ACCADEMIA NAZIONALE DEI LINCEI

ANNO CCCLXXIII - 1976

CONTRIBUTI DEL
CENTRO LINCEO INTERDISCIPLINARE
DI SCIENZE MATEMATICHE E LORO APPLICAZIONI
N.22
i\NI)REW I-L.A.NSO:0J - ]'ULLIO R E G G I ~ -- CLi\UDIO TEI]'ELBOIlVI
CONSTR.Lt\INED
HAMILTONIAN SYSTEMS
CICLO DI LEZIONI TENUTE DAL 29 APRILE AL 7 1vIAGGIO 1974
ROMA
ACCADEMIA NAZIONALE DEI LINCEI
1976
----- - - - . - - - - - - - - - ~ - - - _ . _ - - - , . _ -
(, S
( ~ s-
}{O:\IA, 19i(1
Dott. G. Hardi, Tipografo dell' Accademia :\azionale dei Lincei
OF CONTENrrS
CHAPTER 2. RELATIVISTIC POIXT PARTICLE.
.J.Vo Gauge Constraint. .
B. Gauge Constraint . .
C. Quantuln J11"ecllanics
CHAPTER 3. RELATIVISTIC PARTICLE
A. Revie'zv of Lagrangian to Top.
B. Constraints on Top Lagrangian
C. Dirac TreatlJzent of Top COllstraints
I). Quantum A1echanics.
..c-\.. Forlnal Introduction .
B. !-lanzilton Variational Princzple 'zvitll Constraints
C. Extension to Infinite Degrees of j;reedo)Jz
D. Other Poisson Bracket Surfaces. . . . . .
E. IJynalnics on Curved Surfaces. . . . . . .
F. Quantlun 7lzeory and Canonical l//ariables
CHAPTER 4. MODEL. .
A. System 'Zvithout Gauge Constraints
B. Ortllonorlnal Gauge Constraints
C. Dirac Brackets . . . . . . . .
I). Fourier COlnponents of T/'ariables
E. Quantlan
Page
5
7
8
13
16
18
20
24
27
2i
29
3
1
33
33
3
8
4
47
51
53
57
59
68
7
1
73
74
78
80
82
86
89
89
9
94
97
9
8
lor
102
lS
17
19
Inzproved Hanliltoniall.
1) IRAC'S FOR SYSTEMS
FORE\VORD
CHAPTER 7. EIKSTEI:r\'S THEORY OF GRAVITATION.
A. General Form of tile Hanziltonian
B. The Lagrangian .
C. The Halniltonian . . . . . . . .
D. Flat Space, .C;ur.face Integrals,
Poincare Illvariance at .'::Jpacelike Infillity . .
CHAPTER 5. :YIAXWELL ELECTRO:\L\GXETIC FIELD
A... /?lectnnn'lgnttic [{anziltonian 7oitlzout (7auge Constraillts .
B. Radiation Gauge Constraints
C. xial Gauge . . . .
D. iVull-Plane Brackets.
E. iVull-Plane Radiation Gauge . . . . . . .
CHAPTER 6. YANG-MILLS GAUGE FIELD
A.. Lie Groups . . . . . . . .
B. Systenz lvithout Gauge Constraints
C. Radiation Gauge Constraints . . .
I). ..4lternate Radiation (;auge Techniques.
E. ..4xial Gauge . . . . . . . . . . . .
CHAPTER 1.
E.
-4-
F'ixation of tile Sjacetinlc Coordinates (C'auge)
I. ()pen Spaces:
(a) i\.Dl\I's "1'----T'" Gauge .
(h) Dirac's ":\laxilnal Slicing"
1. Closed Spaces: York's Gauge
. .
Page 115
118
121
1
o-
-)
..:'\PPE\"DIX .:\.
Conventions.
. .
128
A.PPE\"\)IX B.
1. Extrinsic Cur\"aturc (llHl the Etnhedcling Equations of uss and Codazzi
12
9
2. Prouf of Eq. (i. IS) . . . . . . . . . . . . . . . . . 13
3. :VIOll1entull1 and A.ngular :YIolllcntuln of t he Field. 13
4. Relation of Eq. (-1-.16) to Eq. (1. 83) . . . . . . . . . . . . . . 13
2
I<.. EFERE :\CES
..............................
13+
F() R \V() RIJ
]'his \v'ork is an outgro\vth of a serIes of lectures gi ven by one of us Cf.
1\..) under the auspices of the i\.ccaden1ia Nazionale dei Lincei in 1\.0111e in the
spring of 1974. It is intended to help fill the need for a unified treat111cnt of
Dirac's approach to the canonical Han1iltonian forn1ldation of singular I..a-
grangian systen1s. \\7c have attelnpted as far as possible to refer to the original
literature on the subjcct, but there ha ve undoubtedly been S0111C inad \'crtant
0111issions, for \vhich \ve apologize.
\\ie \vish especially to thank Peter Goddard and Giorgio l)onzano for their
essential participation in the for111ulation of the " string lTIodcl " gi \"en here,
and Karel Kuchar for per111itting us to use parts of his unpublished lecture
notes at Princeton in Chapter 7.
C. T'. is grateful to J.... \\lhceler for much encouragcn1ent, and to the
National Science foundation for support under grant GP 30799X to Prin-
ceton lJni vcrsity \vhile i\.. ]. 1-1. thanks the Institute for \'anced Study,
the )J ational Science and the U. S. i\t0111ic Energy COn1111ission
for their support of various phases of this project.
\\7e are indebted to .i\caclen1ic Press, Inc, the publishers of ... of
Physics C:\". \T.), for perlnission to usc various sections of I-1anson and l\.eggc
(1974) and Regge and Teitelboin1 (1974) in this ,,"ork. T\\'o of us (...-\. J. H.
and 1'. R.) arc grateful to the ... X azionalc dei L.incci for the congenial
hospitality enjoyed \vhile this \,'ork ,,'as being prepared.
.. J. I-I A:\SO:\
]'. REGCE
C.
I. FOR
S\TSTElVIS
Constrained canonical systen1s occur \vith ren1arkahlc frequency in phy-
sics. :\Iax\vell's theory of electrol11agnctisl11, theory of gTa vitation,
and nun1crous Inanifestl y Lorentz in\ariant 111echanical S)1'stCIT1S possess
constraints \vhich invalidate the strictly canonical classical systen1s. It is
clear that a correct Ha111iltonian for111ulation of a constrained classical systcln
is interesting in its o-\\'n right, as \\'ell as being quite useful in de\cloping a
valid canonical quantization procedure for the syste111. Our purpose here is
to introduce the reader to the systclnatic trcatll1cnt of constrained I-Ia1l1iltollian
syst elns de vel0ped i11 it ially by I) irac (I 9 50 ) and to ind icatcits rcIat i() n to
quantum 111echanics \vhere kno\vn.
T'his chapter \vill deal \vith the for111al aspects of constrained 1-1 all1ilto-
nian systelns. rrhc ren1aining chapters are devoted to specific applications of
the 11lethods. In particular, \VC \vill exan1ine relativistic spinless particles,
relati vistic spinning particles, the relati \,istic string, vector fields \\,ith
and non-...\helian gauge groups, and gravitation.
rrhe reader \vho v\ishes to acquire a gcncr?l feeling for the
of these 11l,ethods \vithout getting bogged do\\'n in details is adyiscd to skinl
Chapter I, and then carefully study a falTIiliar systenl. (e.g. ("hapter 2, 4 or 5),
referring back to Chapter I \vhen necessary.
\\Te define a singular Lagrangian L (ql 'Yl") as one for \vhich thc \'cloci-
tcs (jz" cannot be expressed uniquely in tern1S of the canonical 11101nenta
pI" = I.:L/2Y
1
o due to the existence of constraints a1110ng the canonical coordinates
and 1110n1enta fol1o\ving fro1l1 the for111 of the Lagrangian alone. rrhe pl-Ohlerll
of developing a consistent classical H a111iltonian d ynalnics corresponding to
a singular Lagrangian systen1 \,'as apparently attacked first by Dirac (1950).
Subsequently Dirac (195 I), ...r\.nderson and Bergn1ann (1951), Bergll1ann and
Goldberg (1954), Dirac (1958a) and })e\'/itt (1959) refined Dirac's original
n1ethods. ...r\.n expanded treatlnent of the general constrained Hamiltonian
systelTI appears in Dirac's lectures on quantun1 mechanics (1964); sec also
Dirac (1969). IZundt (1966) and Shann1ugadhasan (1963, 1973) rcvic\v
some fine points.
Finally, \ve ll1ention several other approaches to the quantization of sin-
gular Lagrangian systelTIs. Schvvinger (1951 (l, I95 1b, 1953) and l)eierls (195 2 )
utilize variational techniques; SY1l1anzik (1971) gi ves an extended treatlnent
(see also the Appendix of rron1boulis (1973)). DeWitt (1967 b), l,"addce\' and
Popov (1967 a, 1967 b) and Faddeev (1969) ll1ake use of Feynlnan path
integrals to undcrstand singular systen1s. No attempt vvill be 111ade to treat
these methods here.
-8-
'1'he forn1ulations of both classical and quantuln Inechanics ha ve under-
gone profound changes in recent tilDes through the usc of modern mathernatical
language and ad vanced techniques of functional analysis. \\"'"e think that
ultin1atcly these concepts should be introduced into our treatlTIent to gi ve a
less heuristic vic\v of the subject than \,'"C have succeeded in developing so far.
V\T
C
have not pursued this lTIattcr in vic\\'" of the practical character of the present
notes. Their usefulness, \,'"e feel, is in providing a set of direct guidelines to
setting up a consistent canonical forlnalis111 for an an1azing variety of physically
significant systelTIS \vhile a voiding lTIany COn11l10n pitfalls.
...\. L I XTRODLCTIO\,
()ur forn1aI discussion of constrained systcnls hegins \\'jth the consider-
ation of an action functional
(1.1)
..b
c= JciT L(qi' q,)
\vhere qi('r) is a canonical coordinate and qi = dq//d-: is a crlnonical \-clocity.
We confine oursel yes to I-Jagrangians ",'"ithollt explicit 7-dependcncc. I)cfining
the canonical n10n1enta as
(1.2)
pi
dL
--
dlo:.
,
Jl
\ve find the equations
(13)
ctpi 8L
c1--r Jqi
0
hy reqUIrIng the variation of the action S to be stationary.
If \,'"e choose for our l)oisson brackets the con\'"entl011
\ve ha ve
813
ciqi 3pi
2:\. 2B
cqi
(1. 5)
__
.I
\\' here 0; is the Kronecker delta. Hereafter, repeated indices \vill be sun1111ed
over unless other\yise stated. The canonical I-IalTIiltonian
(1.6)
then forn1ally generates the I-Iamilton equations of motion
-- f !-]"} -
q - tq, - djJ1:
.' {' H } dH
c
pt == pt , c == -
9
NO\V \ve suppose that I... (qi , qi) is singular, so that there is no unique
solution qz" (q , p) expressing the velocities in terms of the canonical coordinates
and 1110111enta. ..-\ necessarv and sufficient condition that I---t be singular is
'l'his is a sIgn that there exist certain prz'llzalJ! constraints
(1.8) 11l ==-= I " . . , 1\1
follo\:ving fron1 the forn1 of the I ... agrangian alone. "fhe synlhol 0 is read
d \veakly zero" and 111eans that 9m 111ay have non\'anishing canonical Poisson
brackets (1.4) \vith S0111C canonical variahles.
l'he canonical Han1iltonian (1.6) is no\v not unique. \\7C 111ay in fact
replace it by the effecti ve I-I anliltonian
(1.9)
H I-I -L 1! (r'j f
q
lIe.
-- c im I JJl \ 'L .
II generates lle\\' equations of lTIotion replacing (1.7),
(1. 10)
(
"'I
ql' {q/ I-I
l
l',?m
,
j

'dpz"
11
m
lY;i
rrhcse are the 1110st general equations of n10tion consistent \vith variations
'dqi , 'Dpz' \\hich preser\e the constraints 9m o.
In order to ha \e a consistent system, \\"e require the '7 deri \ati ves of the
constraints (1.8) to be zero, or to be linear cOITIhinations of the constraints so
they are \veakly zero:
Yu :--== { 9n , H} {91/ , 1-I c }
I f (1.1 1) is not alread y true as a consequence of the original prin1ary
constraints (1.8), t\\TO possibilities occur. First, \\e n1ay find that Eq. (1. J J
gives no ne\\T inforn1ation, but sin1ply ilTIpOSeS conditions on the forn1 of 11
m
,
Second, Eq. (1.1 1) Inay in1ply a ne\v relation an10ng the p's and q's, independent
of 11
m
, These are seco/lc!ary constraints and n1ust be adjoined to the original
constraints (1.8). NO\V \ve repeat the process, requiring the '7 deri \Tati ves of
the secondary constraints to vanish, and so on, until all independent constraints
and conditions on U
m
have been found. If K additional constraints result,
\ve add them to the IVI prilnary constraints and sUInmarize the complete set as
(1.12) CPa(q ,p) 0 , a =::: 1 " . " K +M == T.
Finally, consistency of all the constraints \vith the equations of n10tion requires
that there exist a solution for U
m
as a function of q and p
-10-
so that Hitself is expressible in tern1S of q and p;
H== H(q ,p) .
I)irac, \VC novv define a function (q , p) as a first (lass quan-
tity if
(1. I 3)
{ , gJa } 0 ,
tl==I
T' ,
1{ (q , P) IS defined as secon(! class if
{ R , gJa 1 0
for at least one d. (Sccond class quantities are obviously anlbiguous up to
linear conl,hinations of first class quantities. :\ote that a second class constraint
squared is first class). of our constraints I. 12) can no\v be di\'ided into
t\VO sets, one consisting of all the linearly independent ,/irst class constrain!.",
(1.15 a)
Yi (q , p) 0 ,
1== 1"",1.
and the ocher of the rcnlaining N" == 'r --- I seeoJhi (lass (:o/lstrllints
I. 1 5 b)
(q , p) 0 ,
Xotc that hoth and epx 111ay include secondary constraints as \yell as
pri1nary constraints.
Dirac (19
6
4) has proven that the second class constraints \\'ill gi\'c 1
4
1SC
to a nonsingular :\ X X lnatrix of Poisson brackets \\'hieh \\'C \\'rite
(I. 16)
(\\?hcn COlllpllting , it is clear that onc 11111st not usc the constraint cquatiollS
until after calculating the l)oisson bracket.) Sincc the dctcrnlinant of an
antisynlnlctric nlatrix vanishes if the dinlension is odd, \\'C conclude that the
nunlbcr N of second class constraints ll1USt be e'veu. Since is nonsing'ular,
its in \'erse exists and satisfies
\Ve novv proceed to construct fron1 any dynatnical variable ...-\. a lle\V variable
...-\.' \vhich has vanishing brackets 'lvitlz all secoJt(l class constraints. \\T
e
define
(I. 18)
and observe indeed that
A
, -;\ {A 1 C--
1
- ...J.. - , f q;",
(I. 19)
-{' m
l
---
f
' mt.-o
- ..I.J.., rY J l ... tl.. , rY J - .
Note that {A', IS not necessarily \veakly zero.
- 11-
No\v \VC sin1ply postulate that the Poisson bracket of t\VO quantities i\ and
13 111USt be replaced by the Poisson bracket of their priJnc(1 variables,
1.20) {A, B'}.
Xote that although .i\ , B' B, the Poisson bracket {A', 13'} IS not
\veakly equal to {A, B}. If \ve define the Dirac bracket as
(r.2r) { A B }
* - {,\ B} --- r " I. C -1 f r B 1
, - 1""1., l .Il.. , CPa J l J ,
then \ve easily see that
(1.22) {
'\ 13 (* (.\' B' I :--..-' {" \' 13 I ;::"'..: r '\ 1) 'l
-Ll.., . f 'l ..i.1.. , ...l.., f '-=- l"'l.., ) f
If all l)oisson brackets are no\v replaced hy IJirac brackets, Eq. (1.22)
tells us that \ve ha\"e e[fecti \'ely chosen to deal only \\Tith first class constraints.
\Ve can set all second class constraints strollg(v to s('ro beca.use the l)irac 1)rackct
of anything \'lith a second class constraint \"anishcs:
F'r0111 Eq. (1.22) and the definition (r. 18) of the prin1ed variahle, \\"C inllllcdia-
tely sec that { B , C }* }* {i\', { B', C' }}, so the Jacobi identity
(1.23) , {B , C }* }* { B , {C , }* }* {C , { ...-\ ,13 }* }* 0
is satisfied \\Tcakly by the I)irac bracket. using the definition (1.21)
one can sho\v directly that (r.23) is actually a strong equation.
\Ve note here also the iterative property of the Dirac bracket. I f the
nUlnhcr of constraints is large, \\"e 1l1ay a \"oid invcrting large 111atriccs by
taking a sn1aller second class subset of the constraints and C0111puting the
prclin1inary bracket (1.21). 1'h('n S01l1e second class suhset of the rcnlaining
constraints is used in (1.21) \\ith all brackets on the right hand side re-
placed by the prelin1inary bracket. Repeating this procedure until all con-
straints are strongly valid gi ves the same results as C0111puting the final IJirac
1)rackct in a single step.
N O\V \\Te are equipped to understand n10rc clearly the nature of the effccti ve
Han1iltonian A in Eg. (r .9). If \ve set
(1. 24) H== fI' == He - {He, cPr.(} c;rl qJ(),
then
(q , p) == - {He , } c;rl
and H(q ,P) is the physical first class replacc111ent for I-I
l
, \vhich coulcl have
1)cen second class.
\Vith the choice I.2-l-) for A, it is clear that the Hatniltonian is still not
cOlnpletely determlned: the equations of motion of the constraints are unalte-
red if \ve add to H an y linear con1binatiop of the I first class constraints
12 -
IS Eq. I . J 5 a). \Vc therefore take as our total !-fani iltonian
(1. 2 5)
'l}' dJ, (q p)
l Tl \' ,
\\here Ii has vanishing hrackcts\\yith all constraints c\yen though it contains
I arbitrary functions '(I", Since the do not necessarily havc vanishing brackets
\\yith canonical variahles \vhich are not constraints, \ve ha vc nc\y cquations
of 111otion
1.26)
Ii i === {q , H } .. { qi , I-I' } + 'lj' { q , Yj }
pi == {pi, 1-1 } {pi, H' } -1- 'lj' { pi , Yj }
\vhich explicitly involve the 'l}," S0111e restrictions on the functional forIn of
the 'l',' and their tin1e derivatives in tcr111S of the rj,. and Pi 111ay fol!o\y fron1
1.26) (see, for cxanlp1e, (5.22)).
l' hero1e 0 f the Yi in Eq. I . 2 5) is to gcnerate infi nitesin1aI contact t ran-
sfornlations of the p's and q's that do not affect thc physical state of the SYStC1l1
',l)irac, 1964; 13erg111ann and Goldberg, 1935). Hereafter \\ye \\'ill refer to
such transfornlations as gauge transforn1ations.
]'he arhitrary functions 'l'i in H occur hecause the original L,agrangian
possessed I /{tllf<g-l' (Iegrccs oj.free{lo711 associated \vith the first class con:;traints
\Ve 111ay fix the \'alues of the 'l'i C:) by choosing explicit fOr111S for each
gauge
Ii (q , P , 7) 0 , I == J " "
and inlposing thC111 as constraints llot follo\ving fr0111 the Lagrangian.
The choice of gauges (1. 2 7) should 111ade in such a \va y that the con-
straints \vill cease to be first class: the Inatrix { , Ii} should be \vell-defined
and nonsingular. Then \vhen \ve replace all brackets in the theory by those
consistent \vith y" === 0 , Ij === 0, the arbitrariness due to the 'l'i in Eq. (r .26)
\vill disappear. \,rc note that S0111e traditional gauge conditions in\'01 \'e velo-
cities \vhich apparently cannot be recxpl-essed in tcrll1S of canonical coorcli-
dinates and InOll1cnta; in such cases, \ve do not kno\\' ho\\r to conlputc the
n1atrix , 'Yi} and it seems that the L)irac 111ethod cannot be used.
In the end, \\'e obvious] y \vant to express the systeln in terll1S of the truly
independent canonical variables alone. It Inay happen that the obvious gauge
choices do not cOlnpletely reduce the phase space a vailable for particle Illa-
tion dO\\'n to the size implied by the Euler equations. constraints
necessary to define cOlupletely the physical system ll1ay occur disguised in
the form of invariant relations. \\7e define (q, p) to be invariant relations if
(r .28 a)
and
at 'T = 0
(r .28 b)
- 13-
(In Eq. (1.28 h), sign 111eans that all other constraints besides the
been set to zero). \\'e thus conclude fro111 Eq. (1.28 b) that vvill re111ain 0
for all if (1.28 a) holds.
Invariant relations differ from constants of the motion in an essential
\\'ay. Constants of the motion are generally used to specify particular
solutions of the equations of n10tion, \vhile in variant relations are instead
conditions \vhich lllust be satisfied in order for a solution to be considered
a physical one.
In \Tariant relations can be as ordinary secondary constraints
by using a Lagrange n1ultiplier )'i corresponding to each invariant relation
O. \\Tc sin1ply express the 1110nlenta in tern1S of the velocities, so that
;/ (q ,P) can he \vrittcn as a nc\v function;i (q and take
as our nc\v Lagrangian. The Euler equations arc
d ClL 8L
+
[\
(ti

)
-f-
l'5q}

-
8q)
l
d),z" y-
c,i""' .
dl: __ ]
+


+
)'i

- 0
cit
"-:z
"".
dt
:l'
all} ell]j
d 8L 2L

I

cit 8),i dAi
- 0
2
These gi ve back the original equations of n10tion, except that no\v the constraints
;i 0 occur as secondary constraints follo\ving from the primary constraints
P;. o. l'he net effect is to make the into ordinary constraints
\\'hile the )'i disappear froIn the dynamics.
I-tet us no\v suppose we wish to in1pose all a vailable gauges and invariant
relations, elinlinate all arbitrary functions froln the equations of motion and
describe the systeln only in terrns of the truly independent phase space varia-
bles. If \ve have properly chosen the gauge constraints and invariant relations,
all constraints vvill now be second class.
Redefining epC( to include all of the constraints, gauges, and invariant
relations, vve should find that the matrix C-x(3 == { epa , ep(3} is nonsingular. Then
Eq. (1.2 I) gi ves the form of the Dirac bracket consistent vvith setting all con-
straints strongly zero and using only the proper variables to describe the Ha-
n1iltonian dynaInics of the systen1. For sufficiently simple systems, this final
Dirac bracket provides the starting point for canonical quantization of the
syStCll1. This point is discussed further in Section I. F.
B. I-IAMILTON VARIATIONAL PRINCIPLE \VITI--I
\Ve no\v develop the ideas of the previous section frorn a slig-htly different
vie\vpoint.
- 14-
We begin by considering the Hamilton variational principle In phase
space, \vith the pfsand qf s considered as independent variables
o = ~ S = ~ f ( ~ 1" dq, - H d-r) =
= ~ f ( ~ ~ CPidq,-qidP')-Hd-r).
\Ve then find I-Ianlilton's equations,
8H
dqi
No\:v consider the entire set of constraints, gauges and invariant relations
\vhich restrict the phase space a vailable for the particle 111otion. rrhe 2 n
canonical variables pi , q/ can be reexpressed in ter111S of the 2,12 independent
variables
Z," , z" == I ,. . ., 2 'IJl
and the 21Z-27Jl contraints
i == 27/'l 1 , ... , 2/l .
l'hus \ve n1a v express pl" and qi as functions of the z's and of 7.,
(1.3 1)
Note that while pi and qi are by definition not explicit functions of 7, the 2
2
"
111ay be expIicitIy 7 - dependent ; the expIi cit ,:' sin I ~ q. (1 ..3 I) are necessar y to
compensate for any 7-dependence of the Zi.
)Jo\:v we consider the z/sand '7 = Zo as a set of independent variables,
neglecting for the 1110nlent the vanishing of the constraints. rrhen \ye find
(1.32)
where
(133)
n 211
~ 2: (pi dqi - qi dpi) == 2: C
a
dz
cx
i=l a=O
\Ve see ilnmediatcly that
(134)
is just the Lagrange bracket of the ne\v set of variahles Za, including constraints,
\vith respect to the old sct of canonical variables qi and pi. If \ve no\v define
-15-
the Poisson bracket as
(1.35)
we find the follo\ving properties:
2n
2: {Zk , Zi } (Zk , Zj) ==
k=l
(137)
Equation (1.37) fo11o\vs fron1 the fact that
(1.3 8)
The equations of nlotion for Zi are no\v
(139)
\vhere (qi , pi ,'1") are treated as the independent variables \vhen conlputing
the right-hand side of the equation.
l'he action principle now can be written
NC\V, let us require that the constraints tJi 0 hold throughout the
variation, so
== OZl. == 0 ,
'The restricted action principle is thus
k == 2 JJZ - t- 1 ,. . " 2 II .
o = as = 0 j C
i
dZ
i
+ (Co - If) Ch) .
The variables Zi, i == 1 , ... , 27Jt are independent variables \vhose Lagrange
brackets are given by Eq. (1.34).
Next \ve sho\v that the Poisson brackets of the independent Z/ are just
the Dirac brackets. If \ve define the Inatrix
a , b == 27Jl 1 , ... , 2lZ ,
\ve find that the Dirac brackets are
(I .43)
21Z
{
t;:I Z t * - {z ry t -- { t;:I t;:I t C-b
I
{ <'":' t;:I t
"'"' i, j J - i , 4Jj J .v i , "-' tl j '(. "-' b , "'"'j j
a,b=:!.nz-j- 1
-16-
No\v \\"C ll1ultiply by the L.agrange bracket, so

(Zi, Zk) {Zi , Zj }* == (z , Zk) {Zi , Zj}* ==
i=1 i=1
2n
a
jk
- 2:

== ;)jl.' ,
\vhcre k, j == 1 " .. , 2,ll and the sun1 can be extended fron1 2JIl to 2n
hy :Eq. (1.43), {Zi' j}* 0 \vhen i == 2JIl I, .. " 2n. I'hus the I)irac
brackets are the in\Terse 0 f the restricted Lagrange brackets follo\ving fronl
(r .42), \vith only 2 III variables; by definition, the Dirac brackets lTIUst there-
fore be the Poisson brackets of the restricted systen1. In other \vords, a simple
restriction in the nun1bcr of variables appearing in the Lagrange brackets
causes drastic chang-es in the inverse of the Lagrange bracket nlatrix; the
canonical I)oisson 1)rackets are changed to Dirac brackets, \vhich can be
expressed in terms of the original canonical Poisson brackets only by using
Eq. (1.43).
It is no\v tri vial to prove the iterati ve propert y of the Dirac brackets
mentioned earlier. Indeed, successi ve restrictions on the range of variables
of the Lagrange brackets gi ve the san1e final restricted Lagrange brackets,
and hence the san1e inverse.
C. EXTENSION TO INFINITE DEGREES OF
\Ve no\v establish our conventions for dealing \vith classical field theories
(see, for exalnple, Goldstein, 1950; Kundt, 1966). The discrete label i on qi (t)
no\v becomes a continulun label x plus additional discrete labels i\, so qi
can be replaced by the field cDA (t , x),
(I.4S) L (qi , qi) --+ 2" ( <I>A(t , x) ;
T'he Lorentz-invariant action functional is then the integral over the
L.agrangian density:
S [<I>A(x)] = rd
4
x2 (<I>A (x) , 21-' <I>A(x .
.i\t this point, \\"C lnust decide on a n1etric convention for treating co va-
riant and contravariant vectors in four-dilnensional space-tin1e. \\7.e choose
(T .46)
-1 0 0 0
o 100
o 0 I 0
000 I
x[J. === (t , x) X!J.. == g[J.'J XV == (- t , x)
(r .48)
- 17-
i\l1 of our conventions are listed in i\ppendix i\., \vhere thev are cornpared
to other conventions in comn10n usage.
I'he canonical I110n1enta arc no\v
and the Euler equations are \vritten
"'\ 32
A)
If the Lagrangian is translation-invariant, \ve find that the canonical
energy-lTIOmentU111 tensor
_
V
c
--
is conserved,
(r .50)
If \ve take the canonical Poisson brackets to be given symbolically by
(1.5 1)
\vhere 0 (x - )1) is the Dirac delta function, then the 11: amiltonian IS
(r .52) H
- /.. d3 eGO - (. d3 (A( ) d<f)A _ (fJ
c - X c - X 1t t ,x 3t e>o-.. )
Using the Poisson brackets (1.5 I) ,\ve see that He generates the titHe c\"olutinn
of the canonical variahIes through functional deri vati \"('s
dnA (t ,x) _ { A H") _ . BH/
--d-I---- - it , c J - - (t
NO\V suppose the s ysten1 has constraints
a = I,"', I',
following fro111 the fOr111 of the Lagrangian and fron1 the equations of motion
of the primary constraints. I'hen the arguments given earlier lead us to con-
clude that the energy-1110tTIenturn tensor is an1biguous. Defining O;w as the
first class energy-1110mentum tensor, \ve replace the canonical tensor by the
total energy-1110mentum tensor
(r .55)
-
2
- 18-
\vhere the Yi arc the I first-class constraints. ...;.\s before, agreenlent \vith the
equations of motion ntay rcstric1- the (x) some\vhat. F'ixing the gauge
degrees of freedom \vill fix the and elinlinate the arbitrariness due to
the fronl the equations of motion
No\v \ve \vrite the ren1aining N == T' --- I second class constraints as
rt. == I,, N
"[hen \ve can define consistent Dirac brackets by conlputing
(I .36) C:l
f3
(.'i- ,y) == { (t , x) , (t , y)}
and its InyerSC, \vhich obeys
(I .57) rd
3
, z) (z ,y) = Jd
3
zCx"y (x , z) c:;l (z , y) === (x -- y).

