NAT10NAAI. WSTITUUT VOOR KERNFYSICA EN

NIKHEF-H/95-050

Propagators and Path Integrals

J.W. van HoltenNIKHEF-H, P.O. Box 41882

1009 DB Amsterdam NI,August 22, 1995

Abstract

Path-integral expressions for one-particle propagators in scalar and ferinioaic field theories arederived, for arbitrary mass. This establishes a direct connection between field theory and specificclassical point-particle models. The role of world-line rep&rametroation invaiiance of the classicalaction avnd the implementation of the corresponding BB5T-symmetry in the qutntuin theory arediscussed. The presence of classical world-line supersymrnetry is shown to lead t/> an unwanteddoubling of states for massive spin-1/2 particles. The origin of this phenomenon is traced to a'hidden' topologies! fermionic excitation. A different formulation of the pseudo-classical mechan-ics using a bosonic representation of 71 is shown to remove those extra states at the expense oflosing manifest supersymmetry.

PilKHEf SECTIE - H POSTBUS 4188',!, 1009 DB A^STERDfW

•UL ? 0 NJ 0 1

KS002045549R: FI

NL96FG754 DEOOSS87924

1 IntroductionBecause of their conceptual simplicity, path-integral methods [1, 2] often provide convenientprocedures to obtain insight in field theoretical problems. In recent work by Strassler,McKeon, Schmidt, Schubert and others [3]-[7] world-line path integrals have been appliedto a reformulation of standard Feynman perturbation theory from which useful informationon the structure of perturbative Green's functions is extracted. Some of these results wereactually first derived in the particle-limit of string theory [8].

A basic question in this context is the representation of propagators in quantum fieldtheory by path integrals for relativistic particles of various kind. In particular one wouldlike to know the classical actions to be used in these path-integrals, as well as the precisemeaning of the functional measure in some regularized form, e.g. by discretization. An-swers to these questions establish firm connections between the so-called first- and secondquantized formulations of relativistic quantum theory.

This paper addresses both of these problems. It pursues them from two complemen-tary points of view. First of all, there is the pragmatic question of how to convert a givenfield-theory propagator to a path integral expression. Such a procedure is discussed in thispaper for both spin-0 and spin-l/2 particles. Starting at the other end, one can also askwhat kind of quantum field theory is associated with a given classical action. The kindof actions considered in this context are usually taken to possess some desirable proper-ties like reparametrization invariance and supersymmetry, which pose however additionaldifficulties to quantization procedures since gauge fixing then becomes necessary. Usingpreviously established BRST-procedures [9,10] this is done in an extended state space witha number of ghost- and auxiliary degrees of freedom. The propagators in these extendedstate spaces are obtained, and it is then shown how to reduce them to standard expressionsin terms of physical variables only.

One result which deserves to be emphasized is that the irreducible path-integral ex-pression for the propagator of a massive Dirac fermion is not based on a manifestly su-persymmetric action, although world-line supersymmetry is realized algebraically in thequantum theory by the Dirac operator. Indeed, it is shown that the manifestly supersym-metric version of the theory contains a 'hidden' topological fermionic degree of freedomwhich doubles the number of components of the physical states. This theory thereforedescribes a degenerate doublet of fermions, rather than a single massive Dirac particle. Apath-integral expression for a single massive Dirac fermion is also obtained (sect.7). As inother cases, the difference between the two models can be traced to the representation of7s-

This paper is organized as follows. In sects.(2)-(6) the free field theory of a scalar spin-0particle is considered. A simple and well-known path-integral expression is obtained, whichis subsequently rederived from the reparametrization-invariant classical model. These cal-culations also help to explain the general procedures used.

In sect.(7) a path-integral for a spin-1/2 particle is derived. For massless particles itagrees with results in the literature [11,12]. For massive fermions a new term is present inthe action which has not been considered before, and which is based on a bosonic represen-tation of 75. The manifestly supersymmetric theory, which uses a fermionic representationfor 75, is analyzed in sects.(8)-(10), and is shown to describe a doublet of spinors if themass is non-zero. Our conclusions are presented in sect.(11).

2 Free particle propagatorsThe Feynman propagator for a free scalar particle of mass m in D-dimensional space-timeis a specific solution of the inhomogeneous Klein-Gordon equation

( - • , + m2) AF(x - y) = SD{x - y), (1)

such that positive frequencies propagate forward in time, and negative frequencies back-ward. An explicit expression in terms of a Fourier integral is

e*p<x-y)

In the limit e —+ 0+ the simple real pole at positive p° = E(p) — -s/p2 + m 2 gives themass of the particle asro = -^(0). The arbitrarily small imaginary part ie implements thecausality condition.

There is a straightforward connection between this propagator and the classical me-chanics of a relativistic point-particle through the path-integral formalism [1]. To establishthis, let us first consider the very general problem of finding the inverse of a non-singularhermitean operator H. Following Schwinger [13], we construct the formal solution

H-1 = Km i r dTe-iT^-ie). (3)e-0+ Jo V '

Here the exponential operator

Ke{T) = e-<S~u) (4)

is the solution of the Schrodinger equation

^P() (5)with the special properties1

1I denotes the unit operator.

K.(0) = I, Km Kt(T) = 0. (6)T—»-oo

If the operator H acts on the single-particle state-space with a complete co-ordinate basis{|a;}}, the matrix elements of the operator in the co-ordinate basis are

= {x\Ke(T)\y). (7)

Completeness of the basis then implies Huygens' composition principle

dDx' Ke,{x - x'\T') Ke»{x' - x"\T") = Ke{x - x"\T + T"). (8)

Note that e = (s'T + e"T")/(T' + T") stays arbitrarily small if e' and e" are small enough.Repeated use of eq.(8) now allows one to write

Ke(x - y\T) = / n dDxn n Ke (xm+1 - xm\AT), (9)J n=l m=0

with AT = T/(N +1), and x0 = y, xjf+i = x. Keeping T fixed, the limit N —* oo becomesan integral over continuous (but generally non-differentiable) paths in co-ordinate space-time between points y and x. (Observe that Ke(x — y\AT) depends only on the difference(x — y) and converges to SD(x — y) for AT —»• 0.)

If the operator S is an ordered expression in terms of a canonical set of operators

* ...p^HZ:;^ ...x«, (10)k,i

then we can expand the co-ordinate path-integral expression (9) further to a phase-spacepath-integral

Ke(x -

/ ;VP(T)VX(T) e{

Here H(p,x) is the c-number symbol of the ordered operator H, and we have tacitlyassumed that the ordered symbol of the exponential can be replaced by the exponential ofthe ordered symbol. This is certainly correct for the main applications we consider in thispaper, as may be checked by explicit calculations.

It is now clear, that one may interpret the symbol H(p, x) as the hamiltonian of someclassical system, and the argument of the exponential as the classical action. Integration

over the momentum variables p(r) then in general leads to the lagrangian form of thisaction

Ke(x-y\T)= r2te(r)c*/. **<*•*>, (12)Jy

where the precise meaning of the integration measure can be recovered either from thephase-space expression (11), or from requiring the path-integral to satisfy Huygens' com-position principle (8).

