KCL-MTH-99-50 December 1999

Canonical Quantization and Topological Theories

Introduction

In this talk the rst steps in a rigorous study of topological quantum theories using canonical methods is described. A topological theory is characterised by a high degree of symmetry, with the number of symmetries equal to the number of elds of the theory. As a result the system has no dynamical degrees of freedom, and the space of elds modulo symmetries, which is trivial at the linear level, is typically a nite-dimensional moduli space which encodes topological information.

As with quantum theories in general, a powerful tool for studying a topological theory is functional integration. The generic expression for the vacuum generating functional takes the form D exp (iS[])
elds/symmetries

(1)

where S() is the action of the eld . Using standard (non-rigorous) methods of quantum eld theory a number of new and unexpected mathematical results have been been derived from topological models, results which in many cases have then been fully proved by more standard mathematical methods, but which would probably not have been discovered without the insights gained from the quantum eld theory. (An early appearance of topological invariants in the quantum eld theoretic situation is due to Belavin, Polyakov, Schwarz and Tyupin [1]. A more recent example of the powerful application of topological quantum eld theory in mathematics may be found in [2], while fuller accounts of earlier work in this eld may be found in the books of Nash [3] and Schwarz [4].) Most functional integrals such as (1), and related expressions with operator insertions, have not at present been properly dened. However, since these integrals have such astonishing mathematical power, it seems that an attempt to dene these objects rigorously should be more than worth while. In this talk we show how this may be done for the simplest topological model, the topological particle, and describe briey some recent work by Hrabak [5] which might lead to progress in the canonical quantization of topological eld theories. Some rigorous results on path integrals (that is, functional integrals in quantum mechanics) are known. The basic classical result (which is described by Simon in [6]) for a particle of unit mass moving in one dimension with Hamiltonian 1 H = p2 + V (x) (2) 2 gives the action of the imaginary time evolution operator exp(Ht) on a wave function (x) by the formula
t

exp(Ht)(x) =

d exp
0

V ((x(s))ds (x(t))

(3)

where d denotes Wiener measure starting from x, and x(t) are corresponding Brownian paths; the potential V must satisfy certain analytic conditions. 2

The curved space analogue of this result for a Riemannian manifold has been developed by Elworthy [7] and by Ikeda and Watanabe [8]. The expression for evolution according to the Hamiltonian H = L + V (x) where L is the scalar Laplacian looks identical to (3), but with x(t) a process depending on metric and connection rather than simply at space Brownian motion. Tangent space geometry plays an essential part in the theory. The present author has further extended these methods by developing a at space theory of fermionic path integrals [9] and marrying it with Brownian motion on manifolds to give Brownian motion on supermanifolds in a suitable form for handling the Hodge-de Rham operator and the Dirac operator on manifolds [10, 11].

The Topological particle

Following Beaulieu and Singer [12] we consider the quantum-mechanical model with elds which are maps x : I M, where I is the interval [0, t] and M is an n-dimensional Riemannian manifold with metric g. The action of the theory is
t

S[x(.)] =
0

i (x(t ))x (t ) dt

(4)

= dh is an exact one form on M with local co-ordinate expression = (x)dx and x (t ) = dx . The action can be expressed in the simpler form dt S[x(.)] = i(h(x(t)) h(x(0))) which shows that the action is indeed highly symmetric, being independent of all but the endpoints of the paths x(t). While Beaulieu and Singer consider the case where = 0, we consider the case where h is a Morse function on M (so that = dh is only zero at isolated points). To carry out the canonical quantization we rst evaluate the momentum p conjugate to x , obtaining L (5) p = = i , x which shows that the theory has n constraints T p i . (6) The Hamiltonian of the theory is as usual dened to be H(p, x) = p x L(x, x), so that, as is generally the case for a topological theory, the Hamil tonian of the theory (prior to gauge-xing) is zero. Since is closed these constraints are rst class and abelian, that is {T , T } = 0 and {T , Hc } = 0. 3

