You are on page 1of 18

Nuclear Physics B267 (19~6) 125 142

' North-tlolland Publishing ('ompan?,

GAUGE FIELD THEORY OF COVARIANT S'I'RINGS

Michio KAKU

( "try College of the ('itv L:mrer~t(l' o / ¥ o ~ York, .V.Y., 10031, L.S4

Received 22 Jul,, 1985


(Revised 8 October 1t.)85)

We present a gauge cot,arrant second-quantiTcd field thcor', of strings which is explicitl?,


in,.ariant under the gauge transformations generated by the Virasoro algebra. Unlike the old field
Ihcorx strings [1] this new formulation is Lorentz cmariant as well as gauge cmariant under the
conth'mou~ group Diff(S I ) and its central extension. We derive the free action: I. = qq X ) ; l ' [ i &
1t, 1}] I'Cl~ ~'1. in the same ~va,, that Fevnman derived the Schr6dinger equation from the path
integral formalism. The action ix manifestl'v mvariant under the gauge transformation 6~(,V) =
V~., i~ ,, 1. ,,q~(X), where P is a pro)cction operator which annihilates spurious states. \\;e gixe
three distinct formulations of this operator I' to all orders,, the fir,st based on extracting the
operator from the functional formulation of the Nambu-Goto action, and the second and third
ba~,cd on inverting the Shapovalo,, matrix on a Vcrma module. This gauge coxariant fornlulat{on
can bc easily extended to the (;reen-Schwar.,. superstring [2.3]
One elegant application of thc:,e meth~x:l> ix to re-express the old Neveu-Sch,aar-z-Ramond
model as a field theory which is manife,stly invariant under space-time super>), mmctric tran~forma-
t{Olls

!. Introduction
The original Nambu-Goto action, which is proportional to the area of a two-
dimensional surface swept out by a relativistic string, contains a remarkable gauge
invariance, the two-dimensional reparametrization group. It is this local gauge
invariance which eliminates all the ghosts in the theory and makes duality possible.
The Nambu-Goto action, however, is a first-quantized action expressed as a
function of the vector X;,(o) which labels the position of the string. As a conse-
quence, it is an awkward formalism in which to study the counting of perturbation
diagrams. In practice, we must sum over all possible two-dimensional topologies
representing the breaking and re-forming of strings. This requires tediously' using the
umtarity rules to avoid the problem of double-counting. No such problems occur in
the second-quantized field" theory of strings, where the interaction terms in the action
uniquely generate all perturbation diagrams with the correct weights.
The string theory can be shown to be equivalent to an infinite component field
theory with arbitrarily high spins and masses (in fact, it is probably the only
consistent infinite component field theory because its interactions are muhi-local). In

125
] 26 '~1. Kal~u , ("¢wartatlt strHl,~

the first-quantized formalism, however, this deep relationship between a string


theory and an infinite component field theory is difficult to show.
Years ago. we [1] wrote down the second quantized field theory of strings defined
in terms of a field ~ ( X ) which is a functional of all possiblc vibrations of the string.
The theory is formally unitary, explicitly contains all interactions of .,,p[itting and
joining strings (which must be put in by hand for the first quantized Nambu-(}oto
action) and the correct counting of perturbation diagrams is achieved without
double counting.
This field theory of strings, however, was defined only in the light-cone gauge,
where Lorentz covariance and re-parametrization invariancc arc both lost. Recently.
a BRST formulation of the field theory of strings [4] has been written down. but
again the Virosoro algebra is missing.
In the zero slope limit, it is easy to show that the light-cone field theory of strings
contains terms like A, 82,4'. Notice that both Lorcntz invariancc and U(1) gauge
invariance are lost. Furthermore. the BRST action yields terms like A,, 8-' A ~'.
Although the theory is relativistic, it does not possess U(1)invariance. What we
desire, how~ever, is a gauge covariant action which yields the Maxwell action

The key to this entire construction is to notice that the Maxwell action can be
re~vritten as
l. = - ~:A P" 8: P"\A x. (1.1)

where P;',, is a projection operator which projects out transverse modes: P~,,,~ 8,,,, -
88,,/8"-. This projection operator guarantees that the action is invariant under
8Al, = 8,~,.
The action for the free graviton can be writtcn in cxactly thc samc manner:

L = p.,p,,.] h,,t,. (1.2)

which is invariant under 8h~,,,= 8 A,. + 8,.A,.


This can be generalized to all spins of an infinite component field theory through
the field thcor\. of strings.