Equation (1.22) is then replaced by
(I .58) {A (t , x) , B (t , y) }* == {A(t , x) , B (t , y) } -
rd
3
zd
3
w{A(t, x) ,rp,,(t, z)} ,w) {rpf;(t, w), B (t, y)} .
J
(1.60)
I'he brackets for the systenl \vith all gauges and invariant relations imposed
can of course be conlputcd in the Stune ll1anner.
It should be noted that the canonical generators of other synl111etry trans-
fornlations of the Lagrangian may also require the addition of multiples of
the first class constraints if gauge conditions are not ilTIposed.
F'inally, \ve observe that one Inust generally supplement the definitions
of the continuous constraints (1.54) viith appropriate boundary conditions,
lest C
x
;3 (x ,_v) become singular. Our treatment of such nlatters here \vill be
purely praglnatic,\vith no attenlpt to rigorously define the nature of the func-
tional spaces invol 'led.
D. OTHER POISSON BRACKET SURFACES
Instead of con1puting the Poisson brackets (1.51) at equal times, \ve may
HI fact choose a variety of surfaces (Dirac, 1949 a; I(ogut and Soper, 1970;
F'ubini, I-Ianson and ]acki\v, 1973). Let
specify a suitable surface. Then the Lagrangian should Le re\vritten in terlns
of d).\ === 2<P
A
(X)/27 and the action functional should be expressed as
S = ( d" d
3
cr!' (cD,\ , <DA , 2<D
A
/2o) = rch L'
__' oJ
-19-
\vhere the (i'S are a suitable set of three variables \vhich span four-space \vhen
con1bined \vith 7. rrhc canonical mon1enta are
(I .6 I)
l'he Poisson brackets thus take the forIn
(1.62)
In order to understand the nature of the Hamiltonian for a system using
a general Poisson-bracket surface (x) =-= '"'=', we must generalizc the treatn1ent
of thc generators of spacetilne transformations. Let (x) generate a spacetin1c
transformation labeled by the index a,
(1.63)
For the present discussion, let us a void the complications of "t'-dependent
Hamiltonians by assulning that (x) is a sYlnmetry transformation of the
Lagrangian, so that \ve ha vc a canonical oether current (x) \vhich is con-
served, 2!J. (x) =-= o. 1'hen the canonical generator of the transforn1ation
can be to l)e
where
(1.65)
dQa
d=r==o.
\Ve remark as usual that before we impose gauge constraints, is ambiguous
and can be replaced by
(1.66)
when generating the equations of motion.
The Hamiltonians or (iynanzical generators of our system consist of those
space-time syn1metry generators \vhich change the Poisson bracket surface
F (x). The other, kine1natical generators are those which leave the surface
F (x) unaltered. For exalTIple, in the conventional equal-time forn1ulation
we have
surface: F (x) =-= .-:r
0
dynamical generators: po , 1\1 (boosts)
kinelnatical generators: pi, lVI(j (rotations).
The Hamiltonian is takcn to be H === po because po generates the transforrna-
tion from one surface x
O
== "r to another.
- 20-
In the null-plane forn1ulation of the dynan1ics (Kogut and Soper, 1970; see
also ...'-\ppenclix A), \VC have
surface: F (x) == :r - ==
2
cl ynan1ical generators: p- ==
kinen1atical generators: P+-
3 0 -Z" I 3 o
2 (1' - l' ) ,M = yz (M' I - M ')
2
M+
i
= /2 (M
3i
+ MO
i
) , M'j',
\vhere i and j take on only the values 1 or 2. The generator M+- == MO:
3
is technically dynamical because it generates a scale change of x"; hovvever,
it is son1ctin1es gi ven special treatlnent since it leayes x === 0 unchanged.
E. DVNAl\fICS ON CURVED SURFACES
Dirac's technique is particularly suited for studying the Hamiltonian
forn1 of a theory in \vhich states are defined on a general spacelike surface
and not just on the special surfaces considered in the previous section (IJirac,
1951, 196-.+). l'hc basic idea is to introduce a systenl of curvilinear spacctinlc
coordinates uP- :=.: (10 , u
1
, Zi
2
, u
3
) into the theory in such a \vay that the equation
U
O
=:-:: constant defines a generic spacelike surface. One then forn1ulates all
the d ynanlics in ternlS of the up instead of the original coordinates x
p
-. Note
that one is also introducing arbitrary coordinates 11 , u
2
, u
3
on the surface.
l'his last step becomes C01l1pulsory if one \vorks (as in Chapter 7) in a general
Rienlannian nlanifold \\?here no natural choice for the spatial coordinates
(such as uf == xz" in flat spacetime) exists.
The procedure (" pararnetrization") used for incorporating arbitrary
spacetin1e coordinates into a field theory parallels very closely the forn1tdation
of the d ynan1ics of a particle \vith respect to an arbitrary tin1e scale. Consider
for sinlplicity the case of a non-relati vistic free particle for \vhich the action is
(1.67)
t I ( dX)2 !'
S [x (t) ] = Jdt 2- m di = JL
,
dt .
1'0 express the problen1 In an arbitrary tinle scale U, one re,vrites (1.67) In
the forn1
(I .68)
, I' I ( dx )
2
( cl t ) -1 Jr.
S [X (u) t (1{)] == du -- JIl ------- ---. == L" du
\ ," ..J :2 du du u,
thereby introducing the original tin1e coordinate t as a nc'Zu (lyJlaflzical l'ar/able
on the san1C footing \vith the position x.
----
21 -
The action in the fornl (r .68) is no\v invariant under reparanletrizations
u --->- feu). .i\s a consequence, the I-Iamiltonian
(r .69)
H _. elt
It - PI du
vanishes identically. (T'hc I110tnenta in (I.69) are defined hy PI == ?L'l)
2(dt/du), etc.... ). ()ne also gets fro111 (1.68) the primary first class con-
straint
(I .70) K [I , x ,Pxl ~ 0
(r .72)
\vhere In this si111plc case the quantity K is just
I( == + P.;/2 11Z
rrhe constraint (1.70) 111USt be added---I11ldtiplied by an arbitrary function---
to the (vanishing) canonical Harniltonian (1.09) to get the total I-Ia111iltonian
\vhich then reads just
H == N
and vanishes \veakly due to (1.70). l'hc arbitrary function N describes then
the rate of change of the physical tin1e 1 \vith respect to the arbitrary para-
111cteru.
1"he steps taken to paran1ctrize a field thcory follcnv the pattern rc\ic\\Ted
for the particle case. One thus introduces the four l\Iinko\vskian coordinates
y1.L as nC\\T fields on the SaI11e footing \\Tith the original fields <VA of the theory
at hand by re\\Triting the action
S [<I>A (x")] = Jd
4
x2>, ($,\ ; 2$J,/2x IJ.J
in the forn1
(173)
\vith
S [$A (u") , Xl' (u")] = Icf4 ug;,
.J
rrhe action \vrittcn in the forlll (Y. 73) is invariant under reparan1ctrizatiolls
z/Y. ~ fX ancl the theory has t;1US Lecon1e " generally co\"ariant ". ...-\.s a
consequence of this invariance the Hanliltonian is
(1. 75)
and \';lJ1ishcs identically just as it did in the particle casco (The momenta
appearing 111 (I, 75) arc defined by Tell. = i) [J J.t;, d
3
U 1/ i) (2x
u
j2uO) , . .. etc.).
- 22-
...t\lso, one no\\7' gets not just one primary constraint as in (r. 70) but four
prilnary constraints per space point vvhich are of the forn1 (4<-)
(r .76) 7tfL(U) + KfL(U) [<P
A
, 7t
A
] 0.
A 11lore convenient, but conlpletel y equi valent, fornl of the constraints (I. 76)
is obtained by projecting thetu into one norn1al c01l1ponent
(r .77 a)
(Here nfL is the unit norn1al to the surface, a functional of the x
fL
) and three
tangential ones:
(r .77 b)
r[he ad vantage of the projected version (r .77) of the constraints is twofold.
of all, \ve replace the high] y arhitrary description of the motion in ternlS
of the coordinates u'X by a description in tern1S of deforn1ations of the surface
parallel to itself (governed by and orthogonal to itself (governed by
\vhich has an invariant geoluetrical rneaning. Secondly, the change in the
field variables under a displacen1ent of the surface parallel to itself consists
only of the response of the field to a change of coordinates in the surface and
has no dynamical content, being determined cornpletely" by the transforn1ation
character of the field. \Ve separate in this the part of the problcrn that
is trivial fronl the truly dynalnical part \vhich is contained in l\loreover,
\vhcn the constraint YtJ. is in1posed as a restriction on the Hanlilton-
Jacobi principal functional S in the classical theory or on the state functional
in the quantum theory then, thanks to Eq. (r .83 a) belo\v, the constraints
follo\v as a cOJZsequence of YtJ. 0, as has been sho\\rn hy IVloncrief
and Teitelboinl. (r 972). 1"'his situation is to be contrasted \vith the forn1ulation
based on (1.76) in \vhich the t\t\TO aspects of the problein are mixed and one
has to deal \vith four equations of the same degree of complexity.
1'0 obtain the extended Hanliltonian \ve add no\v the constraints (1.77)
to the original (zero) Harniltonian (1.75). The Hamiltonian now reads
(I .78)
N Ye.)
1
'The arbitrary functions N
1
(" lapse") and N
i
(" shift") describe the \vay
in \vhich the initial surface is deforn1ed into another infinitesin1ally close one
(fig. I. I). SO, if is an arbitrary functional of the canonical variables of
the theory (\t\Thich include the x
P
-) one has
F[a
t
] - F[a] == ou
o
{F, H}.
(*) In the particle case the quantity K appearing in (1.70) is the total energy. For a
field, the K
p
_ appearing in (1.76) are given by appropriate cOlnponcnts of the energy-momen-
tum tensor. We refer the reader to Dirac (I949b) and Kuchar (1974) for an analysis of
this point.
--.---
23 -
1'he bracket in (1.79) is defined as
(1. 80)
,
r 3 ( of
{F G} == (I u ---------
, .J ' o(I)A (u)
-t-
of
oC; 'F' ("" )
-"--,- -- ( <--)]"!
(It) \ )
In particular, \vhen applied to the canonical variables the111scl \cs equation
(1.80) gives
(1.81 a)
(1.81 b)
(1.81 c)
r _A (0 i) B (0 ,'i\} __
l I\, U ,u ,TC \1 ,1t ) - 0 ,
(ih / 0 ,") __ R ( 0 , I '", I H ( t' , 'i)
l 'vA (J ,U ,I" U, It ) J -- 0 A 0 iJ , U
-
and sin1ilarlv for the x!.L and it
p
_. l'he a-function in (1.8 I c) IS dcf1nccl hv
for an arbitrary scalar testing function! and it lllay be considered to transfornl
as a scalar at u and as a density at u
f
( ...\ctually since a has point suppurt
the only' thing that 111attcrs here is that the sun1 of \veights at u and u' he unity-
the allocation of ',veights to both points is other\vise arbitrary).
8 u0 Nr (u) ( u)
Fig. I. I. IJefonnatll1l1 of a coorclinatized surface. Starting fronl a given surface 0'. of con-
stant uO-tiIne, on \\-hich a coordinate systcnl is defined, one goes to an infinitesill1ally
tlisplacecl surface cr' which corresponds to a snLtlI change ouo by lneans of a clefonll;1tion
N (u) = :\' 1 (u) n (u) :-:,r(u) e
r
, \vhere e) is the tangent vector to the r-th coordinate line.
Nate that the defonnation defines cr' not only in the geoJnetrical sense but also sets a coordinate
systell1 on cr' by the prescription of giving the saIne spatial coordinates to the points at the
tail and at the tip of the deforn1ation.
The hrackets 1.81) are equal tin1e brackets in the tin1C U
O
\\'hich 111('ans,
for cxan1ple, that (D
A
and (DB have \anishing hrackets for an arhitrary spacclike
separation.
Equation (I.Ro) can also be used in principle to find the brackets of the
canonical variahles for nonspacelike separations. '1'0 do this, one first solves
the equations of llH)tion and then expresses (D
A
(u'O) and 7t
B
(1'0) as functionals
of the " initial (D
A
Cuo) and it
B
(u
O
).
Let us finally n1ention a fundan1cntal property and defined
in (r. 77), nan1ely'" their Poisson hrackets (P.B.). It turns out that \vhereas
the form of the varies, of course, from theory to theory, there is one important
- 24-
feature COlTInl0n to all such ,Yt'(J.' naITIely the fact that the P. B. of any two of
thcn1 is a linear combination of the themsel ves and this linear cOllzbinatio71
/s t//t sante for all tlzeories. What \ve arc en1phasizing is not the fact that the
P. B. of any t\\'O constraints is a linear conlbination of the constraints (first
class property)-this merely guarantees the preservation of the constraints
during the c\'olution of the SystClTI. \\That is remarkable is that the coefficients
in this linear con1bination (" structure constants ") are universal. As a matter
of fact one can derive the brackets of the fron1 only tvvo assumptions (Tei-
tclhoin1 1973 a, b), nan1ely: (i) The oY(u. are first class (other\vise the theory
is inconsistent to start with) and (ii) IIamilton's equations are integrable,
that is the change in the canonical variables during the evolution fron1 a gi ven
initial surface to a gi yen final surface is independent of the particular sequence
of intcrtl1ecliate surfaces used in the actual evaluation of this change. (A con-
requirelnent tern1cd by ICuchal" "path independence of dynamical
evolution "). The result is (I)irac, 1948, 195 I, 1964; Sch\vinger, r962 b)
(1.83 a)
(r.83 c)
{ ,YC;. (x) , === (x) 0 ,r (x , x') ,
=== 0 ,s (x ,x')
4J (x) (,- v')
Jl
S
\ 0 ,r ,A,
rrhe quantity .Yt'r in (1. 83 a) is defined by ;Ytf === grs \vhere grs is the metric
of the uO === const. surface. that grs depends on the canonical variables
X!1. \'ia c[rs == "'Jet.r3 (2x'l./2u}') !-;u
5
) , ,vhere "lJcx;3 === diag (- r, 1, I, r).
Lastly, it should IJC cll1phasized that the abo\'c mentioned \vay of deriving
the hrackets (1.83) sho\vs directly that those equations apply equally \vell
for any generally co\'ariant field theory defined on a Riell1annian spacetin1e.
\\le observe that (r .83) hold also in theories \vhich are" already para-
meterized " (i.e. generally covariant to start \vith) such as general relati vity,
for \vhich the generators ,Yf'(J. turn out not to be of the general form (r. 77),
as \ve shall see in Chapter 7.
I:;. rrI-IEOR Y .AXD CANONICAL VARIABLES
()ne of the I11ain 1110ti vations for de veloping thc !-1an1iltonian d ynan1ics
of a constrained classical systenl is the desire to deduce the analogous quantun1
t11cchanical systC1l1; there is often a very close connection bet\veen the form
of the I)irac brackets and the quantu111-111echanical conl111utators. I-Io\vever,
in practice, the usual prescription
I.8-1-)
Inay he plagued hy ordering an1biguities on the right -hand side of the expres-
sion for the IJirac hracket. Suppose, as in Section 1.13, \ve let the Zi (r) be the
2 III independent variables in terlTIS of \vhich the Dirac brackets are expressed.
- 25-
Then, according to general theorenls on canonical systetns (Jost, 1964; Kunzlc,
1969), in the neighborhood of any point in phase space, there exists a local
set of 2 1JZ variables
(1. 85) f/(Z,7)
\vhich ohevs canonical brackets
(1 .86) {
_. - }*
pt, qi == - OJ
.:\ glohal treatnlcnt of phase space raises nontri vial questions \vhich can
be solved only in the fralne",rork of algebraic topology (Abraham and Marsden,
1967). Ho\vcver, here \vere interested only in presenting a heuristic discussion
of the forn1al aspects of the theory, and so \vill assun1e for sirnplicity that
pl' and qt' may be treated as global phase space coordinates. (T'he extended
)J \Vigncr coordinates gi ven in Chapter 3 arc an exan1ple of a situation
in \vhich the glohal prohlct11S can be handled \vith kno\vn techniques).
I n general, the 2 lIZ independent z/ s he certain 7-dependent functions
of the canonical coordinates.
In ter111S of pk and qk, the action principle (1.42) can he \vritten
(1.87) o = oS = o. / (-;- CP d qk q'" dP) .._- Hd 7 )
Repeating the entire argu111cnt of Eqs. (1.32-40), \ve have also
(1.88)
\vhere no\v everything is expressed in terrns of the ne\v qk and jk.
Thus
c === c === I ( - k _ - d;k ).
ex C( 2 P Jz qk Jz
k=l ex ex
H === Co -- Co + H ,
where ty, === 0 " . " 2 1JZ and Zo :=: 1'. T'he integrands of }:qs. (1.87) and (1.88)
at 1110St by an exact differential which can be re1110ved by a suitable
canonical transformation on qk and pk.
A specific choice of the variables qk and pk detern1ines Co and hence,
fro111 (I. 89), H. T'his choice fixes the explicit 1'-dependence of the Zi appearing
in the equation of motion (1.39) \vith (q, P , H) (q , p, H).
Conversely, \ve Inay always perform a suitable 1'-clependent canonical
transformation on the variables qk and pk \vhich changes the Hamiltonian
into any desired function. If the Dirac brackets of the z/s do not depend
explicitly on ":", then the best choice for the I-Iamiltonian is clearly the one
\vhich assigns no explicit "r-dependence to the z/ s.
- 26
'The I--Ianliltonian J-I generates the equations of motion
= {qi ' 11 }* =
11.9)
Since all time-dependence of the canonical variables IS itnplicit, one Inight
call this a "Heisenberg picture" forln of the equations of 1110tion.
T'hc identification of q and p is in general necessary to solve the ordering
proble111s inherent in the transition fron1 classical mechanics to quantum 111echa-
nics \"ia the correspondence principle. 1'he final goal, then, \vould be to find
a co\yering of pbase space by a set of neighborhoods, each one possessing a
sct of regular coordinates qi' j obeying .86), together \vith the canonical
transfor111atiot1s relating the coordinates in the intersection of di fferent neigh-
borhoods .
.:\ Inethod \\"hich s0111etin1Cs \\orks in practice is to seck a c0111plctc sct
of independent d yna111ical variahles ... \\"hich have vanishing [rlnonz"cal bra-
ckets ,\'ith all first class constraints follo\\"ing fron1 the forIn of the Lagrangian.
L.. et us for the sake of the present argurnent that there are no second
class equivalently, that \ve are \vorking \vith a prelilninary
set of hrackets consistent \vith setting all constraints to zero except the first
class constraints and their gauges::' rrhen if \\"C dcnote by the I independent
fi.rst class constraints and by y" the I corresponding gauges, \':c \\;-111t variables
such that
,. \
<I .9I)
:\O\V let
1.92 ;
?rx he the sct of 2 I and y's:
/1' "-
?rx == ('1" 1 , ... , ''iI' Y1 , ... , YI)
]'hcn since the Yz" are first class, the in of C",;3 === { ?'l. , can be \\ritten
in 1;< I blocks of the forIn
-1 [Dll
Car, 11,
-- 21
on1 Eqs. (I.9 I) and (r .93), \\"C ilntnediately see that
{
;\ * .\ * 1* cv { /\ * " * ")
... 'l..(l , .... 'l..b j .Ll..a, ...l..b r
1'"lgar{llcss 0..1 the gauge clzoice. (Xote the contrast \vith the properties of clefi.ned
lJy
In practice there may be a particular gauge choice for \vhich there exist
variables satisfying
(r .95)
-
- 27-
in addition to (1.91). Yet another possibility \vhich son1etinles occurs is that
neither (1.91) nor (1.93) is satisfied, hut C;r3
1
is arranged so that (1.94) holds
for a particular choice of independent variables.
We see that the are a logical set of variables to usc in the
quantization procedure. The ordering probleI11s \\Thich occur in I11aking the
transition froll1 classical brackets to con1mutators are likely to be much less
severe for variables \vhich obey (1.94). t
T
nfortunately, no general procedure
for finding the appropriate variables is a vailable at this tin1c.
different procedure \vhich a \oids entirely the usc of Dirac brackets
is to consider a I-lilhert space :Yt', not necessarily \vith positive-definite n1etric,
of \Jvhich the I-lilbert space of physical states is a subspace. ()n ,Ye' one defines
operators for each unconstrained variable, \vith c0I11mutation relations corre-
sponding to the original Poisson brackets. The constraints epee no\v appear
as non vanishing operators on \vith \?anishing Inatrix elements hct\vcen
any t\\?O physical states \ a , b
< ai
W I b
IIX
o.
'rhe constraint conditions 1l1ay appear also in the stronger forn1s
a == 0 or < a i qJ
x
== 0 .
I30th Eqs. 1.96) and (1.97) arc referred to as suhsicliarv conditions. Such
conditions are inlposed on states of ,Yt)f in order to select the suhspace :If of
physical states \;\,:ith positi \'e norn1. 1"'he resulting theory has a ver:y synlnlctric
appearance, hut the price paid for simple cOlumutation relations is the intro-
duction of a larger manifold states.
our first application of the Dirac approach to constrained Han1ilto-
nian systenls, \ve examine the relati vistic spinless point particle in the mani-
festly Lorentz-invariant formalism. The action is taken proportional to the
path length,
(2. I)
2
S = - m Jcis == - m j
I I
It 1
dx' )2 ,
\vherc \\'"C recall that our rnetric convention 15 I == --gOO == gIl == g22 == g33
I'hen \ve choose an arbitrary monotonic paranleter 7 laheling the particle's
position on its \vorld line and define
(2.2)
dx
P
- (-r)
.
d't"
- 28-
'rhus the Lagrangian is
and the action
L== 'Ill
IS in \-ariant under reparan1etrizations 7 7' I t is easy to check that
Det { :2 IJ'u!J. 2u" } = 0, so that L is indeed a singular Lagrangian requiring
the usc of Dirac's methods to define the Hatniltonian dynan1ics.
I'he canonical 1110n1enta are
'[he equations and their solutions are
(2.6 (7)
or
:2.6 b)

71Z --- ---- == 0
ds
2
})p.
III
s ,
\\-hcre s is the path length defined hy the integral (2. I),
rrhe Poisson brackets for variables at equal 7 are defined as.
so that
(2.8 a)
(2.8 b)
{
(-) (.-\} _
\. ,..1.--'J \ - --- 0v
{
('-'1 p'J (,-\} == 0
\W),
Since we know the solutions (2.6) of the equations of motion, we may consider
Eqs. (2.8 (7, b, c) to give the brackets at ':" == S == 0 and C0I11pute the brackets
for variables at different points. Only (2.8 c) is changed:
(2.8 d)
{
Xl). (0\) x'J (s')} == gP.') __ )
" /' \, / Jll ,
='Jcnv \ve observe that the canonical Han1iltonian 'Zh7Jlz's,!zes,
(2.9)
1IIIIIIIIi_--------------------
-
- 29-
and so I-I
c
is identically the prilned Hamiltonian, I-I
c
== H' == o. 1'he vanishing
of J--I
c
is attributable to the fact that the Lagrangian is h01110gcncous of dcgree
one in the \"clocities. l'hercforc the m0111entu111 p:t is homogeneous of degree
zero in the velocities and no unique solution u!J. (x , P) exists. 1'his is typical
of a singular Lagrangian; the homogeneity of l)f
t
is seen fro111 Eq. (2.5) to lead
dircctly to the first class primary constraint
(2.lO)
'The total Ii alniltonian rnay then be taken to be
(2. I I)
2

H correctly gcnerates I-Ian1iIton's equations of 1110tion In the arbitrary para-
tcr 7:
(2.l2 a)
(2.IZb)
Exa111ining Eq. (z. l2 a), \VC sce that v can be expressed in ter111S of the velocities
uY', so that H finally heco111cs
I-I ==
21lZ
There is, ho\vever, still an arbitrary function in the system because the scale
of 7 (and hence the scale of u
P
-) has not heen fixed relati ve to x!.L.
B. GAUGE
XO\V let us 111ake use of the gauge freedo111 in the action to fix the scale
of 7 and elilninate all arhitrary functions fro111 thc system. \'/e choose
2
7Jl 0
Note that other choices of the gauge ep2 may be equally suitable (Dirac, 1949).
Our old first class constraint p2 1n
2
0 is now second class. We n1ay
thus C0111pute the matrix
(2. I 5)
and its Inverse,
o
Ce<[3 == {epa , CPi3 } == 2 pO
-2 pO
o
(2. l6)
c-
1
ai3 == 2 pO
- 30-
NOvV \VC replace the Poisson brackets (2.7) \vith the Dirac hrackets
(2.17 a)
\vhich are consistent \vith setting the constraints (2.14) strongly zero. In
particular, we have at equal T == xO
(2.17 b)
{ P
lt V}* + fLO Pv
" x == -g g pO
{ P
!.L PV}* == { 1.L v }* ==
, x ,x o.
We can no\v elin1inate the extra variables in the canonical systenl by using
the constraints (2.14) to set
(2. I 8)
The choice of H == !J
o
to replace the original vanishing Hamiltonian is justified
by the fact that it generates I-Ian1.ilton's equations of Illotion in the nc\\? evolu-
tion variable t:
dA.
dt
The velocities are thus gi ven by
JA. -L H 1*
dt i l"" f
(2.20 a)
\vhile
(2.20 b)
(2.20 c)
cL1; == { z" H}* == {. i pO}* == pz"
elt x, x, po
__ == {pt' H J'l.* == 0
dt ' ,
T'hus
pi (t) == I)i (0)
Xi (t) == Xi (0)
pi (0) t
(p2(0) + ,
and the unequal tin1e bracket system IS
{ pz" (0) , xi (t)}* == D1j'
(2.21)
{ pi (0) , pi (t)}* == 0
{xi(o) , x.i(t)}* == 1 JJz'!.) _ pzpi)j(p2 -t-- ?n'!.):)/'2.
-31-
\Vc note for cOlllpleteness that our system is Poincarc-co\yariant, 'fhe
translation generators p,L are constants of the n10tion, as are the con1ponents
11' == tpt" - Xi 11 == tpi - Xi (p2
l\l
ij
== Xi pi _ xi pi
of the Lorentz transformation generators M!-1'J. (Note that {Mo
i
, H}* =1== but
dlYIOijdt == 0 due to the explicit tilTIe derivative in Eq. (2.19).) Defining the
Lorentz group structure constant
\ve verify the I)oincare group algebra:
{ pP., p'J}* ==
{MClr3, p!-L}* == pr3 _ g!J.r3 pC(
{l\l
f
L'J, lVIC(r3 }* == M'r.
C. QUANTUlYl lVIECITANICS
]'he system of quantul1l,-ITIcchanical operators corresponding to the classical
relati vistic spinless particle can be deduced directly fro111 the Dirac brackets
of the previous Section. \Ve first define the Hilbert space nor111 as
(2.26)
(rpJ rp) = Jrp*(x) rp(x) d
3
x.
'Then adopting the convention (1.84), vve take Xi and pi to be operators satis-
fying the equal tinle C0111111utator
(2.27)
(2.28)
t == ,.,yO == parameter == c-number
H == pO == + (p'2 -t- 7JZ2)k == q-number .
J-Ian1ilton's equations hecon1e
(2.29)
dA
dt
i [1-1 , ,
so that the Heisenberg-picture equations of lTIotion are
(2.30 )
- 32-
Since the algebra (2.27) n1ay be realized in the Schr6dinger picture as
\ve find the following nonlocal Schrodinger equation:
Hm (t , x) :::= (-_ n
2
(t ) . 89 (I, x)
T V nz c.p ,x == z 81 .
Iteration of Eq. (2.32) gives an equation of the same form as the local Klein-
Gordon equation.
Requiring the I-Jorentz group generators to be hernlitian vvith respect
to the norn1 (2.26) gives
\Ve find that the J)oincare algebra holds,
'i p!J.] == _gP.rx _+_ g!J.f3 pC(
and in addition
Equations (2.35) and (2.36) are taken by definition to lnean that x
j
transforins
as the space part of a four-vector (Jordan and l\1ukunda, 1963).
rrhe extra ternl in Eq. (2.36) beyond that required by the pure Lorentz
transforlnation properties of x!J. occurs also in the classical Dirac brackets
and is interpretable as follo\vs. I--he gauge choice
is not Lorentz invariant, and yet the Dirac bracket procedure forces the con-
straint equation to be strongly valid in all Lorentz frames. l"'his requiren1ent
can be n1ade physically consistent only if a Lorentz boost to a ne\\7" fraIne,
P'!J. ==:: p!J. w
p

v
Pv
IS accompanied by an infinitesinlal gauge transformation
,- _" ::. (_ ",..) -_ I /-\
/'" \, -
consistent with the equations of n10tion. Classically, the change in x
P
. n1ay
be \vritten
x'p. C':) == x!J.
- 33
Our consistency condition is therefore
and so
Therefore, for a pure boost,
~ . . . dxj '0
(2.41) ox
J
== x'J - x
J
== - d ~ W
Ot
Xi + W
J
X
o
. pi . ( )
== + Wo; t -- H WO
t
x,' t .
Accounting for appropriate orderings of quantun1-mechanical operators,
this agrees exactly \'lith the change in .-:t.
j
generated by the boost operation (2.36).
3. RELATIVISTIC SPINNING PARTICLE
... t\. REVIE\V OF LAGRANGIAN ApPROACH TO Top
We next consider the Lagrangian approach to classical relati vistic spin-
ning particles developed by Hanson and Regge (1974). The treatment gi \'en
here will be slightly n10re general than that in the original paper, \vhich dealt
only with spherical tops. In order to ensure Poincare-inyariance, unphysical
degrees of freedon1 are introduced into the Lagrangian. \\lhen \ve in1posc
constraints to eliminate the un\'lanted yariahles, \ve must use the Dirac for-
malism to find a consistent I-Iamiltonian systen1 and a consistent quantum
system corresponding to the original Lagrangian.
\Ve begin, as \ve did for the spinless particle, by considering a particle
\vorld-line with points labeled by an arbitrary n10notonic paran1eter '7. I-Io\vever
no\v \ve associate with each value of '7 not only a position x!J. ('7) but also a
Lorentz 'lnatrix A!J.
v
('7) obeying
A
I.t AAV _ !J.V
A -g
A
!J.
A
AV _ !l.V
A-g
AO
o
> I.
Denoting 't'-derivatives by an overdot, we \'lrite the most general possihle
action as
't':J
S = Jd"t" L(xll-, :til-, All-v, All-v) .
't't
Now \\re argue that invariance of the systen1 under Poincare transforma-
tions of the external reference fralne severeI y restricts the form of the Lagran-
3
- 34-
gian. I ... et us \\Trite a generic elen1ent of the l)oincarc group as == (x!J., A
P

v
)
and choose the group ITIultiplication conventions
Right transforln: (x', A') == (x , A) . (a , M) == (1\1-
1
x -+- a , A1\!I)
Left transform: (x', A') == (a ,1\11) . (1: , A) == (A-1 a +x , lV1A) .
With the convention (3.3), the right index of j\.!J.
v
refers to the external sjJace-
axes of the system, \vhile the Ifft index refers to the internal body-fixed-axes
of the top. [I f the conventions for right and left multi plication in (3.3) \vere
reversed, the physical lneanings of the indices of \\Tould be interchanged].
I t is no\v easy to see that
A!J.
v
XV } are right-Poincare-invariant
A!J.
A
AVA
(35)
.!J. + AA!J.A- V}
% I,V X
are left-Poincare-invariant
j\I,P. AA
v
Our description of the top \\Till therefore be manifestly invariant \vith
respect to space-axis Poincare transforlnations if we allo\\" the Lagrangian
to depend on the ten \'ariahles (3.4):
(3.
6
) L
(, A :\ - L (A p. ..'J "'\- rxy A[3 \
\, .1: ,.t, - \J l. v .x , -'- _ y) .
Since the derivative of (3.1) in1pIies j\.../\T is an antisymmetric 4X4 matrix, only
six cornponents of j\j\T can be independent. Sin1ilarly, Eq. (3. I) sho\vs that
the sixteen variables are rcally functions of only six independent variables
\vhich \ve \\"rite as CPt', i == I , .. " 6. 1'he canonical momenta may then he
taken as
The Euler equations become
(3. 8)
\vhile the canonical Poisson brackets are \\Tritten
We 'no\\! digress briefly to develop an intuiti vel y appealing notation for
the angular mOlnenta and their equations of motion. First \ve recall that in
- 35
llonrelati vistic 111echanics, the constant body-fixed coordinate x' of a point
in a top IS
\vhere RtJ" is an orthogonal rotation matrix
relating x' to the space-axis coordinate x. x' is taken to be a time-independent
constant, so the tirne derivative of (3.10) gives
(3. 12)
here we define
With CJ:/ == ! 2/
ik
w
jk
, (3. I 2) becomes the usual equation v == w Xx.
If we now take the 7-deri vati ve of Eq. (3. I), \\"e discover the generalized
angular velocity
which has SIX independent components and reduces to (Jij == u/
i
\;vhen
j\'P'
v
is a pure rotation. Let us also define an angular change by
(3. 15) oOpv == J\./'!J. 01\/ == --- Ol\.J!J. J\./,v == -
\vhere
'I'hen vve can express G[J.v and in ter111S of the independent angular variables
as follo\vs:
!J.V [l.V ( )
(J == ai <:Pi,
[.LV ( )
au == ai
or
or
Compatibility \vith Eq. (3.16) den1ands that afv satisfy
dar
V
-Ji--
(3. 20)
generally possesses an Inverse function \vith the properties
!.LV bc<[3 pc< vB p.[3 vc<
ai i == g g --- g g .
Using Eqs. (3. 18) and (3. 19), one can sho\v that

bl.t') b!Y.[3 L __C
flV
c<[3 bC:"C
J d?j J - en' t
- 36-
\vhcrc is the Lorentz group structure constant (2.24). Equation (3.
20
)
n1cans that is a realization of the adjoint representation of the Lorentz
group L.ie algebra (Racah, 1965).
l' '1'
XO\V the Lagrangian depends on CPi only through i\J\ === j\crJ\. l'hus
if vVC define
,ve 111a y ,vri te
T
l dL I
d(j!J..
V
dL
-
dePt"
--
8cr!J.. v
2
I ( ) 5
2 at' CP p.')'
1-1"'ro111 (3. 19), \ve have also
sIt') is thus a con1hination of canonical coordinates and canonical 1110111cnta
son1c\\"hat sin1ilar to the spinless Lorentz group generator p') _ XV p'l .
l;sing (3. 14) to (3.23), ,ve can no,v express the Poisson hrackets
;:3.9) in tcnllS of S1(,\) and l'he result IS

dB 8A. 8B
8..\
8Se<f1 '
\vhere .\ and S arc to he considered as independent \"ariablcs. \\"e then find
the canonical })oisso!l brackets
(3.
2
5)
{ P''', x') } === _
(S'l') 5
IX
[3 I. === + So,,;
l "-, J '0"; "-
{5
1l') I\IX[3 I _ A IX') !l[3 A. v[3
, J J - J\. g -- j\. g
{p
ll P')} == { Jl .') l == { A' la ;\ v[3 I. == 0 .
, x J j \. ,J\. J
]'hercfore S'W obeys the Lorentz algebra by itself and generates the right
(space-axis) Lorentz transformation on j\IX[3 Note that
{S'lV, AO:[3 A
Y
[3} == o.
Thus the brackets (3.24) are con1patible \'lith the strong validity of the con-
straints (3. r) even when all of the s,t') and -L\IX[3 are treated as independent va-
riables vvhen taking partial deri\"ativcs. In effect, we ltazJe sirlestepperf OJle stage
oj tlte Dirac prorc{!ure by starting ,vith brackets \vhich are consistent \vith
(3. I).
\\Te sho\v ho\v to re\vrite the Euler equations (3. 8) in tcrn1S of
and r;!J.'). Using Eqs. (3.17) and (3.22), find
t
I "'I p.'). ...
o == 2- '-'.I ai 9j Sp.,)
I C 'lV Aa [3y
---- ;:)'1.') eli - .. r'3 ai
2 ;
3L
-
_._-
- 37-
N" 0 \V qs. (3. I 8) and (3. I 9) 1et us rcexprcssthis as
(3. 26)
One can verify explicitly that the choice (3.6) for the forn1 of the L,agrangian
in1plies
I I I 8I,
___ s't , -1- cr
ll
. s/ .!t
8j\\
'rhus the spin Euler equation can also be \vritten
(3. 28)
In order to forge a connection hetvveen (3.26-28) and the fcuniliar equations
for the rotator (Goldstein, 1950).
. .
Lspacc === L body
OJspacc X Lspacc === (external torque),
\\'C definc the hc)(l y'"-fixed SpIn
1'hcn \vhen \\'c refer S'l') to the space coordinate systcn1, \ve find that
to (3.26), \ve find that the n10tion of the hod Y'-fixed ill
its o\vn fran1c is
/. '?)