Returning to eq.(l), it states that Ap{x—y) is the inverse of the Klein-Gordon operator(in the space of square-integrable functions). Rescaling it for later convenience by a factorl/2m, we consider the evolution operator

Ke(T) = exp | - ~ - (-D + m2 - »e) | . (13)

In the co-ordinate representation the explicit expression for the matrix element of thisoperator is

Ke(x - y\T) = -i [YJ;} e^*-*?-Y("-&). (14)

The Feynman propagator can then be written as

i f°°

2TTI JO

and using the previous results it can be cast in the form of a path integral [14]

The same expression is also obtained directly by iteration of (14) using the product formula(9). For massless particles yet another derivation, based on the concepts of moduli space,has been discussed in [11, 12]. The result shows, that the scalar propagator is connectedto a classical particle model with lagrangian

related by Legendre transform to the simple hamiltonian

Because H(p,x) is quadratic in p^ and independend of a:'*, the integration measure in thepath integral (16) is just the free gaussian measure of ref.[l].

The path-integral representation of the propagator establishes a simple connection be-tween scalar quantum field theory and the classical point particle model (17). As is well-known, this connection does not only hold at the level of the action or hamiltonian, italso extends to the dynamics, in the sense that paths in the neighborhood of the solutionsof the classical equations of motion give the dominant contribution to the path integral,certainly for the simple model discussed where the path integral is a pure Gaussian.

3 Reparametrization invarianceThe classical equations of motion which follow from the lagrangian (17) state that themomentum is constant along the particle worldline:

p" = mx = 0, (19)

Therefore this classical theory seems to contain less information than the quantum theoryfrom which we started: we have to recover the condition that the momentum should lie onthe mass shell:

pi + m2 = 0. (20)This condition is equivalent to the vanishing of the hamiltonian (18). Since the hamiltonianis the generator of translations in the worldline parameter r, the mass-shell condition isrecovered by requiring the dynamics of the particle to be independent of the world-lineparametrization.

As is well-known [15, 16], this can be achieved by introducing a gauge variable e(r)for reparametrizations of the world line (the einbein). Under local reparametrizationsT —» T' = / ( T ) , let the co-ordinates and einbein transform as

X*(T) -» X'^T') = X»(T), e(r) -> e'(r') = e ( r ) | ^ . (21)

Then a reparametrization-invariant action can be constructed of the form

Scl[x^r),e(r)} = j J*dr ( i«J - e) . (22)

The equations of motion one derives from this action are

I d Idx"edr e dr ( 2 3 )

Identifying df = edr with the proper time, and as a consequence p** = mdx^/df withthe proper momentum, this reproduces both the world-line equation of motion and themass-shell condition.

For the new hamiltonian we can take

This generates proper-time translations on special phase-space functions F(x(r),p(r)) bythe Poisson-brackets:

dF ldF l , F m ,„-*

W = -e-d7 = -e{F'H}- (25)

Note, that one cannot impose the constraint that H vanishes before we compute thebrackets.

The difficulty with this formulation is evidently the additional dynamical variable e(r),for which the evolution is not fixed by the Euler-Lagrange equations derived from 5cj[a;'<, e],and which has no conjugate momentum. The origin of these difficulties is precisely thereparametrization invariance (21). As a result, the hamiltonian evolution equation (25)does not hold for arbitrary functions on the complete phase space, F(x/i,pIJ',e). Clearly,to recover the results of sect.2 it is necessary to fix e(r) to a constant value and changeT —*• f, which amounts to a rescaling of the unit of time on the worldline such that theparticle's internal clock and the laboratory clock tick at the equal rates when the particleis at rest.

Therefore one chooses a gauge

e = 0, (26)

implying that e is constant, and adds this condition to the equations of motion. Butthis can only be done after the variation of e in the action has produced the mass-shellconstraint (20). It is therefore of interest to have a formalism in which one can imposethe gauge condition (26) from the start, and still keep track of all the constraints imposedby the reparametrization invariance (21). An appropriate formalism to solve this problemis the BRST procedure (for an introduction, see for example refs.[17, 18, 19]), which wedescribe here for the case at hand [9].

4 BRST formulationWe impose the gauge condition (26) by adding it to the action with a Lagrange multiplierA; at the same time we introduce a corresponding Faddeev-Popov ghost action, using (real)anti-commuting ghost variables (6, c), in such a way that the complete action is invariantunder the Grassmann-odd, nilpotent BRST transformations

8^ = ex1*, S\e = ce + ce, 6\c = cc,(27)

sAb = - a , sAx = o.After a convenient rescaling ec —> c, turning the ghost c into a world-line scalar densitysuch that S\e — c and S\c = 0, the BRST-invariant gauge fixed action reads

Sgf = 5rf[aj(r),c(r)] + / dr (Xe + ibc) . (28)

Actually, because a partial integration has been performed in the ghost term, the action isinvariant only modulo a total time derivative. This is sufficient. Note, that the Lagrangemultiplier A now plays the role of momentum pe conjugate to the einbein e.

The action Sgf has no local invariances left and can be treated by the usual proceduresof canonical hamiltonian analysis. The canonical momenta are

p» = - i " , Pe =A,

(29)

TTC = — ib, 7cb = ic.

Note that the ghost-momenta (TTC, 7r&) are imaginary. The gauge-fixed hamiltonian is

The equations of motion are given by the Poisson brackets

dF— = {F,Hgf}, (31)

where the brackets axe defined by

(jj? C/\jr OS: t/(jr 0Jr CsKjr (JS? CJ\jr

~dedp~e ~&pe~de(32)

} \dc d*c + dirc dc + db d*h

+ dnb db) '

Here all derivatives are taken from the left, and ap is the Grassmann parity of F.To recover the constraints imposed by local reparametrization invariance, one constructs

the (Grassmann-odd) conserved BRST charge

This BRST charge is nilpotent in the sense that

{O,ft} = 0. (34)

The BRST principle makes use of this property, and states that the physical observables ofthe theory are the cohomology classes of the BRST operator on the phase-space functionsF(x,p;e,pe;c, 7rc; 6, TTJ,) with ghost number2 Ngh = 0 [20, 21]. More precisely, physicalquantities are BRST invariant:

{F,n} = 0, (35)

but this allows an ambiguity in F as a result of (34); this ambiguity is resolved by as-sociating obervables F with equivalence classes of functions differing only by a BRSTtransformation:

F' ~ F & F' = F + {A,fi}, (36)

with A a phase-space function of Grassmann parity opposite to F, and one unit in ghost-number lower. In particular, any function which can be written as a pure BRST transform

G = {A,fi}, (37)

is in the equivalence class of zero and may be taken to vanish. In the case of the relativisticparticle, this happens to be true for the hamiltonian itself:

Hgf = -{e7rc,f2}, (38)

and for the classical hamiltonian as well:

p2 + m2 = -2m{7rc ,fi}. (39)

Hence the classical and ghost terms in the hamiltonian Hgf vanish separately and themass-shell condition follows. With

pe = -i{b,Sl}, (40)

we also recover the vanishing of the momentum conjugate to the einbein e.Before turning to the quantum theory, we draw attention to a peculiarity of the action

(28): it is invariant under SO(2) rotations in the ghost variables (6, c). Therefore thistheory possesses a second BRST invariance with conserved nilpotent charge

} = 0- (41)The algebra of these two BRST charges has the property that

2We assign ghost number +1 to (c, x&), and -1 to (b, 7rc).