2. From first to second quantization

We begin our discussion by extracting out the second-quantized theory from tile
first-quantized theory in exactly the same way that Feynman originally extracted the
SchrtSdinger equation from the first quantized theory. We begin with a point particle
whose action is given by ~mX, 2, where X, is a three-vector which simply extends
from the origin to the location of the point particle. From this first quantized action.
Feynman was able to derive the Schradinger equation, which is formulated in terms
M. Kal~u / ("orurtant.strmg~ 127

a function of a field q~(X) which is a function of all possible values of X. We begin


with the Green function connecting a particle that originally started at point X, and
t, and then propagated to )(1 and t,. We will eventually slice up the interval into
finite intcrvals i, i + 1, i + 2 . . . . . j. This Green function G(X,,t,: Xj, t~) can be
written either in first or in second-quantized formalisms, i.e. either with the position
vector X, as the independent variable, or with a field functional O ( X ) composed of
all values of the position vector. We will use the following to deriPean expression for
L(q~):

fv'DXexpi f L( X)dodr= f Dq~*(X)Dq~(X)rb~(X,)rb( X,)cxpi[f l.(rb) DX].


If we take the lagrangian to be ~rnX,-. then this expression can be explicitly'
I ° "~

evaluated by actually performing all gaussian integrals [5]. We find that this Green
function, in one dimension, is equal to:

[2~ri(t,-t,)lm] i/2exPl~i,,i( X,- X,)2/(t/-,,)]. (2.1

This expression can be viewed as a matrix connecting variables defined at point ./


with variables at point i. As a result, we can use the gaussian integral:

(A /),,-fDOD{b*O,@,*exp(- Y'[{I),,A,,,,,@,,*,]). (2.2)


###11

where n and m label an infinite number of parameters. Using the observation that:

(;( x,.,,, x,.,,} '= ,: (2.3)

wc now can show that:

L ( O ) = q)( X)* [ i / / , - J_~rnV'.a2] q~(X). (2.4)

Notice that we have now smoothly made thc transition from the first-quantizcd to
the second-quantized action, deriving the SchrSdinger equation from the action of a
point particle ~mJ(,'.
The conclusion one draws from this trivial cxample is that, in principle, one can
always rigorously make the transition from first to second quantization if one is
careful to define the path integral correctly. This observation can now be extended
over to the Nambu-Goto string where 1.(X) = l"g& -g~xJgll and g{~ = ) ~ 2 ; , gm =
•' % . . . . . . . . . . . .

XX' and gii = X'2. The application of functional methods to the string model is
straightforward [6, 7], although there are non-trivial complications introduced by the
presence of a gauge symmetry. Unlike the action for the non-relativistic point
12~ .~1. Kaku ,/ (ot'artallt .~trmgs

particle, this action is manifestly invariant under re-parametrizations of the two-


dimensional surface swept out by the string o ~ a ' ( o , r ) , r ~ r ' ( o , r ) . This gauge
symmetry necessarily implies that there are Dirac first class constraints in the theory.
if we define the conjugate momentum

8L 1/2 t
e,.=- ax,'- (-x) [~,,,x;-.~,,~.]. {2.s}

then our first class constraints read:

P/+ (aoX,,)-~-o. e,,aox,.=o. {2.6)

('onverting to the second order formalism, we can show that this Green function can
be written as:

(7( x,. T,: x,. ~,) = f " ' D X DP


A;
I)X l)pA exp i.:tI fi.( p, X)dod'rj (2.7)

L ( P, X ) = P ) ( + ~X( P 2 + X '2) + t ' ( P X ' ) . (2.8)

(3 represents the measure and possible Faddeev-Popov determinants, which we keep


unspecified.) If we functionally integrate over X, p, and P , then we arrive back at
the original N a m b u - G o t o action. This second order action is explicitly invariant
under a local gauge rotation with parameters t ( o ) and rt(o):

ax,,=~.p. +.,x;. a¢.=ao(~x,;+.p.).


8X = - ~ + X'*I - fl'X + p ' t - e'p,

8p = - i l - e'X + X't + p'~l - ~fo. (2.9)

This local gauge transformation guarantees that we can eliminate all ghosts.
]'he transition to the second-quantized theory is now made by slicing up the
interval from i to j into small finite intervals. Labeling each ~" slice by k = i. i + 1.
i + 2 . . . . . . We can now make the transition to quantum operators for each slice:

"
17 DX~ DPkexpi
/, t
[f doP~., x,~' - x,~'., )]

/ 1

= f H <x, lp,) Dx, l)e, <p, lx~ ,). (2.1o)


,'t,I. Kak u / (m'ariunt ~'trings 129

At this point, we will find it convenient to introduce harmonic oscillators in order


to obtain an explicit representation of the ket vectors IX) and IP). We will label the
harmonic oscillators by the index n. Then the ket vectors can be represented by an
infinite product expansion over harmonic oscillators:

. -- -:X~.,,- ,,a;, _,a,,a,,)[0>


n = l n = l

- - _,P~ .,, + '¢2 P~. ,,ai;-'~_a,,a,,'t)10 > . (2.11)


n = 1 n = I

[a,i.a*,, ] =6 ...... (2.12)