\ve ohserve that if \VC define
cll ')
...:-1 ,
thcn the Euler equations (3.8) and (3.28) in1ply

------- == .
d"t'
\Ve no\v sho\v that l\1'l') is precisely the conserved generator of right (space-
axis) Lorentz transforrnations. lJ ndcr a right Lorentz transforn1ation, the
Lagrangian is invariant, so
(335)
_ (P .)t + 'l...
i
.j
- d"t' !J. Ow .1 au) CPu ,
- 38-
"",here \ve used the Euler equations (3.8). Novv the infinitesin1al right Lorentz
transformation Ow has the effect
a
w
x/.L == (Uar3 (xC( g0tJ. - x
f3
gatJ.)
aCJ) == (AP,a g',l';3 _ g",a) .
From Eq. (3. I 7), we find
For arbitrary Eq. (3. 35) in1pIies Eq. (3. 34) and:\IIW is the conserved Noet her
current of right Lorentz transforlnations. By explicit con1putation using Eqs.
(3.25), we see that M
WJ
and obey the Poincare algebra (2.25).
B. CONSTRAIXTS ON Top LAGRANGIAN
We now exan1ine the constraints which must appear in the top Lagrangian
if it is to describe a reasonable physical systeln. First \ve define the set of ten
right Poincare invariants
(A
o .\1 Al .\1 ,\2.',1 A3 .\1
az"== "'x, \lX,J.\.',IX' ',IX,
A
0", Al A0'1 \2 ;-\ 0'1 A
3
A2'1 A
3
A-:-3\1 Al A Iv ,\ 2 )
',I , j \I, .... \. 'I , ',I , 'I , i\. v
conlposcd of the variables (3.4) \vhich appear in the Lagrangian. The deri va-
ti ves of L with respect to these variables will be written
7-reparan1etrization invariance of the action is assured by the requirement that
L be h07nogeneous of degree one in the velocities; L (a) is thus taken to satisfy
Euler's differential equation for h01110geneous functions,
10
L(a) == aiLi == aiLl"
i=1
Since the ten mOlnenta pl-t, S'W are homogeneous of degree zero in the velocities
az", at most nine momenta can be independent; there must exist a trajector.y
constraint relating the momenta,
We will restrict ourselves in \vhat follo\vs to constraints of the form
(3.42 a)
- 39-
\vhere :\1
2
is not a constant but is a function of the body-fixed angular I110n1Cn-
tum S == AS,A.T satisfying
(3.42 b)
l\ '1
2
/-Si
j
-S()z\ -
.lV ('-, )-
rrhe moti vation for this choice \vill becnn1e apparent shortly.
Next, we n1ust force our top to have only the usual three spin degrees
of freedom in the nonrelativistic limit. In order to eliminate three of the six
components of S!).v, UJe hereafter restrict ourselves to Lagrangians jar 'leJhich the
relation
and Eq. (3.42) jollo'lv jront the jornz oj the Lagrangian alone. (Since PtJ.\7
Il
==-= 0,
Eq. (3.43) contains only three independent constraints). By choosing the
constraint (3.43), \ve effectively pick the physical meaning of the coordinate xtJ.
of our spinning particle. .l\ccording to I)ryce (1948), Eq. (3.43) identifies xtJ.
as a body's rest center oj 1/z011Zentul1z, \vhich transforn1s as a four-vector.
(3.43) tTIust for consistency have a \veakly vanishing derivative:
Con1bining this infortnation \vith the equations (3.8) and (3.28), \ve
find
(345)
Putting this back into Eq. (3.28), \ve have
== (external torque),
thus cornpleting the connection bet\vecn Eqs. (3. 29) and (3.3 I).
Now we note the general identity
\vhere
S*f
W
== E!).'JI.a S
2 1.0 ,
and EWJi.cr IS the totally antisYlntnetric pseudo-tensor, \vith 2123 == I.
lVlultiplying by Pi).' \ve find that the constraint (3.43) directly implies
(347)
S
f-L
V
s*
"- o.
Thus, for exan1plc, fron1 Eqs. (3.45) and (3.26),
------
(3.4
8
)
* Au, 8L
Sf-Lv A . --) ,
dA
- 40-
(Nate that
(-srJ-'J ,,* _ I rJ.WJ.r1 A SYO A " == ArJ.(X ,.. sYo) A
V
(3
\"- ) - 2 E:xy (3ij 2
due to .1\'1(3 A)"Y j\08 E:x(3yo == Det ..A. ErJ.'JI,O == ErJ.\Ji'o).
\\'re will not deal here with the problem of finding Lagrangians satisfying
(3.43); procedures for doing so are given in f-Ianson and Regge (1974)
for the case of the spherical top. .i\.. sin1ple exan1ple of a suitable spherical
top Lagrangian is
(3.49 a)
L = - 2-
l
{Au
2
BO' 0' [CA u
2
BO' 0')2_
- 8 B - 2 B Det
\vhere u
2
== UrJ. u[.L , 0' . 0' == O'!.L'J cr
wJ
, uO'cru == u[.L cr[.L'J cr
V/
, u
l
. , Det cr == (O'rJ.'J 0':'1)2/
16
,
and A and B are constants. Direct computation confirms that the constraints
(3.49 b)
follo\v from the form of the Lagrangian alone. Lagrangians also exist for \vhich
the mass and spin are separate unrelated constants:
(3 50)
I S B
2 '-WJ
C. DIRAC TREATMENT OF Top CONSTRA INTS
\\Te no\v consider the dynamics of a relati vistic top whose Lagrangian
gi ves rise to the following \veak constraints among the canonical momenta:
(3.5 I)
"fhe canonical Hamiltonian IS
2 -
M (s)
y'J. == Pv 0 .
He == U!.L T
L p!L I I L
i -- == UrJ. T "- (jWJ - ==
2
and so may be taken as the initial first class Hamiltonian. Ho\vever, the con-
straints (3.5 I) are not all first class, since
{p2 M
2
, \7"1,} 0
{y"", Vv} S[.LV p2 .
-41-
1"'here are nevertheless t\yO first class combinations of the constraints
(3.51), \vhich \ve take to be
We observed in Eq. (3.47) that 92 0 fo11o\vs directly frorTI \7'1 o. ()ur
first class I-I an1iltonian n1ay novv be taken to be
(3 r'r')
\. .))
Since consistency requIres
V
(",)
2 \ T2'
'il
t
=== {VT[.L !-1
v 'J .-....,:, 0 ,
any 111ultiples of \:/t \vhich 111ig-ht be added to H 111ust have vanishing coef-
ficients and there are no secondary constraints.
Defining the tensor
(3 56)
\ve compute the velocities to be
(357)
uy === {x!.L , H } === 2vI p!.L
[.LV _ \I,[.L {A v H} _ r[.LV + ('J
(J - j ), , - VI 4 V2
...\s in the spinless case, \ve n1ay set VI === (- U
2
))(2M), and similarly restrict
the form of V2' Nevertheless, there relTIains an arbitrariness in VI and V2 \\Thich
is eliminated only \\Then \ve choose gauge constraints.
Equations (3.5 I) represent four independent constraints. ]'hc t\'/o con1-
binations (3.54) are first class \vhile the t\VO rernaining independent con1bina-
tions, say
(3 58)
are secoJl{i class:
Clj === {\1 , \Tj} == p [.L pIt
Strictly speaking, \ve lTIUst replace all Poisson L)rackets by the correspond-
ing Dirac brackets (1.2 I) even in our preliminary first-class Hamiltonian
system.
- 42-
O\V let us proceed directly to impose a gauge condition \vhich is effec-
tively the" phase space conjugate" of the constraint V'" 0 on the mon1enta.
Physically, \ve \vant to be a pure rotation in the rest frame of pI", just as
S!J.
v
beCOlTIeS a pure three-angular-n10n1cntun1 in the rest fran1e. V\le therefore
conjecture that
(3. 60)
IS a consistent constraint choice, \vhere
(3.6 I)
and j\
0
0==1\0
0
== . - )\.00 == I for a pure space rotation. Since
only three of the four constraints (3.60) are independent.
:\'" O\V \VC use Eqs. (3.3 I), (3.46) and (3.37) to sho\v
-S:Oi r (1
/
, S-. i _ -SOA 1-. i)
"'C. VI I. . A'
h I
-U.')\.')''X \')(3 h E ( b h ld d 2 (-S) . d
\v ere ' == j . 1"' en if . q. 3.42 ). 0 s an 1\1 IS even un er
SOi _ SOi, lOt' (S) n1ust be ocld:
'"'1M
2
ell.. _ (-Si
j
-SOi, _ -loi /-Si
j
-SO)
---::-- - - -.--- \ ,-L
CiSoi
\\Te conclude that if \ve set
at son1e point 't'o, \VC have also
(3. 66) rO/ == 0 .
l'hen from Eq. (3.63) \ve know that SOi 0 and so SOi vanishes for all T.
Equation (3.65) is therefore an invariant relation provided \VC restrict
ourselves to trajectories obeying Eq. (3.42).
\Ve no\v define
Since Vi 0 implies
-S1)"
Pi 0 ,
we must have
- 43-
Fro111 Eq. (3.57), the equation of Inotion for prJ- is
\\Te no\\?" choose the gauge
pO I) == P
rL

so that
or
ty,. == o.
'Thus fro111 Eq. (3.67), \ve find
'T'he equation of motion of pi IS no\v
Consistency then forces
so that the gauge choice pO P has fixed one of the arbitrary functions In
the Han1iltonian as expected. Furthermore
It _ A!LV P P _ ,,0 p
p - J 1. v 00 - - g
vlhich is exactly equivalent to Eq. (3.50). We conclude that the constraints
consist of one gauge COJlfiz'tion and t'lRJO invariant relations.
\\le may now compute preliminary Dirac brackets consistent \vith the
six independent constraints
\vhich \ve \vrite as
i === (VI V
2
V
3
y2
ep '" s , C:, , s ,
I'he n1atrices needed to compute the preli111inary brackets are
Z === 1,,6.
o
- 44-
and its Inverse
o
We Inay put our brackets into a manifestly covariant forn1 by re\vriting
(C-1),j' as
(C
L T (C-
Laf3
rr
)'i === a ) jf3 ,
\vherc
z',j==: I,,6

(3 7
2
a)
and
(3 7
2
b)
rr,{'!l 0
rr/
ex
==
0
_ pi/pO
0 0
==
_ p2/
p
O
0 -+- I 0
_ p3/pO
0 0
+1
k== 1,2,3
[1.===0,2,2,3
l'he Inatrix Cexr, and its inverse are sin1ply
c
exf3
===
Pg!l\J
/c-1. ex(3
\ ) ==:

p
_SWJ
T'kp. is no\v used to canvert and into \J
p
. and hy using the identities
prJ. V!J. == 0 and == - p;'t as follows:
(3 75)
{i\., yk} Tk!J. {A, ykTk!l'} === {A, V
rJ
.} ,
{A , ;k} Tk!J. {A , l'k!J.} {A , .
Defining the eight component object
ex == (y-'-o \Tl \1
2
V3 ;:3)'
cp ", IX==: I,'', 8,
\ve nlay use Eq. (1.21) to \vrite the prelinlinary Dirac brackets as
(3 7
6
) { A B
) I t:'..I {;\ B } I {A "' l f ;: B l
, f rv .... !J.., - P , v J t , J +
+ .. {A , {y!J. , B} + {i\. , S!J.
v
{ , B} ,
- 45-
\vherc \ve use a prin1c to distinguish the prelirninary brackets fron1 the final
1)rackcts.
cxplicit forn11das for the prin1cd brackets of the canonical variables are
(
) I . { P
rt, x') .... j' _ g'LV
\,377 - \. =
2. ,x'J}' === pCl.)
3. {pl-t, p'J }' == 0
4. {P!L, S'JI, }' == 0
5. {x
tL
, S'JA}' == pl. _ S(LI, P'))/CPC( pCl.)
6. {StJ.'), SCl.13}' == StLC((g'Jr3 _ p'J p13/p). PA) _ SU
13
(g'Jc( __ p'J pcx/PA pi)
+ SCX') (g't
13
_ ptL p13/PA I)A) _ S13'J (gILCX _IJ!t pCl./P
A
pI,)
7. { j\.11-'), xC( }' == (p'J gC(13 __ pr3 gCl.'J)j(P
A
PA)
8. { j\.,t'), p
cx
}' == 0
9. { j\11. '), j\.C(13 }' == 0
10. { ..;\!l'), SC(13 }' === J\!)''X (g'J13 __ p') I>f3
jP
" pA) _ j\.,t
13
(g'Jcx __ p'J petIP), pI,)
-i\P'crpG(gC('J p13 _ g13') PC()/(Pj. P").
brackets are no\\ con1patiblc \vith setting the constraints V'1. and
identically zero. Xote that the position variables x
P
' have nOJlzero brackets
(Pryce, 1948). )Jevcrtheless, the Poincare algebra (2.25) is still satisfied in the
prinled bracket systen1 and X
IL
transforn1s like a four-vector.
C' \"",), 1 ylt 1 I -So/ . h I
..:JIllCC " au( C;' vanI:) 1 strong y, vanlS es a so:
(3.7
8
)
=== j\.o c
lX
r3 !\i == _ P sc<13 '\.1' (_ p2 \-J ===
ex. ,J .. f3 C( .. J 13 ) 0 .
FurtherlTIOre, from (3.64) \ve find
1-\s a consequence, I
cx13
no\v obeys
(3. 80) I
et13
== o.
\Ve can forrn a I-Iamiltonian from the remaining first class constraint,
') -0'
lVI""' no longer depends on S I because of (3.78).
The velocities arc sin1ply
uP. == {X!L, H }' === 2 IV1)'1
[L'J _ ;\,.,1. { \.'J H l' _ r.' 111.'J
() - J . j J, , J ---- LJ
-46 -
'vVc no\v elitninate all arbitrary functions by selecting a gauge constraint
conjugate to the trajectory constraint. IVIaking usc of the iterati ve property
of the Dirac brackets, we choose the final constraints
and ilnpose then1 on the printed brackets (3 77).
The required rnatrices are
.. [0
C
tJ
== { ((). ([).}' == 0
Tt' I} 2P
-1 I [ 0 I]
C --
lJ - 2 po -I 0
Equation (1.2 I) with all Poisson brackets replaced by the primed brackets
(3.77) gi ves the fornl of the final Dirac brackets, \vhich \ve no\v denote \vith
stars. The results are
(3. 85)
2.
3
4
5
6.
7
8.
9
{x'\ x'J}* == (P'''' SO') _ I)') pO) ==
== /1\11
2
_ (P') pi _ S'Ji pi)j(:\I
2

pV}* == 0
{ S'JA}* == 0
{
X''', S'JI'}* == __ .1.. rsf-t'J !J
I
. _ S,l}. p'J) -t- (SOV pI, _ SOl, pv)
M2 \ '- POl\12 '-
SCX
r3
}* == + pv _ + pv
+ Sc('J -1- 1M
2
) __ (g[J.cx -+_ pf-t P
CX
/M
2
)
{
_\'W /\ == _ I (PA Ov _ pv OA) AC( f(r3 +
j 2 po 1\1
2
.J. 1-, g g y
+
__ __ A
CX
(Py Or3 _ pr3 Oy) AtJ. II.,)
2 po M2 Y \ g g I.
- 47-
The Hamiltonian IS
(3. 86)
\vhilc Han1ilton's equations of Inotion are
4-1- - dt + {A , (p2 + M
2
(S'
J
))t}'
Here \ve have used the traditional notation x
O
== t to en1phasize that xO IS
no\v a parameter and not a canonical variable.
]"'he Poincare group generators are no\v \vritten
pO == H
(3. 88)
pi == p'"
M
Oi
== tP'" - H -t- SOi == tp
t
- Xi H sii P.ijI-I
1\1
'
)" == Xi pi _ xi pi -t- S')" .
]"'he star brackets of these quantities obey the Poincare algebra Eq. (2.23)
and the theory is again Poincare-covariant. Applying Eq. (3.87), Vle confirm
that the generators (3.88) are constants of the motion \vith respect to xO == t.
The transfornlation of x
P
' under the Poincare group is altered in the star
hrackets. x
O
== t is no\v just a parameter, \vhile x transforIns as follo\vs:
{ P
o _.i}* - { H i 1,* - _ pi
jH ,J:, - ,X
J
-
{ P
i .. jt* == _
J 0
{
1\.
-1 ij k '\ * i;" j jk l'
1
1
,x)- ==0 X"---o x
rrhe extra tern1 in {lYrO
i
, x
k
}* is just the correction discussed in Chapter 2
\vhich n1aintains the validity of the constraints (3.83) in the ne\v Lorentz
fran1e.
D. QUANTU!vl l\TECHANICS
In principle, the system of single-particle positive-energy quantum-
nlcchanical operators describing the relati vistic top can be deduced directly
froIn the final Dirac brackets (3.85). In practice, however, this can be quite
difficult unless \VC are able to construct fron1 the original set of variables a re-
duced set of independent variables (r .85) obeying canonical brackets. F'or the
top, the appropriate variables correspond to the Pryce-Newton-Wigner varia-
bles (Pryce, 1935: Nc\:vton and ",Tigner, 1949) supplemented by the
angles. We define
sz' == -A- siik Sjk
- 48-
and take our ne\v variahles to be
qO === t
s x P . sOt
q - X -r
t
- - H (H + lVl) -I-I"+KT
J
=== M S + p (P . S)
H H(H + M)
R
ij
== A
ii
_ j\iO pilCH + lVI)
/j A
ik
pk pi
=== j\' - H (H + M) .
p===p
pO === H == + lY1
2

2 -i' .
Here IVI (S J) can he taken to be a functIon only of
J'" === R"j]i
because
Here we haYC observed that R
ij
is an orthonormal I11atrix,
R
ik
R.i
k
=== R
ki
R
ki
===
and used the identity E
zjk
R
ai
Rbi R
ck
=== subr Det R === S"bc.
l)efining
\ve find by direct C0I11putation fr0111 E:qs. (3.90) and (3.85) the follo\\'ing brackets:
(395)
I . { pi, r }: = -
{ I-I , qJ} === - 1).1I H
2.
{ q, qi }* === 0
3
{ p" }* === 0
4
{ Jj } * === 0
5
{ qi, ] j }* === 0
6.
{ ]i,
Ji 1.* _ -::-ii
k
]k
J - .....
7
{ R
iJ
, qA' }* === 0
8.
{ Rii, I)k }* === 0
{ R
ii
, H }* === -
I
R
ik
sP'/ II
:2 H.
9
{ R
ij
, RIm }* === 0
10.
{]i 1.* _ -::-ijl, Rfl,
, j -- \000
-49 -
Equations (3.95) now forn1 the basis for our quantum mechanical system.
vVe begin by choosing the Hilbert space norn1
(<I> , <1 = Jd
3
x<l>* (x , t) <I> (x , t) .
We then take
t == qO == c-number
2 . 2 -
M === - p ~ l P(1. == IVl (J) == q-nun1ber,
and the Hamiltonian
where the t\vel ve independent operators are
q, pi, Ji and three of the R
il
.
Starting froIn the Dirac brackets (3.95), \VC postulate the comn1utation relations
(399)
i [Pi, qj] == ~ ij
i [pi, Ji] == i [pi, pj] == i [pi, R
z
O
k
] == 0
i [qi, qi] == i [qi, ]i] == i [qi, R
ik
] == 0
i [J i, ] i] == - Siik Jk
i [], R
li
] == _ szjk R
1k
i [R
zi
, RIm] == 0
We also observe that
(3. 100)
i [J, Ji] == 0
i [1, Ji] == + si
k
Jk
i [Ji, R
il
] == +Siik R
kl
,
(3. 101)
so f COIntTIutes \vith right vectors and transforms left vectors. The Han1il-
tonian (3.98) has the comtTIutation relations
i [H , q] = pi
JH
i [H , pi] == 0
i [H , ]] = 0
We omit the explicit computation of i[H , R
Zi
] due to nontrivial ordering
problems.
l'he Poincare group algebra is generated by the operators I-I, P, and
lVI'i == qi pi _ qi pt siik Jk
(3. 102)
- 50-
l'he Han1ilton equations of tTIotion for an operator i\ are
ciA. 8A
----- == z [H A] + -- .
dt ' 8t
i\n explicit realization of the algebras (3.99) and (3. roo) in terITIS of dif-
ferential operators is the follo\ving:
})1 == ~ _ ? __
i 8ql
rro get explicit expressions for J and J in terI11S of the three independent
Euler angles, \VC first define the 3 X 3 I11atrices
(3 lOS)
realizing the algebra of the J
l
,s, and then write Rtj" as
(3. 106)
R(j ( ~ ) - iaLa - if3L2 - iyL3
a, ,y==e e e
cos (I., -slna 0 cos ~ 0 sin ~ cos y ~ s l n y 0
SIn 7.. cos 7.. 0 0 0 sIn y cos y 0
0 0 _ ~ sin ~ 0 cos ~ _ 0 0
cos rt.. cos ~ cos y --sin a sin y -cos 7.. cos ~ sin y -sin a cos y cos (/.. sin ~
sin (/.. cos ~ cos y -1-- cos 7.. sin y - sin rt.. cos ~ sin y cos rt.. cos y sin C( sin ~
-sin ~ cos y sin ~ sin y cos ~
Then \ve see that J lTIay be \vritten
F = i (- ~ ~ ~ ~ :x +- sin y ~ -+ cos y cot ~ :y)
'J ( sin y 8 -L d. d )
J'" == Z -.- ---- I cos Y 8S- - sIn y cot ~ ~ (
SIll ~ 8rx I I ,
\vhile J IS
-1 . ( 8. 8 cOS'Y. 8)'
J == I - cos 7.. cot ~ ~ - -SIn rt.. -:y:- + ---:----(.-\
ox C!i3 SIn ,j oy
(3. 108)
) ( d
J:' == i -- sin (/.. cot ~ ~
-:3 . d
J === ljrx
-51-
Thus the Schrodinger picture \:va ve equation which IS the counterpart of the
I--Ieisenberg picture equation (3.103) is
?] 1(!J
i\ -1- .... (J \) 2 ! == .1. T
lV \.) Z 'at '
\vhere j is taken to be the operator (3.14) or (3.108). (We make no atten1pt
here to resolve ordering problen1s \vhich n1ay arise \:vhen 1\1
2
(J) is expressed
as an operator). (3.109) is non10ca1; a local equation, \vhich cculd
be the starting point for a quantu111 field theory of relativistic tops, I11ight
take the form
(3. 110)
We ccnclude \vith the relnark that [)irac's procedures can also be carried
out for the electrically charged rclati vistic top. I-Io\vever, since this systen1
becolnes exceedingly c0I11plicated and is still ilnperfectly understood, \:ve \vil1
not atten1pt to treat it here (see Hanson and 1974).
4. STRING MODEL
Our examples thus far have dealt on1y with point-particle mechanics,
\,\tThere all canonical variables depend only on a single parameter 7. As our
first example of a field theory \vith continuous degrees of freedom, \ve exan1ine
the relativistic" string n10del " (Goddard et al., 1973; for a revic\:v, see RelJbi,
1974). V\Te take as the action functional the reparametrization-invariant
expression for a surface elnbedded in a ])-din1ensionallV1inko\vski space, ana-
logous to the reparametrization-invariant line elenzent (2.4) chosen as the
point-particle action. '[he Nan1bu action is then
"t'2 G
2
(T)
S [xP<] == - N dTJ" _ (dXr-t )\l/:!
d"r d(J / d"r v,!, 2cr dcr JJ .
T"l Gler)
N is a normalization \vith dimensions of Planck's constant divided by length-
squared, and one convcntionally chooses
'[he 1) canonical coordinates (7 , cr) no\v are t\vo-dinlcnsional field densities
vlith continuous dcgl-ees of frcednI11 labeled by cr. .l\ny transforn1ation of the
fornl
.... (-- )
(j -r G \" , cr
leaves the integrand of unchanged.
- 52-
rnetric is induced in the (':-, (5) space by its embedding in
Ho\vever, \NC can avoid explicit use of this ll1etric by \vorking exclusively
\vith '7, (J, and the quantities
(4.4 a)
and
dXP-("t',O')
== === ---
d't'
("t' , 0')
x'!J.=== === ---
dO'
(4.4 b) E === v
2
, G === u
2
, F===uv.
rrhen \ve ll1ay \vrite the Lagrangian density of the string as
OJ _ N {( )2 2 2 }1/2
uv -v u
=== - N { F
2
- EG }1/2.
The canonical 1110mcntun1 &'1 is then \vritten
=== - N
2
Gv!-1 + N:!
I t is convenient to define also
N0\V \ve require the variations of the action functional (4. I) to vanish.
For the 1110111ent, \NC continue to allo\v (51 and (52 to depend on T, thus obtaining
-r2 02(-r)
o = (')S = IdT JdIT (')Xf' (- -
't'l 0 1(:)
02(-r)
+JdIT ((')x1" ,9
1
") 1::::
Ol(-r)
't"2
+JLIT {(')Xf' (a2) [ TIl' (IT2) - giL (IT2)] - (')X
1
" (ITl) [IT" (ITl) - giL (ITl)])
rrhe equations are therefore
53 -
while the boundary conditions are
("rl , a) === (72 , a) =-= 0
i === I
'?
, ..., ,
Hereafter, \\ye \vill treat n1ainly the con ventional case \yith 7-indcpcndent
boundaries in the da integral and \vill choose
'[he houndary conditions that \VC need to use then ])CCOn1C
Il
tJ
- ("r , a === 0) === 0
Il
fl
("r , a === n) === 0 .
i\.. SVSTE!vf 'Vrr:f-lOUT GAUGE
Froln Eq. (4.6), \ve find the follo\ving t\VO prin1ary constraints,
(4 13 b)
A d ynalnical relation,
. 1 === f!lJ. f!lJ
=== f!lJ. u
,>
N'""u u 0
o.
follo\vs fro111 Eq. (4.7). Equations (4 12) and (4.14) in1ply that 1,2 =-= 0 at
a === 0 ,n. If \VC define our canonical Poisson brackets to be
{ f!}J!J. ("r , a) , XV ("r , a')} === - g!J.
v
- a')
{gJ!J.("r, cr), pjJV(7, cr')} === 0
{ xi! Cr , cr) , x') (r , a')} === 0 ,
then \ve find that and are class:
{ (-, , IT) , 1 (, , a')} = -+ (Y2 \' , a) + 'h (, , a')) -;r; a(a - a') a
1
(4 I 6) { Y2 (,,: , a) , Y1 (": , a')} === 1 (,,: , a) Y1 Cr , a')) (a - (J') 0
{
,L (- ,L (- - (- a) + tlJ (- a')) -- 0
't'2 "', \.J) , 't'2 \... ,v - \.. T 2 \. , 'T2 I., 6() 0 v v ro.,;:,
(See .l\.ppendix B for a discussion of the relation bet\veen Eqs. (4.16) and Eqs.
(1.83)). Hereafter, the explicit dependence of the canonical variables on T
will be dropped if no confusion arises.
- S4-
NO\V \VC cxarnine thc canonical " energy-monlentU111 tensor" in (1", a)
space. If we let zit == vt-L == .i
P
. U)L == 1IY' == ':r'1J. \vith uY == (utJ. 1/)1) the
1 ' :! " , aI' 2 '
tensor may be taken as
[6 r = - fI!
c b
V
p
__ 2
IT'l VtJ.
Formally, the canonical tensor IS conserved,
o.
Since [()c](lb is obviously first class, ,ve Inay add linear conlbinations of
the first class constraints (4. I 3) to forn1 the density ,vhich is integrated to gi,'c
the first class generators of gauge-like transforn-lations in 7 and cr. Thus \VC
define the total "energy-Inomentllln tensor" in (1" , cr) space to be
n
a
Uia,f, -L Va I
U b == I b 't'l j b !f2 0 ,
so the generators of and cr transformations arc
1'C
HT = Jda 0\
o
1t
H
a
= ( cia 0\.
-J
o
l-L-; and I-I
cr
ha ve \veakly vanIshIng brackets ,vith one another becausc the Yi
arc first class. rrhc brackets \vith J::1J., ho\vever, are nontrivial and \ve I11aV
place restrictions on Uab and '.la
b
by requiring
"1.'!I:= 'l _ { JL H t - "JUl .. OlJp,
---u- X, crj-'" 2;-:;r
Thus the definition (4. 6) of g;'" tells us that the choice
VI == I .
2
reproduces (4.20).
In an ordinary scale-invariant t\vo-dinlcnsional field theory \vith the
canonical tensor [8[]a
b
0, there is an infinite class of symrnetry generators
found by integrating combinations of [OJa
h
over an arhitrary function. These
(nonvanishing) generators produce all of the transforlnations 111apping the t\VO-
- 55-
dinlcnsional space pseudo-confonYlally into itself (see, e.g., r'"'ubini, I-Jansoll and
]acki\v, 1973). We 111ay construct a similar class of generators for thc string
using the zt1eakly zero tcnsor (4. 18).
We begin by defining z'" (cr) and (a) for - 1t a 0 l)y
uY (j) === --- U tL ((j)
a) ==
(Rebhi, 1974). rrhen \ve rnay Vvrite the generator of all T and cr rcparall1ctriza-
tions as the functional
it'
(4. 23 a) L[f]= (cla{[f(a) __ a)](Y'2
fN
2Z/H-2N[j(a)-j(-cr)jY. u}
.1
o
Tt'
-1:\ Iciaj (a) ([J}J NU)2 ,
\vhere
rrhe functionals [fJ forn1 a closed algehra in the Poisson brackets,
{ L [f] , L [g]} == L [fg' - f' g] .
'[he fan1iliar Virasoro functionals are essentially the 1
1
"'ouricr cOlnponcnts of
(4.23) and are defined as
it'
L,,== L [ei",,] = -4k (cia e
ill
" (.9' + Nu?
-Tt'
Classically, the LJl ohey the algehra
{ L
m
, I-'Jl } == -- i (7ll JI) L,1 -+-
m