V (42)TO v l " ' ~

Note that both factors on the right-hand side are separately conserved, and vanish on thephysical hyperplane in phase space.

5 BRST quantizationHaving formulated a complete, BRST-invariant (pseudo-)classical mechanics for the rela-tivistic point particle, we now return to the quantum theory and study the relation betweenthis model and quantum field theory. Various aspects of this problem have been discussedin [22, 23, 9]. To obtain additional insight, we construct a quantum theory correspondingto the classical hamiltonian (30) in an extended state-space, compute the propagator andestablish the relation with the usual Feynman propagator (2), thereby showing the physicalequivalence of these different formulations of scalar field theory.

The quantum theory of interest is obtained by replacing the phase-space variables(x'i,pfj,;e,pe;c,Trc;b,i:i,) by operators, and postulating (anti-)commutation relations be-tween them in direct correspondence with the Poisson-brackets (32):

[e,pe] = i,

(43)

In the co-ordinate representation these algebraic relations hold on making the identification

- _ • d - - -d

(44).d . .6

Xc = ~*d~c> Xb = ~ldb-The bosonic momenta are self adjoint, and the fermionic ones anti self-adjoint with respectto the inner product

($,$) = iJdPx'dejdbdcW, (45)

where wave functions like *P are polynomials in the ghost variables, with co-efficients de-pending on the remaining co-ordinates (sc'z, e):

, c) — rl> — ibrpb + aj;c- ichp^. (46)

The inner product (45) then reads in components

(#, *) = ^V* + «*e + £ * + &*• (47)Clearly this form is not positive definite. However, with respect to this inner product thehamiltonian operator,

and the nilpotent BRST operator,

" = 2^(-D + m ! )+ 'a5p °° = 0' <49>are self adjoint. The indefinite metric implicit in this inner product is a general andnecessary feature of a model with a nilpotent self-adjoint operator like tl [10].

It is also possible to define a positive-definite inner product on the state space: introducea duality operation [24]

(50)

and define

(*,#> = (M)(51)

With respect to this positive definite inner product the fermionic momenta and the BRSTcharge are no longer (anti-)self adjoint. In particular c* = ZTTC, b* — ITTJ, and

$) = <$,6*¥), (52)

where the adjoint BRST charge, also known as the co-BRST operator, is

JT = -L (-D + m2) | - + »!- . (53)

Like the BRST charge itself, O* is nilpotent; however, it is not conserved:

[£l\Hgf] = 0, (54)

where fi is the operator corresponding to the dual BRST charge we have encounteredbefore in (41):

10

We also note in passing, that the BRST operator fi itself is obtained from the commutatorof the hamiltonian with the adjoint of the dual BRST charge:

[d\sgf] = -n, (56)with

We can now show how the physical states of the scalar particle are reobtained in the BRSTformalism through the cohomology of Q. One identifies the physical states with equivalenceclasses of states which are BRST invariant and differ only by a BRST-exact term:

fitfpfcj,, = 0, Vphy, ~ %hya = Vphya + &A. (58)

To obtain exactly one representative of each BRST invariance class, consider the zero-modes of the BRST laplacian [10, 9]

&BRST =

. Q2 (59)

This operator is positive definite, and its zero-modes are zero-modes of the two terms onthe right-hand side separately. Therefore

&BRST Vphy* = 0 & I ( - • + TO2) $pfeya = 0 A Q^phy, = 0 1 . (60)

Hence the wave functions ^phya indeed satisfy the Klein-Gordon equation and are indepen-dend of the gauge variable e.

6 Gauge-fixed propagator and path integralThe next step is to construct the propagator for the scalar particle as described by thewave functions (46) in the BRST-extended state space, using the formalism of sect.(2).Let us label the co-ordinates (s**, e, b, c) collectively by Z. Then the Schrodinger equationfor the kernel Kgf(Z; Z'\T) in the co-ordinate picture becomes:

i-^Kgf(Z; Z'\T) = (^ (-D + m2 - ie) + i-^~) Kgf(Z; Z'\T). (61)

This kernel is required to satisfy the initial condition

11

Kgf(Z;Z'\O) = S(Z-Z')(62)

= SD(x - x')6(e - e')S(c - c')5(b - b1).

The solution to this equation is

Kgf(Z; Z'\T) = -iT6(e - e') ( W ^ " ete(*-*')2-^-«)H(>-»')(c-C')]. (63)\27reTV

This is the direct counterpart of eq.(14). It is straightforward to verify that the initialcondition (62) is satisfied, as is the composition principle:

dZ' Kgf(Z; Z'\T') Kgf(Z'; Z"\T") = Kgf(Z; Z"\T' + T"), (64)

where the integration measure reads

JdZ = i JdDxdefdbdc. (65)

The propagator in the extended state space then becomes

A,,(Z; Z') = £- J~ dT Kgf{Z; Z>\T\ (66)

It is the solution of the inhomogeneous extended Klein-Gordon equation

- « + - ^ ^ l *.t& Z>) = 6(Z - Z'). (67)

Substitution of the expression (63) for the kernel Kgf(Z,Z'\T) in eq.(66) for the gener-alized propagator gives a closed expression for the propagator in co-ordinate space whichis conveniently represented in terms of a momentum integral. Switching to the Fourierrepresentation of the co-ordinate part and performing the integral over T yields

J(68)

From this expression it is straightforward to obtain the Feynman propagator by integrationover the unphysical degrees of freedom:

AF(x-x') = jXde f dbJdcAgf(x,e,b,c;x',e',b',c')

r dDp eM—0J (2x(2x)c f + m2 - ie'

12

Alternatively, repeated use of the composition principle can be used to construct a con-figuration space path integral representation of the propagator. Introducing the Fourierrepresentation for the delta function 6(e — e') repeated use of eq.(64) gives in explicitnotation:

Agf(x,e,b,c;x',e',b',c') = 2 1 f°dT f [Vx(r)V\(r)Ve(r)Vc(r)Vb(r)] ei

Zm Jo JT

(70)where Sgf is the classical gauge-fixed action (28) and the functional integral is over all pathsF in the configuration space between (x',e',b',d) and («,e, 6, e). From the construction aconsistent discrete regularization, specifying the measure to be used, is obtained:

[Vx(T)V\(T)Ve(T)Vc(T)Vb(r)\

N * AT / m \ ®fe\

p ^ f c * ( J j x (n)

where again AT = T/(N + 1), with T fixed. Note that the measure contains a factor\/m/2TcieAT for each integral over a co-ordinate a:*4, and a factor \/*AT for integrationover a ghost b or c. Eqs.(70) and (71) give a precise meaning to relation between thepath-integral representation of the propagator in the extended state space and the BRST-invariant gauge-fixed classical action (28).