( P I X ) = f l 2-L/2exp(-iP,,X,, ) , (2.13)
tl = 1

(PIP')= f l (wa( P,i- P,[),


tl=l

( xIx') = f l d-rra( x,i- X,;), (2.14)


li - L

....
1= rr ~/21P,1)dP, l(P,, I = w "-IX,i)dX, i(X,i I
# = tl=l

= ~ ~,l;<qO)(Ola~(q?) ' (2.15)


n=l q~l

e,,IP) = 2-'/2(.,, + a,tl)lP).

x,ilx> = i2 '/'-(a,,- a?;)lX), (2.16)

a,,[X) = - i 2 '/2( X,,- O,,,)IX ) , a},lX) = i2 ,j,2( x,, + a,. )Ix>.

a,,IP) = 2 ~/2( p,,_ Or,,)lp)" a~ilP) = 2 ,J-'( p,, + a,, )1/>>. (2.17)

(To simplify these expressions, notice that we have explicitly deleted the space-time
index ix.) To prove relations like (2.15), one can power expand the ket vectors in
terms of polynomials of the harmonic oscillators, and then explicitly perform the
integration over X. Other relations can be proven by power expanding these
130 ~.1. K~t/,,u / ( .vartant ~trmg.~

expressions in terms of Hermite polynomials:

H,,(x) - ( - 1)'%xpx 2 O,qexp( - x 2) = exp( ' , x - ' ) ( x - O , ) " e x p ( - ~,x-' ). (2.18:

(Ol.,,"lX,,) = [--,~lll'T
: 1 q( XI,-- a. ") <OIL,> = • ,, [ - l "~ . Ill 21 ]"cxp( I ~X,~)ltq(X,,).

(2.191

& r S ( P , , - - - t , )) • = (P,,IP,,)i = q p
(P,,la,*,'qO)(Ola,,lP,,)(q!)
q= 1

= ~. v " 2 2 " ( q ! ) ' H q ( P , , ) e x p ( --, ! P,;~ ) II,( P,,' )exp( - I,P,,': ). (2.20:
q= I

expi(P,,X,,) = ~. iu2 u
l/2(q!) 1It,( P,, )exp( - ~.P,, ) tt,( X,, )exp( - I, X,~).
q. I

(2.21 ',

Sandwiched between the various slices, of course, we must insert 3( p2 + X,2 )3(PX')
which represent the first class constriants taken at each slice. Now let t,s find an
explicit operator expansion of q)(X) itself. Let us define Iq~) as an arbitrary state
vector in the Hilbert space of harmonic oscillators. Obviously. this means

t 't i ' . i i
Iq,) -- ~(-,,o)10) + A ( x , , ) a l , o) + g," (.~,,)a,, a~,. +- -.- . (2.22,

where the state Iq~) contains the tachyon, the photon, a massive graviton, etc. (It
rcf. [1]. we also introduced the vacuum state for the individual quantum field.,
denoted by 10)). To be precise, the state Iq~) must be powcr expanded around a
vacuum state which is the product of I0) and 10))).
If we now take the matrix element of (~1 in (2.22) with I X ) in (2.11). then wc
arrive at the usual expansion of the field q~ in terms of Hermite polynomials [1]:

H o ( X ..... ) e x p ( - 2
it = l
~ = 1.26

+A,'(xo)H,(X,¢)exp(- ~X, 2 }

FI no(X,,.o)exp(- + .--. (2.23)


•~l = 1
a= 1,26
.~t. Kaku / ('or'arrant string.* 131

Notice that the definition £5( X ) - ( ~ l X ) allows us to make a rigorous transfor-


mation from the first-quantized formalism to the second quantized formalism
starting from Feynman's original path integral expression. There is no more guess
work or intuition involved in going from the first-quantized formalism to the
second-quantized field theory.
At each time slice labelled by i, the local gauge invariance guarantees that we have
enough symmetry to eliminate the ghost states. In Dirac quantization, we can apply
the first class constraints directly onto the state vectors:

f d o e x p i n a [ j b 2/, + 2 ~ X ~,+ X,211~) ___L,,lq~ ) = 0 (2.24)

We now use standard Gupta-Bleuler techniques to complete the covariant quanti-


zation of the Hilbert space of the string. The important observation made by
Gupta-Bleuler was that the spurious particle operators L ( + ) and L ( - ) , which
remove the unwanted gauge mode of the Maxwell field, cannot both be applied onto
the physical Hilbert space. In fact, they found that applying both operators
simultaneously destroyed the entire Hilbert space. In particular, they found that only
applying the destruction or negative frequency part of the gauge operator was
sufficient because it still preserved Lorentz invariance while killing the ghost states.
Likewise, only using one set of Virasoro operators still preserves conformal invari-
ance while eliminating the ghost states.
Functionally, the origin of this choice is that we only demand that the expectation
value of these operators vanish, where the L's are sandwiched between states, so we
are free to apply negative frequencies to the right and positive frequencies to the left.
This means that we are allowed to extract out all the spurious states out of [@)
with a projection operator that we will call P. We will now give three distinct
formulations of this operator P:
(i) Using the methods pioneered by R.C. Brower and C.B. Thorn. in writing down
a Hi]bert space projection operator* P = Y ~ i R , ) ( R , ] summed over all real states
which satisfy L , , ] R , ) = 0 (Notice that, because this operator is explicitly off-shell, we
cannot use the D D F states and the Brink-Olivc-Corrigan-Goddard projection oper-
ator constructed out of such states because they are not Lorentz covariant.)
(ii) Using the techniques pioneered by B.L. Feigin and D.B. Fuchs, in writing
down a power expansion over Virasoro operators [9] composed of arbitrarily large
products over the elements of the L's, P = E L , F , , ( L o ) I . , , + . . . .