\Ve clnphasize that the I-'1l are :::,'),"0 hecause they are fOrIllCd froll1 the
first class pritnary constraints (4. I 3).
\Ve no\v turn to the generators of the Poincare group tranSfC)rnlations
of the fields x
P
. (7 ,cr). Since the x!l- (7 , a) , l-L == 0 , .. " D - I, are D separate
scalar fields in the two-dinlensional (7, a) space, syrnmetries acting on
the index !J.. are IT10re like internal isospin symn1etries than spacetirne S}"111111C-
tries of a convcntional field theory.
\Ve con1putc the currents generating Poincare transfornlations
of the x!-L by cxalnining the corresponding variation of the Lagrangian density,
\vith boundary conditions (4.10):
o == 0P == DV
p
+ Tpl. DUll-
___ 8 / ))j!t 8 (nIt )
---- -:;,- (.T oXrl- + .r---- ox".
, c'cr r
56 -
If ox!J. is an infinitesimal translation,
we find the usual Euler equations
dqJfL dill-!'
+ === 0 .
The --r-independent "charge" generating infinitesimal translations of x
p
IS
therefore
1t
pl'Cr) = Idcr&J!LCr, cr) = ( dcr&J!'('t', cr).
J
crIer) 0
\Vc revert for a lTIOment to --r-dependent boundaries for (j and note that both
the Euler equations and the boundary conditions (4.10) enter into the deri\?a-
tion of the T-independence (conservation) of pl-t :
If ox!). is all infinitesimal Lorentz transfoI11ation,
(4.31 a)
we find
(4.31 b)
d ( 11, f]lJv V Y'!t) I d ( I" 11
v
v TPl) _
-r- X -x -r 2cr x -x -0.
'I'his equatioll also follo\vs from Eqs. (4.6-4.7) and the equations. '1'hc
generator of infinitesimal Lorentz transformations is thus
it'
M'L>' = {dcr (x" fJJv - XV &J") .
.J
o
Using Eq. (4.31 b) and the boundary conditions (4.12), \\?e find that lVI,tV IS
conserved:
cl:Vll-t",
d-r--
== 0:
The canonical Poisson brackets (4.15) I11ay nc)\v be used to sho\v that the
full l)oincare algebra is satisfied,
{ pC( , pf3 } === 0
{M'tv , pC( } === p" _ gV'l. p'"
{MI-tV , lVf
xf3
} === M-r
- 57-
\Vherc is the structure constant (2.24). It IS also clear the xtJ. trans-
forms as
(435)
{ pIt, XV (-r , O')} == _
{
j\ 'l!tV C( ("'I"" )} _ !let ." (...,.. ,,_ Vet.!-L ( "
.iV ,x w, 0' - g r't ., (J) g x
so that p'l and IVr
lv
indeed generate the infinitesin1al IJoincarc transfortnations
of X'I (7 , 0'); no integrals of the first class constraints (4. I 3) need to be added
to the canonical expressons for p,t and M!IV.
13. ORTIIONORl\fAL GAUGE CONSTRAINTS
\\Te \\Till no\v choose a convenient set of gauge constraints to fix the scale
of ':" and (J relative to x
tJ
-. \\rc begin by taking an arbitrary constant vector I,p.
and examining the consequences of the gauge choice
(4.36 a)
(4.36 b)
(
pp. )
AtJ. xtJ. , (J) - N':- 0
N is defined as usual as the normalization appearing in the action (4. I) and
p,t is the translation generator (4.29). Note that the conditions (4.36) are
nonlocal in a.
If \VC differentiate (4.36 a) \'lith rcspect to ':" and a, \\e find
1\' o.
Equation (4.36 b) and the Euler equations in1ply
so the boundary condition (4. I 2) requires
A . IT , (J) 0
for all (J, not just (J == 0 , n.
Next \\c observe that the definitions (4.6) and (4.7) of ,3'" and nIl, together
\\ith }=q. (4.37), in1ply that
A. {lj> - (A . v) Gj2
- (/\ . f?/J) XG 0 .
In addition, Eq. (4.39) requires
A . IT (/\ . v) N
2
FIfE 0 .
- 58-
\\TC first conclude that
(4.42 a)
'ZJ' O.
Secondly, (4.40) tells us that NGj!f/ I, so \ve discover that
(4.42 b)
By definition, (4.42) sho\v that the gaugc choice (4.36) has resulted in an
ortlzornor7Jzai jJara7J1etrzzatioJl of the surface s\vept out by the string.
I'he value of the Lagrangian can novv be \vritten
NT /F
2
EG)1f2 NG + N' 1....
",z;===-l'( --- -i I ..L ,:..
N (1-- C-"
2 --T)
so the follo\ving slgn con vention results:
fIlL NU'L == -- Nx',t.
'I'he equations and boundary conditions novv take the forIn
(4.45 a)
(4.45 b)
cr==o,7t,
\vhere \\'e nlust continue to bear in Blind the orthonornlality conditions (4.42).
7he solutions of (4.45) IZd've 710 JJlea71ing unless they are cOllsistent
l'itlz tIle constraints.
It is \\'orth\vhile to ren1ark that in1posing alone docs not fix
the gauge completely in thc \vay that (4.36) cIoes. By 111aking a rcparanlctri-
zation
such that
(4.46 a)
'1-' ':J-'
ocr C''t''
d(1 J-r
':\'""
2;; 0(1
J";
'\
ocr
\vhere
=== 0 ,
- 59-
\\'C 111<1 Y prcscr\'c (4.42 ) ,
\vhere J'P' === .y1J. C:- (7 , a) , a (T ,G )). l"hus \ve ll1ay rCll1ain in an orthonortl1al
gauge \\'hile ll1aking a different linear cot11bination of the x
P
' (-:- , a) proportion-
al to -:- in (4.36 a).
C. DIRAC BRACKETS
\\Tc no\\' proceed to develop a set of Dirac hrackets c0111patible v\'ith the
prin1ary constraints (4. I 3), the gaugc choice (4.36), and the boundary conditions
(4. I 2). Only then can \\'C take (4.45) to be strongly valid. "le first ex-
tend the canonical \'ariables to all \'alues of their argull1ents defining a
nc\v periodic Poisson bracket
(447)
\vhere
and
{ t (1" , a) , XV (1" , (j')} == _ g' t V (a , a')
(a , a') r (a , cr')
00
(0 , a') == 2: [0 (a - a'
11=-00
2 U1t) 0(a + a' + 27Z1t)]
0/
()
da (a, a')f(a) == f(a') , o a's 7t .
\\le \,vill need the follo\ving properties of ,
(0' ,0") _ 2-1_ (cr , cr')
---d-O'- - -
and
-

at a - 0, 1t
-

at a'== 0, it
L-/- 2cr'
-

at
(J
-

, it

2
=-1=
at 0, a - 1t
l'hus \ve see that (4.47) is COll1patible \vith the boundary condition (4.45 b).
- 60-
()ur full set of constraints is no\v
CPl (a) === f!JJ. It 0
cP1 (a) === I. . x 'TA . !)IN1t 0
rrhe Inatrix of Poisson brackets is
(4.50 ) Cij (a , a') === { CPi (0") , CPj (a')}
o
o
o
o
Ap
o ". tl(cr,cr')
T: C(J
o
A'P d
TI: -6cr'..i (cr , a')
') '" 1)
... A A ( ')
---.:.....1cr,cr

oJ (' A ( , I
L1\(J,cr) - .....
I",
o
o
o
o
I-Ierc \Ve ha ve 111ade use of the constraints (4.49) after C0111puting the lH-ackets,
as \VC are al ,vays entitled to do.
To the Dirac brackets, we need the in matrix Ci/ (v , (;')
defined by
.J cIcr" C
lk
(cr , cr") C,;/ (cr", cr') = 1'1 (v , v') =
o
(a , a') 0 0 0
0 L1 (a , a') 0 0
0 0 L1 j (a , a') 0
0 0 0 (a , 0")
I'he appearance of in the first diagonal entry of the generalized unit 111atrix
Iij (a ,a') and in the ren1aining positions is dictated by the boundary
conditions. If \ve differentiate (4.5 I) \vith respect to a and evaluate the result
at a == 0 or 0' == TC, Eqs. (4.50) and (4.48) imply
d J
1
" d "C ( If) c 1 (" ') J I (') L
Ja a 14 a, a 41 a ,0' i (j = 0,;: === :}cr 11 \ () ,a (J = 0,;: =j 0 .
The corresponding expressions for the other diagonal elements of I
ti
vanish
and they accordingly 111ay be set consistently equal to L1 (a, a').
To determine Ill' \ve first observe that the solution of
Jcicr" C
41
(cr , cr') Ci"l (v", cr') = 1
44
(cr , cr') = (cr , cr')
- 61 -
IS
C
- 1 ( ') 1': D ( ') 1': ( ') cr )
/41 a, a === ""), ' P a, a ===), . p a, a 1':.
Here the generalized step function 0 (() , a') 0 beys
a > a' }
, 0 :s;: a :s;: 7t ,
a<a
8
-80- 0 (a , a') === (a , a')
o(a , a') === I
o(a , a') === 0
while the periodic sa\vtooth functions
o :s;: a' < 7t ,
(4 55)
2: (a , a') === I", 1: (eiJl(o-O') ein(o+o')
2 Ttl n=fcO n
2: i- (a , a') === 0 (a , a') - :
ha ye the properties
'\ - ,-, (' )
f- a, a ) - "'-J_ a, a
(4,56)
2: (a + 2 7t , a') === 2: (a , a')
2 2: (a a') === +- (a a') -
dcr ' " 1':

c ( ') _ /\ ( ') _ c (' )
..;...J_ a, (j - L...l_ a, (j - - '1 ",;;,..,j i_ a, a .
ccr ' ocr
But no\v Eq. (4.53) and the requiren1ent that
- 1, -1, Tt ( ,,, cr' )
C41 (a , a ) == - C14 (a, a) == 2:__ (a , a ) - -i:-
Tell us in1mediately that
III Ccr , crt) = Jdcr" :0' Ll, Ccr , crt) (L:_ Ccr", (J t) -:) = Ll_Ccr , crt).
rrhe other eletuents of the inverse matrix are
Ci2
1
(a , a') == - (a , a') == 2;:' P (a , a')
(4.5
8
)
1t
C
-1 ( ') - Tt C-1 (') I Jd "C-1 ( ")
13 a, a - 2),' P 14 a, a - a 14 \ a , a
o
Tt
2
( ')
-2-(A-,-P)-2 ",;;,..,j i- a, a
1 1 Tt
2
C3i (a, a') === - Ci3 (a', a) == .) 2:_ (a , a') .
2 (A, Pt
62
The final result for Cij
1
is therefore
C
--
1
( ') 1t
'zj' a, a == A' P
o
o
x
_ 2: a')
2A'P _v,
2:;_ (a , a') _ (1'
7t
o
o
(a , a')
2
o
(a, a')
2
o
o
f (a , a') + :
o
o
o
Equation (1.58) can no\V" he used to con1pute the I)irac brackets compatible
\vith the constraints (4.49). 'I'he results are
(4.60 c)
- .)
{x!J. (00) , X" (a')}* = u!J. (a) ['Lv -t\ g;v (a')] .l":j- (a , a')
+ A [AI' - g;1' (a)] U
V
(a') (a , a')
+ NA P (AIL f!jJv (a') _ A
V
f!jJ!L (a)) ,
{ g;" (a) , XV (a')}* = (-g!J.'1 iJjJv (a')) (a , a')
+ [- .1:
2
&>!t (cr) (a , a') (--- A2 .qjJ" (cr') I,v ), . P)
(A . p)2 it
)"
2
1t'2 N 2 u. ')' , 't'N lL ,,]
- . u (cr) u (cr) '"-J_ (a , u ) - It' (a) A .
(A. p)2 A'
The constraints (4.49) are strongly valid in the Dirac bracket system.
We may therefore solve the equations of motion (4.45 a),
(
82 3
2
) u. 8
2
(,,: , a)
, .) -- -.,. x ('r, a) ==-= 4 "( t ) 3()== 0
, C"';" / 0 "r - (1 't'--- (j
in the orthonornlal systen1 subject to the boundary conditions C4.45 b). rrhc
n10st general solution consistent vvith the reflection principle (4.22) takes
the forn1
;tJl (- a) == (Qf
l
(- -1-- a\) -1- Qll (7 -- u))
, \., :2 \ \ ..
- 63-
\vhere \ve 111ay identify
as the coordinate of the end of the string. rrhc momenta arc
\vhere
O
'f-t (Cl) =--: (8) _ ",1 /Ji)f-t (8 )
""" v- d8 -N<7 ,0.
(Q"l is often \vritten " })11" in the literature; \ve reserve this syrnbol for the
translation generator (4.29).) By cxanlining
\ve find that the boundary condition
requIres
Q'!l (0 + 27t) -- Q'P (8) == 0 .
Next, \ve notice that the four constraints (4.39) on X!1 and can be ex-
pressed as constraints on Qll ,
1(1 == Q,2 (0) == 0
.., P
"',') == /... 0 (0) - 0 == .
-- NT:
ate also that
... 0' 0\ I, . P
A, (') ==='---;----
'-' / .i\ it
rro find the brackets of Qll (0) c0111patible with the strong constraints
(4.69), \ve insert Eq. (4.62) for xlJ. c-r , (1) into both sides of the Dirac bracket
(4.60 a). i\fter identifying ter111S of the satne functional fornl on hoth sides
of the equation, \ve find that
-( QIl (0) , Q" (O')} === - g!J.'1 E (0 - 0')
\)
-t- (), . P)-l Q'i{(O') - 0' A" (0)]
(A
Y
Q'f' (0) (A P) Ai' Q'v (0') (), . P) .- A
2
Q'i' (0) Q'Y (0')).
- 64-
Here we have defined the functions
p (8) === ~ _ ~ I eillO
11. n=:pO 7l
(4.72 a)
E: (8) === p (8)
8
\vith the periodic properties
p (8 + 2 1t) == P(8)
2 U1t)
(4.72 b)
E: (8 + 2 1t) == E: (8) + z
~ (8 + Z 1t) == 0 (8) .
One may verify that all of the brackets (4. 60) follow from (4.7
1
), (4.
62
)
and (4.64). The brackets (4.71) are also con1patible \vith the constraints (4.
6
9).
In fact, it is alTIusing to note that if the " canonical" bracket of Q ~ L (8) is
taken to be
{ Q ~ (8) , QV (at)} == -R- g!J.') E: (8 - G
t
) ,
Eq. (4.7
1
) is exactly the result one gets by using the Dirac procedure with the
constraints (4.69).
Examining Eqs. (4.60) or (4.71), one sees that the Dirac brackets look
like a plausible basis for a quantunl theory provided
).2 == 0
It" === 0 for the D-2 independent degrees
of freedoll1.
\Ve are free to choose the gauge in such a \vay that the theory is as simple as
possible. Therefore \ve take
(4 75 a)
i == I " . " D-z
and define also a c()1nplementary vector I,,: with
A ~ === - !"r)---l === - 1IV z
(4 75 b)
so that

(4 75 c)
)..... A* == 1 .
- 65-
Hereafter, any vector be split into a transverse part Y i, i == I " .. ,
D-2 and
-)(-
(4 7
6
)
y+ == J,P. == (\7"D- 1 \TO) /V
2
== y_
y- == 'A:V!l == (V
D
-
1
- VO)/V2 == y+
In accordance \'lith the null-plane conventions listed in Appendix A.
\:Vith the gauge choice (4.75), the nonzero hrackets in Eq. (4.60) becon1c
{ gi (a) , xi (cr')}* == - azj (cr , cr')
{ x+ (cr) , x- (cr')}* == - -J;;
{x- (cr) , x- (o')}* == P+ [u- (0) L:+ (cr , a')
IT
(,0jJ- (a
f
) - ,cr (a))]
{
- () ( ')}* - ( ( , \ ") ( !'
X 0, X \ cr - p + \U a) \ 0 , (j )
( /Jl;- () (}lii ( ')}* _ N 2 rr: i ( ) dL1 ((J , (J')
l. ;:r a,;:r a - - P+- It \ a --8-;'-
(!)Ji (a'))
Nrr "
{ eJ- (a) , x- (a
f
)}* = -;-i-' [,OjJ- (a
f
) Ll (a , a
f
)
+}cr (0jJ- (a) L_
1
(a , a
f
) -- u- (a))]
( c)/J- () _i ( ')}* _ rr: (}1J i ( ') A ( ')
l.J 0 ,x \a - p+J cr Ll o,a
For reference, \ve gi ve also the nonvanishing brackets (4 7I) \vith the gauge
choice (4 75):
(4 7
8
)
{Qi (8) , Qj (O')}* == - oij S (0 - 0')
{ Q+ (6) , Q- (6
f
)}* = - :-i
{Q- (0) ,Q- W)}* =F-i-- [d- W) (p (0 - Of) :) +
+ ci (6) (p (0 - Of) -
{ Q- (0) , Qi (Of)}* = ;+ Qf
i
W) (p (0 --- Of)
- 66-
l'he action of the translation group generators (4.29) IS novv
(4 79)
{ P-, x- (cr)}* == ,q;- (a)
{ P
- i ( )}* _ rr tJ]J i ( )
,X "cr - p+ <:T cr
{ pi , x
j
(cr)}* == _
\vith all other brackets vanishing. Since
Nrr
'T == -p-.+- x+
and p- is the canonical generator of x'+ displacements, \ve may solve the con-
straint Eq. (4. 13 a) for to gi ve the generator of 7-displacenlcnts,
p+ p-
I-I == -
.:\rr
..
2
1
:\ rdrr (9. & +N
2
U u) .
()
We veri fy t hat Eq. (4. 80) gi \TCs I-I atl1i1ton's equat ions 0 f 111() t i() n In '7 fo r the
independent variables Xl and f!J/:
(-t. 81 )
\Ve next examine the Lorentz group generators (4.3 2 ). rfhe rotation group
generators l\l
ij
transform Xi as a three-vector.
{ M
ij k l.* _ i _ i
1 ,Xj-OX oX
'[he boost generator 1V10
i
, on the other hand, adds gauge transforn1ations to
the variables so as to prcserve the constraint Eqs. (4.36) in thc nc\v Lorcntz
fran1c. We find
{ M
-i ( )}* _ - ( ) + rrN dxJ (0') ( ,. () _'r
Pi
)
1 ,.:t cr - -0 x cr cr -1-- Nrr
7t
+
!f.j (cr) Jf" d 8x
i
('"r, x) (
P+ 3cr lJ. d'"r \ (X , cr) .
o
'rhe first tern1 is expected from Lorentz covariance, while the rernaining terlns
arise froln the fact that a gauge transforlnation (a (7, cr)-reparan1etrization)
must be Inade to preserve the gauge in the new Lorentz fran1e. If the infinitc-
sin1al Lorentz transformation is written'
p'!!' == p,L
- 67-
then the change in x!-L (7 , a) under a Lorentz transformation will invol ve also
an infinitcsinlal rcparanlctrizatioll 7 , ('j ;; (7 ,a), G(7 , a):
",.rJ. ('-. " ",.'(1. (.-r- == ",-!1. (;::. (.- ;:;. (.- + ,.\!1.') 'V (.-
,v " .. , V) -r A, \, , u) A I",,, \ .. ,v ,v \," , V) UJ .t"J ", v
X\J ('7 , a)
l'he value of -17 is fixcd by requiring
-r
(P-t I +" I) )
(t) v'
Equation (5.85) then immediately gives
Once Ll-:- IS fixed, Lla is con1pletely deten11inecl by the requirenlent (4.4
6
)
that the orthonorn1ality of the coordinate systen1, Eqs. (4.42 ), be preserved
by the rcparan1ctrization,
(4. 88)
,
0cr
]'he solution of (4.88) IS
a
N1t(u+
V
(' aPv I" , ( ,\)
== ----.'--- - da 'l' I '1" a)
P+ l\rr 'J\ ,
,
o
(I
TC
- - J!" d ',., (.- ') ( ')
-- p + . (j (Jv \ ,a + (J ,a .
o
Thus when a I.Jorentz transformation is made, the required change of x
tJ
('1" , a)
\vhich preserves (4.86) and (4.42) is
x'p ('1" , a) == x
P
. (-r , a)
U.'J ( )
(0' Xv '7, (J
N1tw+
V
J
- /til- (-r , a) da' Vv (-r , a') L:+ (a , a') ,
o
Vvhich agrees \vith the result (4.83) given by the Dirac brackets.
- 68-
D. FOURIER OF \TARIABLES
\Ve begin the Fourier analysis by observing that the a-a veragcd coordi-
nates and n10Inenta of the string arc
n TI
q" = : JdG Xl' (0 , a) = 2
1
7t JdO QI' (0)
o -n

pi' = Jda &pI' (0 , a) = N Jda v" (0 , a)
o 0
=== N (TC) - Q't (- TC)).
Then "\ve Il1ay \vrite
vvhere one Ina)' deduce fronl (4.68) that I'" is periodic,
Then, [ronl (4.62), the canonical coordinates x
P
' can he written as
fl.. ft I 't' pf" I fl ft .-
X (7, a) == q -;---Ni- +--2 (1 ('"r -t- u) +j (w -- a)).
Next, \ve expand 1
ft
in harn10nic oscillators using the DDF Fourier
C0111pOnents (-1(-) (Del C;iudice, Di \Tccchia and Fubini, 197
2
)
(495)
TI n
a;:, = IdO Q/f' (fl) exp [im ),. Q(fl) Nn/), P] JdO Q'" (0) /mO
TC'
I rcia (.qJI' (or , a)
2 J
-n
It (.- \) im(cr-t--:)
l'U \",U) e .
Fron1 (4.92), we see that Inay be \vritten
7t
!t _ pIt --_. N J' dO'jf
l
(0') JmO
I
am - ( m,O l1Jl..l. 2 {;
-n
(*) A modified fonn of these oscillators can be used to quantize the stri ng even \vhen
the gauge is time-like (Goddard, I-I a.nson and Ponzano, 197..).).
- 69-
1\1ultiplying by e-
im
o/nz and sun1ming over 17Z
to find an explicit expression for f!L:
0, \ve n1ay ell1ploy (4.72)
f
It (n\ _ i "l I IL - illl 0
\J) - -_ .. ---- am e .
1\ 1ll-!: 0 17Z
Thus the c0111plete Fourier expansion of x
P
- is
and &It == N 2x
lt
/27 , u
lt
== 2x
l
"/2cr.
Not all of the Fourier components q''', pIt and are independent. sing,
for exa111ple, the expressions of the Fourier components in tern1S of Qll and
the Dirac brackets (4.71), we find the following nonvanishing Dirac brackets,
{
It V}* _ . N / It V
am ,a
l1
- - z 7t lJzg 0m,---ll
{p'+-, q-}* == - I
{pi, qi}* == ()l)"
r p- -}* _ p-
I.. ,q - p+
( 1>- 1'}* == pi
I.. ,q p+ '
\vhere It IS the gauge paran1ctcr (4.75) and \\'C bear 111 n1ind that
'rhcse brackets arc c()1npatible \vith the strong constraints (4.49) or (4. 69),
\vhich translate into the follo\\Ting constraints on the c0111ponents:
qf- ==
T'he expression for a: in terms of the other variables can be deri vcd, for
example, by solving the constraint Eq. (4. I 3 a) for ,q;-,
-70 -
expanding g>'" and u'" in the Fourier cotnponents (4.98), and identifying the
coefficients of cos ncr e-
iwr
.
Now \ve may take as our independent variables
Z =-== I " . . , }) ----- 2 ,
Z === 1,"',1)--2
1l == -- CX),"', CX)
\vith the nonvanishing brackets
(I)i .i}* _
l ,q _.. - 0
Using (4.101), \ve easily confirn1 that the brackets (4.99) directly. Note
that
In the literature, one often sees the notation
(4. 106)
p+
2! ==--
m NT':
\vhere f,Jl obeys an algebra z'sonzorphic to the \/irasoro algebra (4.26) (Bn)\;ver,
197
2
):
( UJ (/ 1* __ . / . , ) (/:J
l'x
1Jl
, c-Z
ll
f - - Z (Jlt - n J..
1l
-
1
-
m
.
rrhc in1portant difference, is that fE,n is Jlollz1anis!zillg, \vhile the \Tira-
soro operators vanish in the Dirac bracket algebra.
We tnay no\v usc the independent variables (4. 103) to \vrite explicit
expressions for the canonical variables and Poincare group generators of the
full y constrained systetTI. The canonical coordinates are
(4. 108)
i ,-, I .
----- .t:..J - a- cos ncr
Nit II FO n Il
i I -il11'
..J am an - m COS ncr e
2N1tP+ n=l=O allm n
-71-
2: [ama_- m
JJl
I ) -, !: _-- /Il '7
+ ...,. /-A (71/ COS l!o e :
I. IILU
\:\Tith S0111e labor, one 111ay check that these expressions for and repro-
duce the Dirac bracket algebra (4.60) \'lith the choice (4 73) for )t.
rrhe Fourier decomposition of the Han1iltonian (4.80) is
I-I ==
00 (
I -., I ')
);_ ..2: am a_-
m
== -- p'"
J. lJlc.:=-N I. \
2 amoa '11)
lJl=-=l
and continucs to generatc the cquations of 111otioll (4.81). ]'he I--torcntz-
in\'ariant mass-squared can be expressed In the form
ex:>
1\1
2
== --- 2 p+ P- - p2 == 2 2: am' a_
m

111=1
rrhe translation generators are P+, 1
Ji
, and the dependent \'ariable P- gi ven
in (4.101). lTsing the definition (4.32) of the Lorentz transforn1ation generator
l\I'lV together \vith (4. 108), \ve find
1\
-1- If: I pk /... IJj
1 i' == q' --- q
Ivrr /..: == _ qk p+
IV1+ == - q- 1)+
lVr-
l
. =--== q- 1)1.' __ qk p--
\
Ill)
In the classical systen1, the Poincare group algebra (4.34) continues to hold.
rrhis question is n10re subtle in the quantuln n1echanical systelTI, \vhich is
treated in the next section.
IV1ECI--TANICS
]'he quantunl mechanical systen1 corresponding to the classical string
rnodel described in the previous section has been treated extensively in the
literature (Goddard et aI., 1973; 1974). We vvill give here only a
brief summary of the results.
-72 -
The follo\ving variables can be taken as operators \vith con1111utators corre-
sponding to their Dirac brackets, using the convention {A, B}*-7 I; [A, BJ:
z
[
i k] N ~ i k ~
am ,a11 === I 7t 1no 0l1,-m
i [P+, q-] == I
i [Pi, qi] === () zj'.
i\JI other commutators vanish, with the prOVISO that a === pt".
The dependent variables are
P
- _ - + N1tllO _ I
-an ----
p+ 2 p+
(
: 1; a
m
o
a_
m
:-2N7tiX
O
)
111=-00
a,l. ==-
2 p+
00
~ aman - m
111=-00
N1t if
p+ 11'
12 o.
Here the colons : ( ): in P- lnean that the expression is to be norn1al-ordered,
\vith the destruction operators a ~ \vith 1n > 0 al\vays on the right. Since !J-
contains noncommuting operators, \ve may need to acId for consistency a
zero-point energy which \ve haye \vritten as N7tcxo/P+. ]'he Lorentz-invariant
mass operator is thus
The requirement that ao be expressed in norn1al ordered form causes
the C0111mutator of canonically conjugate pairs [ a ~ , a=m] to pick up a
Schyvinger term vvhich \vas not present in the classical theory.
Other COmlTIutators are unaffected:
(4. I 16)
-73 -
'-fhe Ha111iltonian is no\v
p+ p--
H == ----
l\T:
and generates I-Ian1ilton's equations of 1110tion in '7 for the dynatnical varIa-
bles. \\Te 0bserve that T is dimensionless and
(4. I 18)
3A(--r,cr\ 8A
. [H 1,'] pl/NT
Z ,q == li1t'
due to the constraint on x+.
equations of motion
We find that the Fourier con1ponents have
so that the fields x
k
(T , cr) and 1110111enta g/.: (T , IT) obey
i [H , x
k
] == g;kIN == 0'1/':
i [H , g;k] = N -::2 x
k
(-r , cr) = :jk.
l'he I110st unusual feature of the quantized string is of course the fact
that the L,orentz group algebra does not close unless certain conditions are
111Ct. \,re \\Till not go through the calculation here, but 111crcly state the result.
One finds the because of hern1iticity and norn1al ordering requirements, the
comn1utator of the quantu111 operators 1\1/ - and :VI
J
- takes the for111
i [M
1
-, M/] = - 2 1; [172 (1 --- 2
1
(D -- 2))'
111=1 4-
I (1 )] i j j i
- ----- (D - 2) - Cl.
o
(a_
m
am - a_
m
am) .
'J7Z 24
l'his C0111111utator Inust vanish if the Lorentz group algebra is to be satisfied,
so \ve conclude that
D == 26
(Xo == 1 .
'-[here are various n1ethods of relaxing these restrictions, but the nor11'1 of some
state al\vays becomes negati ve for D > 26. No clear intuiti ve reason for such
a phenomenon is understood at this time.
5. lVIAXWELL
The classic exan1ple of a theory \vith an invariance under an Abelian
gauge group is l\1axwell's theory of electrolnagnetism.
-74 -
I-IAl\1ILTOi\IAN \VITI-lOUT C--;'AUGE CONSTRAINTS
rrhe application of the Dirac nlethod to electronlagnetis111 has heen cxa-
lllined by Dirac (1951, 1964). (Sec also Anderson and Bergn1ann, 195 I, and
IZundt, 1966). \\Te shall begin by considering a field A!L (t ,x) transforlning
as a four-vector on the index [.1 and possessing continuously infinite degrees
of frecdonl labeled by the spatial coordinate x, or sonle suitable cOlnbination
of the :t
Y
depending on the type of d ynanlics chosen. The 1110St general fornl
for the action is assulned to be
s = rd
4
x2(A"(x) , 2Al'(X)(2x').
N O\V \\'e \vish to nlake ...-\!L descri be a 111assless \'ector field, \vhich has on1y
t\VO independent degrees of freedonl, not four. 1'he tinle-tested nlethod for
doing this is to alIo\\'" 2! (up to total eli vergences) to he a function onl y of
has the \'irtuc that it IS unchanged by the spacetinle dependent gauge
transfornlation
'rhe freedo111 to choose .:\ (x) Inay be used to clirninate the unphysical C0111-
ponents of (x). Note that with the convention (5.2), the usual E and B
fields are
13'" ===
I '"/; '1 . OJ;
. slj " .F'.J
f
< === slj"
:2
=== (V X A)1 .
\\1estill ha vc SOI11e latitude a vailablc in choosing the Lagrangian; one
unconventional forIn \voldd be the Born- I nfeld electrocl ynarnics
(55)
2===
\vhich agrees \vith the l\/Iax\vell theory only in the \veak-held lin1it (see Dirac,
1960). In \vhat follo\vs, ho\vever, \ve shall be content to in\'estigate the usual
l\laxwell Lagrangian,
(5. 6)
in Lorentz-Heaviside units.
if === - F
4 (1.')
-75 -
\\Te first nlake the traditional choice of slJecifying the cl ynanlics on con-
stant-tiI11c surfaces. the canonical I110nlenta are
'I'he spatial conlponents of nIL are
(5. 8)
:r (x) == -- E (x) == A-t ,
\vhile the tinlC conlponent vanishes \veakly, gi \-ing the prl111ary constraint
(59)
D( \ evO
n X) I"V
rrhe canonical I)oisson brackets of n'" and l\.v arc given by
(5. 10) {
-.T''' ( ) _\ ( )" ( __ '
II. t, x ,- ') \t , Y J - Ov 0 X Y)
and are ob\-iously i nconlpatihle v.:ith (5 9)
\-ariation of the fi.elcls ... r\" (x) gives the usual equations, \vhich
nlay be only \veakly valid:
(5. 11 )
') CiS!? == F,t", == A..",!l __ L\'l," () .
L ') 8 (c\) .A.. pJ 'J'" ') ... - ')
The zero C0I11pOnent can be \vritten
(S12a)
.A) o.
LTsing (5.7), \ve this in the for111
(SI2b)
y.;r

