On the deconfining limit in (2+1)-dimensional Yang-Mills theory
aa r X i v : . [ h e p - t h ] N ov On the deconfining limit in(2+1)-dimensional Yang-Mills theory
YASUHIRO ABE
Cereja Technology Co., Ltd.3-1 Tsutaya-Bldg.5F Shimomiyabi-choShinjuku-ku, Tokyo 162-0822, Japan [email protected]
Abstract
We consider (2+1)-dimensional Yang-Mills theory on S × S × R in the framework of aHamiltonian approach developed by Karabali, Kim and Nair. The deconfining limit in thetheory can be discussed in terms of one of the S radii of the torus ( S × S ), while the otherradius goes to infinity. We find that the limit agrees with the previously known result for adynamical propagator mass of a gluon. We also make comparisons with numerical data. Introduction
The Hamiltonian approach to (2+1)-dimensional Yang-Mills theory has been known over adecade as a novel framework for non-perturbative analysis of the theory [1]. Technical elab-orations might be necessary in the formulation of the Hamiltonian approach, in particular,with regard to regularization processes [2], but the upshot of the Hamiltonian formulationis quite simple; namely, it gives rise to (a) an interpretation of an origin of the mass gapand (b) an analytic calculation of the string tension for SU ( N ) gauge groups [2, 3]. Theseresults, obtained by Karabali, Kim and Nair (KKN), are remarkable not only in consequenceof a (conformal) field theoretic framework but also in comparison with lattice simulationsof the string tension [4]. Recent developments relevant to the Hamiltonian approach can befound in [5, 6, 7, 8, 9]. There are also various other analytic approaches to the theory; forrecent progress, see, for example, [10, 11, 12].In the present paper, we follow the Hamiltonian approach to consider the deconfining limitin (2+1)-dimensional Yang-Mills theory. In the Hamiltonian approach, confining propertiesof the theory can be shown by the following steps [3]:1. matrix parametrization of gauge fields;2. calculation of a gauge-invariant measure on the configuration space;3. evaluation of a vacuum-state wave function Ψ and its inner product in terms of gauge-invariant variables;4. calculation of the vacuum expectation value of the Wilson loop operator h Ψ | W ( C ) | Ψ i ;and5. reading off the area law or the positive string tension from h Ψ | W ( C ) | Ψ i .The goal of this paper is to rephrase each of the above steps for the theory on S × S × R such that we can discuss the limit of vanishing string tension in terms of one of the S radiiwhich may correspond to a parameter of finite temperature. For such a parameter we willuse Im τ , where τ is the modular parameter of a torus ( S × S ) in the three-dimensionalspace S × S × R of our interest. A radius corresponding to the other S is taken to belarge (or equivalently the corresponding winding number is taken to be large, with the radiusbeing finite) so that we can identify the theory on S × S × R as a planar theory at a finitetemperature in such a limit. This implies that one of the S directions corresponds to thetime coordinate.In the KKN Hamiltonian approach, one takes the temporal gauge A = 0 and analysesare made entirely on complexified spatial dimensions by use of conformal properties. Inthe present paper, we however take a different gauge, say A x = 0, to make an analysis ofgauge potentials on torus which include a time component. Such an analysis may containsubtlety in discussion of physical dynamics but for argument of static properties it maystill be useful since the theory of interest is relevant to the one with imaginary time or theEuclidean metric. In this paper, leaving that subtlety aside, we focus rather on construction2f the vacuum wave function Ψ in pure Yang-Mills theory on S × S × R (with one of S directions corresponding to an imaginary time coordinate), following the framework of theKKN Hamiltonian approach. We shall not deal with