7 Fermions

The Dirac-Feynman propagator for a free fermion is the solution of the inhomogeneousDirac equation

(7 • dx + m) SF(x -y) = SD{x - y), (72)

with the same causal boundary conditions as for the scalar particle. The Dirac matrices 7M

form a D-dimensional Clifford algebra and are normalized to satisfy the anti-commutationrelations

{7^,7"} = 2 iT- (73)

The standard Fourier integral representation of the Dirac-Feynman propagator is

13

{-;T\+m) e"'-»'. (74)p2 + m2 — is

It is possible to use the Schwinger procedure described in sect.2 to construct a path-integralrepresentation of this propagator as a Clifford-algebra valued object. There is howeveranother method that is often preferred in applications, especially when interactions areintroduced into the model. This method uses anti-commuting variables to represent theClifford algebra [25, 26, 27]; the application to spinning particles has been studied forexample in refs. [12,15,16], [28]-[33]. The use of this method here allows a straightforwardpath-integral representation of the Feynman propagator for fermions in terms of bosonic co-ordinates x* and a matching set of fermionic (Grassmann-odd) co-ordinates if)1*. However,as the details of the procedure depend on the number of dimensions, we will from now onchoose D — 4, and limit ourselves to that physically relevant case. Modifications to treatthe same problem in another number of dimensions are straightforward to make.

In order to achieve the transition to anti-commuting variables, we define a representa-tion of the Clifford algebra in terms of two anti-commuting variables (£*,£2) by

(75)

In addition to the 7M we also need the usual pseudo-scalar element of the Clifford algebra:

i \

(76)

Note that in contrast to the other 7M, in this realization 75 is represented by a Grassmann-even (bosonic) operator. Therefore we refer to this representation as the bosonic form of the75. In later sections we will also encounter a representation of 75 in terms of a Grassmann-odd operator, which is appropriately refered to as the fermionic representation. Now thefollowing algebraic relations are satisfied by these operators:

{7", 7"} = 217"", (77)

and

7s = 1, (75,7M} = 0. (78)

14

These operators therefore realize the Clifford algebra of the Dirac matrices including 75.The operators (7^,75) act on spinors $(^1 ,^2), here defined as functions with a Grass-

mann polynomial structure

where all co-efficients are functions either of co-ordinates or momentum. One can definethe usual positive-definite inner product for spinors:

<*,*>= £ # # - • (80)a

This inner product can be written in the Grassmann representation as

where a second set of Grassmann variables £1>2 has been introduced as argument of theconjugate wave function, and the star denotes complex conjugation of the c-number co-efficients of $(£) plus reversal of the order of the £k. It is straightforward to check thatw.r.t. this inner product

d u

($,7--r$) = (£*$,*). (82)d£k

This result implies that in contrast to the other 7* (i = 1,2,3) and 7*, the time-like operator70 is not real w.r.t. ($ , $ ) . This is of course to be expected from the lorentzian signature ofspace-time. On the other hand, hermiticity can be restored by defining a (lorentz invariant)indefinite-metric scalar product {$,#) using the Pauli-conjugate spinorial wave function

t*) = « + e# - e# - eetiIn general the co-efficients of $ are independent complex numbers, in which case theyrepresent the components of a Dirac spinor <f>a. Irreducible Weyl spinors can be obtainedas eigenfunctions of 75. As

e) = *(-e,-e), (84)it follows that the Grassmann-even components define a Weyl spinor of positive chirality,whilst the Grassmann-odd components define a Weyl spinor of negative chirality. In therepresentation chosen here the chirality can therefore be identified with the Grassmannparity of the spinor $ .

It is also possible to represent Majorana spinors by requiring C $ = $, where $ is thePauli-conjugate spinor and C is the (anti-hermitean) charge-conjugation operator

15

( 8 6 )

The Majorana constraint results in the component relations <f>4 = <fo and <f>$ — —<^\. In thefollowing we consider Dirac spinors unless explicitly stated otherwise.

With the above definitions, the Dirac equation can now be transcribed as follows: let

ey, (86)then the components of $' are then related to those of $ by

<f>'a = [ ( -H-P + m H s U , ^ , (87)in the representation of the Dirac matrices denned by

7 5

(1 0 \

^ \ 74 _ • 0 _ f 0 1 \

(88)

Since the matrix 75 is unitary, the equation $ ' = 0 is completely equivalent with the freeDirac equation in momentum space, and we infer that SF{X — y)^ can be identified withthe inverse of the Dirac operator (—17 • d + m-ys) in co-ordinate space. In the process oftranslation we have obtained the following correspondence

Y »-» - i7 M 7 5 , 75 «-»• 75- (89)

The next step is to write the inverse of the Dirac operator in the Grassmann co-ordinaterepresentation. This is achieved by introducing the ordered symbol for the operator [25,26, 27]. A quick way to derive the necessary results is the following [34]. For a singleGrassmann variable £ the most general form of a differential operator is

A = ao + a^ + at-;- + (o3 - a0) £-57. (90)

Consider the action of this operator on an arbitrary function /(£) = /o +

= a0f0 + a2f! + (ajo + a3/i) £• (91)

The operation of A on / can be represented equivalently in terms of an integral. Firstobserve that

f(0 = fd?6(? - 0f(O = fd?dZtSV-*>f(?). (92)

16

Here we have introduced the Fourier representation of the anti-commuting 5-function interms of a conjugate Grassmann variable f. The ordered symbol A(£,() of A is defined bythe relation

(93)

Our construction shows, that it is obtained by replacing every d/d£ in A by a £:

A(Z> 0 = <*o + ai£ + a2£ + (a3 - a0) & (94)

In the following it is useful to have an expression for the symbol of a product of operatorsin terms of the product of their symbols. The relation is given by the equation

, 0 = / tfd? e(?-O«'-0 itf, ? )£(£', £). (95)

It is straightforward to generalize this to a product of an arbitrary number of operators.In that case a symbolic notation for the integration measure is used, of the form

= [ f[ d&dt*. (96)

Returning to the case of a spinning particle in 4-dimensional space-time we add two con-jugate variables (^j^2) and introduce the symbols for the operators (7^,75), rescaled byfactors l/y/2m for later convenience:

(97)

and

( 9 8 )

Note again the non-standard feature that, whereas the if)1* are Grassmann-odd, the ^5introduced here is Grassmann-even (bosonized).

In terms of these quantities, one can rewrite the Dirac equation in momentum space as

f dH'd2ZeW-V [p • + + m 5̂] (£, 0 $ ( 0 = 0. (99)

As in our representation the chirality equals the Grassmann parity, it follows that termswith different Grassmann parity in the wave function are mixed by a non-zero mass.

17

To obtain the propagator in momentum space we have to invert3 the Grassmann integraloperator in (99). As a first step, we observe that the product rule (95) implies the usualidentity

(100)

The inverse of the Dirac operator in momentum space can then be written in an extensionof the Schwinger representation as [35]

y/2m(101)

p2 + m2 — is '

where T is the usual Schwinger proper-time parameter, and a is a Grassmann-odd coun-terpart. Eq.(lOO) then implies that

f *€*? e(f-«)•«•-« \p ̂ + rni>5] (£, ?) [SF %} (pM; ? , ft = 1. (102)

In order to get the expression in the co-ordinate representation, we have to compute theFourier transform. We also redefine the Grassmann variable, making it proportional toproper time T: a = Tx- Now introduce an integral kernel:

l ^ e e £ ( )(27r)

(103)

\2icTj

Then the following results hold:(i) The Feynman propagator for a Dirac fermion can be written as

oo A'TA'T f

-Tf dXKx{x - y;U\n (104)1 J

(ii) Huygens' composition principle holds in the form3 Of course, this inverse is to be interpreted in the usual sense of distributions to deal with singularities.