\vhich is indeed a \veak equation because it is inconlpatible \vith the l)oisson
brackets (5. 10).
Let us no\v begin to apply I)irac's approach to our syste111. First \ve recall
that the lVlax\vell Lagrangian (5.6) is invariant under the group of
spacetinle transfornlations [in fact, it is in variant under the full confornlaI
group of \vhich the Poincare group is a subgroupJ. \i\1e list belo\v the effect
of each Poincare group transfornlation on _y'L and the held (x):
(SI3
a
)
(5.13 b)
where
I f Il (ItB (3 Itx\
== _.lorentz trans orm: ox' == (U
x
i3 g' 1 -- ,,'tot g' )
;\!L _ [( /X Af3 _ . Jj I_X),"). Afl . i.. !lC( 1\[3 _ lJ.]
- (.0:x[3 \x g X g '-I. T g... g
[
lJ. ALL [3 All. ex + ,,----ex[3 I.ll A'V]
== (.0.:x[3.X ',-- x ' , k.i),'J g I
- 76 ---
Note that the tensor indices of ... t\'" transforn1 oppositely to those of x'" since
it is the inverse transforn1ation which relates the old tensor indices to the new
coordinate systen1.
.i\ canonical conserved Noether current is associated vlith each transfor-
lnation. The canonical currents and the corresponding conserved Poincare
group generators are given in the equal-time convention by
and
(5.14 b)
t{lf.!lV _ . elf.V _ O!f.[L __ 2:[1;V\O_
.A - X c C ClAj"lf. Ao'" -
== -xv l;lf.A ...
NO\V \VC take the canonical 1-1 an1iltonian
(5 I 5)
LI -- 1)0 - I" i
3
0
00
- { i
3
( !l\
r c - - (. x c - (. X n ... lJ..
p
..
.. '
and add, according to Dirac's prescription, a n1ultiple of the sole prln1ary
constraint nO o. We get
(5. 16)
H
O
= He .1 d
3
x VI (X) nO (X) =
- rd
3
(I 2 -+ I B
2
n i\ 0 I 0( ')
- .. X 2:T - 2- - :T' V ..tl.. I VI _1) it :l); .
where \ve have used the identity
\Ve look for secondary constraints by con1puting the bracket of I-I
o
\vith the
pritnary constraint nO 0 and requiring the result to be 0 for consistency:
(5. I 8)
Our t\VO (obviously first class) constraints are thus
== nO 0
== \7 . :T 0
-77 -
We no\v add both and to the Hamiltonian with arbitrary coefficients.
Using (5. 17) and doing an integration by parts, \\Te find

H = He Jd
3
X (VI nO V
2
V. :r) =
= rd
3
x +B
2
vI nO (V
2
+ AO) V. :r) -
- J ds :r A
O

Soo
I-Iereafter, we \vill neglect surface integrals such as the last term In (5.20).
By evaluating the brackets
(5.2 I)
AO == {AD, H} == -VI
A
a
__ {,, H} ",0, ""1/""1"-
== .... ""J.., == 1t -.l"""'l. - .:x
\ve find that \ve can set VI == - ... == AO,D and V2 == 0 to give the final result
(5. 22)
that AO is a basically arbitrary function in the Hatniltonian which is
elin1inated fron1 the equations of n1otion only after \ve choose a gauge. 'fhe
last t\VO tern1S in (5.22) are in fact the generating functional for infinitesin1al
gauge transforn1ations with i\ (x) == ...t\0 (x).
l'he functions VI and V2 are not necessarily the san1e for all components
of the energy-n10Inentun1 tensor. \Ve 111ay in fact write
Arguments parallel to those which led to Eg. (5.22) give
p = fd
3
x(-:r X B + nO VAo A(V :r)),
so
We see that Eqs. (5.22) and (5.24) are the appropriate modifications of Eg.
(5. 14a) to generate the spacetin1c translations of the theory \vithout gauge
constraints:
(5.
2
5)
{P
It AV()'l. .
, X j ==.: ...
The other generators (5. 14) of spacetilne sytnn1etries ll1ay be treated In a
similar fashion.
-78 -
I). RAI)IATION GAUGE
V\re no\v sho\v that the Dirac brackets in the radiation gauge lead directly
to the traditional quantutn-Inechanical structure (e.g. Bjorken and l)rell,
19
6
4).
Our objecti ve is to use the gauge freedorTI (5.3) in our systenl to fix t\VO
conlponents of .i\f.t so that the first class constraints (5. 19) beCOlTIe second class.
Since ito 0, one logical choice is to set
0
o. rrhis is accoll1plished hy
the gauge transfornlation
(5. 26)
XO
( dt AU(t , x) ,
o
for \vhich the infinitesinlal generator takes the forIn of the first class, \veakly
vanishing addition to Eq. (5.22):
t 3 /r i
3
" fl '"'\ 1\
L[Al =.1 d' x(A(x)nO(x)-A(x)V . (x =. { lIn ell,!
No\v the Euler equation (5. I 2) becolnes
(5.
2
7)
\7. A' 0,
so that a second tinle-independent gauge transfornlation can be ll1ade to
fix A'. ]'he radiation gauge, \7. A" 0, is a con\Tcnicnt choice cOtllpati hIe
\vith (5.27), and is achieved by lnaking the gauge transforll1ation
A' A" (x) = A' (x) + Vx ( d
3
y 4 IT ; _ y V y ' A' (xo , y)
(5.
28
) .1
\ 0' !\ 0" ( ) I d r/3 I n A' (() )
11.. == 0 -+ ....l.. X == 0 -1'"'\ {y I IVy' x, y == o.
oxo 4 T: X - Y
Here A' is given by (5.26) and \72 (I/4n I x - y I ) == - 0
3
(x - Y),
\vhile the vanishing of ... ;.0" fo11o\vs directly fron1 (5.27). Dropping all pritnes,
\ve find the set of constraints
C?l == nO 0
CP2 == V . 0
C?;{ == A
O
0
?4 == VA 0
In the ne\v gauge.
The matrix of Poisson brackets of the constraints (5.29) IS
o
o
--1
o
o
o
o
o V;
o 0
o -\7; 0 0
- 79 --
Inlposing the boundary condition that the fields vanish at cx), \ve find that
the in \-erse of (5.3) exists and takes the fonll

0
__0
3
(x --- y)
0
0 0 0
(5.3 1) Ci;.1 (x ,_v) ==
47":
x--y
0
3
(x - y)
0 0

0
0 0
47t"
1'hus the Dirac brackets may be conlputed frOITI Eg. (1.58) to be
{rc
fl
(t , x) , A
V
(t , y)}* == (- gf
lV
_ g'lO gVo) 0
3
(x y)
{
,t . V}* _ {A!l A
V
1.* _
7t,1t - , j-O.
Equations (5.3
2
) are COlnpatible \vith setting the constraints (5.29) strong-
ly to zero. 1'herefore, only t\VO of the and t\VO of the nz" are independent
variables. \ve n1a y sol ve \7 . A == 0 and \7. == 0 for the rerrlaining
t\\TO dependent variables if required, this is unnecessary; in the process of
in verting the nlatrix e
ij
to fInd the I)irac brackets (5.32), \ve ha ve autolna-
tically given the dependent \-ariables the SaI11e hrackets they \vould ha \-e
\vhen expressed in terI11S of the independent variables.
\'7e now verify Hanlilton's equations of I110tion in the fully constrained
system. 1'he Hanliltonian is just
(5 33)
1'hus the equations of ITIotion are
(5 34)
{A(x) , H}* == .1 (x) == A
, H}* == - \7 X B== - E.
In deriving Eq. (5.34), \ve have assunled that E vanishes sufficiently rapidly
at CX) to drop a surface integral and have nlade use of Eq. (5. 1 7).
Using Eq. (5.14), \VC find the renlaining Poincare group generators to be
I"
P = .J ti
3
X(-1/ VA')
(535)
- 80-
In the star brackets, pIt and :YIlt'V continue to obey the I)oincare algebra (2.25).
Under the action of translations and rotations, the ..:-\i transfornl as
{pi, Aj (x)}* = = 8
i
Ai
t
{lVIii, A
k
(x)}* == (Xi 3
i
- x.i 2
i
) A
k
(x) -t- a
ik
Ai (x) - a
jk
Ai (x) .
usual, surface integrals have been dropped in Eq. (5.36). Under boosts,
however, extra tern1S occur \vhich cannot be neglected. \\7e find
(537)
rrhe extra terlTI is required in order for lVI()z" to have vanishing star brackets
\vith V A. It nlcans that under an infinitesinlal boost paralnetrized by
(VOi' ..:-\,t (x) undergoes a gauge transfornlation of order (VOi in order to keep
...t\0 == 0 and V A =--= 0 in the ne\v Lorentz franle:
rro sho\v that A'o == 0, \ve have used the fact that \vith our gauge choice
(5.29), the Euler equations (5. I I) reduce to
and integrated by parts.
c. ...J\XIAL GAUGE
.i\nother an1using gauge \vhich can be treated in the canonical forn1alisn1
is the axial gauge (Kumnlcr, 1961; ..:-\rnowitt and Fickler, 1962). We begin
by using the gauge invariancc of the theory to set
We see that if \ve choose 'V
2
== 0 in (5.21), then Eq. (5.40 ) \vill hold
for all tin1es anI y if
Equation (5.41) is sinlilar to a secondary contraint follo\ving from the
gauge constraint (5.40); it replaces the condition V A 0 used in the pre-
- 81-
ceding section. The full set of constraints is then
(5.4
2
)
CP2 == \7 .:r ~ 0
CP3 == A
3
~ 0
m 3 A
O
,3
T4 == n - ~ o.
The Poisson bracket matrix is
0 0 0
(8/8x
3
) 03(x_.y)
0 0 --(8/2x
3
) 03(x- y) 0
0 -(3/3x
3
) 03(x-_y) 0 + 03(x-y)
(2/2x
a
) 03(x-- y) 0 -03(x-- y) 0
We choose boundary conditions so that the inverse of (5.43) is
0 -g (x, y) 0 f(x, y)
eti
l
(x , y) ==
g (x, y) 0 - f(x, y) 0
(5 44)
0 - f(x, y) 0 0
f (x , y) 0 0 0
\vhere
3g(x ,y) ==f(x, y)
dX
3
and
(5.46)
g (x, y) == ~ - 0 (xl_yl) a(x
2
_y2) \x
3
- y
3
1
f (x , )1) == ~ - 0 (xl - yl) 0 (x
2
- y2) E(x
3
- y3) .
Here E (x - y) == algebraic sign of (x - y).
When we insert Eq. (5.44) into the Dirac bracket formula (r .58), \ve find
(5.47) { n ~ (t, x), n
V
(t, y)}* == 0
(5.4
8
) {i\'l (t , x) , ~ ~ v (t , y)}* == (glLOg3
v
-t- gVOg3
1l
) f (x , y)
(5.49) {nit (t, x), A
V
(t, y)}* == (_gllV _gllOg"O) 0
3
(x- y)
It3 d f ( ~ )
+g -3x
v
x , Y .
6
- 82-
The nonvanishing brackets are
(5.50 ')
{ 7t
1
, A
1
} * == {7t
2
, A
2
} * == - (x - y)
{
7t3 A1}* == -t- ,yt
, dX1
{
7t
3
A2}* == -+_ 3/(x,zL
, d.t-2 ,
so 7t\ 7t
2
, A
l
and ... t\.
2
still obey the canonical brackets (5. 10). brackets
are 111anifestl y con1patible with all the constraints (5.42) and can be used as
the hasis of the axial gauge quantization schen1e.
'rhe Han1iltonian in the axial gauge is
\vherc \ve have set the constraints (5.42) equal to zero. i\.
0
is a dependent vari-
able found by sol ving the equation
, 2
L'2 7t
to gi ve
All (x) = - I Ix
3
- I (-\ n
1
(tx
1
x
2
+ 2
2
n
2
(tx
1
x
2

oJ
Using just nl, 7t
2
, A \ A
2
as the independent variables, \ve easily confirn1
all of the brackets (5.5). appropriate boundary conditions, \VC
find the Han1iltonian equation of n10tion
{7t
i
, I-I}* == - (V X B)i == - :It
i
,
{Ai, H}* == i, I === I , 2 ,
for the independent canonical variables. ]'he properties of thc rest of the
r>oincare group generators are treated as usual.
D. :N"ULL-PLANE BRACKETS
an illustration of ho\v one can define a Han1iltonian field theory on
a surface other than the constant-tin1e surface, \ve exan1ine the null-plane
treatIl1cnt of Kogut and Soper (1970). One interesting feature of null-plane
- 83-
dynamics is that the nunlber of independent canonical variables is reduced
to half the nunlber prcsent in equal-tinle d ynanlics.
We begin by establishing our notation (see also .l\ppendix ...~ ) . vVe
define
(5.5
1
)
') ) _ ")_ _ _I_ ( ~ 3 _ ":'0)
~ o - ~ . - _ v CJ
Vz
x === x z == 1 ,2,
so that the nletric IS
-0
0 0
0 0 0
(5.5
2
)
glJ.V ==
0 0 0
0 0 0
1_
\vith tL == (+,-, r, 2). Latin indices of vector quantities (e.g. x) \vill take
on only the values i === (r ,2). rrhe variable x+ \vill now be used in place of
x
o
to define the canonical lTIOlllcnta and the evolution of the systenl. rrhe
I)oisson brackets and sYlTImetry generators of the theory can then be \vritten
as indicated in Section 1. D.
1'he canonical nlomenta nil- are conjugate to the x t- derivatives of ...J\.(J.'
and are therefore defined as
Lt
n'
\vhere fi? is still gi ven by (5.6) because the volun1e elenlent is unchanged.
1'he canonical Poisson brackets are given on the surface x+ == constant, so
== - o ~ 0 (x- -y-) 0
2
(x - y)
== - o ~ 0
3
ex -y).
Since the nletric is gi ven by (5.52), A-is the canonical conjugate to 7t+, rather
than A+.
1'he first thing \ve notice is that the definition (5.53) of the canonical
1110n1cnta gi yes rise not to one constraint but to three pri17lary constraints
- 84-
\vhich are incon1patible with the canonical brackets (5.54):
(555)
(5.5
6
)
(557)
7t+ ~ 0
1 1"'\ A+ + ""I Al
Tt - 01 c_ ~ 0
The definition of the fourth momentun1
(5.5
8
)
is not a constraint but a dynamical relation between the velocity
the canonical variables Tt- and A-. We observe that
A+ and
so the constraints (5.56) and (5.57) are second class.
\Vith null-plane dynamics, the forn1al expression for the canonical energy-
n10n1entun1 tensor is
\tvhere the 111etric IS gi ven by (5.52). The canonical conserved translation
generators are deduced fro111 Section I. D to be
(5. 60)
../
and formall y generate the transfornlations
(5.6 I)
Since P
r
== P- generates x+ displacements, the natural object to use as the
Han1iltonian is
".
He = - p- = -./ dx- d
2
xot- (x) .
where
Cl +- __1 ( -)2 _ 1 F,12 F
12
_ ( -) -L i).\ i\-
Vc - :2 Tt:2 Tt \.- ITt\. 1) ~ ' " " " " l .
The minus sign in (5.62) is dictated by the convention that the Han1iltonian
be positive definite and the fact that our null plane nletric (5.52) has no n1inus
signs as did the ordinary metric (I .46) with gOO == - I.
Now we search for secondary constraints. We forn1 a new Han1iltonian
H by adding multiples of the constraints (5.55) - (5.57) to the canonical
Hamiltonian,
-- 85 --
Then we require that H have vanishing canonical brackets \vith all of the
constraints (5.55)-(5.57):
(5 .65) {
+ H.... } - ') - -L ') i 0
7t , - ,,- 7t i \. z' 7t
Equation (5.65) is a genuine secondary constraint, while Eg. (5.66) in1poses
conditions on the u/. Of our final set of constraints (555)-(5.57) and (5.65),
the onI y first class ones are
=== n+ 0
Y2 === 3_ 7t- 2 7t 0 .
1"he relTIaining t\VO,
(5. 68)
1.....,1 ;\ +
Xl === 7t - ..tl..
2 .....,2." ..'--
1.2 == 7t ...'"""1. I

2_ A
2
0
are second class due to F:g. (5.59).
NO\V \ve in vert the bracket lllatrix {I./, Xj} gI \Ten In Eq. (5.59) for the
second class constraints the result
(x ,y) == [1 0] I S(x- _ y-) 0
2
(x _ y) ,
. 014
\vhere s (x- - y-) == sign (x- - y-). Next \ve take H to be the first-class
Hamiltonian analogous to (1.24),
It is easy to verify that (5.70) contains the explicit solution of Eq. (5. 66)
for the uJ, as discussed in Section 1 . 11"'inally, we use Eq. (5.69) to conlpute
the preliminary Dirac brackets, which \ve denote with a prime:
(5.71) {AIL (x) , A" (y)}' = (g'"l i
v
g'"2 g2'') : E(x- - y-) a
2
(x - y)
{]tf1(x) , ]tv (y)}' = -l"+gVi : E(X- - y-) a
2
(x - y) +
I ( "z' .....,:r
2 g g \.'z'
,,+ III ,.....,.t ( Iv
g g g g
2V) -"X) ( . \
g g di 0 .X - YJ
.l\.ll equations of nlotion are no\\ conlputed using the preliminary brackets
(5.7 1) cOlTIpatible with setting the second class constraints (5.68) strongly
to zero.
- 86-
vvhere the ITIultiples of Xi in (5.70) can he dropped if \ve use the prin1ed
brackets. Hamilton's equations becon1e
{
A+ H}'==J,A+--:-' 1J
, \..-, v_ 2
{A-, H}' == VI
{A
i, H}' == Ai 1 ""t."
- 2 \,.' 'l'2
{ 7t+, H}' == 2_ 7t- + 2
i
tt
i
0
(5.73 a)
Before choosing any gauge constraints, \ve nlay take our total Hanlil-
tonian to be (5.70) plus arbitrary nlultiples of the first class constraints (5. 67);
(5.7
2
)
{
- H}' "" -
7t , == 7t
{
i H}' J i
7t,
(S 73 b)
\Ve n1ay therefore choose to express VI and V
2
as
VI == 2 A-
'fhe generating functional of the gauge transfornlation is thus
NULL-PLANE RADIATION
NO\V let us eliminate the arbitrary functions from our theory by using
gauge invariance to fix the reITIaining t\VO degrees of freedon1 corresponding
to the first class constraints (5.67). First it is convenient to recall that a
straightforward deri vation of the Euler equations gi ves the following results:
(575)
NO\:v let us choose the null-plane analog of the CouloInb gauge by setting to
zero the field A+ == A- conjugate to 7t+ o. A gauge transformation \vhich
accomplishes this is
(5.76)
o't+
\ It.. 1\' II .\11 /) ....,' t. rI A-,/ . --- )
...'-1.' I == r ' \X ( "r \7 ,x ,x
o
where
A'- o.
(577)
- 87-
F'r0 n1 qs. (5. 75) and ( 5.58) , \vC fi nd t hat
2 (2 .i\'+ 2;" i)
2: (Tr.- 2i ...A.. /1) 0 .
Exactly as in the equal-tin1e case, \VC n1ay find a gauge transfonnation \\-hich
- I ;\ ,,; I '1 . .. '\ ,,- d E ( , rI' h
sets TC -:-2;" ...'"""\. ueto 5.77). c
solution is
(5.7
8
)
A' -- A" = A' (x) - Vx Jd
2
y III (Vy A' (y) n- (y))
A
'+ i\ "+ _ A'+ ('\ jl'd
2
III (x - ( f; (,.- )\
-+.1""1. - X) - Y 47: TC .:; I1. Y )
,'" 2
A
'--+ :i\"-::..:..= 0-1 d
2
y In(x-y) (- ':)YA'I'( + .- \)
---- d T- \ n iX, X ,y)
, 41t'
,.;
O'
I)ropping the prln1cs, \ve find our final set of constraints
CPl === n+ 0
CP2 == 2_ n- 3; n 0
CP3 == == 0
CP4 === n- -t- 2; ...L\. 0 .
ITsing the prelin1inary hrackets (5.7 I), \ve con1pute the lnatrix of hrackets
of the constraints (5.79) to he
(5. 80) {
m. (-) ( ,\ t.! -- C'" .J,. !"-
T I \ ..1 , TJ \ J )J -jI J \ ,J) -
o 0 --I 0
o 0
000
o 0 0
y).
rro specify the boundary conditions, \\c ohseryc that \vhen the (strong) con-
straints (5.68) arc con1bined \vith CP4 and inserted into the Euler equation
CP2' Vie fin d
n
2
A+ --, (..., .\ 1\
V
We then choose our boundary conditions (and exhaust all renlalnlng gauge
freedon1) so that the solution of (5.8 I) is
\ ;- ( ) 8 I' In (z -- x)
.t"l. == 2 ( Z ----
ClX 41t'
.. '
8.\i (x+, x-, Z)
dZ
1-'hen the inyerse of (5.8) exists and n1ay be written
o 0 0
3
(x -- y) ()
o
o
o
G (x ,y)
o
o
o
-(; (x ,y)
o
o
- 88-
\vhere
G(x,y)== I
ln
(x-y)2a(x--y-)
4
7t
V; G(X ,y) == a
2
(X - y) 0(x- - == 6
3
(x - y) .
.For use in the Dirac brackets we also define
(5.84 b)
d 3
F (x ,y) == a (x - y)
CJX
Plugging Eq. (5.83) into Eq. (1.58) with the primed brackets (5.71),
yve find the Dirac brackets
(5. 85) { (x) , n
V
(y)}* === - g\!+ V.; F (x ,y) +
+! [__ gV1 +gP2 gV2) gVi +
+ gV+ 0
3
(x __ y)
, A\' (_v)} * == gV1 + gV2) F(x , )') + g"- 2
x
G(x ,y) +
!--Li v_ x x G( )
+g g 3
i
J_ x ,y
1 p+ vi x F ( )
- g 8
i
X ,y
vj ""x '""Ix G( \
g g di OJ x ,y) -
\vhich are nlanifestly compatible \vith all the constraints (5. 68) and (5.79).
Using Eq. (5.82) cOlnbined with the constraints (5.68) and (5.79) it is
easy to verify that all of the Dirac brackets (5.85) follow directly from the
nonlocal Dirac brackets
(5. 86)
of the two independent canonical variables A
l
and A2.
Equation (5.86) serves as the starting point for the quantum theory.
.l\pplying all of our constraints, \ve find that e+- is simply

.8-;) G+- - - 1 ((") r\ 2_ ':'1 i\ 1\2 -+- (':'1 A

1
-L ") A
2
)2)
_J - 2 \.- 1 1""1. \., 2 ..... I \., 1 I\.-2
and the Hal11iltonian is
(5. 88)
The Han1iltonian equations of motion then become
- 89-
Differentiating by x-, \ve find
A
i - {":\ At" H}* - _ 1 ":\ ":\ ;\
- - 2
so that Eq. (5.89) can be understood very sinlply as the x- integral of the
Klein-Gordon equation
(5.90 )
i\ i _ ( """\)
L" 11.. L ..... "'"i.. - 2 0 : L" __ ":\ A..
1
== 0 .
.....
We leave as an exercise for the reader the development of null-plane
electrodymanics in the gauge A+ 0 chosen by Kogut and Soper (1970).
We remark that the choice A + 0 a produces a theory very similar to the
equal-tinle electrodynalTIics in the axial gauge A
3
O.
6. Y.l\NG-l\1ILLS GAUGE F"'IELD
We now turn our attention to theories with a set of vector fields whose
Lagrangian is invariant under a gauge transfortnation, as pro-
posed by )lang and IVlills (1954) [see, e.g., vVeinberg, 1973; i\bers and Lee,
1973]. Since the fields in the lllultiplet Inix anlong thelllsel yes under the gauge
transformation, this type of theory is silllilar in SOllle wa ys to the theory of
gravitation gi yen in the next chapter. '1'he yTang-lVlills fields are in fact self-
interacting fields, again like gravitation. Since vve have until no\v restricted
oursel ves to free s ystellls, this Vo/ill be our first exposure to the difficulties of
interactions.
i\. LIE GROUPS
We begin with a brief digression on Lie groups (Racah, 1965). Consider
an n-parameter Lie group f and the associated Lie algebra !F (f). Let '1--
a
,
a == I , ... ,'JZ, be a cOlllplete set of linearly-independent elenlents of if (f).
The composition rule in f (f) for these elements is
(6. I)
The antisymmetry of (6. I) and the ] acobi identity inlply that C
ab
c
satisfies
(6.2)
(63)
We can now define quadratic polynomials on the enveloping Lie algebra
using the symmetric second rank "metric tensor"
Det I gab! is nonvanishing provided the group is senli-sinlple (has no ...
invariant subgroups). We novv define

L
(65)
- 90-
lIsing (6.2) and (6.3) \\TC can sho\v that is totally antisymrnctric
in its indices.
Next, we note that any given Lie group \vill have a specific nunlber of
in variants G,. \vhich are polynomials in the elen1ents I'll and \vhich
are In the center of the enveloping algebra:
(6.6)
[ Ta , Gt'] == 0, for all a and all i.
'fhe nl0st fanliliar Casilnir invariant is the quadratic one, 'shieh can be
written
(67)
\vhere gab gbe == 0; and the validity of Eg. (6.6) follows fronl Eq. (6.5).
F'inally, \VC note that the JZ X n nlatrices
(6.8)
forn1 the (l{ijoint representation of the Lie algebra. satisfles Eq. (6. I)
tri\"ially due to the Jacobi identity (6.3).
13. SYSTEM \VITI-IOlJT
1"'or any senlisin1ple Lie group J- \vith structure constants c
a
be
, \ve consider
an action functional
,vhere the II \'ector fields ... (.1:) , a == I , ... ,ll, are taken to transfornl as
follo\vs under an infinitesin1al gauge transforn1ation:
(6. 10)
== C be /\ F
P
'
v
o aa .... b {'.
It is a standard result (Dtiyanla, 1956) that although does not transfornl
as the adjoint representation of f on its a-index under the gauge transfor-
Ination, the quantity does:
F
!J.
v
- A
V
_ A!.L _ be ;\ !.L, \'
a - v a vag /(l oL'-l.b .L'-l.{O
(6. I I)
For convenience, we define
13
" =--= 1 L,jk
. a c,.. r a
Our objective is no\v to find an action (6.9) \vhich is in\"ariant under
the gauge transformation (6. 10), so that t\VO components of each are
unphysical, and the describe massless spin one particles. In principle,
- 91
any Lorentz scalar function of the yvith the a-indices conlbincd in the
fornl of the Casin1ir invariants (6.6) \\Till do. 1'he traditional fornl is
(6. I 3) 2==-
and includes only the terln analogous to the quadratic Casinlir invariant
(67)
I n the rest of this chapter \VC will restrict oursel ves for convenience to
the Lagrangian (6.13) and to the group SU (2). Since SU (2) has only three
generators and
\vhere c-abc is the total] y antisynlnlctric tensor in three ditnensions, \ve can
novv redefine everything in such a \vay that \ve can forget about raising and
lowering group indices. All of \vhat follows can easily be generalized to an y
I...iie group by the interested reader.
If \ve no\v specify the d ynanlics on constant-tinle surfaces, the canonical
11lonlenta are
(6. I 5)
\Ve sec that
(6. 16) :'T == --- E === --t-- A -,- "AO
(l a I a r Va
l'his is not a constraint equation hut serves to define the d ynanlical proper-
ties of .A
a
. On the other hand, for [1. == 0, \ve find the expected n prinlary
constraints:
(6. I 7)
The canonical I)oisson brackets are
(6. I 8)
( p. ( ) A" ( ) t [J. v ( )
l. 1ta t, X, b t, Y J == -- g 0ab 0 X - Y
1'hese are inconlpatible \vith (6. 17).
Noyv SInce
8Ie _
(;lAa - a
..1-\.l-1.,V
\ve find the Euler equations
(6.20)
1"he zero component can be written
(6.2Ia)
92 -
as usual, IS also inconlpatible \vith the Poisson brackets. 'rhe space
components of (6.20) are
1
(6.2 I b)
(6.22)
Due to the confornlal space-tinle synlmetry of the systenl, conserved
Noether currents exist \vhich generate the full algebra of the confornlal group.
I-Iere we shall concern ourselves only with the canonical energy-nl0mentunl
tensor, \vhich takes the form
8!J.v == _ ..,:\ . v + [J.V se ==
c 8Aa aI., g
I" !.J..
1"he canonical Hamiltonian is then
(6.23) I I)
- B
a
B
a
- --. ..Ia:T
a

2 2
We nO\\7 add arbitrary 11lultiples of the prinlary constraints (6. 17) and use
(6. 16) to get the preliminary Hanliltonian
...-\s usual, the secondary constraints are found by conlputing
(6.25)
which agrees \vith (6.2 I a). Our 2n first class constraints are now
(6.26)
l
a 0
tj)1 == :Ta 0
(6.27 a)
\\1e add theln to H! with arbitrary coefficients and integrate by parts to get
the Hamiltonian
H = He J =
= rd
3
x (+ :ra :ra ++Ba B
a

+ (V .la gO-abe :r
b
. A
e
)) -
- ( ds . :r
a
.

