space would not help to produce a
unique Schrodinger equation. The correspondence
Equation (20) is the "Schrodinger equation" using limit also does not uniquely determine a Schrodinger
Feynman's path integration formulation of quantum equation. Feynman's path integration formulation of
mechanics. quantization, however, does produce a unique equation,
which is Eq. (20) above. The "usual Schrodinger
m. DISCUSSION equation" appears to be foundationless.
(a) Equation (20) above is different from the "usual
ACKNOWLEDGMENTS
SchrOdinger equation" in which the term n2 R/6 is ab-
sent. Notice that both equations are convariant under The author wishes to thank Professor C. N. Yang for
any coordinate transformation ql • •. qN ~ Ql ..• QN. his encouragement and guidance and Dr. C. S. Hsue
for many useful discussions.
(b) In case the curvature R vanishes, one does not
have to discuss which of the two equations is to be APPENDIX A
preferred, since they are the same. Such is the case
when the kinetic energy is that of a collection of non- In this appendix, we want to expand S(q(t + E), q(t» as
relativistic particles in Euclidean space where N :::::: (3 a power series of t::..q. The equations of motion are
times the number of particles).
g 1 (a g mj + __
". :::::: _____
. qJ ag ma
._ - - a j ).qaqJ
ag - '. (AI)
(c) If R ,c 0, it may seem at first sight that canonical mJ 2 aqa aqJ aqm
quantization rules will yield the "usual Schrodinger
equation." That is incorrect! In fact, only in the case
g ij :::::: constants are the canonical quantization rules
or qk:::::: -1:/3\ gagB. (A2)
[p i ,qj] :::::: - ino i j Via Eq. (A2) it is not very difficult to prove
d 1 '. '.
unambiguous and independent of coordinate transfor- - ('2g .. q'qJ):::::: O. (A3)
dt 'J
mations (if they maintain g ij :::::: const). If R ,c 0,
"canonical quantization rules" are ambiguous. That is,
(d) The limit Pi ~ 0 of both equations the same give 2 S(q(t + E), q(t» :::::: J/+€ Ldt
results as classical mechanics, since the term
- n2 R/6 is an equivalent potential energy and app- : : : [t gij(q(l + E» qi(t + E) qJ(t + E)l E. (A4)
roaches 0 as Pi approaches O. Now if we know qi (t + E) as a series of t::..q, we know
(e) If R ,c 0, one can always embed the coordinate S(q(t + E), q(t». In order to find out qi(t + E), we need
space ql '" qN as a curved subspace in a Euclidean ·q·k. From Eq. (A2), we find
J. Math. Phys., Vol. 13, No. 11, November 1972
Downloaded 06 Apr 2011 to 129.2.41.235. Redistribution subject to AIP license or copyright; see http://jmp.aip.org/about/rights_and_permissions
QUA N T I Z A T ION 0 F A G ENE R A L D Y N A M I CAL SYST E M 1725
';i k = - (o!r 1:/3 ~ -
k
21m /3\ 1;D qaq8qr • (A5) (B3)
and
Now for small E and I~q I:; El/2, we can write mn
_ag _ =_gmagn8 ~ ag (B4)
aqk aqk
qi(t) = qi(t + E)- Eqi(t + E) + (E2/2!)'qi(t + E)
Equation (B4) can be derived from
- (E3/3!) qi (t + E)+ ..• (A6)
or _0_ (g gna) =_a_ (on) = o. (B5)
oqk ma oqk
qi(t) = qi(t + E) _ Eqi (t + E) _ E2 )/ i (qaqB
Tf'
2! 0!J3\
(c) lJ,.
+ E3 (_0 j it _ 2 ~ i 11 m ()qaqBqr
3! oqY /aJ3 ~ lm/3\ ay \
+ ....