a more involved regularization program,either. Note that (2+1)-dimensional Yang-Mills theory on torus in spatial dimensions hasbeen studied before in a different context [13].Apart from what have been mentioned, our main motivation to consider the deconfininglimit is currently available lattice data on the deconfinement phase transition in (2+1)-dimensional Yang-Mills theory [14, 15, 16, 17, 18, 19]. In the Hamiltonian approach, thestring tension is obtained for a continuum strong coupling region where e /p ≫ e and p being the coupling constant and a typical momentum scale, respectively. So weshall limit our analysis in this region and do not discuss the nature of deconfinement phasetransition. What we aim at is, however, to obtain a critical temperature of deconfinementtransition which can be compared with the numerical data.Study of (2+1)-dimensional Yang-Mills theory on S × S × R is therefore physicallywell-motivated. In order to execute the study in the Hamiltonian approach, however, thereis a key mathematical concept to be reminded of, that is, physical states of the planar Yang-Mills theory in the Hamiltonian approach can be described in terms of holomorphic wavefunctionals of Chern-Simons theory. In the next section, we briefly review this relation. Oncewe understand how this relation arises, consideration of the toric theory would be clearer byuse of the so-called Narashimhan-Seshadri theorem [20], which we also mention in the nextsection.In section 3, following the references [21, 22, 23], we present a matrix-parametrization ofgauge potentials such that it is incorporated with the zero modes of torus. This section coversthe first two steps of the above list. In section 4, we deal with the third step. Technicallyspeaking, the objective of this section is to find holomorphic functionals of Chern-Simonstheory on torus in a context of geometric/holomorphic quantization [22, 23]. In section 5,utilizing the results of section 4, we consider the last two steps of the above list. We obtainan expression for a deconfinement temperature and make comparisons with lattice data. Inthe last section, we present brief concluding remarks. In the Hamiltonian approach, the gauge potentials A i ( i = 1 , ,
3) are parametrized by theelements of SL ( N, C ). The gauge group we consider is G = SU ( N ); A i can be writtenas A i = − it a A ai where t a ’s are the elements of SU ( N ) represented by ( N × N )-matrices,satisfying Tr( t a t b ) = δ ab and [ t a , t b ] = if abc t c . Note that under the temporal gauge A = 0the gauge potentials are described by A z = ( A + iA ), A ¯ z = ( A − iA ) where z = x − ix , ¯ z = x + ix are a complex combination of the spatial coordinates ( x , x ). Matrixparametrization of the gauge potentials is given by A z = − ∂ z M M − A ¯ z = M †− ∂ ¯ z M † (1)3here M ( z, ¯ z ) , M † ( z, ¯ z ) are the elements of SL ( N, C ). Gauge transformations of A z , A ¯ z canbe realized by M → gM , M † → M † g − with g ∈ SU ( N ). A gauge invariant matrix variableis given by H = M † M . The parametrization (1) corresponds to step 1 of the list in theintroduction.Let A denote the set of all gauge potentials A ai . The gauge-invariant configuration spaceis then given by C = A / G ∗ (2)where G ∗ = { set of all g ( ~x ) : R → SU ( N ), with g → | ~x | → ∞} . (3)The calculation of the gauge-invariant measure in step 2 leads to the following result, up toan irrelevant constant factor [1]: dµ ( C ) = dµ ( H ) e c A S WZW ( H ) (4)where c A denotes the quadratic Casimir of G for the adjoint representation, c A δ ab = f amn f bmn .For G = SU ( N ), this is equal to N . S W ZW ( H ) is the action for a G C /G Wess-Zumino-Witten(WZW) model where G C = SL ( N, C ) is the complexification of G = SU ( N ). Explicitly,the action is given by S W ZW ( H ) = 12 π Z d z Tr( ∂ z H∂ ¯ z H − ) + i π Z d xǫ µνα Tr( H − ∂ µ HH − ∂ ν HH − ∂ α H ) (5)where dz (or dx ) is a real two-dimensional volume element which is equivalent to dzd ¯ z/ i .For description of a coset model in connection with gauged WZW models, one may refer to[21, 24, 25, 26]. The inner product of physical states in general, not only for vacuum states,can be evaluated as correlators of the G C /G -WZW model; h | i = Z dµ ( H ) e c A S WZW Ψ ∗ ( H )Ψ ( H ) . (6)where Ψ ∗ ( H ) indicates the complex conjugate of a wave functional Ψ( H ). The wave func-tionals for the vacuum state can be given byΨ ( H ) = 1 . (7)Equation (6) provides an essential setup for the KKN Hamiltonian approach.It is known that current correlators of a WZW model can be generated by holomorphicwave functionals of Chern-Simons theory [22, 23]. The wave functional Ψ( H ) can then beinterpreted as a functional which arises from a holomorphic wave functional ψ [ A ¯ z ] of Chern-Simons theory. In a context of geometric/holomorphic quantization of Chern-Simons theory[22, 23, 28], a polarization condition is imposed on the functional, i.e. ,Ψ[ A ¯ z ] = e − K ψ [ A ¯ z ] . (8) It is well known by the study of knots in terms of Chern-Simons theory [27] that the conformal blocksof current algebra on Σ (or the current correlators of a WZW model on Σ) correspond to the sections ofholomorphic line bundle over M (or the generating functional of Chern-Simons theory on M ), where Σ isa two-dimensional compact space, M is a moduli space of flat connections on a G -bundle over Σ and M isa three-dimensional closed manifold with its boundary being ∂ M = Σ. K is a K¨ahler potential associated with the phase space of Chern-Simons theory in the A = 0 gauge; K = k π Z Σ A a ¯ z A az = − kπ Z Σ Tr( A ¯ z A z ) (9)where Σ indicates Riemann surface and k is the level number of Chern-Simons theory. Theintegral over Σ is taken for dz = dzd ¯ z/ i . In terms of the parametrization (1), ψ [ A ¯ z ] canbe written as ψ [ M † ]. The flatness of the gauge potentials, which is required as the equationmotion for A (or the gauss law constraint) of Chern-Simons theory, must be satisfied on theholomorphic wave functional ψ [ A ¯ z ], i.e. , F z ¯ z ψ [ A ¯ z ] = 0 where F z ¯ z = ∂ z A ¯ z − ∂ ¯ z A z + [ A z , A ¯ z ].This leads to ψ [ M † ] = e k S WZW ( M † ) . The inner product of the gauge-invariant physical statesis then given by [23] h | i CS = Z dµ ( C ) e − K ψ ∗ ψ = Z dµ ( H ) e (2 c A + k ) S WZW ( H ) (10)where we use (4) and ψ ∗ = ψ [ M ] = e k S WZW ( M ) together with the Polyakov-Wiegmannidentity [29] S W ZW ( H ) = S W ZW ( M † M ) = S W ZW ( M † ) + S W ZW ( M ) − π Z Σ Tr( M †− ∂ ¯ z M † ∂ z M M − )= S W ZW ( M † ) + S W ZW ( M ) + 1 π Z Σ Tr( A ¯ z A z ) . (11)Comparing (6) and (10), we find that the inner product of the vacuum wave functionalson the planer Yang-Mills theory can be obtained by the inner product (10) in the limit of k →
0. The theory defined by the correlator (10) with positive k is known as Yang-Mills-Chern-Simons theory [30, 31, 32].It is interesting that the physical states of (2+1)-dimensional Yang-Mills theory can beobtained in terms of the holomorphic wave functionals of Chern-Simons theory. A simpleexplanation of this relation is that, as shown in the first reference of [1], under the A =0 gauge commutation rules among the gauge potentials A i ’s and the electric fields E i ’s(which are canonical momenta of A i ’s) can be interpreted as two copies of the Chern-Simonscommutation rules among A z ’s and A ¯ z ’s in the same gauge.For physical states other than the vacuum, the holomorphic wave functional ψ [ A ¯ z ] ofChern-Simons theory may be expressed as ψ [ M † ] = e kS WZW ( M † ) F [ M † ] in general where F [ M † ] is a matrix function of M † . Thus ψ [ M † ] may not lead to the gauge invariant functionalΨ( H ) in (6). However, as shown in [8], one can in fact take a suitable gauge choice suchthat F [ M † ] depends on the current of the hermitian SL ( N, C ) /SU ( N )-WZW model, J a = c A π ( ∂ z H H − ) a . In terms of this gauge-invariant current, the flatness of the gauge potentialcorresponds to an equation of motion of the hermitian WZW model, ∂ ¯ z J a = 0. In theHamiltonian approach, this indicates the vanishing of magnetic fields which act on ψ [ A ¯ z ].Note that magnetic fields do not necessarily vanish when acted on Ψ[ A ¯ z ]. In the presentpaper, effects of magnetic fields will not be discussed in constructing the vacuum-state wavefunctional. 