18

t? eP-eHe-Q Kx{x - *'; £, £'|T') Kx(x' - x"; ?, f\T"(105)

= Kx(x-x";U;\T' + T").

(in) The integral kernel satisfies the boundary condition

= 8\x-y). (106)

Apart from giving the explicit expression (101) for the Dirac-Feynman propagator, eq.(104)can be used to construct a path-integral representation of the propagator by iteration ofeq.(105). Indeed, repeated use of this equation leads to the result

Kx(x-r,U\T) =k=i

N

«=0

where x0 = y, ajv+i = a, £o = £, £N+I = € and AT = T/(N + 1). Furthermore

(107)

61 AT) =

[ J(108)

In agreement with our earlier definitions, the symbols rl>£ here are a short-hand notationfor

with similar espressions for (iftkirfk) m t e r m s of (£fe+n£ifc)? a n ( i ̂ l16 bosonized ^>| given by

In the continuum limit (N —>• oo, T fixed) the free fermion propagator can now be writtenas a path integral

19

(* ~ V\ e, t) = -J=r^F [ dX Vx{r)Vi{r)V({r) eiSf-4^)^M, (m)V2m Jo I J

where up to boundary terms

Finally it can be cast into a manifestly Lorentz-invariant form4 by replacing (£*(r)>l*(r))by the vector-like variables •0/i(r) defined in eq.(97):

Sferm [X{T), ^ ( r ) ] = J J^ ^ [i£ + ty - $ + lix • X i> + 2iX V>5 ~ l] • (113)

This expression resembles closely the one usually encountered in the literature [15,16], [28]-[33], which is based on the realization of a (gauge-fixed) local world line supersymmetry.However, in contrast to the standard approach the present construction uses a Grassmann-even ip5, and as a result inclusion of the mass-term violates explicit supersymmetry at theclassical level, even though it is realized on the operator level in the quantum theory, asshown by eq.(lOO). In fact, for non-vanishing m the pseudo-classical action SfeTm does noteven have a well-defined Grassmann parity; eq.(84) and the discussion following it makesclear that this is a direct consequence of the non-conservation of chirality for massivefermions.

It is shown below, that the discrepancy between the result (112) and the manifestlysupersymmetric spinning particle model has its origin in a doubling of the number of degreesof freedom in the supersymmetric case, which is regularly overlooked. More precisely, ifone preserves supersymmetry of the classical action by taking ^5 to be Grassmann-odd,and if the classical algebra of Poisson-Dirac brackets is mapped in a straightforward wayto the quantum (anti-)commutation relations, then one does not obtain the representationof the quantum operator ^5 by the 4-dimensional Dirac matrix 75 = —i/4! eAu/«A7At7I'7't7^.Rather, a Grassmann-odd (fermionic) representation of the operator ^5 is to be used, anti-commuting with the other ^>ti. Then the equivalent matrix representation of this operator

4There is a subtlety concerning the Lorentz-invariance of the continuum limit, as it seems to requirethat the main contribution to the continuum path integral comes from paths which are smooth in the sensethat for AT -+ 0 one has ||&+x - & | | = O((AT)5+P) with p > 0; from the calculation of explicit examplesdirectly in the continuum limit this seems to be correct. Some arguments for a consistent continuumpath-integral for simple spin systems, related to the Duistermaat-Heckman theorem, have been advancedin [36].

20

is a 75 taken from the Dirac algebra in six dimensions. This implies a doubling of thenumber of degrees of freedom of an irreducible spinor.

On the other hand, for free massless particles manifest supersymmetry is not violatedeven in the present theory; this is because chirality is conserved, and so only terms ofeven Grassmann parity occur in the action. One therefore sees that the apparent violationof manifest world line supersymmetry is a result of mass generation, and has the samedynamical origin.

8 Worldline supersymmetry

In this section we turn to the point-particle model with full classical super-reparametrizationinvariance on the world line, to compare it with our treatment of Dirac fermions in sect.(7).To realize complete off-shell supersymmetry5 it is necessary to introduce three differenttypes of super-multiplets (superfields):

1. The gauge multiplet (e, x) consisting of the einbein e and its superpartner x-> a^so

refered to as gravitino; under local worldline supersymmetry with parameter £ themultiplet transforms as

Se == -2ieX, SX = £. (114)

2. Matter multiplets (x1*, V>M) describing the position and spin co-ordinates of the par-ticle. The transformation rules are

1 r

= - t e ^ " , 8^ = e - ^ - , (115)e DT

where the super-covariant derivative is constructed with the gravitino as the connec-tion:

Dx" dx»_ _ = _ +

3. One or more fermionic multiplets (77, / ) with Grassmann-odd 77 and even / ; it is usedin the following as an auxiliary multiplet and its transformation properties underlocal supersymmetry are

5The term off shell implies that the supersymmetry algebra is realized without using dynamical con-straints like the equations of motion.

21

6V = ef, Sf = - i e 1 - 1 ^ . (117)

The super-covariant derivative is formed as before:

= $-*/• (118)

It may be checked that in all cases the commutator of two supersymmetry transformationsgives a local translation (reparametrization) with parameter a = (2i/e)eie2> plus a localsupersymmetry transformation with parameter e' — — (2t'/e)e1e2X'

The minimal free particle action invariant under these one-dimensional local supersym-metry transformations is

= Y J dr ( j « i + ty -i> + jXi> • & + *W + %iXV + ef2 - 2e/) (119)

Obviously, the bosonic variable / appearing quadratically without derivatives does notrepresent a dynamical degree of freedom and may be eliminated using its algebraic equationof motion / = 1 (which is equivalent to completing the square). Then the classical actionbecomes

i & ) (120)

In addition local super-reparametrization invariance may be used to fix the gauge multiplet(e,%) to constant values. In particular, if the proper time r is rescaled by e, so as to bemeasured in the same units as the laboratory time (effectively setting e = 1), and theconstant value of % is rescaled by the same amount, then comparison with the fermionicaction (112) shows that Sferm is formally related to 5 w n , by 77 —>• 0, rj —> fa. However, thislast relation is frustrated by the mismatch in the Grassmann parity of the two quantities;moreover, contrary to / the variable 77 is a dynamical degree of freedom, associated withan additional two-valued parameter labeling the wave-functions of the particle. As itcouples to the gravitino, and therefore appears in the first-class constraints of the theory,its dynamics cannot be taken into account by simply equating it to a constant in the action.At this point we observe however, that one could add more fermionic multiplets (r)k,fk)whose auxiliary scalars have vanishing classical value:

) (121)

22

Eliminating the fk by their algebraic Euler-Lagrange equation fk = 0, there remain onlyfree fermionic degrees of freedom rjk which decouple from the dynamics because they donot interact with the other physical or the gauge degrees of freedom. In fact these ad-ditional fermions define a d = 1 topological field theory, in the sense that their action isreparametrization invariant without involving the metric (represented by the einbein e).It is also easy to see, that they do not contribute to the hamiltonian of the theory.