* Nee Browcr and Thorne [SJ. where the on-shell projection operator i~ constructed I am inf,,~rnlcd b',
Thorn thai the,, also considered the off-shell ca~,e a~ ',.vdl, and that in that paper they originally
introduced the concept of the Shapovalov matrix in their work (private ccmmmnication).
132 M. Kal,u / Cm'ar/ant ~trtngs

(iii) In terms of string operators by extracting the projection operator directly


from the F e y n m a n functional at each r slice.

3. First method: projection operators for each level


It will be helpful to first define some mathematical concepts from the theory of
a ffine Lie ( K a c - M o o d y ) algebras [10]: let the highest weight ~'acuum eector be a stale
JR) which satisfies L,,IR) = 0 for all positive n. We know that w e can reproduce the
entire representation associated with IR) if we apply all possible lowering step
operators. Similarly, let the universal era,eloping algebra U(l. ,,) consist of all
possible products that we can form from L ,,. A typical element of the enveloping
algebra is L / = I.~,11.~'-2... I.~_,,,,~ U ( L ,,).
This construction is a simple generalization of familiar techniques found in the
theory of finite Lie groups. For example, for 0(3). the highest weight vacuum vector
is simply Ihn) where m = / . Notice that the application of a raising operator on this
state yields zero: 1 . Ill) = 0. Now. by applying I. onto this state, we can extract
all m e m b e r s of the multiplet of states (singles, triplets, etc.) The universal enveloping
algebra for 0(3) only consists of the lowering operator L .
N o w let p ( N ) equal the number of partitions that we can make of the integer N.
Let 1 equal any integer from 1 to p ( N ) . If we let N = I11 = Y",' it?t,, then 1 simply
counts all possible partitions of the integer N = Ill. Now. the set of all states created
by applying the universal enveloping algebra (the set of all products of lowering
operators) to the highest weight vacuum vector IR) merely creates the entire
muhiplet, called the Verma module: V(h. c) = U ( L ,,)IR) where L,~IR} = h l R )
and c is the dimension of space-time appearing in the universal extension (Schwinger
term). Notice that all the vectors in the multiplet or the Verma module are spurious
states except the highest weight vacuum vector I R). Using the notation of Feigin-
Fuchs. let us now define I I ) - L /10). Then we can define the Shapo;'aloe matrix as
[111:

s,~(p-') = ~11J) = ~01L,/..,10) (3.1)

for ] I I = IJI = N and h = p2. Given the Shapm,aloe matrix, we can now write down an
expression for P to all orders. Let us define pR,, as the projection operator which
projects onto real states at the nth level, while pS, is the off mass shell spurious
particle projection operator at that level. For example.

P ~ " = 10>~01. (3.2)

pR, = ll _ pS:. psi = ( 2 p -~) li. IPR"I.t. (3.3)


,'vl. Kaku / ('ot'ariant strings 133

One can now take this to the second level:

pR: = 1, - pS,, (3.4)

VS:= [4p 2 + ~D]aL 2 ,PR"L~--6p2A[1. 2 ,PR"L 2 + 1. eP"'L}I


+4p2(1 + 2p2)A[L_ 2P"'q.2] + (2 + 2p2) l L xpR'LI

2
= y" ]1)2{S l)2tj(p2)(J]2+(2+2p2)L ,PIqL l, (3.5)
l.J=l

a - I s . I -~ I 1 ) , - / ~ - ' - 110), [2)2~ L 210), (3.6)

S~j=[ 8p4+4p2 6p2 ]


(3.7)
6p 2 4p 2 + , 'O =(II2lJ)2"

It is now straightforward to take this to the third level:

p~, = 13 _ pS,. (3.8)

3
pS,= y' 11>3(S l ) r ; ( p 2 ) ( j [ 3
I,.1= 1

2
+ E ;. . ,P",L,(S '):,,( p 2 + I J + ( 2 p 2 + 4 ) L l P a L l (3.9)
IJ= 1

s,~, = ( I I 3 1 J ) 3 . 11)3 ~ L 3 ~10). 12)3 = L l L 210). 13)3 -= g 31o),


(3.1o)

24p2(1 + f ) ( 1 + 2p 2) 3 6 ( p 2 + p 4} 24p 2
36(p2+p 4) 8p4+(34+D)p2+2D 16p2+2D (3.11)
24p 2 16p 2 + 2D 6p 2 + 2D