5ow let us return to the toric theory. It is known that there exist flat connections (or non-trivial gauge potentials that lead to vanishing curvature) for any compact two-dimensionalspaces Σ which have complex structure. The relation between the current correlators ofa WZW model and the holomorphic wave functions of Chern-Simons theory, given by theexpression of (6) or (10), therefore holds for any Riemann surfaces Σ including a torus. Forthe study of WZW models on Riemann surfaces and on torus in particular, see [21, 33, 34].A main purpose of the present paper from a mathematical perspective is to clarify the struc-ture of the inner product of such holomorphic wave functionals when the Riemann surfaceis given by a torus.As seen in the next section, one can in fact incorporate zero modes of torus into a matrix-parametrization of gauge potentials, analogous to the form in (1). What we need to do istherefore to obtain a toric version of (6) or (10) by use of such a matrix parametrization. Amain claim we like to make in this paper is that, in the toric case, the level number appearedin (10) no longer vanishes since it is now incorporated with nontrivial zero-mode dynamics.If the toric level number, say ˜ k , vanishes, there will be no topological differences from theplanar case. We discuss contributions of zero modes later in section 4. Torus can be described in terms of two real coordinates ξ , ξ with periodicity of ξ i → ξ i + n ( i = 1 ,
2) where n is any integer. Complex coordinates of torus can be parametrized as z = ξ + τ ξ where τ = Re τ + i Im τ is the modular parameter of the torus. There are twononcontractible cycles on torus, conventionally labeled as α and β cycles. By use of these aholomorphic one-form of the torus, ω = ω ( z ) dz , can be defined as Z α ω = 1 , Z β ω = τ (12)where the normalization of ω is given by Z dzd ¯ z ¯ ω ∧ ω = i τ . (13)Equivalently, this can be written as R Σ ¯ ω ∧ ω = Im τ with Σ = S × S . Note that ω is azero mode of ∂ ¯ z . In construction of matrix parametrization of gauge potentials on torus, wetherefore need to take ω and ¯ ω into account.We shall denote a as a complex physical variable of zero modes for the moment. We mayregard a as an abelian gauge potential corresponding to the zero modes of torus. Note that a and ¯ a satisfy the periodicity a → a + m + nτ and ¯ a → ¯ a + m + n ¯ τ ( m, n ∈ Z ) where Z denotes integer. m and n correspond to winding numbers of α and β cycles, respectively. In this case, the moduli space of flat connections on a G -bundle over Σ can be identified with the modulispace of a stable holomorphic G C -bundles on Σ. This mathematical fact is known as Narashimhan-Seshadritheorem [20]. τ = 0and choose mn to be a large integer such that m ≫ n = 1. Note that Im τ and m can beinterpreted as radii for the two circles of torus ( S α × S β ) corresponding to α and β cycles,respectively.Let us consider a change of variables for the one-forms ω and ¯ ω in terms of ξ and ξ .Note that ξ i ( i = 1 ,