Soo
- 93-
As before, the surface term will be dropped. {i\'\ H}, we find
that the final forn1 of the Halniltonian can be taken to be
(6.27 b)
o J3 ('I I
H === P === d x 2:T
a
:Ta + 2 Ba Ba -
Parallel argunlents gi ve for the translation generator
(6.28) p= J
A
a
(V . :T
a
+gSabc :Tb . AJ) .
Unlike the IVlaxwell field, the present theory has additional Noether
currents due to the local invariance of the Lagrangian under the non-Abelian
Lie group. Suppose we Inake a space-tiI11e independent transforI11ation with
n infinitesimal paraI11eters }'1' so that (6.10) beconles
Then the standard argunlents tell us that the conserved current IS
(6.29)
With constant-tiI11e Poisson brackets, the time-independent generators of
the synln1etry transforlnations are
(6.30 )
rrhe Qa obey an algebra in the Poisson brackets isonl0rphic to the Lie algebra
(6. I),
(6.3 I)
We see also that the fields transfornl under Qa as the adjoint represen-
tation of the group:
(6.3 2 )
We lnay easily verify that the infinitesinlal generator of the full gauge
transforlnation (6. 10) is
(6.33 a)
- 94-
and obeys the algebra
the last t\VO ternl,s of the I-Ian1iltonian (6.27) consist of another
gauge-like tra!lsformation
(634) K [A
a
] = g-l Jd
3
X 20 A
a
-- (V :T
a
+gZabc :Tb' A
c
) A
a
] 0
\vith A
a
=== - I-J [./\a] and IZ [A
a
] differ by a functional of the forn1
:r r 3 . 0 0
Q [A
a
] === L [A(/] - I'.. [A
a
] === g-l I d x (V . (:T
a
A
a
) +g Eabc j\a 7tb A
c
) .
,.i
If vanishes like r/r at (X) , Q [-- = 0 and L [-gA?zJ === K [-- o.
If j\a does not decrease rapidly enough, L [AaJ and Q [l\aJ \vill nol
vanish \veakly. nonvanishing charge (6.30) generating the symn1etry
group algebra is jn fact seen to be a particular case of Q [J\al, namely
Q" = Q [A
b
= = g 1 rd
3
x (V .la +gZabc .
,.1
C. RADIATION GAUGE CONSTRAINTS
The problem of finding a suitable canonical quantization schenle for
the Yang-Mills field in the radiation gauge is \vell-kno\vn to be quite conlplex
(Schwinger, 1962a, 1962b; Mohapatra, I97Ia). \Ve now proceed to attack
this problenl by inlposing the gauge constraints in the context of the I)irac
111ethod.
\Ve begin by choosing the traditional radiation gauge condition
We will find it useful to define a Green's function Gab (X ,y ,A) solving
\vith the boundary condition that Gab fall like I/r at infinity. While no exact
solution is kno\vn for Gab' it can be conlputed as a po\ver series in g:
G b (x Y A) ===
a , , 4r:lx-y,

3 I . 8
-g d z ------ E b
.., 41t' x-y ac 4it' z-y!
\Ve now take as our prelinlinary constraints the second class pair
(6.3 8)
(639)
- 95-
The Poisson brackets of the constraints (6.38) and (6.39) are
{ (x) , cpi (y)} === - gEl/be (V:T
c
(X) -i-- :Tel(X)' A
e
(X)) 0
3
(X - y) 0
{ (X) , (y)} == (aab V.: -r- gEl/be A,. (X). Vx) 0
3
(X - y)
{
a b"( 2
CP2 (X) , cpl (Y)J === - (Dab Vx
{ (X) , (y)} === 0 .
The inverse of the l1latrix ,Y) = cp5(Y)} can thus be chosen
to be
[
0
[C
--1J abC ,
ij X,)/) ==
. Gab (.-1: ,)' , .t\)
\vhere C;ab(X ,y is the solution to Eg. (6.36) introduced at the beginning
of this section.
Equation (6.4-1) is then used to C0111pute the prelirr1inary Dirac brackets
c0111patible vvith the strong constraints (6.38) and (6.39):
A
i
( )] 8G
,lC
gE
bcd
;/.Y
f\ i ( ) _.I (. )}' _ ( _ )
l I, X , Il.b ",I ,y - 0 Dab () X Y
f A
P
'( A
V
}'_
l llt,X), b(t,y)-O
{
() ( ) v ( )}'
7ta I, X ,7tb \1 , Y === 0
{7r;; (I , X) (t, y)}' == _gOV Dab 0
3
(X - y)
X) y)}' == _._gOf
L
Dab D
3
(x-y)
NO\V \ve l11USt see what becomes of \vhen we in1posc the strong con-
straints (6.38) and (6.39). l'aking the divergence of Eg. (6.16), \ve find
- 96-
Using Eq. (6.36), \ve solve Eq. (6.43) for
(6-44) (x) R:! - rd
3
yG
ab
(x ,y , A) gEfJed :T
c
(y) Ad (y)
... '
Equation (6.44) is seen to be a 'lveak equation because the right-hand
side has vanishing brackets (6.42) with while does not. It is clear
that our final set of constraints is
(6.45) xi == (x) 0
= (x) +g Jd
3
yG
ab
(x ,y , A) E
bed
:T
e
(y) Ad (y) R:! 0
The prin1ed brackets (6.42) of these constraints for111 the 111atrix
== { (x) , (y)}' == 0
(6.4
6
) == { Xl: (x) , (y)}' == - { (x) , xi (y)}' == + Oab 0
3
(x -- y)
=== { (x) , (y)}' == M
ab
(;l: ,y)
where M
ab
(x ,y) can be explicitly COlTIputed as a power series in g if desired.
We need not COITIpute M
ab
(x ,y) here hecause the inverse of ,y) is seen
by inspection to be
We thus find that the final Dirac bracket system is
(6.48) { (t , x) , (t , y)}* == lVl
ab
(x , y)
{ (t , x) , Ai (t , y)}* == 0
{ (t , x) , (t , y)}* == - gG
ac
(x , y , A) E
cdb
(y)
; 3
- g.J d zG
a
, (x , z ,A) E
red
A
e
(z). \7.
[3r7 ajb +gEjbg (y)] Gdj(z ,y , A)
{ (t , x) , (t , y)}* == { (t , x) , (t , y)}* == 0
{ 1t: (t , x) , 1ti' (t ,y)}* == (t , x) , 1tl (t , y)}'
{ (t , x) , Al (t , y)}* == { (t , x) , Ai' (t ,y)}' .
- 97
We n1ay no\v take the I--Ia111iltonian to be
In the Dirac bracket systerrl (6.48), H indeed generates the correct equations
of 1110tion:
(6.50 )
Sch\vinger (I 962b) has argued that the brackets (6.48) produce a Lorentz-
covariant theory. We leave this as an exercise.
D. RADIATION GAljGE TECIINIQUES
four sets of constraints (6.39) and (6.45) hold strongly In our Dirac
bracket systen1 (6.48). Hovve\er, the field-dependent tern1S on the right-
hand sides of the I)irac brackets 111ake it extren1ely difficult to use these bra-
ckets as the basis for a canonical quantization schen1e. Several procedures
are a vailable for CirCl1111venting these problenls. ()ne exan1pIe is the direct
path-integral nlethod and Popov, 1967a, 1967b) for developing a
diagran1 expansion consistent \vith the constraints. The" ghost"
particles vvhich occur in this procedure correspond precisely to the integTals
\vhich appear in Gab on the right hand sides of our Dirac brackets.
popular procedure (Sch\vinger, 1962a; IVlohapatra, 197 I a) which
\\;e \vi11 sketch for con1pleteness gives up dealing \vith the strictly canonical
1110n1enta One ignores cOlnpletely and transfornls the fields to the
radiation gauge
(6.5 I)
Then :T
a
is split into a trans'verse part P
a
and a longltudi1zal part equal to the
gradient of a scalar field <1>a ,
(6.5 2) :T
u
== P
a
\7<P
a
,
\\here P" and <P
a
have their canonical properties {lefined by
(653)
(654)
7
-98 -
l'hus the Euler equation (6.25) and the gauge choice (6.51) imply
(655)
VpLJ o.
T'hen one con1putes the Dirac brackets consistent with (6.5 I) and (6.55) find-
ing that Pa and All obey the fan1iliar quantizable l\!laxwell field I)irac brackets
(5.32). i\t this point
J
one C2n use Eqs. (6.16) and (6.2Ia) to find the follo\ving
expressions:
(657)
ITsing (6.36), one then sol ves for <I>a ,
(6.5 R)
and expresses in terms of <I>a ,
o ( 3 2
An (x) = . d yG
nb
(x, y ,A) Vy <l>b(Y)'
Since has been ignored, Eqs. (6.58) and (6.59) are taken to define the
properties of <Da and ... l'he Han1iltonian can then be written
(6.60)
\,"here the Y<1>u y<I)a tern1 is con1pared to the instantaneous Coulon1b energy
\vhich appears In interacting lVIax\vell electrod ynan1ics in the radiation
gauge.
\\Te prefer the techniques of the previous section, \vhere the variables
are all treated in a strictly canonical fashion. l\/loreover, there is a gauge in
\vhich the canonical procedure can be used directly to define the quantun1
theory: this is the subject of the next section.
E. i\XIAL GAUGE
We no\v explore the properties of the Yang-Mills field in the axial gauge
(6.6 I)
introduced by .l\.rno\vitt and Fickler (1962) (see also IVlohapatra, 1971 b;
IZonetschnyand KU111111Cr, 1975). Using the I-Ialniltonian (6.27 b) \ve find that
(6.62)
- 99-
Therefore Eq. (6.6 I) \vill not hold for all tinle unless
\Ve no\v adjoin (6.6I) and (6.63) to the original constraints (6.26) to forn1
the set
'fhe Inatrix of l)oisson brackets is found to be
(6.65) (x ,y) == { (x) , (y)} =
o
o
o
Oab
o
o
o
o
- Oab
o
\\Tith appropriate boundary conditions gi ving the explicit solutions for the
dependent variables to be \vrittcn later, find the in verse of (6.65):
(6.66) (x ,)1) ==
r
lab (x ,--,V) - 0abG (x ,y) H
ab
(x ,)1) 0ab F (.. 1: , y)
Oab G (x ,y) 0 - O'-b (x ,.Y) 0
H
ab
(x ,y) - 0llb F (x , .1
/
) 0 0
0ab F (x ,y) 0 0 0
functions F , G , H
ab
and lub obey
(6.67) F (x ,y) == G (x ,y)
F' (x ,y) == 2
3
G (x , y) == 0
3
(x - y)
- 100-
and are chosen to have the explicit forn1s
(6.68) C; (x ,y) === I x
3
- .-v
3
I 0
2
(x - y)
F' (x ,y) === E (.1'3 _y3) (x - -,v)
lI
ab
(x ,y) == I
ilh
(x ,y) --l{2
11bc
(x) c; (x ,.-V)
\vhere E (x -- )1) is the algebraic sign of - y). .Lr\ll ren1aining gauge free-
donl is exhausted by the choice of boundary conditions in1plied by (6.68).
T'he I )irac brackets corresponding to Eq. (6.66) are
{
.\ tJ. (\ i\ v ()}* ( !to
...'l...a X) , .... '-1.b _v === 0ab ,g g
o ( !lO vz'
- ab g g
ll<) "0) F ( )
g,.Jg x,y
II '" '1'0 /\ l' ( )] G ( \
g' g ...'-1.
1
' X J" X, )')
ItO '1'0 I ( )
g' g , )'
(6. 7:J)
(6.7 1)
{
v. () "( )}* (' It z' z' () vi i ( )) F ( )
7t;z x , 7tb Y === g2
ab
[ --g' g 7t[ X g' g 7t( _V X,_V
{ (x) , ... Cv)}* === (_g,t" _g'lOgVO) Oab 0
3
(x- y)
g"'" Oab 2";-- (x ,y) - g!t1'g"OgSabL' (x) G (x ,.-V)
v\le nO\\l take as our four independent \Tariables Ttl: , ...'\(: , i == I ,2, \\Thich
ha \'C the follo\ving I)irac brackets anl0ng thenlsel \Tes:
(6.7 2 )
{ (x) , }\{ (y)}* === 0
r i () ;\ J ( )}* __ iJ ( )
l 7ta X ,ll"b Y - - 0ab 0 X - Y .
]'hese are ranoniral brackets, \vith no fields on the right-hand side. 'l'he
strong constraints , and can no\v be used to show
(673)
so that ... f\;; IS a dependent variable:
(674)
r
..., i ( 0 .1 2 Y)
. ,-' l' 7t
1l
\. X , .1' , X , C;
rrhen 91 == 0 also itnplies
,... l' 1 .2 .'1;) ;\ i (0 .1 ';)1
/.[c"abL' Ttb \:{, ,x ,x , '-) ...'-1.
c
X ,:t ,J: ,":' .
(675)
1
----------------------------------------
-
- 101 -
'fhe Hanliltonian in the axial gauge ITlay be \vrittcn
(6.76)
and \ve Inay conflrnl that it generates the correct equations of J11otion of
the independent variables in the I)irac hrackets:
(6.77)
\Ve \vill forego the usual discussion of I)oincarc covariance, etc.
V\Te that the independent canonical variables in the axial gauge
have canonical brackets equal to their I)irac brackets (6.72). are no
fields on the right hand side, and no ghost loops in the 1
1
"eynnlan rules. ()n
the other hand, the canonical variables in the radiation gauge fields
on the right hand side of their I)irac brackets and ghost loops appear in the
11"eynnlan rules. It therefore appears that the canonical quantization proce-
dure is best carried out in the axial gauge.
f--inally, \ve renlark that it is also instructive to analyze the
field in the null-plane fornlalisnl. '1'he gauge choice = 0 has properties
sinlilar to the radiation gauge and has gl-:ost loops. 'The gauge = 0, on
the other hand, sinlplifications parallel to those of the axial gauge C1'on1-
houlis, 1973). In particular, the independent variables have field-independent
I)irac brackets and no ghost loops result. \'le lea ve these analyses as exer-
cises for the interested reader .
... our last exanlple of the application of I)irac's Inethod to physical
systen1s \vith constraints, \ve treat f:=instcin's theory of gra vitation. the
references giving elen1ents of the Han1iltonian forlnulation of general relati\'ity
are Pi raniand Schi Id (1 9 50) , Bergn1ann, Penfi eld, Schi 11er and Zat kis (1 9 50) ,
Pirani, Schild and Skinner (1952), Dirac (1958b, 1958c), 13ergmann (1C)62),
Deser l\lisner (1962), Hojnlan, Kuchar and l"eitelhoinl (1973,
1974), and Regge and Teitelboinl (1974a, b). a n10re con1plete account of
the vast anl 0 unt 0 f lit erature, see 1'-. ucha r (1 974) . \\'e 0hser ve t hat in Einst ei n ' s
theory of gra\Titation, it is the expression of the theory in nlanifestly cova-
riant forn1 \vhich causes constrair:.ts to exist clIllong the canonical variables:
the gauge-like frceclo111 to 111ake r;eneral coordinate transfornlations can then
be exploited to reduce the nU111hcr of independent degrees of freedon1 on a
spacelike surface to t\VO pairs of canonically conjugate variables per point.
- 102-
Our starting point will be the Hilbert action
(7. I)
for the pure gravitational field, where for convenience the units have been
chosen in such a "vay that G, Newton's gravitational constant, takes on the
value (I 61t)-1. \Ve deal with a hyperbolic Riemannian spacetime of signature
(- , + , + , +). Four-dimensional spacetinle quantities carry an upper
left index (4) whenever it is necessary to distinguish them from the three-
dimensional quantities to be introduced later. Greek indices run from 0 to
3 and Latin indices from I to 3. Four-dinlensional covariant differentiation
is denoted by a semicolon, while its counterpart on a three-dimensional
h)"persurface is indicated by a slash. The Rienlann tensor is defined as
vvith
'The Ricci tensor and the scalar curvature are RrJ.'J == R ~ : t \ ' and R == R : ~ ,
respectively. 1'he determinant of the nlctric is denoted by the letter g.
The appearance of the scalar curvature in (7. I) n1akes the action
invariant under general coordinate transfornlations
which change g!J.') (x) to
I'he theory is thus "already paranletrized " in the sense of Section I. E.
.Note that the arbitrary coordinates denoted there by utJ. are called x!J. here
to conform vvith standard practice.
I'he action (7. I) describes the pure gravitational field ....-\ddition of n1atter
is straightfovvard provided the matter action density contains no deri vati \ ~ e s
of g!J.'J, which is actually the case for particles and for the electronlagnetic
field.
A. GENERAL FORM OF THE HAMILTONIAN
Einstein's equations are a second order partial differential systen1 in the
ten nletric components g!J.'J of a Riemannian spacetime. When looking at
this system from a Hamiltonian point of vic\v, one focuses his attention on
a three-dimensional h ypersurface enl bedded in the four-din1ensional space.
The state of the system is then gi ven by specifying the value of certain fields
defined on the surface; by means of the Hamiltonian, one is able to calculate
the change in the field variables induced by a deformation of the hypersur-
- 103-
face. IfF is an arbitrary functional of state \ve expect its under a
deforn1ation N!t. of the hypersurface to he of the forn1
(7. 2)
i
O
1
-., I d
3
"TIL '\f ( \
=== Xl." \x) p. x) ,
according to the general discussion on curved surfaces in Section (I. I-Jere
frJ. (x) is sonle other functional of the canonical variables, the precise forn1 of
\vhich \vill be exanlined in Section 7.C'.
We can no\v use (7.2) to obtain infornlation about the fornl of the I-Ianl-
iltonian by relating the NIL to the ga';3. ]'his is 1110St easily done \\ith the help
Fig. 7. I. 'The relation bct\veen the :\!J,. and the gG(r3 is obtained by eyaluating the spacetin1C
interval bet\veen the points ,vith coordinates (xi, t) and (xi -f- dx
i
, t +- elt) in t,,o eli fferent
\\'ays and comparing the resulting expressions for arbitrary dx
i
, clt. This leads to Eqs. (7.5).
of Fig. (7.1) \vhich tells us that the spacetinle distance het\veen the points
(Xi ,t) and (Xi dx
i
, t -t- dt) can be expressed as
ds
2
=== - (N
1
)2 dt
2
(N
i
1
g/j (t
On the other hanel, kno\v
.. i
dx
1
) (N.! dt --+- dx') .
Upon con1parlson \vith (7.3), Eqo (7.4) sho\vs that
(7 sa)
and
(7.5 b)
g N
I jJ_
g
Ii'" - Oi'
Conversely \ve can express K!l in terms of the gWJ as
(7
6a
)
(7. 6b)
N
J
== gii gOi ,
l\T1 ( 00) - .t
1" === -g
- 104-
that g''i in (7.6a) is the inverse of the spatial n1etric gj (i.e. gk! glnl ===
and is in general different fron1 In fact the follo\ving useful relations
hold (Arno\vitt, Deser and lVIisner, 1962):
(4) z"./ === ij _ '-T I' N.T jj(NT1):2
'g g v
(7. 8)
(7.9 a)
(7.9 b)
(4)
[(N
1)2 Ni NT;.J
goo == - 1
1
-- gl.! .... 1 ,. _
I. 1 1
( __ (4)0-)2 - N.T g2
\ b _.- ...
(7.5) tell us that goo and /[Oi nlust enter in the I-Ianliltonian
theory as arbitrary functions. 'I'his is so because the N,t nlust be prescribed
fronl the outside in order to spccifiy the defornlation of the hypersurface;
consequently one cannot expect I-Iall1ilton's equations to restrict then1 at
all. ]'hus \ve learn that the non-tri vial degrees of freedonl are contained in
the gij and their nii.
It \vill save a little \vriting to \vork \vith the defined h)" (7.6) and the
/tl'} as independent variables instead of using the goP. and the gij. '-[his change
of \Tariables is pern1issible because on account of (7.S) and (7.6), one can go
back and forth hct\veen both sets of variables. If \\'"e denote l?y 7t!J. the conju-
gate to N
f
\ then the fact that the Nfl are arbitrary tells us that lt
iL
ll1uSt enter
the I--Iall1iltonian n1ultiplied by an arbitrary function /,'l which \vill correspond
to the tin1e derivati\Te of :0Jf
l
]'his renlark, together \vith (7.2), indicates that
\ve should expect the total I-Ianliltonian to be of the forn1
re
I-
J - I d
3
. ("'\""1 7/) [ _i
j
]
L -, \.1." ./l 1 g ij , I ..
1
Since the 7r[L are nlultiplied by arbitrary functions, \ve ll1USt also expect
the (first class) constraints
(7. I I) 0
to hold. Finally, since (7. I I) nlust be preser\Tccl under surface defornlations,
n1ust have {TI
fl
, I-I} o. Fronl :Eq. (7. IO), \ve find that {n'\ I-I} 0
in1plies the additional constraints
(7. I 2) J([L o.
The constraints (7. I 2) rnust be first class in order for the equations of n10tion
\vith arbitrary to he consistent. '-rhus the constraints (7. I I) \vould he
prill1ary constraints \\Thercas (7. I 2) \voldd be secondary constraints.
Note that the abo\'e discussion applies generally for any relativistic
I-fan1iltonian theory in a Rielnannian space. It fo11o\\7s then that \vhen n1atter
is present the grayitational and n1atter parts of the H alniltonian should he
separately of the form j (supplemented of course by any extra first
class constraints characteristic of the ll1atter at hand).
105
\\le sec that yve ha \"C ohtained son1C insight into the structure of the
1-1 an1iltonian on general grounds. (7.1 I) tells us that \vhen \vorking
fr01l1 the Lagrangian do\vn to the Han1iltonian, \ve should try to add suitable
di vergences to the action density in (7. I) so that the resulting I__ agrangian
\vill contain no ti1l1e derivatives of the (...-\nderson, 1958; De vVitt, 1958;
I)irac, !958b; ... Deser and lVIisner, 1962). Furthernlore the Lagran-
gian should ha \"C a \-ery si1l1ple dependence on the Nfl thenlsel \"es in order
to get a J-Ian1iltonian of the forll1 (7.10). We thus need a \vay to analyze the
spacetiIne curvature \vhich clearly distinguishes the dependence of the cur-
vature on the g"j froll1 the dependence on the N
P
. Such an analysis is best
perforn1ed \,"ith the help of the enlhedding equations of Gauss and Codazzi
,,"hich are re\-ie\;vcd in ..:-\ppcndix B.
Follo\\"ing- Kuchal- (1971), \VC no\v dcri\"c the desired Lagrangian hegin-
ning \\'ith the I-liJhert action (7. I). (}3.5),
relates the extrinsic cur\-aturc to the velocity ga!I,O of the d ynan1ical coordi-
nate gab and to the functions N
1
and N" \vhich describe the deforn1ation of
the hypersurface. Note that gab itself also appears in this expression for K
a
!)
through the Christoffel syn1hol hidden in the covariant deri\"atives Nal
h
N
b
/
a
1'herc are, ho\vc\'er, no tin1e derivatives of N
l
and N" in (13.5.). rr'hus
if \ve succeed in expressing the action density in ternlS of glib and K
ab
, \VC \viJl
ha\"C achie\"cd our goal of elin1inating the \'elocities N
l
and N/. Let us \vork
in this direction \vith the help of the Gauss-Codazzi equations. In the basis
(n , eJ used in (B. I), \VC have
(7 I 3)
(J)R - - (-t)R
a1
(nR
lb
lb
(l)R
lIl
__
- - ab al ---
== (Ol(ab
ab
- 2 U)R
a
lal
===
- (-l)R"b __ (I)!yet
-- ab 2 \.. letl
lIn (7. 1 3), \ve ha\"c cxploited the antisyn1n1ctry property ==
== of the Rien1ann tensor]. O\V Vie use (13.12) to transfOrITI (7.13) into
K
r2 __ K I
rab
-- ') (-OR
et
ab '- ... letl
\"ith K
1'he first three tern1S in (7.14) are alread y of the forn1 sought. \\,.e need
to he concerned on1y about the renlaining tern1 (-!)R . Our first step is to
apply to the norn1al nJ. the con11nutation rule for the second covariant deri-
vati ves of an arhitrary vector, thus ohtaining
- 106-
ext note that ll1ultiplication of (7. 15) by n1J. and subsequent contraction In
the pairs (tL, and ((l, A) can the right hand side of that equation
into the desired quantity,
(7. 16)
:\0\\' \YC rearrange the left hand side of (7. 16) by 111ca11S of the identities (see
PPC11 cI ix B for a proof)
(7
1
7
a
)
(7 I 7 b)
18 a)
(7.18 b)
r3 y __ (. y \ .. (3 . y
n n ;i3y - n n ; ;y -- Il : y Il ; r3 ,
r3 y K K
ab
Il :y n ;i3 == - ab ,
r3 K
IZ == - .
I
...... Jl h ( (I))J J I l' hI
(eca Ing t at _. g 2 == 1: t 1en 0 )taln t e re atlon
(/ _ ( (-t))J (t)R __ c/},\rrt.j"rt.
-z HILBERT - \-- g - eL - ,::t,
\vith
(7. 20)
and
(7.2 I)
(7.20) sho\vs that by adding to the Hilbert action density the
eli of (7.2 I), \VC obtain a n10dified action density !f \vhich contains
no tinle deri vati ves of N
1
and N i and \vhich contains only first tinlC deri
ti of gij' 'rhus the Lagrangian
(7. 22) L c= rd
3
x !f = ,. d
3
x N
1
i (K
ab
K
ab
- K
2
R)
../ ..:
\vill be our starting point for getting the Hanliltonian. Ho\vcver, before doing
that, let us first take a closer look at the di vergence in question. By Ineans
of Eqs. (7.5)-(7.9) and (B. 3) one can re\vrite the quantity \;rrt.,c< as (DevVitt,
19
6
7
a
)
vet: == 2 (gJ K'! _ [gQ (I{N _ gij N
1
0)] 'o.
,C(. / ,0 ,; ,Z
(7.23) shows that both types of un\vanted quantities, narnely the
first tin1e cieri vati yes of N1, N" and the second time deri vati \'es of gij, enter
the Hilbert Lagrangian
L
j
d
3x (_ (J)R
HILBERT == ..
107 -
In the fornl of a total deri vati ve,
and can therefore he elinlinated altogether by omitting (7.25) fronl the
.
rIhe fact that one can elinlinate the second tinle cleri vati yes of fronl
the action is not surprising and \vas kno\vn long before the I-Ianliltonian for-
nlulation of general relativity \vas investigated. One can actually do even
better and elinlinate all second cieri vati ves (te111poral and spatial) of the ten
,f{!J_'J by adding a suitable di vergence to Y
H
ILBERT equations arc of
second order after alL). ()ne arri yes in this \va y at the so-called Einstein or
" ga111111a-galnnla " Lagrangian (see for exanlple Landau and Lifshitz, 197 I)
\vhich \vas used by Dirac (195 8b) as a starting point for finding the gravita-
tional Hanliltonian. What is 11luch 1110re re111arkahle and vital for the I-Ianlil-
tonian fornlalism is the elirnination of the jzrst tinle deri vati yes of N 1 and N i
[recall Section 7.i\J. This possibility \YClS explored hy De\Vitt (1958),
...-\nderson (1958), and Dirac (195 8b).
]'HE I-IAMILTO:\ IAN
\\1e proceed no\v to cIeri ve the exact fornl of the I-I cuniltol1ial1 for the
gra vitational field starting fronl the L,agrangian (7.22). Since no tinlC
of the NIt appear \ve inl111ediately get the (first class) prin1ary
constraints
(7. 26)
as expected. '1'hc 11101nenta 7rJi conjugate to i[ij arc bv definition
" oL
'"T'IJ --
i .. - .
H.. ecalling
(7. 28)
(B.5), \ve find after a sin1ple calculation that
nz'i == _ gQ (K
ij
_ Kg
1J
) ,
that the conjugate to the first fundanlental fornl gij is closely rela-
ted to the second fundan1ental form K
ij
. Equation (7.28) can be inverted
to express K
zj
as a function of gij, \vhich anlounts to expressing the velocities
as functions of the 1110rnenta and the d ynalnical coordinates. ()ne thus
gets fron1 (7.28) that
I
T ij _ i (...... 1
j
1..... i
j
)
- -g J'.. -:2 /I.g ,
\vhere \ve have defined
(7.3)
i z'j
n == n i == gij 7t
There are therefore no prin1ary constraints other than (7.26).
-- 108 -
:\ext, using (B. 5), (7.22), (7.26) and (7.29) \ve 0 bt ainfor the Ha111i1tonian
\\'ith
11:
1
/' 13 ( N'I.
, C X 7t!J. 1 ..
and
(7.32 b)
--yt/l' =::.: -- 2 (y '/..' rr/<'J . - (2 (y L' . - (Y b' .\ 7tJ,j == ---- 2 7t /, ..
<-"> / ,j \ b /,/ . ./ ("-' '<.7,,) 1 !.J
In arnnng at (73 I), we ha \T dropped the surface integral 2 f SJ Ttl} ::'\ I
fro111 the right hand side of that equation.
'1'0 obtain the total I-Ian1iltonian, \VC ha\'e to add to (7.31) the first class
prI1l1ary constraints (7.26) 111ldtiplied hy arhitrary functions. \\lc therefore
ha\'c
I-I
.,.
I cl0x