To construct the canonical quantum theory corresponding to the supersymmetric actionSstuy, we employ the BRST procedure used before. Besides the reparametrization ghosts(b,c) and the lagrange multiplier A we introduce commuting supersymmetry ghosts (a,/?)and an anti-commuting multiplier s. The nilpotent BRST variations of the full set ofvariables read:

1. For the gauge multiplet:

2. For the matter multiplets:

• in Tjtp^1

C II ' 11 i ILL C I il I tX I **•*• x > » * / /lOOX

e DT

3. For the fermionic multiplet:

Stf = cq + i<*f, hf = cf + - —^. (124)e DT

4. For the ghost variables:

6Ab=-i\, SA(3 = s,

= 0, 8As = 0.

23

In the classical theory defined by Ssuay we wish to impose the gauge conditions

e = 0, x = 0- (126)

Two write down a relatively simple BRST-invariant gauge-fixed action, it is convenient tomake the redefinitions:

ec —* c, a — icx —> OL. (127)

Then the BRST variations are covariantized and simplified; for example

&A.X = i&, S\c = -ia2, 8Aa = 0,

cDx»=~^ + a r W = l 7 + 7 7 (128)

c Dn . , _ , c Df a Dve DT e Dr e Dr

In terms of these new variables the BRST-invariant gauge-fixed action reads

fT%

Sgf = S^, + f dr (Ae + isjc + ibc + pa + 2iabx) , (129)

where we take Sjuaj, as in eq.(120), in which / = 1 has been inserted. This is consistentwith the BRST variations (128) and their nilpotency, provided the equation of motioni} = x is used.

The canonical momenta for the physical and gauge variables in this theory axe

^

Pe = A, 7TX = -is,

This is to be supplemented with the ghost momenta

TTC = - i6 , nb=:i(c + 2ax),(131)

Pa = Pi PP = «•

A minor complication is the appearance of the second-class constraints for the fermionicmomenta (TT^TT,,). These are resolved in the standard way by Dirac's procedure, and oneends up with the gauge-fixed, unconstrained hamiltonian

24

Hgf = — [pi + m2) - %x (V1 * P + mV - 2iairc) - i-Kh-Kc + pap^, (132)

plus the following Dirac-Poisson bracket for functions (F, G) on the unconstrained phase-space spanned by ( z M , ^ ; ^ *

i^£*L _ <¥-?£- dZdG[ _ dFdG*• ' ^ dzKdpn dppdxv^ dedpe dpe de

dFdG _ dF^dG d£dG dFdGd d d d dpdadpa dpada d/3 dp& dppd/3

(133)I % OJC OKT % OH OLX Or O\jr Or O\jr \ *

m dj>»d^ + m drj dr)

(_l)aif(d£dG_ dF^dG dF_dG_ dF dG\( ' \dcdirc

+ dircdc db dvh d-Kh db) '

The equations of motion then take the canonical form

~ = {F,Hgf}. (134)

Similarly, the BRST variation of a phase-space function F can now be obtained from

6AF = (-iy*{F,n}, (135)

where Q is the nilpotent, conserved BRST charge

ft = -— (p2 + m2) + a (j> • V> + rnrf) — mbpe + ippnx - ia2icc,2m (136)

{a,Hgf} = o, {n, a} = o.Like for the scalar particle, the full hamiltonian Hgf is BRST-exact:

Egf = {-e*c + ixpa,Sl}. (137)

In the BRST cohomology the hamiltonian is therefore equivalent to zero and the physicalparticle motions are on the mass-shell.

25

9 Quantum supersymmetryThe BRST-invariant classical action for the supersymmetric theory can now be taken as thestarting point for the construction of a canonical quantum theory for a free point particle.Later this quantum theory is then used to construct a propagator, and it is shown that itcan be expressed in path-integral form using precisely the classical action Sgf. Thus thecorrespondence between the classical and quantum theory is established in both directions.From the canonical formulation it will then be entirely clear, that for TO ̂ 0 this theorydescribes a degenerate pair of Dirac fermions, and that this is completely independend ofthe BRST quantization procedure; the origin of the doubling of the degrees of freedom canin fact be traced to the fennionic dynamical variable t\.

The first step in defining the quantum theory is to postulate a set of (anti) commutationrelations in correspondence with the classical Dirac-Poisson brackets (133). In addition tothe operator (anti) commutators (43), we introduce the operator algebra

— Vt»<> {^6,^e} = —,

[&,pa] = i, [P,Pji[ = i.

(138)

The notation $6 (rather than the more usual ^s) has been introduced for the operator corre-sponding to 77, to emphasize the Clifford algebra structure of the fermion anti-commutators,but also to avoid thinking of this operator as the product of the other ^ as in the caseof the Dirac fermion in sect.(7). The above commutation relations can be satisfied bychoosing the following linear representation of the operators: the *̂* are realized in termsof two Grassmann variables (£x,£2) and their derivatives, as in (75) by

As before, the operators tpo are hermitean only w.r.t. to the physical indefinite-metricinner-product involving Pauli-conjugate spinors.

Furthermore we introduce Grassmann variables £3 and %, as well as ordinary real vari-ables (a,/?) and define

(140)O

Pa = ~*"HT' P& = ~l!Ta'

26

A

Note that we have a fermionic (Grassmann-odd) representation of ip6. As a result, havingat our disposition the variable £3, one can define at no expense the additional operator

which anti-commutes with the other I}>M• This operator does not appear in the hamiltonianof the theory, which we define in correspondence with the classical hamiltonian (132):

5?(142)

A d d2 d2

= eH0 - ixQ+ + 2 +c oboe oaop

and a nilpotent BRST operator

«2 /)2 Q

»_-*—- a2-.

d2 o2 d

(143)

The short-hand notation in the second line of these equations has been introduced with aview to the supersymmetry algebra

Q\ = Ho. (144)

At this point a remark of caution is appropriate: for m / 0 the operator l?o has zero-modes only in Minkowski space; but for lorentzian metrics the operator Q+ (essentiallythe Dirac operator) is not hermitean w.r.t. the positive-definite inner-product $*$ (onlyw.r.t. the Lorentz-invariant indefinite inner product W$). Therefore the zero-modes ofHo (the physical states) are not necessarily zero-modes of Q+: one also finds back thenegative-energy states associated with the vanishing of Q+. Moreover, there is a similarrelation for the Lorentz-invariant operator Q- obtained by replacing m —> — m:

Q2_ = Ho, Q- = (-4 • d - mfa) . (145)

The resolution of these difficulties lies of course in the full quantum-field theoretical treat-ment; it is of interest to see how the BRST procedure is implemented in this context, andas a first step we construct in this paper the free propagator, in sect.(10).