The innitesimal gauge transformations generated by T are (x) = i ( (x) + (x)(x)) (7)

where (x) is a wave function and quantization is in the Schrdinger pico ture with p = i . (Below, when ghosts are introduced, we will nd that we require p to be represented as a covariant derivative i .) The explicit form of the gauge transformations suggests that representative of gauge equivalence classes may be obtained from the condition X = 0 where X = g (p + i ). The validity of these gauge-xing conditions will become clear below. The BRST quantization scheme will now be applied to this system; to do this anticommuting ghosts together with their conjugate momenta are introduced. (The phase space is now a (2n, 2n)-dimensional supermanifold, with odd coordinates , transforming as indices suggest.) Poisson brackets on this extended phase space are dened by the symplectic form 1 + dx dx R , (8) 2 (where denotes covariant dierentiation using the Levi-Civita connection); this is a special case of the symplectic form introduced by Rothstein [13]. Quantization is carried out by introducing states represented by wave func tions (x, ) and momenta acting as p = i , = i . The wave functions (x, ) are functions on the (n, n)-dimensional supermanifold SM with local coordinates x , , and the explicit form of the action of the covariant derivative on a wave function is given by dp dx +
(x, )

= (x, ) +

(x, ).

(9)

The BRST charge Q takes the standard form Q = T = i ( + ). (A covariant derivative is not required here because of the symmetry of the connection.) The gauge-xing fermion also takes the standard form = X = ig ( ). Alternatively, if we make the natural identication of forms on M with wave functions (x, ) then Q = ieh deh , = eh eh where d denotes exterior dierentiation of forms and = d is the adjoint operator. The gauge-xing Hamiltonian is then Hg = i(Q + Q) = d + d + g i( ) 4 2h . x x (10)

This Hamiltonian has appeared in the literature on other occasions; for instance, when h is constant, it is the Hamiltonian used by Alvarez-Gaum e [14] to prove the Atiyah-Singer index theorem. (A rigorous version of this proof may be found in [10].) It is also the rst supersymmetric Hamiltonian used by Witten in his study of Morse theory [15]. It is evident that the choice = X is a good gauge-xing condition, satisfying the essential conditions derived by the author in [16]. First, as observed by Beaulieu and Singer in the constant h case, the standard theory of harmonic forms shows that the gauge condition determines a unique element of each Q cohomology class. (The observation that these arguments extend to all functions h is due to Witten [15].) Also, the zeros of Hg coincide with these representatives of the cohomology classes, while the eigenvalues of Hg tend to innity, so that this Hamiltonian does regulate the non-physical states. The path integral formulae for this Hamiltonian can be put in rigorous form using the methods of the author in [10, 11], and used to establish rigorous results for this model. The key idea in this approach is to use Brownian paths xt , t in the supermanifold SM with local coordinates x , . The Brownian paths are dened by the stochastic dierential equations
t

x = x + t
0 t

e dba , a,s s e e (xs ) dbb a,s b,s s

e = e + a,t a
t

a t = + t e a,t

+
0

1 a ( s e dbb t de + s R (xs )s a e ds), s a,s s a,s b,s 4

(11)

where bt is at bosonic Brownian motion and (t , t ) is at fermionic Brownian motion. The measure corresponding to this process incorporates as the kinetic term the heat kernel of the Laplace-Beltrami operator (d + )2 , so that these Brownian paths are appropriate for the analysis of the Hamiltonian (10). Further details will be found in [17], where it will be shown that Wittens approach to Morse theory [15] can be put on an entirely rigorous mathematical footing. (Some parts of Wittens analysis have been proved rigorously by Simon et al [18] and by Mathai and Wu [19]; however the explicit modeling of the manifolds cohomology via critical points and instanton calculations does not appear to have received a full mathematical treatment.) 5