IS3jI = 24p21S~;l[3p4+ ( D - 7)p 2 + D + 2]. (3.12)

By examining the iteration in lowest orders, we can now easily generalize our results
to the arbitrary case: calculating the nth level from the n - 1 level (levels are ranked
134 M. Kaku /" ('ocariant ~tt'mg.~

by the eigenvalue of the number operator on the Hilbert space):

P= ~. PR'= ~ 1.v-PS', (3.13)


V-O \' =0

k" p IN - .~.1 )

E E c (s "( p : + ,14)pR"Lj. (3.14)


.~! = 1 I. J = 1
I l l = i J I = ~.t

It is easy to show that p2 p, PlR) = iR), ( R i P = ( R [ , and that (RII,,,P =


=

PL ,,]R ) = 0 for any highest weight vacuum vector ]R ). Notice that iJ the Shapova-
hm matrix is known, then we know P to all orders. Since the divergences of the
S h a p o v a i o v are well-known and have been isolated by Kac [12]. we know that it is
possible over a wide range of parameters p2 and c to construct the @shell
projection operator.
It is easy now to see that, at least for the free field theory, this reproduces the
gauge invariance of the Maxwell field and the graviton field (when we generalize to
closed strings). This is because L i = k"al,. If we let [q)) = A(x.)~,a]"10) + - . -
and I ~ ) = A ( x o ) [ 0 ) + " " , then the relation

ale5 = ~ ~ ,,L ,,i,b5 (3.~5)


tt=l

to lowest order becomes:

81q~)=aA"ai lO) + . . . . e ,L_llb ) + .... L l&'l(xo)10)+ .... (3.16)

This implies that 8A. = O~A for the open string and 8h,,, = i),A,. + 0,,.'l for the
closed string case. In other words, the symmetry generated simply by L ~ repro-
duces the gauge symmetry of the free Maxwell and graviton fields.

4. Second method: power expansion in U( L .)


Rather than expand P in terms of bra and ket vectors ~'IR,)(R,I. we can also
e x p a n d in polynomials in the L ' s themselves, i.e. in all orders in elements of the
universal enveloping algebra and its conjugate. Following the method of Feigin-
Fuchs. who created a generalized Casimir operator by this same power expansion in
L's. we can show that P = 1 - pS and

P=I-I. ~(2L,,)'LI-L._2(41.o+2)AI.2+3L..AL ~

+ 3 L ~ A L . ~ + I. 21[[4Lo(lq,+ l)] '-(4Lo+ l,_D)(21.o'a)]L~ * . - . (4.1)

A-(16Lo+2(D-5)L,,+D) ' (4.2)


M. Kaku / C m , a r t a n t string.~ 135

This easily generalizes to:

p S = ~,2 P~"' }
~. L ,/-).~( L o ) L J . (4.3)
N=I l,J=t
II1= J l - : %

Ftff(L{,) --- [ S H . - A)i 2 t( Lo)]{S t)t).,(L,,). (4.4)


V 1 p ( ,if )
E ~ L ,£)M(Lo)I.jL .,.]R) =-L_pA)~&.' ( L I , ) I R ) . {4.5)
'~! = 1 1, J = 1
I! = I J , = s

whcre St.~( Lo)IR ) -~ L~LjIR ) and A x is determined iteratively from A x 1. Notice


that we have, to all orders, an expression for the projection operator in terms of a
simple power expansion in operators composed out of the universal enveloping
algebra and the Shapovalov matrix.

5. Third method: extraction of the operators from the function',d

We only sketch the last method, which consists of extracting directly out of the
Feynman functional the operator expression which corresponds to the delta func-
tions we found arising from the first class constraints, the Virasoro conditions.
Since the transition from Feynman path integrals to harmonic oscillators is not
that transparent, it will be first instructive to do a simpler case, where wc have
explicitly fixed the gauge and extracted out the field theory lagrangian. In this way,
wc can rigorously construct the field theory lagrangian from the Feynman path
integral. Let us now, as an exercise, break the Virasoro gauge invariance by choosing
a gauge.
Choosing the conformai gauge X = 1. p = 0 in our prcvious calculation of thc
gauge invariance of P and X, we can calculate the Faddeev-Popov determinant
associated with the gauge invariance given by (2.9). We are left with - 8 X = ,~ + r/'
and - 8 p = ~ + g. When we calculate the Faddeev-Popov ghost associated with this
gauge, we find a new term in our action:

Jvpfd~dna[X{~,n) - 1 ] 8 [ p ( ~ . n ) ] =- 1. {5.1)

8X 8p

A I'P ~ d e I
8X 8p

-0~ -3.]
-- det
-0o -3,
{5.2)
whcre 1.(0. Po) is given by:

p~{ a. + i~)o' . pg;( a , - ao)0". (5..~)