'Y(l
.J
,/f;,
In agrcc111cnt \vith the general discussion In Section (7..:\).
X 0\\' to obtain the lllost general n1otion \\'hich is physically pCrI11issihlc,
\ve ought to add to (7.33) the first class secondary constraints (7.12)
\vith arhitrary coefficients lIY' (x). 1-1o\vc\'er, since the :.J" are arbitrarv to
start \vith, \ve do not get any additional freedo111 by doing this and \ve n1ay
as \ve11 set lIY' == 0 \vithout loss of generality (*). rrhe 1110st general Ha111ilto-
nian is then just (7.33).
Equation (7.26) tells us that the degrees of frecdo11l descrihed hy the \'aria-
hies (7t!.L' N!l) arc not physically i11lportant (7t!L is constrai11ed to he zero and
is arbitrary). vVe can then drop these degrees of freedo111 fro111 the phase
space altogether and treat N
I
henceforth as an arbitrary function e
f
(x) \vith
vanishing hrackets. 1
7
0r111a11y, such a procedure a1110u11ts to i111posing the
(*) Ceolnetrically speaking \vhat R(v's on here is the follc)\ving: Suppose SOlne spacetirne
coordinate systern .r
tJ
is gi\-en. Then the :\!.L 1'0). part of the generates the changes
in the canonical \'ariables which correspond to passing froln the surface xl) t to the surface
X
O
t +. dt. :\ O\V the addition of the extra tenn u!.L .1'[!). would tell us that \\'e are not forced
to choose X
O
= t dt as our next surf-lee, but that \\-c can go frorn .ro t to any inflnitesi-
rnally close surface. .Howc\'er since the spacctinH> coordinate SystClll x:
J
is arbitrary, \ve can
rnake .l-O t dt correspond to an y gi \'en surfa.ce, and conseq uentl y the freedorn inherent
in the u!J tenn ca.n be considered to be already included in the arbitrariness of :\!)'. A. totally
sinlilar situation occurs \\,ith .:-\0 and the function Vz. in the ::\laxwell held (Section 5..:\). The
analogy is established by associating .\0 \\'ith :\t.L , V2 \vith uP. and by 111aking gauge
rnations (Ai j-\z' .+- J
i
.A) take the place of surface defonnations.
- 109-
second class constraint
which n1akes the originally first class equation (7.26) heco111c second class.
rrhc constraints (7.12) rcn1ain first class. ()ne then passes to the I)irac bracket
hy the usual procedure (\vhich in this sin1ple case an10unts to vvorking just
\vith the gz') and It
ii
) and takes (7. 26) and (7.34) as strong equations.
this stage the Poisson hracket is therefore
\\1hen Eq. (7.35) is applied to the fundan1ental canonical variables then1sel Yes,
\ve find
{
r .. (. k I / ')} _ .. k I ( ,'" ')
l, I} J ,It X -- a Z} a ,x ,
as the only non-vanishing Poisson Lracket. l'he syn1bol Oijkl In the right
hand side of (7.36) is a shorthand for
and the Dirac o-function is defined as in (1.82) 'luitflout recourse to the JJl{?fric,
by
for an arbitrary' scalar testing function.
I f (7.20) and (7.34) are taken as strong equations, the FI an1iltonian
reads
(7.3
8
)
and vanishes vveakly duc to the first class constraints (7.12).
The rate of change of an arbitrary functional of gz'j and Te
i
) IS there-
fore given by
so that Eq. (7.2) IS explicitly verified.
D .... FLAT SPACE, SURFACE INTEGRALS, IIVIPROVED
HAlVIILTONIAN, INVARIANCE SPACELIKE INFIN1TY
We arrived in the last section at expression (7.38) for the gravitational
I-IaI11iltonian. \\;"'e can no\v check whether I-IaI11ilton's equations correpon-
ding to H o (together \vith the constraints (7.12)) reproduce equations.
I t turns out that this is indeed the case provided one neglects certain surface
tern1S. For a closed space one is certain that no c0l11plication could possibly
- 110-
arise fron1 neglecting surface integrals, bacause every surface integral vanishes
identically \vhcn the 111anifold has no boundary, So, for a closed space, !-lo
given by (7.38) is the correct Ha111iltonian as it stands. Ho\vever for an open,
asynlptotica11y flat space certain subtleties arise and one comes to the con-
clusion (see below) that in such a case I--I o lnust be supplenlcnted 1)y the addi-
tion of certain surface integrals at infinity in order to gi ve the correct equations
of n10tion (Regge and rreitelboinl, 1974a, b). T'he addition of the surface
integrals also pIays a crucial role \vhen one inlposes gauge conditions \vith
the purpose of arri ving at a canonical systenl having just t\VO independent
degrees of freedom per space point.
see the roll' played by surface integrals, \ve start by observing that
an essential requirclnent which Inust he 111et by an acceptable definition of
the phase space of a d ynanlical systen1 is that all physically reasonable solu-
tions of the equations of nl0tion n1ust lie inside the phase space. If this is
not true, the variational problen1
aJ(p,. qi - H) dt = 0 .
has no solutions because the extren1al trajectories are not admitted alTIOng
the original " conlpeting " of the variational principle. Once the re-
quircnlent of containing all extrenlal trajectories is lnet, one can enlarge
the phase space at \vill but one caJlJlot n1utilate it arbitrarily.
In vaCUUln general relativity, after conditions (7.34) are in1posed, a point
in phase space is represented by t\vcl\'c function variables (gz'j, n
k1
). No\\?
any solution of Einstein's equations representing a physically reasonable,
asyn1ptotically flat spacetinlc behavcs at spatial infinity in the Sch\varzschild
forn1
(739)
1
') (' 1\1 )' 1')
(s:'" - 1- " ,.. ct-
r 8 rrr
(
0 ' . _L 1', I -li .:rj )' d
'.I I 8TC' 1'3 C .1 X.
1'he phrase "physically reasonable" here nleans essentially that the total
lnass-energy of the systen1 ll1USt be finite. 1'his assulnption \vi11 he satisfied
if the systenl has been radiating (gravitationally or other\vise) during a finite
tilne only. It is quite plausible, ho\vever, that one can adnlit a nlore general
situation in this context; one could allow a systenl which has been radiating
during an infinite tin1c, hut doing so in such a \vay that the total amount of
radiation renlains finite. 1'he precise forn1 of the line elenlent (7.39) can be
altered by a change of coordinates, but no coordinate systenl exists such that,
\vhen 1\1I 0, all conlponents of the Inctric and its first spatial
can be made to decrease at infinity faster than r-
1
and r-
2
respectively. It
fo11o\\7"s therefore that any definition of phase space lztls to contain 111etric
functions such that
(7.40 a)
and
(7.40 b)
- 111 -
\\1e will need later a nl0re precise definition of phase space, but the essential
point is that one cannot do any better than (7.40).
Let us return now to Harnilton's principle, keeping the asynlptotic beha-
\'ior (7.40) in nlind. When one deals \vith a continuous systenl like the present
one, I-Iall1ilton's equations read
(7.41 a)
ancI
(74
Ib
)
itj (x) === -- 0 (I-Ian1iltonian)jag
ij
(x) ,
l'he functional cIeri vati ves appearing on the right hand side of (7.41) are,
by definition, the coefficients of Ogij and on
ij
in a generic variation of the
Han1iltonian, i.e., if
then
(7.43 a)
and
(7.43 b)
(I--Iamiltonian)jog,'.i === i\i
j
rrhus, in order for I--Ianlilton's equations to be defined at all, the variation
of the Hanliltonian nlust he expressible in the form (7.42) for an arbitrary
change in the phase space point (g,") ,n
kl
). \V"e shall see now that oHo (\vith
11
0
gi ven by (7.38)) cannot be put in the forn1 (7.42) and that the Hanliltonian
has to be alnendcd by the addition of a surface integral in order for Hanlil-
ton's equations (7.41) to coincide \vith Einstein's equations. Introducing
expressions (7.32) into (7.38) one gets for the change in Ho, keeping all terrns,
i d2 (' ijl.: I ('T 1 " T1 )
- f SIJ ,J:ogij -

\vhere
(745)
1'he coefficients ... r\ii and B
ij
need not be explicitly \vritten here [they may
be found, for exanlple, fronl the right hand sides of Eqs. (7-3.15 a ,b) of
.i-\rnowitt, Deser and Misner, 1962]. We need only to observe that in order
for I-Ialnilton's equations to reproduce Einstein's equations, one Inust identify
Ai} and B
ii
in Eq. (7.44) with the variational derivatives appearing in (7.43).
- 112-
This tneans that 1/ the slojace integrals /n (7.-+4-) 'Zt'oul{! van/sIt tllen H
o
would
the correct HaJJlilton/an. For a closed space this is indeed the case; hovvever,
for the open, asyn1ptotically flat situation, son1C of the surface tcrn1S in (7.44)
do not vanish due to the slo\v asyn1potic decrease characteristic of the gravi-
tational field.
1"'0 deal vvith the surface integrals in (7.-+4-) one needs a nlorc con1plete
specification of the asynlptotic behavior than the one given by (7.4-0) .....t\n
exhaustive discussion of the boundary conditions in question vvould take
us avvay fron1 our t11ain linc of dev
T
elopn1ent; vve refer the reader to Regge
and 1"'eitelboin1 (197-+a , b) for a detailed treatnlent of this issue. We indicate
explicitly here only that the lapse and shift functions Nfl are assun1ed to behave
asymptotically as
(7.4-
6
)
with
It
rx'
r - 70v
== -- .
[The Greek indices in (f..f
l
and are raised and lovvered vvith the lVIinkowskian
n1ctric == diag (-- I , I , I , I)].
}=.quation (7.4-6) says that \\TC allovv asyn1ptotic spacetinlc translations
(rx
fl
) , space rotations and boosts anlong the pern1issi hIe defornla-
tions of the hypersurface. \\1hen (7.46) is inserted into (7.4-4-) and the
corresponding- boundary conditions on l:,j and are used, one finds
vvith
1). .
'.I
07:/.1 (x) 1
(7.4-8 a)
(7.4-8 b)
1'-+00
MIS = -- 2 f d
2
Sl (x' re
ls
- X
S
n!T)
1'--;0..00
(7.49 b)
I-Iere v:ve have defined.
(7.50 a)
so that one has
(7 sob)
== ---
L....- __ _
-- 113 -
From (7.47) \ve see that H
o
has well defined functional derivatives only
\vhen the surface integrals (which are in general different fron1 zero) are absent
from the right side of that equation, i.e., when == == o. I'hus H
o
is a
good Harniltonian only for those defornlations of the hypersurface which
invol ve neither spacetime translations nor spacetime rotations at infinity.
However, Eq. (7.47) itself suggests ho\v to deal with the more general case:
one simply passes the surface integrals in (7.47) to the left hand side to obtain
a new "inlproved " Hanliltonian
The functional H defined by Eq. (7.5 I) has well defined functional derivatives
and generates the correct equations of motion even when asymptotic Poincare
transformations are allo\v"ed arTIong the pern1issible deformations of the
hypersurface.
Now H
o
vanishes weakly on account of (7.12) inlplying that for any
solution of the equations of nl0tion, the nunlcrical value of the H anliltonian
(7.5 I) is
(7.52 ) H - rxf-L P!l- + ! lVI
wJ
This expression IS In general different from zero. When == rxt" == 0 and
rx
O
== lone has H == PI, which identifies (7.48a) as the energy. Sin1ilarly
proceeding along the familiar lines of Noether's theorem one identifies expres-
sion (7.48b) as the linear n10mentunl and (7.49) as the angular momentunl.
(See Appendix B.3). It is important to realize that the identification of P!l-
and M
po
has been achieved without recourse to any special decomposition of
the canonical variables or to a specific fixation of the spacetime coordinates,
which makes obvious the "gauge independence" of these quantities (compare
Arnowitt, Deser and Misner, 1962).
Having written expression (7.48) and (7.49) for p
ft
and lVl
po
' one asks
hirnself inlnlediately \vhether these quatities have the proper behavior under
asyn1ptotic Poincare transforn1ations. The straightforward approach to
ans\vering this question would be to introduce the lapse and shift functions
(7.46) into the equations of nlotion (7.41), but such a procedure would be a
bit awkward. \Ve prefer to follo\v instead a different route which sheds light
on other aspects of the prohlern. Since the spacetinle is asymptotically Min-
kowskian, \ve can introduce a system of rectangular spacetime coordinates
at infinity. These coordinates will be denoted by y.\ (A == 0 , 1 , 2 ,3). The
surface on \vhich the state is defined \vill have, asymptotically, the equation
yA == a
A
b'\ x
r
, with bAr bAs == Drs (this ensures that the coordinates x
r
hecol1le rectangular at infinity). N0\\'", according to Dirac's procedure for dealing
with curved surfaces described in Section (I. E), one should consider the
" surface variables" describing the location of the surface on the same footing
as the" truly dynamical" variables then1selves. In the case of the gravi-
tational field, the nonasynlptotic parts of the" surface variables" are alread y
8
-- 114
in1plicitly included in the six canonical pairs (g/j; n
ij
) , which exceed by four
the required nunl her for a theory \vith t\VO "trul y d ynanlical " degrees of free-
doni per point. a consequence, one has the constraints .ye!J. 0 (Baierlein,
Sharp and \Vheeler, 1962; IZuchar, 1974). Ho\vever, the aSY1TIptotic location of
the surface, governed hy ten independent quantities arnong the sixteen a'\ b"\
\ "
(renlelnber b'r b.\s == Ors) [anJlot be {!i?!('rJllin('(! fronl a kncnvledge of glj and n
l
}.
One nlust therefore introduce (l'\ and b \. together vvith corresponding conju-
gate J110nlenta 7:.\ ,itA
r
as additional canonical variahles on the samc footing
as the g/j ,re
IJ
. .i\fter this is done, one \vill havc a flanliltonian fornlalisnl \\!hich
is nlanifest1y covariant under transfornlations at infinity.
\Vhen \ve introduce ten new pairs of canonical variables \ve nlust gain,
una \'oidahly, ten ne\v constraints. J:1=ach of these constraints \vill enter the
\vith an arbitrary Lagrange 111ldtiplier. 1"he Inultipliers in question
\vill descrihe the anlount of hypersurface defornlation at infinity an(l \vill he
gi\'cn precisely by the and appearing in (7.39). 1"he ne\v, extended
Hal11iltonian, 11
E
, \vill then be of the form
HE = Jd
3
x N
i
' (x) .1t;Jr) r/' (P!L - P!J.) +J (m!,"
== 1--I PrJ. 'Ill!J.')
I n addition to the .yt;J. 0 equations, \\"c find the ne\v constraints
and
(7.54 b) 'tIl!J.')
l\1tJ.
v
R::j 0 .
1"he quantities p!J. and 'In!J.') arc constructed froIn the asynlptotic surface varia-
bles ([.\ b'\ and their conjugates. 'l'he details of this construction are actually
irrelc\'ant \vhat Inatters are the Poisson brackets that Prj. and 7Jl!J.') satisfy.
()ne nlay in fact shcnv that the integrability conditions (1.83) of a Hanlilto-
nian theory on curved surfaces inlply that these quantities satisfy the algehra
of the infinitesinlcd generators of the I\)incare group (I\..cgge and rreite1bDinl.
1974tl ,b). One thus has
(7.55 a)
(7.55 b)
55 [)
\"here lJ!J.') IS the :\1 inkovvski nlctric.
Once \ve kno\v that PP. and JllrJ.v obey the Poincare group brackets, it fol-
10\\lS inlnlecliately fronl (7.54) that 1)!J. and 1\1
1
J.') transforrn in the correct \vay
under an asynlptotic Poincare transfornlation; the only \\"ay in \vhich the
constraints C7. 54) could be preserved under surface deforl11ations is for Prj.
- 115-
and P p. (171 tJ. 'J and -- Ivr rJ. 'J) toe\' () Ive in the sanl e fa shi() n . l' hercadcr 111 a y
\vonder at this point \\"hy one does not just calculate the I)oisson hrackets
of the p p. and 1V1;:0 atllong then1sel ves instead of \:vorking \vith the extra
variables PIJ. and J71!J.'J" 'fhe ans\yer is that these Poisson brackets do not exist
for the san1e reason that the functional derivatives of H
o
given by (7.3
8
) do
not exist. It is only the sunl (7.5 I) \vhich has \vell defined Poisson brackets
rrecall (747) 1 Geotlletricall y speaking, the non-existence of functional deri-
\'ativcs of PrJ. and lVl
t
: o follo\vs fronl the itl1possibility of continuously defonlling
a surface at infinity \vithout altering its shape else\vherc. (See Section (7
belo\v for a discussion related to this issue).
\\7e finish this section \vith the ohser\'ation that thc f-I anliltonian II L
gi ven by (753) is to open spaces \vhat H
o
gi ven by (7.5 I) is to closed spaces.
I n fact, both quantities vanish \vcakly and their argunlents contain, in each
case, the full specification of the surface on \vhich the state is defined.
spondingly, it is the product of the space of the l:"j, Tel) with the space spanned
hy the aSytllptotic surface variables a'\ b":. and their conjugates \vhich plays
for open three-spaces the role that the (glj, 1t
1
)) space alone pIays for C0t11pact
three-surfaces.
'1'he equatiot1s of tnotion associated \\'ith the I-Iatlliltonian (7.5 I dcscrihc'
the c\olution of the SystCt11 under arbitrary defort11ation of the hypersurface
on \vhich the field state is defined. ()ftcn one \\Tants to ans\vcr a 1110rc lit11itcd
question, such as ho\v the Syst(,111 c\'ol\"es along a onc-paralnetC'r fcltllily of
sllrfaces. ]'hcn one cIoes not need the \vhole po\ver of the hased on
the general f-Iatniltonian containing four arhitrary functions and :\ /; a
(*) To verify that the constraints (7.5-1-) arc nrst one treats all of then1 at OllCC,
together \\"ith the ,:10). 0 equation, by \\Titing
IJ E (N)
(The subscript (N) in "Y,Y- and r3
tJ
." is intended to retnind us that these quantities are related to
through the aSYll1ptotic fonnula (7.-1()).) TIle statclnent that all constraints (I.e. 0
and (754r arc first class then reaeIs
for any gi\"cn and Ii and sorne;..Actually for the theory to be consi::itcnt with the !\i('lllClll-
nian structure of spaectinlc (i.e. for the c\'olution to hc path independcllt in the sense of
Section (I.E.), ; 11111St be gi\'en by (Tcitclboin1. J()73 fl, b)
which is an alternative way of writing Eq (I .83). is the starting point for deriving (7.55).
116
reduced Hamiltonian \vhich is able to give only information one is asking
for suffices. The reduced I-Iamiltonian theory is obtained by freezing the
degrees of freedom corresponding to an arbitrary deformation of the surface.
It has no constraints and a lesser nun1ber of degrees of freedom than the
full theory. The gauge-like freedom to make arbitrary transformations of
the spacetime coordinates is thereby destroyed.
It turns out that the procedure for fixing the coordinates is quite dif-
ferent depending on whether the three-space is open or closed; we will there-
fore treat each case separately.
I. OPEN SPACES
In this case the gauge freedon1 corresponds to the possibility of making
arbitrary deformations of the surface while keeping its asymptotic shape
unchanged.
We then \vant to impose additional constraints so that the previously
first-class equations J/P!J. 0 \vhich constrain the defornlation generators to
vanish become second class. After the additional constraints have been imposed
the Hamiltonian will become just a surface integral
(7.5
6
)
I === - el't, P!J. + ! M!J.v
according to (7.52). It will be possible to use (7.56) as a meaningful Hamil-
tonian because, as we shall see belovv, the surface integral (7.56) acquires
\ivell defined Poisson brackets (the Dirac brackets) after gauge conditions
have been imposed, even though its functional deri vati yes \vith respect to
gz"j and 1t
ij
do not exist (*).
Instead of immediately evaluating the Dirac brackets for the gauge
conditions of interest from Eq. (r .58), we shall follo\v instead a son1e\vhat
indirect procedure. rrhis will enable us to n1ake contact with, and use of,
results already available in the literature. The schenle (which includes all
cases found so far in practice) runs as follo\vs: One aSSUll1es that the variables
gt'k ,1t
ik
can be separated into two sets, (x) , 1t
cx
(x)) and (x) , 1t
A
(x)),
by a bijective, time-independent, functionally differentiable canonical trans-
formation in such a way that
(a) the surface integral (7.56) depends only on the cp'X and the 1t
cx
(b) when the 1tc( are prescribed as functions pC( of x in such a way that (**)
(7.57 a) pC( === 0 ,
(*) After coordinate conditions are imposed the deformation of the hypersurface is
globally determined by C/y. and (the freedoln of making arbitrary deformations in the in-
terior has I t then becolnes Ineaningful to ask for the generator of such a defor-
mation. The generator is precisely the surface integral (7.56). This is the geometrical reason
why (7.56) acquires \vell-defined Poisson brackets after the gauge has been fixed.
(**) Strictly speaking Eq. (7.S7a) is the \veak equation nC( (x ,t) - jJex (x) o. How-
ever in the follo\ving \ve shall not insist on the weak equality sYlnbol unless confusion could
anse.
- 117-
the constraints -nJ. 0, 0 can be sol vcd to express the ([)'X as functionals
(7.57 b)
of the rCll1aining canonical variables. functional deri vati ves of fX \vith
respect to yA ,itA are assumed to exist.
If the above conditions arc true, then Hanlilton's equations for the
I-I at11iltonian
(7.5 8) H
reduced [,' A ] - I [, rj ]
y ,itA - g'l' Tt = joe
Tt'oe = Poe
together with Eqs. (5.57) are equivalent to equations In the parti-
cular frame defined by it): == PY. .
Proof. Recalling that Poisson brackets are invariant under canonical
transfortnations and that the H anliltonian is unchanged if the canonical
transfornlation IS independent of tit11C, \ve ha ve
(759)
()n the other hand
(7. 60)
1-1 [c.p'X ; Tt
x
, ; itA] . == I [cp'X ; it
IX
] = Hreduced ; itA]
?oe=j): -.?):=joe
i:oe Poe 'iT:oe = PIX
::\ ext, differentiating (7.60) \vith respect to 'ITA we get
(7. 61 )
?('J.=joe
-;:oe = Poe
SHredllced

I-Io\vc\Tcr, by Eq. (7.5ia)
icc< (x) = - = 0
:'-?):==-=joe
IT:('J. = Poe
whence
SHreduced
STt'A (x)
yoe=j('J.
Tt'oe = Poe
Equation (7.62) inserted back into (7.59) sho\vs that Hredllced generates
the correct equation of nlotion for yA. In a cot11pletely analogous way one
shows that the correct equation of I110tion is also obtained for itA' The evolu-
tion of cp''X as calculated from (7.57b) will agree with the one given hy the
unreduced formalism because the constraints are preserved in time.
- 118-
Note that \VC ha ve not for an explicit dependence of I on Tt,\
i.e. a dependence other than the one induced by == fX) because such a
1 11
. ] h" I l' . "LI reduced I'" , \
epCl1r ence \\'OU C In genera. cause t e varlat10na (er1 vat1yes or /0',/ ,
;)I-IrcdulcdjoTC:\ to be undefined. 'This fo11o\\'s fronl the fact that 1-I
o
and I do not
separate] y ha \'C \vcll--definecl functional deri vati Yes, only the sun1 flo I
cl()cs. F'or an explicit exanlplc of the issue discussed here see the discussion
of (7, 19) belo\\-.
In practice, the new phase space coordinates (ep'Y. ; TCcJ , ; /CA) are not
canonical. Neverthe]ess, one can easily verify along the sanlC lines as the
I f I H
rcduccd, 'II 1 ' 1 1 ' , 1 l' f h
a )()\'C' proo t 1at IS st1 0 )ta1ne( )y InsertIng t 1e so ut10n 0 t c
constraints into I [glj ,reI)] provided that the following restrictions hold:
(h) ]'he "111atrix " {7't
x
, q/) (x':;} is in vertihlc, I.e.
(7. 6.+) I d:
3
x' { (x) , iJi
j
(x')}frj (x') == =?frj == .
.J
If equations (7,63) and (7.64) hold, then lIrcdllced gi\'en hy (7.3
g
) \vill generate
t he correct equations of 1TIotion
i .\ {,,\ Hrcduced'l
'y == 'Y , f
o. -.- {..,.,.. I..J reduced 1
re. \ - ,.. A' -1 J
pro\'idcd that the Poisson brackets are conlputcd using the general forlnula
C)( l;(
3F
)

2Qa 2Pb
\vhich holds for a general (not necessarily canonical) set of phase space coor-
dinates (Qil , Pa)'
\\TC no\v proceed to apply the abo\'c procedure to the gauges that ha vc
heen proposed in the literature: the ...-\.rncHvitt, Deser and :VI isner (1962) "'[--'1'
gauge" (\:yhere "'1'-T " stands for " transverse-traceless ") and the so called
" nlaxin1al-slicing " gauge condition proposed by Dirac (I95 8c).
(a) ADlVf's (;auge.
r\rno\vitt, Descr and 1\lisncr 1962) separated the canonical yariahlcs into
t\VO sets as follo\vs: (i) (gT; /C" ; 7:'1' ; g;) and (ii) (independent) pairs
(g-;/ ; TC
1jTT
). 1"'he nc\v variables are defined by applY'ing to both g".i --- 0/.1
anel 7:
lj
the deconlposition.
(7. 67)
(7. 68)
f/ == . or,) [f,j,/ - I '2 fl.:j'kil] '
2y'
- 119-
and
(7.7)
Here 1!v:!. is the inverse of the flat space Laplacian, \\lith appropriate houn-
dary conditions at infinit:y. r1'he fixation of coordinates is achie\'ed by inlposing
(7.7
1
)
(7.72 )
rr;T == 0
T'hus gT and rr;i correspond to the ep'X of Eq. (7.
6
3) and rr; T and g" to the T:
x
.
]'he role of ; rr;:\) is then played by (g-}/' and rr;ijTT.
"[he Poisson brackets of the nc\\! variahles are
(7 73)
(7.75)
{g
T ('lA) ...... T I' ( - 2 (x 'lO')
A ,'" \A- J -- () \ ' ,A ,
{ n
i
(x) ,gk (x')} = I " (aa, - 2
i
2,) a(y , x')
2 \7- 2 \7-
(x) , rr;kITT (x')} == (x , x') ,
all others heing zero.
Th 1 I
TTl!. /) h' l' a . 1
e synl)O 0 lj In (,,7.75 represents t e Intcgro-c 111erentla operator
\vhich projects a synl111etric tensor onto its "1'-'T " part, nanlely,
(7.76)
ai/
ki
[(ili! (ajk. - + (il", __ (il
j
!__ ._
_ (a
kl
- (ail __ .
T'he operatot- (7.76) satisfies the foIlc)\ving relations
(777)
(7.78)
(779)
(7. 80)
(7. 81 )
fTT'j' === fTTl!
Jl ,
"TT.U TTl!
Oii == 2/ ==-= 0 .
(777) expresses the fact that there are only t\VO independent cano-
nical pairs JrkITT) per space point. Eqs. (7.73)--(7.75) one easily
checks that conditions (7.63) and 7.64) are satisfied. ['[he only result that
is not inl111ediately obvious is
:" (a
Zk
- 2i 2,.) f' = 0 => f' = 0 ,
'This is easily checked by lllaking appropriate contractions and recalling that,
\vith zero boundary conditions at infinity, (I Iv"!.) (0) == 01.
(7. 82)
- 120-
Next, let us \vrite down more explicitly the reduced Hamiltonian corre-
sponding to this gauge. Once gauge conditions have been imposed, the part
of the surface integral (7.56) obtained by setting ct.z" == == 0 and ct.
1
== I,
namely PI, is what one usually calls "the Hamiltonian". V\Te now write
I .
down P and leave PI and 1\;1 p.') to the reader. In the new variables the
surface integral defining pI is
I ., 12 l'
P [gij, n
l
)] :.: J d Sk (gk,/ - gil',k) == - J d Sk g ,I.:.
Following Arno\vitt, Deser and IVlisner (1962), we assume that, when conditions
(7.71) and (7.72) hold, the constraints can be solved to give
(7. 83a) gT == iT , n
klTT
] ,
(7
8
3
b
)
Then the reduced generator of pure tilne translations at infinity IS
pI reduced = - d
2
Sk gT
k
[gTT , n
klTT
] .
J ' I)
The rate of change of the dynamical variables under a surface deformation
which IS aSYlnptotically a pure time translation is
.1'1' 1''1' I reduced 1'Tkl 8pI reduced
glj == {gtj' ,P } == at)"
(7. 86)
8pI reduced
. k/TT == { k/TT pI reduced} == _ aT:rkl _
'it n, Z) 8g!.T
I)
with a;;Tkl and pI reduced given by (7.76) and (7. 84) respectively. The presence
of the" projection operator" in (7.85) and (786.) ensures that the right
hand sides of those equations are transverse and traceless. Equations totally
similar to (7. 85-86) hold of course for the changes generated by pz" reduced and
Mreduced
..I.. po
Finally, let us evaluate the Dirac brackets of the original variables gij,
n
kl
in the gauge (7.71 -72). The original definition (1.58) of the Dirac bracket
in1plies that, in the phase space coordinates (g1', n
T
), (gz" n'), 1t
kITT
),
the bracket is obtained sin1ply by dropping the terms involving (gT, 1t
T
) and
(gi' n') fron1 the original Poisson bracket (7.66). Sin1ultaneously, one should
consider the second class constraints (7.7 1), (7.72) and (7.83) as strong equa-
tions. This implies that the Dirac bracket between the 1'-1"' variables is given
just by (7.75). To find the Dirac brackets of the original variables, one then
returns to the deco111position (7.67) and proceeds as follovv'"s: ...\fter taking
(7.7 1), (7.72) and (7.83) to be strong equations one has fron1 (7.67):
'1'1' l'
(7.
8
7) gz"} == gu + it)" + 2 aij ,
(7.88) 7:
kl
=== 1t
kITT
-T- fk,l -t-fl,k r:;kITT -+- fUt
k
)
- 121 -
where /:5 is related to fT by (7.o9). rrhis gives, for exanlplc,
]
(7. 89)
f TT (, ') fTC )1* + {fT jT
-- t gnUl ,X , tj X J' /j \"1), Nm" ) J
where
(7.90 )
and
(7.9
1
)
Equations (7.89}--(7.91) express the Dirac hrackets of the g/j \vith the111-
sel ves in tern1S of the functionals fT defined by (7.83). The renlaining
Dirac brackets, \vhich \vill be left to the reader, contain also the functional fi'
The functionals fT and i/ 1l1ay be obtained in practice to any desired
accuracy hy the rnethod of successi ve approxi11lations. 1'he 1'-T' quantities
in the right hand side of (7.89}-(7.91) are to be expressed in ter111S of the
g/j , n
k1
by 111cans of the projection operator (7.76). The equations of
nl0tion (7.85)-(7.86) can be re\vritten in ternlS of the bracket as

\vhere H is gi ven by (7.56) and F is an arbitrary functional of the gl"j , n
kl
.
It is understood that in evaluating the right hand side of (7.92) one takes
the surface integral appearing in H outside of the bracket and perforn1 s the
integration at the end.
(b) Dirac's" hfaxinzal .Slicing" Gauge
The one paraI1leter fanlily of surfaces on \vhich the evolution IS being obser-
ved I1lay be fixed by the " I1laxinlal slicing" condition (*)
7t == nZ: 0 .
1-
(*) Equation (7.93) is calleel the" Inaxin1al slicing condition ", because on account of
(7.29) it can be equivalently \vritten as
K=o
where K is the trace of the extrinsic curvature tensor. NO\V the change under a deformation
dt of the volu111e \T = I d
3
enclosed in a three dill1cnsional region is
. .
d = d) (gl/2) d
3
X = - )1\1gl/2 Kd
3
x ,
by (B.S). This shows that if K = 0, the YOIUlnC of the three-surface is unchanged by a sur-
face defonnation. It lnay be shown that for a spacelike hypersurface ernhcddecl on a hyper-
bolic H.. iclnannian space one is actually dealing \vith a InaxinlUIll.
9
- 122-
C'ondition (7 is in \'ariant under changes of coordinates in the surface and
so docs not restrict the spatial coordinate s ysten1. 'The I-I an1iltonian \vi11
thercfore be only parlially reduced by condition (7.93). It \vi11 still contain
a term / N'If; and will still possess the three constraints .:If', 0 corresponding
to the frecdo111 to 111akc arbitrary tangential deforn1ations in the surface.
'1'0 achieve total reduction one can still in1posc three coordinate conditions,
hut \VC shall not \vorry ahout that here (see Dirac, 195
8c
).
})irac (I958c) realized, one 111ay rearrang-e the canonical variables
in the follo\ving \vay: (a) ()nc pair (ep ; rc) and (b) other pairs (g/i; if/i).
'1'he quantities under consideration are defined as fo11o\\'s:
(7 9-t)
I
loc:rg
J h.'
'rhey ha vc the foll()\vi ng l)oisson brackets:
{ ep , it (x')} =--= () ex ,X')
\vith
(7. 1(0)
and
(7
10r
) {
:::"i; (1'\ ::"'J:l (r'\!
II.' \_t) ,1\. '\" )j ,x') .
other Poisson brackets arc zero.
The quantity l:-/.i appearing in (7.96) and (7.97) is the in\'erse of the con-
fon11al 111etric 1;','.1' i.e. glm kim 6:, and is related to the full 111etric by
g,J =:= g,j. 'fhe" conforlnal Kronecker delta" defined hy (7. 100) has
the follo\ving useful properties:
(7. 102)
(7.
1
3)
(7roS)
o
123 -
Equation (7.102) says that there arc only fi\'e independent canonical pairs
(1:
1
.1 ' 7:(i) per space point, In accordance \vith the conditions
(7.
106
)
(7. 107)
f1"'rol11 (7.101) \ve see in1ll1ecliately that the ne\\7 phase space coor-
dinates satisfy thc conditions (7.6-t) of the reduction thcore111 ...:-\ccor-
ding to (7.66) and (7.98)-(7.101) the Poisson bracket is COll1puted In
the n('\\' \'ari a hIes usi ng the equation
(7. 108)
"
.:3 A [ of oC
{ F' , G } == d.1 00' oJ:-
6/i -
0(; of'
09 077
oC of'
- )
I (_ ....-.. 1,/ - 1,/.- ") ... 01.".. gA ..,...1\ (TIJ ..
-:;- \. \. . ---- i\. (,:>' - f:::"'z'i o:;:..ld
,.J Iw J '"
One can check fronl (7.108; that It and ic ha ve zero Poisson I)rackets \vith
everything and, consequently, they can he set strongly equal to unity and
zero respecti \'ely. 'rhus one can preserve the synl1netry in all indices e\'en
thougb one is dealing \vith n10rc variahlcs than needed [ren1en11>er s.
( 7. I 06) ancl (7. I 7) ] .
I n order to reduce the HCll11iltonian, \ve ha\'C to express the surface in-
tegral (7.36) as a functional of the the ne\v variables. \ve \vill just \vork
out pI and lea \'e pi and as an exercise for the reader. ()nc gets:
1,"'ron1 (7. 1 9) \ve sec that to ha \Te \vell-defined \'ariational deri \'ati \'CS of
pI reduced \\Tith respect to gl).:, \Ve ha \'C to get rid of thc gik,I' tern1. In other
\\Tords, the reduction of the I-I an1iltonian by 111eans of the nlaxin1al slicing
condition i: == 0 is not possible unless one chooses a n1Dre restricti\Te aSYll1p-
totic spatial coordinate condition than in (7.-t0b); \\'C shall require
(7. 110)
rrhe asyn1ptotic forn1 (7.39) IS thus not allo\\Tcd In this context, Inlt the linc
clerncnt
(7.1 I I)
(is .---.-----_.... _---+ -- (I __;\l __ ')
r--;>-CV , KT:r
(I
::\1 ) d -1' -1 .J
) o,'J' (.l',
KT:r / -
\vhich is obtained fro111 (7.39) hy a change of coordinates
(7. 112)
satisfies (7. I 10).
(7. I 13)
- 124-
Taking into account (7. 110), the energy heconles
I
Jl - 2 d
2
1/:3
- -- J Skg ,k'
()ne ll1a y check that, \vhen inserted into (7. I 13) the n1etric (7. I 1 I) correctl y
gi vespl == ]"'he final step in the reduction procedure is to solve the Han1il-
tonian constraint Xl. == 0 and express gl/:3 in (7. I 13) as a functional of the
cR"/j and ir/
j
. l'his leads to the follo\ving equation, first exan1ined by Lichne-
ro\vicz (194-+) and recently extensively studied by Choquet-Bruhat (1972,
1973) and hy 0' lVlurchadha and ''"ork (1972 , 1973):
(7.
11
-1-)
\'lith <D ,g-1/12. (Here Tclj = ff! gil.' ir/
k
and R, are, respectively, the cur-
vature and the Laplacian in the n1etric g/). 1"'he reduced H anliltonian IS
therefore gi ven hy
(7 I 1 5)
1
_Trectu(:Pd == _ ])"J.l'CclllCC(1 -t- 9"" "'1r,c<!uced + r 1
3
"z" ( ) r./} ( .\
1. J. , P H :J. 'J .. (J..i. " "x ./( z" \ A )
. h !)ll'CclllCP(}. 1 / '\ C' '1 . h Id f 1)z" d 1\1 l"'h
\Vlt gl ven )y (7. I 13). ,lllnI ar expressIons 0 or' an 1 / f'
constraints
(7.T 16)
still hold as first class equations. 'l"'he tangential generator appearing in (7. 1 IS),
(7.116) is also understood to be reduced by the condition 1t==0. It n1ay be
\vritten in the forn1
(7.
11
7)
\vith the covariant deri vati ve being taken in the nletric (g-z"/.
\ve \vrite the equations of n10tion for il') and fc(j In a n10re
explicit fornl. Fron1 (7.108) these equations are seen to read
(7. 118)
reduced
;. _ { "'" HredUCed} _ oH
gl'} - gj, - a l } oft
kr
--
..,
""' kl == { .., kl Flreduced l == _ 0 +
1t 1t, J lJ ocr"
0 1)
. h d H reduced. 1 ( ) d () . 1
\Vlt U z"j an gl yen )y 7. 100 an,7. I I 5 respectl ve y.
()ne n1ay ohtain the Dirac bracket corresponding to the gauge condition
(7.93) in a \vay totally analogous to the one followed for the rr_T gauge. 1"'he
situati,:)n is sitnpler here because the gauge condition T: == 0 is just an algebraic
equation. ]"'he I)irac bracket then turns out to depend on the solution of
(7. I 14), \vhich as \ve said above has been exan1ined in detail in the literature.
- 125 -
()nc thus considers (7.93) and (7.114) as strong equations and \\Titcs the l)irac
bracket as
(7. I 20 a)
\vhcrc
(7
120b
)
and
(7
120r
)
( -
l lIz}'
{ (pt
(I)l ( . "\ ") x-
, ,.l ) J
-.
Jd:1y (y)
I r X')