The two-dimensional topological sigma model

To conclude, a brief indication of some developments in a two-dimensional topological model will be described. The model, which was rst proposed by Witten [20], concerns the geometry of J-holomorphic curves (or pseudoholomorphic maps) u : M from a Riemann surface into an a 2m-dimensional almost-Kaehler manifold M. The other elds of the theory are a bosonic set H and two fermionic sets and . (Here = 1, 2 are indices on while = 1, . . . , 2m are indices on M.) The elds H and satisfy constraints P H = 0, P = 0 where P = J is a projec tion operator with the complex structure on and J the almost complex structures on M. Witten constructs by hand a set of supersymmetry transformations (beginning with u = i ) on these elds, and then an invariant action. The model is used to derive deep geometric insights into the moduli space of J-holomorphic curves on M. Recent work of Hrabak [5] shows that the rather complicated and seemingly ad hoc supersymmetry transformations of the model can be derived in the canonical setting as BRST transformations; the novel feature of Hrabaks work is that the formalism used is not the standard canonical formalism (in which time plays a special rle) but the o multisymplectic formalism which is manifestly covariant; corresponding to the elds u : M there are multimomenta p (which after projection re late to Wittens H ), while in the ghost sector the ghosts have momenta P which relate to Wittens . The BRST symmetry obtained by Hrabak corresponds directly to the J-holomorphicity of the embeddings. Recent work by Kanatchikov [21] on quantization in the multisymplectic framework suggests that it may be possible to use Hrabaks approach to carry out a full canonical quantization of Wittens interesting two-dimensional topological model.

References
[1] A.A. Belavin, A.M. Polyakov, A.S. Schwarz, and Yu S. Tyupin. Physics Letters, 59B:85, 1975. [2] E. Witten. Monopoles and four manifolds. Math. Res. Lett., 1, 1994. [3] C. Nash. Dierential topology and quantum eld theory. Academic Press, 1991. 6

[4] A. Schwarz. Quantum Field Theory and Topology. Springer, 1993. [5] S.P. Hrabak. On the multisymplectic origin of the nonabelian deformation algebra of pseudoholomorphic embeddings into strictly almost kahler ambient manifolds, and the corresponding BRST algebra. Preprint mathph/9904026. 1999. [6] B. Simon. Functional Integration and Quantum mechanics. Academic Press, 1979. [7] K.D. Elworthy. Stochastic Dierential Equations on Manifolds. London Mathematical Society Lecture Notes in Mathematics. Cambridge University Press, 1982. [8] N. Ikeda and S. Watanabe. Stochastic dierential equations and diusion processes. North-Holland, 1981. [9] A. Rogers. Fermionic path integration and Grassmann Brownian motion. Communications in Mathematical Physics, 113:353368, 1987. [10] A. Rogers. Stochastic calculus in superspace II: dierential forms, supermanifolds and the Atiyah-Singer index theorem. Journal of Physics A, 25:60436062, 1992. [11] Alice Rogers. Path integration, anticommuting variables and supersymmetry. Journal of Mathematical Physics, 36:25312545, 1995. [12] Beaulieu and I. Singer. The topological sigma model. Commun. Math. Phys., 125:227237, 1989. [13] M. Rothstein. The structure of supersymplectic supermanifolds. In C. Bartocci, U. Bruzzo, and R. Cianci, editors, Dierential geometric methods in theoretical physics, proceedings, Rapallo 1990, volume 375 of Lecture Notes in Physics, pages 331343. Springer, 1991. [14] L. Alvarez-Gaum. Supersymmetry and the Atiyah-Singer index theoe rem. Comm. Math. Phys., 90:161173, 1983. [15] E. Witten. Supersymmetry and Morse theory. Journal of Dierential Geometry, 17:661692, 1982.

[16] A. Rogers. Gauge xing and BFV quantization. Classical and Quantum Gravity, to appear. [17] A. Rogers. The topological particle. (In preparation.) [18] H.L. Cycon, R.G. Froese, W. Kirsch, and B. Simon. Schrdinger opero ators. Springer, 1987. [19] S. Wu and V. Mathai. Equivariant holomorphic morse inequalities. 1: A heat kernel proof. ICTP preprint IC/96/29, 96. [20] E. Witten. Topological sigma models. Commun.Math.Phys., 118:411, 1988. [21] I. V. Kanatchikov. On quantization of eld theories in polymomentum variables. To be published in the proceedings of International Conference on Particles, Fields and Gravitation (Devoted to the Memory of Professor Ryszard Raczka), Lodz, Poland, 15-18 Apr 1998., AIP Proceedings 1998 hep-th/9811016 .