(1) ! J exp (~ g ij b.qi~qj\ Ii d(~q)
A 2nE }
(A7) = (inl'iE)N/2 (B6)
From Eq. (A 7) we get A
(ji(t + E) = b.qi _ ~ j i { ~qm~qn (2) !J exp(~g"Mi~qj)
A 2IIE IJ
X Ii
E 2E lmn\
~ (-; 1 i t + j it j a t) ~q m~qn b.q X [_i_ gi j i ( j m (~qa~q8~qy~q6
+
6E oq mn\ tall /mn\
I + • •• •
(A8) 2IIE Y la/3\
(om\
Substituting Eq. (A8) into Eq. (A4), we get
S(q(t + E), q(t» = iE gij [b.qib. qJ -l~n\ b.qj~qm~qn
+ ! lip
j (b.qmb.qnb.qa~q8
4 mn,IO!J3\
+ 1 ( 0 j i (+ j i/ \ a ()~qj~qm~qn~ql+ .• J.
3 oql/mn\ /0'11 /mn\
(A9)
This is Eq. (6).
APPENDIX B
In this appendix, we calculate the coefficients of 1/1,
01/l/oq, 021/1/aqoq by using Eqs. (12)-(14).
2
(a)
oqmaqn 2A 2nE
J
0 1/1 : ~ exp (~g ij~qi~qj)~qm~qn·./id(~q)
= (irrliE)N/2 (iIIE) gmn (B1)
A 2
(b) a1/l : ! J exp (_1_ g i' ~qi~qJ) fi (~\
aqn A 2nE J 2liEJ
x g \ i ( b.qn~qjb.qab.q8d(~q)
+ ~ g iygj61~J3\ l~n\ (ai3yomn)(.
'J (O!{J\
(B7)
+ -1 J' exp (- i gi,~qlb.qJ ' ) afi
_, ~ql~qll'
A 2nE J aq' In these equations, (a 1,01 2 ..• 01 2 m) stands for (gal a2 ..
ga2m-la2m) + terms with other possible permutation
= (irrnE )N/2 (_ ilIE) _1 1i \ g" { gnjga8 of (01 1 01 2 ••• a 2m ).
A 2 af3 IJ
The term
+ gnagjfJ + gnB gja} + (irrnE )N/2 (iliE )gna 1/3 I
A af3\
igiygj61a~} l~nf (a/3yomn)
= (irrn~N/2 (_ ilIE) (~gaB 1:/3\ + gna l~D
= t g iygjOl~J3H~n} [gmo (ai3yn) + gno (/3yom)
+ (irrIIE)N/2 (ifiE)gna \ {3 (
+ga6(mnyf3) +g/3o(mnay) +gyo(mna/3)}
A /a/3\
i
= (irrn~N/2 (_ ilIE) ~ ga61:/3\ = tg iy ia J3\ j~~\ (ai3yo) + igiy l~J3H:o\ (a/3yo)
_ (i7T1IE )N/2 (''''
_
A
)1
ZrtE - -
1 -0- (V-
2 ~ oqm
g g mn) . (B2) + igiy In:/3H~~ \ (O!i3yo) + igiy 11~/3H~~ \ (a/3yo)
In obtaining Eq. (B2) we use the identities + igij \ i 1j j 1 (ai3yo).
/a/3\ /yo\
J. Math. Phys., Vol. 13, No. 11, November 1972
Downloaded 06 Apr 2011 to 129.2.41.235. Redistribution subject to AIP license or copyright; see http://jmp.aip.org/about/rights_and_permissions
1726 K. S. C HEN G
Thus Eq. (B7) becomes + 1 (I {3
2 Im{3\ Ina\
(I a (+ a
aq"
I (3
Im{3\
t) (mn)lJ
(a{3yo) (B8)
• Supported in part by NSF Grant GP32998X. 2 B. S. DeWitt, Rev. Mod. Phys. 29, 377 (1957).