0<1>1 (.1"

-t- l-ilh(. '\ ::--(r!
C)' n \JJ) g
3
0<1>4 iX')
oi:(t! '
\\-here 101) and (7.108) have heen used. Cfhe righthand
side of (7. I 20) n1ay he expressed as a functional of the original \-ariahJes
it!'j ,r;/i hy n1cans of the equations gij === gz'j and Tel) =-= cg-
I
nil). '1'he
other Dirac hrackets arc ohtained in a sin1ilar rnanncr and \\-ill he left to the
reader. rrhc equations of n10tion (7. I 18), (7.119) then reaeI, in tert1ls of the
original \-ariahles,
(7. 121 ) nl'j == { TI
IJ
, H }*
\vith H gi\-cn hy (7. 115) ...:\gain in (7. 121) as in (7.92), one deals \\-ith the sur-
face tern1 in H hy taking the surface integral outside of the Poisson hracket.
'1'he constraints (7.116) ren1ain first class in the starred hracket.
2. CLOSED SPACES: York's G'auge.
\\7hen the three-surface on \vhich the state is defined is cOlnpact, all
inf()rlnation ahout the" location" of the surface in the en veJoping spacctilllC
is contained in the CR"ii ,TI(i). rrhere is in this case no aSYl11ptotic region and
no set of surface \-ariahlcs tZ"\ b'\; consequently no surface
integral ever enters into the H anliltonian, \\-hich reads sin1ply
(7. 122)
there arc no problcn1s no\v \vith H
o
not ha \'ing \vell-defined \'ariatiol1al
deri \-ati \cs.
I t is clear that the procedure used for fixing the gauge in the open case
will not \vork for the C0I11pact case since there are no surface integrals to deal
with. l\Ioreover, a gauge condition such as TI == 0 \vill not be a satisfactory
- 126
condition in this case as it \vould lead to no dynanlics at all (-1<-). To elinlinatc
the frcedolll to lllakc arhitrary defornlations of the surface, one nlust usc
second-class constraints \\'hich arc explicitly tin1e-depenclent. I-Io'Never,
there is no general schenle a \-ailable for deri ving the reduced Han1iltonian
in this case. ()ne kno\vs only that \vhen the gauge condition is of the forn1
qO -: then the reduced I-I anliltonian is -- Po a s111ay be \'erified by ohserving
that the right equations of nlotion are obtained.
F'ortunately, the only gauge condition so far proposed for cOlnpact spaces
is of the sinlple forn1 qO '7 and one can then carryon thc reduction proce-
dure \vithout difficulty.
York (1971, 1972) has proposed the condition
(7. I 23)
to fix the spacetinle slicing. ?\ ote that, as happened \\-ith the 7t 0 condition
for open spaces, (7. I 32) is in \'ariant under changes of coordinates in the
surface and therefore the constraints 0 \vill still relllain first class. ()nc
nlay get rid of these constraints hy ilnposing three nlore conditions besides
(7. I 23). \Ve \vill not carry out this step
In order to deal \vith condition (7. I 23) one needs a slightly n10difiec1
versIon of Dirac's \'ariables (7 94)--(7.97); one uses
T===
.2
({-II:! 7t
3
(7.
12
5)
p === _ gt:!
instead of ? and It gi \-en by (7.94) and (7.93 '1'he rest of the \'ariables renlain
the sanlC'. One then has
(7. 126)
{ '1' (x) , P (x')} === a(x , x')
instead of (7.98) and again, all other brackets relllain the sanlC. 'rhe })oisson
bracket is then gi \-en hy (7. 108) \\Tith ? replaced by T' and 7t replaced hy P.
'1'0 elinlinatc the extra degrees of freedonl one sol ves the equation .f(l 0
and expresses P as a functional of the renlaining canonical variables and of
the tin1e x
O
(the feasibility of this step is the rnain test of \vhether or not (7.123)
is a good gauge condition). Before gauge conditions arc inlposed the Poisson
bracket is gi ven by (7. 108) \'lith ? l' and 7t P. \Vhen the gauge condition
C*) The eq uation 0 ,,ould becoIl1e a second class constraint after the T: 0
condition is irnposed. :\ o"\v since the gauge condition is tilne independent, the lIalniItonian
would be unchanged; this Ineans that the of lnotion \\oldcl be giycn by
II' /' '\' / 4/ 1*
l " (X J(. /j ,
showing that there \\'oldd be no d ynalnics left except for changes of coordinates on the surface.
Thus 7': 0 cannot be usecl to fix the gauge for cOlnpact spaces.
- 127-
(7 I 23) IS enforced, one drops the P and '1' tcrn1S and the hracket 1)('C()l1H'S
\7.
12
7)
oC 81" ,
8;", cx:::./!)
,.... './ 0 ,
'1'hc reduced I-Ianliltonian then reads
(7.1 2 R)
I
1
:3 " T t ,u'
( X \ ,/{ z'
uation (7.
1
28) tells us that thc d ynalnicall y ilnportant part of the II anl il-
tonian is the \Tolun1c of the surface ./(;. is connected only \vith changes of
coordinates in the surface). In (7.128) gJ is a functi of ,{://
and 7:/:/, and dcpl\nds explicitly on the tin1C .1'0.
In order to express as a functional of the relnaining- \"ariahle's, one
re\\Trites the ,YfJ. 0 constraint in the fonn of a non-Ii ncar partial di ffcrenti al
equation:
\vhich is a generalization of (7. I I-l-), (\ork, 1972; 0' :\Iurchadha and \ork
1972 ') .
...-\s \ve cIllphasized at the heginning, the' constraints ./(, 0 still hold
as first class equations and can still he expressed in the fonn 7.
1
Ii;. '1'he
l)irac hrackets corresponding to the gauge condition (7. 123) arc the sanle
as those exenlpliflcd hy (7.120) corresponding to the it' 0 condition. '1'he
only change is that no\v the functional <t> appearing in (7.120) sol \cs (7. I 29)
instead of (7. 1 14).
- 128-
i\.PPENDIX i\.
:\lE'I'RIC CO:\\T:EN'fIONS
()ne of the lTIOSt ,"cxing problell1s encountered In \Vrltlng about relati-
vistic physics is the choice of a ll1etric convention. \Ve have chosen to usc
. 11 ')'):J;3 00 .. .
the ll1etrlc 1 === g === == g == - g In the 1naln text because the tranSI-
tion between nonrelati vistic and relati vistic canonical 1110111enta seems n10r('
straightfor\vard. This convention also agrees \,'ith n10st of the current litera-
ture on general relativity. On the other hand, it disagrees \vith the latest edition
of Landau and Lifshitz's Classical of Fieltls (1971) and \vith Bjorken
and Drell (1964). VV"e therefore hope that the follo\ving table sho\ving the
relation of our con\"entions to those of Landau and Lifshitz and of 13 jorken
and Drell \vill help in S0111e slnall \vay to n1ake this \vork 1110re useful to those
\vho prefer other ll1etric con ventions. Our con ventions for general relati vity
agree with those of lVlisner, '"[horne and \Vheeler (1973) pro\"ided Nc\vton's
gravitational constant G is set equal to (16n)-1. r1'he reader 111ay also find in
this reference an exhausti ,"e con1parison \vith other con ventions used in the
literature.
=== 0,1,2,3
0 0 0
,tV
0 -I 0 0
g -
0 0 -I 0
0 0 0 -I
13 1\:\1> IJRELL
(Li\:\l)j\l' LIFSI-IITZ, IF I)IF-
FERE:\T) }-IERE
== 0,
1 ,2,3
-1 0 0 0
ILV
0 0 0
g'
0 0 0
0 0 0
p === (_po pI p2 p:3)
lJ \ ",
p == (pO _ pi _ p2 _ pa)
p. \' , ,
'\
C
ds ===
U d ,1')
dx' xp.J-
d
( , d II d \
S == \ i- X - x
p
)
..:\.",'1 __ \'
I Ezjk Fjk == Bi
2
zli
A
- F
ik
= B
1
(+ zli
k
F
ik
= _ B')
- 129-
\Vc hayc also chosen a different null-plane nlctric frol11 that of I(ogut
and Sopcr (1970). ()ur convention has the ad \'antage that all cOll1ponents
of g!J.\J are positi yc even though \VC arc in a :\Iinko\vski space.
I-IERE
!-L
-- I , 2
0 0 0
IlV
0 0 0
(g-'
0 0 0
0 0 0
pll P!J. == 2 p; p-
IlV
/t
p
pP p,
) .
... B
KOc;CT A:\lJ SOPER
1.1
- I , 2 , -
0 0 0
0 --I 0 0
0 0 0
0 0 0
2
2 pip - -- = ;
Consider a spacelike hypersurfacc clnhcdded in a four-di1l1ellsional
spacetin1e of hyperholic signature. Let the unit norn1al to the surface he n
P
'.
Denote hy the suffix 1 the projection of a spaceti111e tensor of any rank 011
the norn1al (i.e. "'-\1 == ... A..!.L n!l == _ ...-\
1
and sin1ilarl y for higher order tensors)
and hy latin indices the projection onto the surface. rrhe foIlo\\'ing relations
the hold in the basis (n , eJ \vhere e/ denotes the tangent \'cctor to the i-th
coordinate line on the surface:
(B. I a)
(!)R I
abet! == \..ubed
(B. I b)
I-i:quat ions (B. I) are kn0 \Vn as the Gauss-Coda zzi equat ions (see, forexan1pIc
Eisenhart, 1926).
l'he syrnnletric space tensor K
ab
appearing in (B. I) is calle the extrin-
sic curvature or Sl!COJll! jUll{!afJZenta! jorJll of the surface and it has, fron1 the
I-Ian1iltonian point of yie\v, the inlportant property of being detern1ined con1-
pletcly once the surface is giyen. rThe extrinsic curvature is geon1ctrically
defined by the paraJIcl transport of the norn1a1. n
P
' along the surface. Since n!J
has unit length, the difference I)n het\veen n:J. at the point Xl' dx
l
' and the
vector ohtained hy parallel transporting Jl:J. fro111 Xl to Xl d,,1.
J
: along the
surface lies on the surface. One then defines K
ab
hy \vriting:
(B.2)
- 130-
'rhe left-h an:l side of (B. 2) n1a y be expressed in terl11S of the lapse and
shift functicH1s as fo11o\vs: First one nDtices froll1 that
ll, =
[.I.
0 ,0 , 0)
rninus sign In llo ensures that l/J. points in the direction of increasing Xu),
ITsi ng s. (7. 5 7.8) \\'(' fou ncl that (B. 3) inl plies
I
)' 0 )1-) 0 )1..,0. I} ---
. llb - lib llO C.1 --
=--= (2 :\1)--1 [(-- gab,O gOa"S l{Ob.il) _:\c gab." /Slil,b gch,a)J =-=
== (2 :\1)-1 (-- gah,O :\11.b -- 2 NJ =--=
=== - (2 l\d)-l (--- <-f{ab.O -i- N(/,b Ilia) .
(B5)
,!JIlII. ()
N
1,b
\\'hich In particular sh()\vs that 1'(//1 IS syn1111ctric.
'rhe follo\ving relations hold In the hasis ,n, ell):
(8.6)
(B.7)
r).(1 I). (J{/ II (J
g === C;l e - It J!,
ll, llt). == -- I
). ,
uation (13.7) is just (13.2) \\'rittcn in general coordinates.
f---roll1 (1).6), \ve get
(13.10)
Since the co\'ariant derivative of (13.8) is gi\'cn by
(B. I I)
(B. 10) reduces to
(B. 12)
'rhe last step in (B. 12) follc)\vs fron1 (B.7). 'This pro\'es (7.18a). Equation
(7.18b) is proved In an entirely analogous manner.
3. Jl0Jll.'lltUJJZ auel jl0lJlflllUJlI of the F/('!e!.
an illustration, let us sketch ho\v is iclentificcl as the linear
1110111cntuln. rrhc linear 1110lnentU111 is norn1allv defined as the conserved
quantity associated \vith in \'ariance of the action under translations. tran-
- 131 -
slation is an operation \vhich is defined only for a 11at space it a1110unts
to displacing the fields fron1 the point x
k
to the point x
k
2
7
', \vith Sk fixed
(independent of x) and the are cartesian coordinates. 'rhe correspon-
ding (nlore general) operation for a cur\"cd space is to displace the fields fron1
the point x
k
to the point x
A
' (x) in a coordinate-in\"ariant \,"ay. 'l'his is
achieved by substracting fronl a field quantity its L-ic deri \"ati \"C along the
vector field (x). O\V the action
c __ Ii" d \ )1' 13 ( 1)' " T' I ,// '\
.-, --- . tiC ..1: 7t g/j ---- ... , ,/7 11-)
./ ..
is invariant under such a transfornlation c\"en if behaves as r
i1
at infinity
(\vhich is the case for an asynlptotic rotation). :\ote that \ve ha ve set === 6
i
l
and == 0 in (7.) I) in order to get just a pure generating changes
corresponding to a l\Iil1ko\vskian tinle c1isplacclnent at infinity.
Kno\,"ing that the action is in \"ariant, the next step is to follo\v the lines
of Xoether's theorct11 and rearrange the ,"ariation of the action in the fornl
I 1 cl I 1
3
( if a
: C t elt , ( x\7t gil?
II
1f \ve then insert
(13. I 3)
into (B. 13), \ve obtain
tcnl1S vanishing hy the equations of 111otion/ .
'1'he second tern1 in (13.14) vanishes due to constraint -Yf;, o. \'That IS left
n1ay he transfornled into a surface integral so that as reads
1 5)
If \ve set Sk, a constant, \,"C arc dealing \vith an aSYlnptotic
translation. \\7' e can then \vrite
(B.16)

1 d
2

. === 2 J Si it .
Fron1 (13.16) \ve identify pl giycn hy (B. 1 r) as the total lincar 1l10nlcntu111
of the systenl. 1t is a constant of the nlotion hecause the action is in variant
under the transforn1ation in consideration, so DS in (B.16) is zero. It is iln-
portant to realize that only Cl == ((X)) provides nontri vial inforn1ation.
- 132-
'1'hc in\'ariancc of the action \vith respect to arbitrary (x) in the interior
is inlrnaterial in this context; enters in the equations al\vays nlultiplied
by ,10> as in (B.14), \vhich again reflects the close relation betv:een the invari-
ance of the action under arbitrary changes of the spacetime coordinates and
the occurrence of constraints in the Hanliltonian theory. The fact that only
(ex 11latters is the reason that the linear nl0111entU111 in general relati vity
is referred to as the conserved quantity associated \vith /f spatial translations
at infinity". Sinlilarly one speaks of the energy as being associated \vith
tinlC translations at infinity, and one associates the angular 1110111entU111
\vith spatial r()tations at infinity. 'fo find the angular nlonlentuln Lone
just r ... S;"/I oqi xi in the ahove reasoning and replaces (B. 16)
hy == oepi (L
i
(/
2
) - L
j
(11)). 'l'his discussion sho\vs also that the concepts
of energy, nl0111entunl and angular 1110111cntunl have no nleaning for a
closed uni \erse.
-to Relation /!.q. (4. 16) 10
'r0 reI ate J qs . (4. I 6) t 0 qS. I . 8J) 0 nc not iccs fi rst () fall t hat the func-
tion is the generator of reparall1ctrizatiollS a f (a). 'rhus in the notation
of (I.R3) \\ye ha "C
(B. I R)
On the other hand the generator of dcfonnatiolls along the nonnal to the string
eli ffers fronl hy a factor \\'hich is dctcrlnincd up to a convcntional sign
frOlll the condition:
for an arhitrary function !\ (a) .
1'his gi "es
(B.20)
\vhere :\ IS the nor111alization factor appearing in (4. I).
'l'he functions ,.IIi. and ,Jf'l ha ve the follo\Ning equal-7 hrackets
,
(13.2 I) {J. (a) , ,1{1 (a')} == (a) ,.leI (a) + (u
2
)-1 (a') (a')] a: 0(a - a') -
-- 2 :\---1 (0) ,le
l
(0") (a) -t-
22)
(13.23) {dili (a) , (a')} === [-.IIi (a)
':)
(a')] 0(0" - a') .
ocr
- 133-
Eqs. (B.22) and (B.23) have the san1e forIn as (r .83 b ,c). l'his should l)c the
case since these relations just characterize .YlJ. and as heing respccti\-ely
scalar and vector densities in the one-dinlcnsional space of the string. If \ve
notice no\v that the n1etric tensor along the string has just one c0111ponent
gll given by
11)-1 2
gIl == (g == u ,
\\7e see t hat (B. 2 I) agrees \\Ti t h (1.83a) but does so 0 UJeak . 1'hc discrep-
ancy con1es fron1 the second terITI in the right hand side of (B.2 I):
(B.25)
"'\r-
1
[( 2 ( ,,/J ( ) ,,/fl ( )
- 2 ..L' \U (J Jl 1 1 (J
') , , ,d ,
+(u- ((j )) .... ( (j ) yt'l((j )] 2cr 0((j - (J )
Note ho\vever that espression (B.25), being quadratic in the constraints, has
\veakly vanishing brackets \vith c\-crything. 'rhus the hracket relations
(B.2 I )-(B.23) still ensure that thc d ynanlical evolution of the string is "path
independent" in the sense of Section (I. ..:\. sinlilar analysis sho\v5 that
one can replace the factor in (!).20) by - (obtained by
solving == 0 for u
2
), and get an equally good ./fl. Such a 111odification in
eJfl. results in fact in the addition to the right hand side of (B.2 I) of another
ternl quadratic in the constraints.
- 134-
l{EFERE:\CES
:\HERS E. S. and 13. \V. LEE (H)/3)-- .. Physics Reports", 9C, I.
:\BRAH.\\I R.. and ]. E. \IARSI>E:\ (I9()/)- 1
1
(JltJldatioJls of J/et!lrlJlit's, BenJaInin, ;'\ew York.
.\:\IH:R"'O:\ J. L. (H)5 8) .. Phys. Re\'.". IIf. 9()5
:\:\DER."'O:\ J. 1... anll P. (H)51)- .. Phys. Rc\'.", 83, 101X.
:\R'\O\\'ITT R., S. l);':-';E;{ a'Ll C. \V. -- (;rruJitdioJl: ,-Ill Inlrodllt'tioJl to CUFri'J!t
l?eS(\lrr!z 1.1... \VittC'!l. EJ.I. \\'iley. ).;c\\' York.
:\R:\O\VITT R. and S. 1. FICKLER (19()2)-- .. Phys. Rey. ", I27, 1821.
BAIERLEI:\ R. F. (1962) -- .. Phys. Rc\'.", I26, 1864.
BERC\L\:'\:\ P. C. (1962\' -li.'J/()'t"!ojJedia of IYlzvsit's IS. Flliggc, Eel.] \'01. I V, 203, Spring"(r.
Berlin.
BERC.\L\:\:\ P. and I. COLDBER(; (H)55) .. Phys. R.e\'.", 98, 531.
BER(;\L\.\',\ P. C.. R. PI:-\FIELl>, R. SCHILLER and I-I. (1950) -- "Phys. Rc\'.", Su. 81.
BJORKE'\ J. U. and S. I). f:H)f)4) - I?('!(dh'istir ()lIflJlt10Jl J;it'ld.,', \Ic(;r<l\\'-l-lill, Incl..
).;cw York.
BRO\\'ER R. C. (H)72) - .. Phys. Re\'.". U 6, 1()55.
CHOC)l'ET-BRl'H.-\T (1<)/2\ ., C. R..\cac.l. Sc. Paris ". 27-f., ()82.
CHOQl' r:-r- BRl' H.\T ([ ()/ 3) .. S YlnpD:-iia n, 12, Istituto ).; azionalc .\lta \late-
Inatica, Bologna.
1)EI. E.. P. DI an(l S. Fl'BI.\'I (I<)/2) .. ;\nn. Phys. (:\. Y.) ", 70 , 3/ R.
UE\VITT B. S. (I<)5R) unpublished.
1)E\VITT B. S. (1959) unpublished.
DE\\'rrT B. S. (196/a) - .. Phys. Re\'.". 16o, III 3.
})E\VITT B. S. (I9()/b) ;, Phys. Re\'. " I62. ['95.
DIR.\C P. ..\.l\:1. (I<)4R) - .. Phys. I{c\'.". 73. 1092.
D I RAe P. A.. \1. (I 9-1-9a) .. Rc\'. \/Iod. Ph ys.", 2 I. 392 .
IJIRAC P..\. \1. (I<)49b) .. The Dynalllical Theory of Fields, Classical and Quantunl"
Canadian :Ylathclnatical Congress. Secontl SUIntllCr SClninar, he'lei at the L'niversity
of British Colulnbia (A.ugust. 1<)49);
IJIRAc P. :\. \1. (1950) .. Canael. J. :\lath.
n
2, 12<).
l)IR:\C P.:\.\1. (1951) -- "Canad. J. \lath.",]. I.
DIRAC P. :\.\1. (195Rai .. Proc. I{oy. Soc.", ,12-f.6, 32().
I)IRAC P. :\. \1. (195Rb) -" Roy. Soc", .12-f.6. 333
DIRAC P.\. \1. (H)5Rci .. Phys. Re\'.", II-f.. 9 2 4.
UIR\C P. :\.:\1. (19ho)- .. Proc. Roy. Soc." ...1257, 52.
DIRAC P ..\. \1. (I9()4) [Jfl'tun's OJ! ()uflJlhIJIl Jlt,t!;flJlics, Belfer School of Sci(ncc.
V' eshi \'a l' ni \'crsity, cw York.
1)IRAC P. :\. \1. (I9()9) in IJI!ysics: ]'ri('stf .\VJJljJosi!1Jl1 19h8, \'01. T, 53C). In-
ternational :\tolnic Energy .\gcncy. Vicnna.
L. P. (I<)2h; - RipJlZ'VlIlirlJl Princeton l'ni\'ersity. Princeton.
F..\DDEE\' L. (196C)) .. i \latelnaticheskaya Fizika". I. 3: rEnglish transla-
tion in and \Iathcinatical Physics ", I, I (1f)70il.
F.\n{)EEV L. and \T.:';. POPO\' (I<i)/a) -- .. Phys. L,ett.", 251/, 29.
FADDEE\' L. and POPO\' [()6/b)-" Perturbation Theory for Fields".
Kif'\' preprint ITF ()/-3() I-English translation ct\'ailahle as :\.\L-THY-5/ by D. (;orclon
and B. \V. LeeJ.
FUBI)[I S., :\. J. HA.:'\so.\' anp R. ]ACKI\V (19/3) - Phys. Rev.", IJ 7, 173 2 .
(;O[)DA.RD P., J. C. REBBI and C. B. THORXE (1973) - ":';ucl. Phys.", H56, loq.
GODDARD P., f\.]. H:\:'\so:\ and G. Po:\z..\:\O (1975), - '" Phys.", B89. 76 .
- 135-
II. C!a.l'.I'ita! .If{,t!laJlirs.
IL\.\"so\" .\. J. and T. RYLCE 1<)74\'-" .\nn. C\. '{ 87.
HO.l:\L\\' S.. K. KLCH.\{{ and C. TEITELBOI:\I (1()73a) .. \'aturc Phy". Sci.. 2-/-5, <)7.
II0.l:\L\\" S., K. KI'CI-L\R and C. TEITELBOI:\I ;:1(74) .. .. ..:\l1n. (\'. \ ... 111
JORI);\.\" T. F. and \'. (1()()3) - "Phy'-\. Re\'.". 1]2,
18
4
2
.
JO..,T R. (lC)()4)- .. lZc\'. \lud. Phy,-\.", ]6, 57
2
.
K.oca'T J. B. and I). E. SOPER (1<)70) .. Rc\'.", D 29
0
I.
\\'. and \\". 11<)75 .. :\ucl. Phy..;. 11100. lOr).
Kl'CI-I:\.}{ I":'. (197 1)- Pri1H'('tO!1 l'lli\Tr,-\ity Lecture \'otes, unpublished.
I":'I'CHA]{ K. C,l'lnlita! (!Ul.'ltiZ,ltioJl (;),rl7Ji/.1' , in of the 1<>72 B:llltl
SUnlll1Cr School. l)orclrec!lt. Reidel.
K.L'CI-I:\R I":'. (1974) -- .. Sourcc'-\ ill in preparation.
Kl',\L\IER \V. (J<Y()I .. ..:\cta I-/-, I-+().
\V. (1C)()6) -, .. Springer Tract'-\ in \Io(lern Physic'-\", 40, 107.
K.L: H. P. (I<i)<j) -- .. ;\nl1. I nst. I-{en ri Poi ". ,.J I 1, 3()3.
L.\XDAl' L. I). and E. \1. LIFSHITZ (1<j71) _.. il/(7 Classic: llleO(J' }'ir'!dI' , tran:-;latcd
by IV1. !LlIlwnne'-\h. ..:\cldison- \\'eslcy, Reading. .\Ll'-\sachusetts and London.
..:\. (19-+-+)- .. J. \:lath. Pure ..:\ppl.", 37.
IVIISXER C. \\'., K. S. THOR:\E and J....\. \\'HEELER 1<)73) --- (I'rfl7JitatioJl, \\'. Ii. Fn'cIn:lll
anel Co., San Francisco.
\:IOH.\.P.\TR:\. 1<.. (1971 a) -- .. Ph ys. Rc\'.", l) 4. 37
H
.
lVI0I-L\PATRA (197Ih) .. Re\'. ". l) -/-' 221).
:\10XCRIEV \'. and C. '1'1<: ITELBO I\1 1(72) .. Ph y'-\ Rc\'.", !> 6, q()().
\'E\\'TOX T. D. and E. P. \\'IC\:ER I ()4()) .. Rc\'. \lod. Phys.", 2I. -too.
O'\.lURCI-L\DI'L\\'. and J. \\'. YORK IC)73) .. J. :\Iath. Phy'-\.", [4. 1551.
PEIERLS R. E. (1952)- .. Proc. Roy. Soc. ", .12I-/-. 143.
PIR,\XI F. :\. E. and .\. SCHILD (U)50) .. Phys. R(,\,. ", 79. 98(>-
PIR.\.\"I F. :\. E., .\. SCHILD and IZ. 1952: .. Phy:.;. Rc\'. ",87, -+5
2
.
PRYCE \.1 .H. L. (1<)35) .. Proc. Roy. Soc. ", .J 15u. I()().
PRYCE \1. H. L. I<)4H\j .. Proc. Roy. Soc " . .. J 195, ()2.
R.\.CAE (19()5) .. Springer Tracts in \.Iodcrn Phy:-;ics", 37,28. Springer-\'crlag, Berlin
REBBI C. (1<)74) -- .. Phy:-;ic'-\ Report,-\", 12C. I.
REGCE T. and C. TEITELBO!:\[ -- ...\nI1. Phys. \'. Y "88, 2S().
SCH\\'IXGER J. (U) 5 Ia)-- .. Ph Re\'. ", 82, <) 1-+.
SCH\\'I);GER J. (195 Ih) - .. Proc. :\atl. .\cad. Sci. (T. S. ".37, -+5
2
.
SCH\\"I\"GER J. (1953) -- .. Phys. Re\'.", 9[, 713.
SCH\\'IXGER J. (H)()2a) .. Re\'.". 1 ..!:=;.
10
43.
SCH\\'I\"GER J. (IC)()2h) .. Phy:-;. Re\'.". [27.3
2
4.
S. ([<j()3) --- .. Proc. C:unh. PlJilo:-'. Soc, ".59. 743.
SHA:\\Il"GADHAS.\.:-: S. (U)73) -- .. J. \lath. Phys. ", LJ.. ()77.
SV\IA);ZIK K. (IC)! r) - ,. on L:lgrallgian ()uantu1n Field Theory", unpllhlished
IJES\l Internal IZeport T--7 1--I.
'rEITELBOIl\I C. (1973;1; -" The I-I:zllniltonian Structure of Ph. I). Prin-
ceton t" ni yersity.
TEITELBOI:\! C. (IC)73b) ...Ann. Phys. (\'. Y ",79.54
2
.
TO\IBOCLIS E. (1973)- " Phys.Re\'. J) 8. 273(>.
l'TIYAl\JA H.. --- Phys. Rc\' .. lOT, 15<>7.
\VEI);BERG S (IC)73)- Phy'-\. Rc\'. ", J) 7, Io()R.
YA);G C. X. and 1<,. \lILLS (195-+) - .. Phy'-\. Rc\',". 96, I<)J.
\TORK J. \V., (1972) - Re\'. Lett. ", 28, 1082.