27

For our present purpose it is however sufficient to note, that the BRST-cohomology canstill be used to characterize the physical states, in the following way: since the hamiltonianHgf is a BRST-exact operator:

E" = {'h-ixii'ti}' (146)

one can pick a state from each equivalence class of solutions of the BRST condition

fiVy = 0> (147)

by requiring the subsidiary condition

for arbitrary values of e and \i as suggested by super-reparametrization invariance. Thisamounts actually to two subsidiary conditions:

d dj - c %hy, = 0, — %hyt = 0. (149)

One may think of these conditions as defining the ghost vacuum. It now follows automat-ically from (146) that

Hgf%hy, = 0. (150)

Combined with the BRST invariance of the physical states expressed by (147), this givesthe physical states precisely as solutions of the Dirac (and consequently Klein-Gordon)equation:

(-ii> • d + m^6) Vphya = 0. (151)

The wave functions $pfcv, can be decomposed as

where each of the terms $1,2(^5 £2) is a 4-component spinor of the type (79). Therefore thephysical states have eight spinor components rather than four. Working out the action ofthe Dirac operator on these wave functions, eq.(151), it can be written explicitly in matrixnotation, in terms of ordinary four-dimensional Dirac matrices, as

f i P m ) [ ! i ] = / £ r ^ + m r 6 U = 0, (153)

28

where the TM are an 8-dimensional representation of the Dirac matrices in 6-dimensionalspace-time. Thus the supersymmetric spinning particle can be thought of as a reductionof a 6-dimensional massless spinor to 4 space-time dimensions, by compactification on acircle in the 6th dimension (p6 — m) and trivial in xs (p5 = 0). It follows, that the statesof this theory represent a degenerate pair of 4-dimensional massive fermions.

10 Supersymmetric propagatorIn the co-ordinate representation, the quantum states of the supersymmetric particle inthe full extended state space (including the ghost degrees of freedom) are represented bywave functions depending on the variables Z = (xM, £fe, e, x-> ci &j «)j8), with fi = 0,..., 3 andk — 1,2,3. To obtain the second quantized propagator for this theory, we first constructthe evolution operator associated with the hamiltonian Hgf, eq.(142).

It is actually convenient to do this in two steps: first we construct the expression forthe matrix elements of K(T) = e~tTB'f restricted to the space of gauge- and ghost degreesof freedom z = (e,x;c,b;a,f3), leaving the 'matter' content (a?**,^) unspecified; only thenwe specify the precise model for the physical degrees of freedom. The advantage of thisprocedure is, that the results are easily applied to other models, for example particlesinteracting with background fields.

Our starting point is the general expression (142) for B.gf-

d d2 a2

Hgf = eH0 - ixQ+ + 2iaX- + . ' — - — ,

where J7o = Q\- For any such Ho and Q+ not depending on the gauge- and ghost degreesof freedom the following results hold:

1. The equation

i-fr K(z, z'\T) = Hgf K(z, z'\T),

with the initial condition

K(z, z'\0) = 8(e - e')S(x - y!)6(c - c')S(b - b')S(a - a')8{P - p) I, (155)

has the solution

K(z, z'\T) = i - 8(e - e')8(X - x') «*^W-/n-(«+-0(*-V)x-ir (6-6')(C-c')

(156)-*T(eHo-txQ+)

29

2. This solution satisfies the composition rule

j dz' K(z, z'\T) K(z', z"\T") = K(z, z"\T + T"). (157)

We observe that the gauge (e, x) — const, is manifestly realized in the expression (156).Let us now first derive an operator expression for the propagator in the general case (beforespecifying Ho and Q+):

*./(*, z') = £- jH dTK(z, z'\T), (158)

which is a solution of the generalized Klein-Gordon-Dirac equation

HgfAaf(z,z') = ~S(z-z')l (159)

To get an explicit expression for Agf we use the following intermediate results:

e =27T

and

e =

Note that the factors of T in front of these ghost kernels cancel in the expression forK(T), as expected from supersymmetry. It is now straightforward to obtain the operatorexpression for the propagator in the ghost-momentum space:

- iXQ+ + P«Pe

(162)The matrix elements of the inverse operator inside the square brackets now have to becomputed. Again, we first consider the evolution operator. For the free particle we use theelementary result

J

30

In addition we replace the fermionic operators (^,•06) by their symbols as in eq.(97),supplemented by

) ( 1 6 4 )

where we have returned to the original notation for the additional fermionic degree offreedom. Similarly, for later convenience, we also introduce the symbol of the presentlyredundant operator ^>5, denoting it by 771:

( 1 6 5 )

Carrying out the integration over momentum pM then gives the expression for the matrixelement K(Z,Z'\T):

Ke(Z, Z'\T) = —S(e- e')6(X - x')2"K

( 1 6 6 )

Note that after reseating T and x by a factor e (or equivalently, taking e = 1), the expressionin brackets involving the physical degrees of freedom is almost identical with the expression(103) for the kernel KX(T) of the single Dirac fermion, except for the replacement of theGrassmann-even V'5(£*>f*)» (* = 1>2), by the Grassmann-odd T / ^ 3 , ! 3 ) -

The expression for the propagator Afff(Z, Z') in the full ghost-extended state space issimilarly obtained by taking the matrix element of A(z, z'); this amounts to making thereplacement

-— \eHa — ixQ+ + paP8 - w J * = ^- I dTe~Zm

(2TT)2 p2 + m2 — is + ^ [—ix(p • i> + rni}) - tir6irc +

(167)

in the r.h.s. of eq.(162), and interpreting {^,7}) as the above symbols.The construction of the path-integral formula for the propagator now repeats the steps

for the Dirac fermion, with the difference that instead of the symbol of the physical partdepending on two canonical pairs of Grassmann variables (£fc,£*), there are now three suchpairs. Then of course there is also the difference that we have included here additionalghost degrees of freedom. However, these do not cause any difficulties. As a result weobtain

31

Ke(Z,Z'\T) = j\fldZk] eli:r=1[(^-^-O^-^(^

N

j(168)

Ke(Zs+1,Z,\AT),*=0

where the exponent involves 3 types of (£*,!*) and as before AT = T/(iV + 1).Representing once more the delta-functions by their Fourier decomposition, and thereby

including integrations over lagrange multipKers (Ak,Sfe) in the measure Y[k dZk, we can useeq.(166) to evaluate the product of iTe(^,+i,Z,|AT)-factors on the right:

xe

xe

ra-ic)+i^X.+^..(^^

Here the >̂£ are defined as in eq.(109), and we take similarly(169)

( i ro)

Also, in the front factor on the right-hand side of (169) we have included a contribution(eo)2^"1"1) in the denominator, instead of IIfcLo(efc)25 kv making use of the fact that ek isconstant: in the integral its value remains the same from time step to time step.

Finally, the exponent of the finite differences in (£*, &) in eq.(168) can in the continuumlimit be rewritten in terms of the *̂*, rj and 7/1 and their derivatives:

E [(6 - 6-1) • 6 - 6 • (6+1 - &)] -™ i> + vv + (in)

As a result we can now construct the propagator of the theory in the full ghost-extendedstate space as

°°Agf(Z, Z1) = £- J dT I DZ(T) « ], (172)

where integration over the lagrange multipliers (A(r), s(r)) is to be included in the measureT>Z(T), and modulo fermionic boundary terms the action Sgf is that of eq.(129) with the

32

addition of a single topological fermion of the kind (121). As has been discussed before,at the classical level this additional fermion is completely harmless, whilst in the quantumtheory it signifies the doubling of the number of components of the spinor wave-functions,or equivalently the doubling of the number of degrees of freedom in the propagator.