1 R. P. Feynman, Rev. Mod. Phys. 20, 327 (1948).
An "H-Theorem" for Multiplicative Stochastic Processes
Ronald Forrest Fox
School of Physics, Georgia Institute oj Technology, Atlanta, Georgia 30332
(Received 7 June 1972)
In a recent paper the author showed how multiplicative stochastic processes lead to a potentially comprehensive
theory for nonequilibrium phenomena. In this paper an" H theorem" is proved from results obtained using
multiplicative stochastic processes.
INTRODUCTION where S2n is the symmetric group of order (2n)! The
properties given by (3)-(6) are those appropriate for
In another paper,1 the theory of multiplicative stoch-
a purely random, Gaussian, stochastic matrix.
astic processes was explained, and it was shown how
such a mathematical theory leads to a formalism for A denSity matrix representation for the Schrodinger
nonequilibrium thermodynamics: In this paper the equation is obtained in terms of the density matrix
thermodynamical" H function" will be introduced, and PaS' which is defined by
a proof of an "H theorem" will be presented.
(7)
RECAPITULATION
The Schrodinger equation for nonrelativistic quantum If LaSa'S' and Lasa'S,(t) are defined by
mechanics may be written in matrix form as
LaSa'B == 0aa,MS8' - 15 BB ,M:a"
L- aBa'B' (t) =- 15 aa,MBB'(t)
- - 151313 , M-*
(8)
(1) aa, (t),
where M aa' = M:a , which is the condition of Hermit- then Eq. (2) may be used to directly verify
icity, and~a C:(t)Ca(t) = 1, which is the condition of .d ~" -
conservation of total probability. The Hermiticity of t dtPaB(t) = L.J L.I [LaBa'B' + LaSa'B,(t)]Pa,s,(t). (9)
a' B'
M aa , in (1) is necessary and sufficient for the con-
servation of total probability. Suppose that a fluctua- This is the density matrix equation. By averaging
ting contribution to the Hamiltonian is considered. over the stochastic contribution by means of pro-
Then (1) becomes perties (3)-(6), an equation for the averaged density
matrix, (PaB(t), may be obtained, although only after
.d ~- significant computation l :
Ul
= ~Maa,Ca,(t)
t-=tCa(t)
a'
+ L.J
at
Maa,(t)Ca,(t), (2)
where Maa,(t) = M:'a(t), and the follElwing properties ddt (PaS(t) = - i~ ~LaBa'B'(Pa'S,(t)
a! B'
hold for the averaged moments of Maa,(t)l:
- ~ ~RaBaIB/(Pa'B,(t). (10)
a' B'
(Maa , (t) = 0, (3)
The matrix RaSa'B' which appears in (10) is defined
byl
(Maa,(t)MSs,(s) = 2Qaa'SS,o(t - s), (4)
(M""(tl)···M,, " (t2n-l)=0 forn=I,2, ... ,
R aBa •B ·= 0aa·6QBeeB'
e
+ °BB·6 Qeaa'e
e
.-1 1 '-2n-1 2n-1
( 5) - QBB'a'a - Qa'aSB" (11)
(M".-1 " 1(tl)·· ·M"'-2 n v2 n (t2n) It is also provable that for arbitrary complex mat-
_ 1 ~
2nn! pE s2n
IT (MI'p(2J-I)
j=l
. " .
p(2j"1)
ricesX all ,
( 12)
x (tp (2j-l))Ml'p(2j)Vp (2j)(tp(2 j )))
pE~n tl
(13)
- 2:nl 2QI'P(2j-l)"P(2j-I)l'p(2j)"P(2j)
Conditions (12) and (13) lead to irreversible behavior
x o(tp(2j-ll - tp(2j))' (6) in (10) with the equilibrium state being proportional
J. Math. Phys., Vol. 13, No. 11, November 1972
Downloaded 06 Apr 2011 to 129.2.41.235. Redistribution subject to AIP license or copyright; see http://jmp.aip.org/about/rights_and_permissions