We see. therefore, that the net effect of choosing the conformal gauge is to allow two
ghosts to propagate in the theory. The two Faddeev-Popov ghosts associated with
the conformal gauge therefore effectively reduce the 26 dimensional excitation.~ of the
string ~h~wn to 24 excitations. Because the ghost particle now contains a "r derivative
in the action, we must also let it contribute to the bra and ket vectors when we time
slice the functional. Our Green function now becomes G( X,. 0,: X,. 1.0,. 1) which
can be rewritten as:

(X,O,[Plpo)fOP, Opo(Plps[X,. 10,. l)exp i.(f d~ [ p,2 + (0X)2+ po, pOoO,]f.

where we have combined the spinors imo one Iwo-dimensional spinor and p is the
Pauli spin matrix o.. In order to actually perform the integration over the canonical
momenta, we will find it useful to first calculate the matrix element between two
ordinary string states:

( x, lx,') = 1-I J~, 8( x,,,- x,;,) = ( X,l~,) f D~,,*, Dq~,,exp - ~¢b,*q~,, (qJ[A',')
tl It J

Now modify this matrix element by inserting e ';tjT between the state vectors:

( X, le'"JTI X,. 15 = f Dq~*( X ) l)q)( X ) q~( X, )~ q~( X,,, )

× e x p i [ f DX { ~ i ( X , ) [ < b ( X , ) - ~ ( X , . l)]

+ i~)( x , . , ) ' l n x,., )~( x,,, ) J~ }].


Notice that in the continuum limit, this expression reduces to the lagrangian
for the Schr6dinger equation for an infinite number of free fields: 1. =
cb(X)t[iO~- H] ~ ( X ) where the expression for the hamiltonian is just L . - 1 and
where we add in the contribution of the two ghost particles in the hamiltonian.
The point of this lengthy exercise for gauge-fixed string actions was to show that
the transition from first-quantized to second-quantized actions is a straightforward
.~/. Ku~,u / Corarium .~'ml~.~ ! 37

and rigorous process if we are careful in manipulating the path integral formalism.
In summary, we can convert any expression in the path integral into operator
language via these techniques, using the realization that ~ ( X ) = ~ I X ) and using
the identities (2.11) to (2.21).
In the case of the gauge covariant theory, we proceed in the same way, systemati-
cally converting c-number expressions into operator expressions level by level via
(2.11) through (2.21). These equations make possible the rigorous transition from
c-number quantities defined in the path integral approach and q-number operators
appearing in the second-quantized field theory.
In the gauge covariant theory, we cannot simply invert the Green function,
because it is singular. However. the important observation is that the net effect of
having the terms a [ P : + (OoX)2]6[PO~X] sandwiched between state vectors is to
eliminate the ghost terms level by level, which simply reproduces the projection
operator. In effect, we can take the hamihonian in the new second-quantized field
theory to be the Dirac delta functions, which have the property of eliminating the
spurious states from the theory. We find that the transition from the first-to
second-quantized field theory is accomplished by converting the Dirac delta func-
lions, which are written in terms of c-numbers, into q-numbers. Notice that when we
introduce a complete set of intermediate states into the theory f J P ) D P ~ I ' I and
f l X ) I ) X ~ X I , the presence of the Dirac delta functions restricts the integration
region of I)P and DX. Therefore, the Hilbert space is smaller than before. If we now
insert a complete set of interdiate states given by ./Iq~)D~b~I times a measure term,
then we see that the state I~) must also be restricted. In particular, Dirac tells us
that we must apply the restriction L,,[q)) = 0 onto the state vectors of the theory. In
practice, however, converting the Dirac delta functions from c-numbers to q-num-
bers is quite difficult, and has only been accomplished for the lower lying spins.
Although the problem is, in principle, a straightforward one, in practice the
problems arising from normal ordering are quite difficult when extrapolating to
higher levels and higher spins. We have all the identities, from (2.11) to (2.21), to
make this transition, but in practice the calculation i,s quite tedious and not very
illuminating. In practice, the transition from c-numBers to q-numbers for the Dirac
delta functions kills the spurious states let,el hy let,,el in the theoo', so in essence we are
simply reproducing the projection operator that we found earlier. Rather than give
these teciious results, which in effect reproduce the projection operator level by level,
we will give motivating reasons for believing that we can reploduce the projection
operator to all orders. We first notice that, following Dirac, we can apply the first
class constraints onto the state vectors of the theory. This means, of course, that the
functional integration over the states IX) and IP ) is restricted. The integration DX
and D P does not extend from minus infinity to plus infinity, but is constrainted b~
the delta functions.
Furthermore, we know from experience with gauge theories that the symmetry
generated by (2.9) means that the integration over X and P will create an infinite
138 M. Kal, u / ('ovartant ~trtn.~

number of identical copies of the integration region, which must be truncated by a


gauge-breaking term.
The net effect of this discussion is the realization that care must be taken when
making the transition to the second-quantized action in the presence of first class
constraints, which tell us that the integration region for X and P is no longer
unrestricted.
"l-he mathematical expression of these statements can be summarized by the
statement of Dirac that the state vectors of the theory must vanish when acted on by
the positive frequency parts of the first class constraints, as in (2.24). This means, of
course, that the integration over X and P is indeed restricted. That is the main
difference between the covariant string and the light-cone field theory. In the
covariant field theory, the restriction of the integration region over X and P can be
mathematically translated into q-number language as the statement that we must
project out ghost states out of the theory.
From the path integral, we can extract out the hamiltonian which performs two
functions: not only does it propagate the string in -r space an infmitessimal distance,
but it also kills spurious states in the process. In other words, we find that the
hamiltonian generated by the path integral is given by:

H= f <q, IX> D X <XI~I Y> DY <YlCl,) = f D X DY ~)( X )*H( X. Y) ~(Y).

where tt( X. Y ) satisfies the Dirac condition:

[- c)~- 2iX~'~~ + X '~]H( X. Y) = O.

H ( X . Y). because it moves the string in "r space, must also contain the usual
propagator of the theory, given by 1.o - 1. We now see that the expression I1( X. Y)
is given by

H(X.Y)=<XIP(L,,- 1)PI Y3.

where P is precisely the same projection operator found earlier. This is just the
operator version of the c-number expressions generated by the path integral. We see.
in conclusion, that the statement that the Dirac constraints must be applied to all
intermediate state vectors of the theory is equivalent to using projection operators
for the theory.
Our next task is to generalize our previous methods to superstrings. For the case
of Green-Schwarz superstring, the reader is referred to rcfs. [2.3]. For the even G
parity sector of the Neveu-Schwarz-Ramond model, how'ever, we easily generalize
our methods to the case of anti-commuting operators. Using the same methods as
.'~.f. Kal, u / ('m'ar/attt.~trtttgs 139

before, we find easily: G , - GX't/~(';x'i~..._


. . G~';, and S u-- (01G,G jI0) where

[a,,,.c,,l. = 2 L ...... + ~ Z ) ( m : - ' ) 8 ....

[Lo,.a,,l =(~m-,,)a ...... . (5.5)

pR.= 10)(01. (5.6)

pR, : = 1172 _ pS, : (5.7)

pS,_.=(2p2) l G ~ ~_I())(OIG,/:" (5.8)

p~: = 11 _ pS,. (5.9)

PS'=(2p2) 11. - l l 0 ) ( 0 l L l + ( 2 p 2 + l ) ' G 1,:2 pR, :Gt'2" (5.10)

pR,_. = 1 ~ : _ p S , . (5.11)

pS,:= ( 4 p 4 + 2pZ) AG .v2IO)(OIG~:2 + (2p2 + D) AG ~.,2L llO)(OIL,Gt ~

-4p2A[G V210)(OIL,G,:+G 1:21. ,10)(01G ~,,2]


+(2p2+ 1) IL 1P R' :L 1 + G I:~PR'G .:.

,a t = 4 p ' - ( 2 p ~ + ( D - 3 ) p : + ~.D). (5.12)

We can now easily generalize our result to the arbitrary case:

=
,v=o~ ,, -
i
.w=I.'2 'z
/../=
I/, .Jl ~'1
(; , ( s ~ ,, p:+M),°R',%
I
,. (5.13)

w:here p represents the numbcr of distinct partitions of a half-intcgcr or intcgcr.


Notice that the generalization to the Neveu-Schwarz model is straightforward, since
G is the square root of L. For the supersymmctric case. notice that wc can easily
restrict our projection operator for the even (; parity case.
T h e same methods apply to the R a m o n d model. The algebra in this case is given
by:
11.,,,. F,,] = ( ! . m - n)F, ..... .

[ & . ~;, ], = 21. ...... + :, t),,,~-~ .... .

[I.,,,.I.,,] =(m-n)L ...... + ( ' , l ) ) m ' 8 .... ,,. (5.14)


140 M. Kal~u / (Tmartant ~trtng~

Using the same methods as before, the first few terms in the projection operator are
given by:
pR,,= 10),/01,.. (5.15)

pR, = 1, - p S . etc. (5.16)

With these definitions, we can now write the N e v e u - S c h w a r z - R a m o n d model in a


concise fashion:

Z. = + ( X. V ) " P [ l a . - (Z.,,-,~,,)] P + ( X. r )

+ ~(x. 1")P[~a~- (F,,- ~,.)] Pq,( x. t') (5.17)

where 1"(o) takes on different boundary conditions depending on whether it is


describing bosonic or fermionic strings. Notice that this action can be easily
modified so that the sum over G operators only contains even G parity states.
Written in this elegant fashion, we see that it is an obvious generalization to assume
that the above action is invariant under the supersymmetric variation:

6CI) = A ~ , 6q ~ = Bq). (5.18)

where A and B are related to the strange matrices [13] found earlier when studying
the fermion vertex function which interchanges the Hilbert space of h operators for
the Hilbcrt space of d operators. This shows that the N c v c u - S c h w a r z - R a m o n d
model can, at least in principle, bc written in a fashion which is manifestly
supersymmetric. This is quite ironic, because supersymmctry (at least in 2 dimen-
sions) was first discovered explicitly in the string model [14] and later generalized to
four dimensions, but it was only years later that it was suspected that the Neveu-
S c h w a r z - R a m o n d model itself was invariant under a space-time supersymmetric
t r a n s f o r m a t i o n [15]. In this formalism, however, it is relatively easy to show that the
N c v e u - S c h w a r z - R a m o n d model is supcrsymmctric if we can form the matrices A
and B. Details of this construction will be presented in a forthcoming paper.
We should also note that the integration over r found in (5.17) can easily be
eliminated by simply taking the Fourier transform of the equation. ()f course, r,
being nothing more than proper time, is merely a remnant of the original derivation
and hence no physics can depend on it. As is well known, exactly' the same "r
derivative emerges when we make the transition from the first-quantized theory of a
relativistic point particle to the Klein-Gordon equation. In the point particle case.
the field crl.x) that emerges from the transition from first to second qt, antization is
actually a function of two variables qv(x.r). However. the dependence on r
is trivial: q>(x, r ) = g , ( x ) c '~:', where E, being the conjugate of a fictious variable, is
also fictitious.
M. Kaku / Coeariant strm.v,s 141

Putting this into the modified Klein-Gordon equation:

( i a , - [] + rn=)w(x, r ) = ( - E - [] + m 2 ) q ) ( x , 7") = (--[El + h-'i2 ) qp = 0 ,

we find that E simply shifts the mass of the point particle. For the string case, the
variable E is same for all excitations of the string, and hence it can be re-absorbed
in a trivial re-definition of the slope.
(Yet another way of eliminating of the z derivative in the Klein-Gordon case is to
observe that the r-dependent Green functions can be written as:

DI 2
., e t t,...l( ~.l '., ),i('rr r, ) :/2
G ( x f , "rf: 3:,, ~'i) ~
("re--

But by simply integrating out over "r, we obtain the usual Green function:

i
- <xfl

However, since this integration is performed directly on the Green function, rather
than the action, this can only bc done after the gauge degrees of frccdorn arc
eliminated.)

This work was supported in part by NSF PHY-82-15364 and C U N Y - F R A P - R F


13873. We are happy to acknowledge many stimulating and fruitful conversations
with Bunji Sakita on the functional formalism for strings and with Joseph Lykken
on the quantization of the superstring. A summary of this paper was presented on
March 29, 1985 at the Argonne Symposium on Anomalies, geometry, and topology,
where we met Tom Banks. Upon comparing our results, especially the power
expansion techniques, we found great similarities between the work of Dr. Banks
and his collaborator, Michael Peskin, and ours, although our approaches differ in
emphasis.

References
Ill M. Kaku and K. Kikka~a. Phys. Rev. I)10 (1974) I l l 0 . 1823
[2] M. Kaku and J. Lvkken. Supergauge field theor', of covariant superstrin~s. (.('NY preprint
[3] M. Kaku and J. 1..',kken, Supergauge field theor,, of superstrings, Proc. Argonne Svmpo~,iunl on
Anomalies, geometr3..', and topolog.,,, March 28 30, 1985
[4] W. Siegel, t-'h,,s. Lett. 148B (1984) 556: 149B (1984) 157, 162
[5] R.P. I:ex'muan and A R . Hibb.,,. Qt, antum mechanic~, and path integrals (Mc(iraw-Hill NY. 1%5)
[6] C S . Ilsue, B. Sakita and M A . Virasoro. Ph~s. Rex'. I)2 (1970) 2857
[7] J.I.. (iervais and 13. Sakita, Phys. Rev. l.ett. 30 (1973) 716
142 ,~L J~'?l~ll / ("?}l'[lt'l~lll{ "~[I'IlIL~

IX! R ( ' I}rov, cr~ind CI}. Thorn, Nucl. Ph'~s. 1}31 (1971) 163
JgJ BI,. l:ci~m and I) 1~ t'uch~,.Sov. Math. I)okL No. 2. 27 (19~3) 465
ilO I V K:lc. Infinite dim,,nsional |..it algebras (Birkh;:iuscr, t}o~,ton, lgF.))
[11] N N ,Shapo'.alov. l:unc A n a l AppI 6(19,'72) 307
[1~] V K~tc. l.ccl Not~:~ in l~ll~. ~ol. 94 (1979) 441 5
[I 3] 1 ('orrigan and I) (,)live, Nuovo ('ira 11A (1972) 74~;
[14] J 1. (icr,.aisimd B Sakita, Nucl. Ph,~,. 1}34 (1971) 477
[15] I (ihoz,'i,J. Schcrk and D.O]ixc. Nucl Ph',s. B12211077) 253

You might also like