11 Conclusions

In this paper we have studied the propagators of free spin-0 and spin-1/2 particles andconnected them to classical relativistic particle-mechanics through the path-integral for-malism. It has been established that starting from the known field-theoretical expressionsis advantageous, as it can specify the representation of certain operators to be used, some-thing which is usually not possible from the canonical 'Poisson-bracket to commutator'quantization prescription. In the case of Dirac fermions, this has been used to argue infavour of the bosonic representation of the operator ^5, implying that non-manifestly su-persymmetric models are prefered to avoid doubling of the spectrum of states. Anotherway to resolve this problem would be to project out half of the states by additional con-straints. Such an approach, in which a superselection rule is invoked to restrict the physicalmatrix elements to half of the spinor degrees of freedom, has been attempted for examplein [37].

Our results can be generalized to the case of particles moving in certain backgroundfields: scalar, vector (e.g., electro-magnetic) or gravitational fields can be included in thepath-integral expressions for the propagators. The constructions we have presented havebeen designed in such a way that only minimal additional work is needed to cover thesemore general cases. An example is the inclusion of scalar fields, which by Yukawa cou-plings may give rise to mass generation through the mechanism of spontaneous symmetrybreaking. The contribution of the scalar interactions to the classical action (119) is ob-tained by replacing the terms for the fermionic multiplet (77, / ) , which for free particlesread 17777 + 2ixv + e/2 — 2e/, by

ALac = irfq + ef - 2gef<p(x) - 2iger}ip(idtl<p(x) + 2igxv<p(x)- (173)

Here <p(x) is the scalar field, and g is the Yukawa coupling constant. Eliminating theauxiliary variable / by splitting off a square, this becomes

A L S C = 27777 - eg2tp2 -

This result, here obtained through multiplet calculus, agrees with the result derived bydimensional reduction in [38]. Taking <p(x) = const. ^ 0 returns us to the original actionfor a free massive particle, and shows how masses are generated by spontaneous symmetrybreaking. It is quite straightforward to apply these results to the other actions for scalar

33

and Dirac particles, by either removing all fermionic degrees of freedom, or by keepingthem whilst replacing the fermionic 77 by the bosonized ^>5.

Interactions with the electro-magnetic or other vector fields can be introduced, e.g.through minimal coupling, whilst gravitational interactions result from covariantizing theexpressions with respect to general co-ordinate transformations. A quantum treatment ofthe point particle in curved space has been presented in [7]. A general discussion of theinclusion of background fields will be presented in a separate paper [39].

Acknowledgement

The research described in this paper is supported in part by the Human Capital andMobility program of the European Union through the network on Constrained DynamicalSystems.

References[1] R.P. Feynman, Rev. Mod. Phys. 20 (1948), 267;

R.P. Feynman and A.R. Hibbs, Quantum Mechanics and Path Integrals (McGraw-Hill,1965)

[2] A.M. Polyakov, Gauge Fields and Strings (Harwood), 1987

[3] M.J. Strassler, Nucl. Phys. B385 (1992), 562

[4] D.G.C. McKeon, Ann. Phys. (NY) 224 (1993), 154

[5] M.G. Schmidt and C. Schubert, Phys. Lett. B318 (1994), 438;Phys. Lett. B331 (1994), 69

[6] J.W. van Holten, Z. Phys. C66 (1995), 303

[7] J. de Boer, B. Peeters, K. Skenderis, and P. van Nieuwenhuizen, Nucl. Phys. B446(1995), 211

[8] Z. Bern and D.C. Dunbar, Nucl. Phys. B379 (1992), 562

[9] J.W. van Holten, Int. J. Mod. Phys. A7 (1992), 7119

[10] W. Kalau and J.W. van Holten, Nucl. Phys. B361 (1991), 233

[11] A. Cohen et al., Nucl. Phys. B267 (1986), 143

[12] J.C. Henty, P.S. Howe and P. Townsend, Class. Quantum Grav. 5 (1988), 807

34

[13] J. Schwinger, Phys. Rev. 82 (1951), 664

[14] R. Casalbuoni, J. Gomis and G. Longhi, Nuov. Cim. A24 (1974), 249

[15] L. Brink et a l , Phys. Lett. B64 (1976), 43

[16] L. Brink, P. Di Vecchia and P. Howe, Nucl. Phys. B118 (1977), 76

[17] T. Kugo and I. Ojima, Suppl. Progr. Theor. Phys. 66 (1979)

[18] M. Henneaux, Phys. Rep. 126 (1985), 1

M. Henneaux and C. Teitelboim, Quantization of Gauge Systems (Princeton Univ.Press, 1992)

[19] J.W. van Holten, in: Functional Integration, Geometry and Strings, eds. Z. Haba andJ. Sobczyk (Birkhauser, 1989)

[20] E.S. Fradkin and T.E. Fradkina, Phys. Lett. B72 (1978), 343

[21] M. Henneaux and C. Teitelboim, Comm. Math. Phys. 115 (1988), 213

[22] M. Henneaux and C. Teitelboim, Ann. Phys. 143 (1982), 127

[23] S. Monaghan , Phys. Lett. B178 (1986), 231

[24] J.W. van Holten, Phys. Rev. Lett. 64 (1990), 2863; Nucl. Phys. B339 (1990), 158

[25] F.A. Berezin, The Method of Second Quantization, (Ac. Press, 1966)

[26] L.D. Faddeev, in: Methods in Field Theory, eds. R. Balian and J. Zinn-Justin (North-Holland, 1976)

[27] B. DeWitt, Supermanifolds (Cambridge Un. Press)

[28] F.A. Berezin and M.S. Marinov, Ann. Phys. (NY) 104 (1977), 336

[29] R. Casalbuoni, Phys. Lett. B62 (1976), 49

[30] A. Barducci, R. Casalbuoni and L. Lusanna, Nuov. Cim. 35A (1976), 377; Nucl. Phys.B124 (1977), 93

[31] M. Halpern, P. Senjanovic and A. Jevicki, Phys. Rev. D16 (1977), 2476

[32] M. Halpern and W. Siegel, Phys. Rev. D16 (1977), 2486

[33] J.W. van Holten, in: Proc. Sem. Math. Structures in Field Theories (1986-87), CWIsyllabus vol. 26 (1990), 109

35

[34] J.W. van Holten, Grassmann Algebras and Spin in Quantum Dynamics, lecture notesAIO School Math. Phys. (Un. Twente, 1992)

[35] E.S. Fradkin and D.M. Gitman, Phys. Rev. D44 (1991), 3230

[36] E. Ercolesi et al. Path integrals for spinning particles, stationary phase andDuistermaat-Heckman theorem, Bologna preprint (1995)

[37] A. Barducci, F. Bordi and R. Casalbuoni, Nuovo Cim. 64 B (1981), 287

[38] M. Mondragon, L. Nellen, M.G. Schmidt and C. Schubert, Phys. Lett. 351B (1995),200

[39] J.W. van Holten, preprint NIKHEF/95-055

36