2) take real values in 0 ≤ ξ i ≤ ξ i = 0, 1 being identical. Let ω and ω be ω = ( d ¯ z − dz ) / i = − Im τ dξ ω = ( τ d ¯ z − ¯ τ dz ) / i = Im τ dξ (14)where we assume Re τ = 0. Since an integral part of Re τ can be absorbed into m of theperiodicity of a → a + m + nτ , this assumption is equivalent to Re τ being an integer, whichmay not cause obstacles in the following discussion. With (14), holonomies of torus can berewritten as I α i ω j = (Im τ ) ǫ ij (15)where ǫ ij denotes a Levi-Civita symbol and α , α denote the the alpha and beta cyclesdefined in (12). Note that we can set ω ( z ) = 1 in (12) with identification of the alpha andbeta cycles by loop integrations of the variables ξ and ξ , respectively. Normalization for ω and ω is given by Z dzd ¯ z ω Im τ ∧ ω Im τ = 1 . (16)We now introduce a new set of variables corresponding to ω , ω by a = ¯ a − a , a = τ ¯ a − ¯ τ a . (17) One might argue the assumption of Re τ = 0 is relevant to the imaginary time formulation as discussedin the introduction but the relevance is not entirely clear at least for the author. a → a + m + nτ and ¯ a → ¯ a + m + n ¯ τ , a and a vary as δa → ( − i Im τ ) n ,δa → (2 i Im τ ) m . (18)From (15) and (18), we findexp I α πω Im τ δa Im τ ! = e − i πm , exp I α πω Im τ δa Im τ ! = e − i πn . (19)For a nonabelian case with an SU ( N ) gauge group, physical variables a , ¯ a are given bythe following matrix-valued quantities [21, 23]: a = a j t diag j , ¯ a = ¯ a j t diag j (20)where t diag j are the diagonal generators of G = SU ( N ) in the fundamental representation( j = 1 , , · · · , N − G . a j are complex variablessatisfying a j → a j + m j + n j τ with m j and n j being integer. In the expressions of (20), sumsover j should be understood.Nonabelian versions of a and a can also be given by (17) with a and ¯ a now defined as(20). By use of such a and a , we can express matrix parametrization of gauge potentialson torus, which is analogous to the planar case (1), as e A ξ = A ξ + M (cid:18) πω Im τ a Im τ (cid:19) M − , A ξ = − ∂ ξ M M − e A ξ = A ξ + M †− (cid:18) πω Im τ a Im τ (cid:19) M † , A ξ = M †− ∂ ξ M † (21)where ∂ ξ i denotes ∂∂ξ i ( i = 1 , ǫ ij ωa j combinations appeared in(19) such that we have invariance of e A ξ , e A ξ under the transformations of ( a, ¯ a ). In termsof a , ¯ a , the parametrization can be expressed as e A z = − ∂ z f M f M − = − ∂ z M M − + M (cid:18) πω Im τ ¯ a (cid:19) M − , (22) e A ¯ z = f M †− ∂ ¯ z f M † = M †− ∂ ¯ z M † + M †− (cid:18) π ¯ ω Im τ a (cid:19) M † (23)with f M and f M † now defined by f M = M exp (cid:18) − π Im τ Z z ω ¯ a (cid:19) ≡ M e γ z , f M † = exp (cid:18) π Im τ Z ¯ z ¯ ωa (cid:19) M † ≡ e γ ¯ z M † . (24)Equations (22) and (23) agree with previously known matrix parametrization for gaugepotentials on torus [21, 23]. 8 gauge invariant measure and decomposition of f H Let us now consider a gauge invariant measure which incorporates the zero modes oftorus. From the calculation of the planar case shown in (4) and from Narashimhan-Seshadritheorem, we can define the toric measure as dµ ( e C ) = dµ ( f H ) e c A S WZW ( e H ) (25)where f H = f M † f M = e γ ¯ z M † M e γ z = e γ ¯ z H e γ z . The change from H to f H can essentially be takencare of by replacing ∂ ¯ z with ∂ ¯ z + π ¯ ω Im τ a and similar for ∂ z .Decomposition of a , ¯ a out of f H may be achieved by imposing[ a, H ] = a j [ t diag j , H ] = 0 ( a j = 0) (26)This is a strong assumption we would like to impose on H later in the next section. In terms ofmatrix configuration, this basically leads to diagonalization of H for arbitrary choices of a j ’s( j = 1 , , · · · , N − a or ¯ a , i.e. , at least one of a j ’s should be non-zero. For example,we can choose a l = 0 ( l = 1 , , · · · , r ; 1 < r < N −
1) and a m = 0 ( l = r, r + 1 , · · · , N − H does not get fully diagonalized but has a block-diagonal structure with two blocks,one of which being an r -dimensional block. Under the assumption of (26), we can expressthe gauge-invariant measure as dµ ( e C ) i decom. = dµ ( H ) dµ ( a, ¯ a ) e c A S WZW ( e H ) (cid:21) [ a,H ]=0 = dµ ( C ) dµ ( a, ¯ a )] [ a,H ]=0 (27)where dµ ( a, ¯ a ) = Q N − j =1 dµ ( a j , ¯ a j ). In the last step, we neglect ( a, ¯ a )-contributions of S W ZW ( f H ),which may be absorbed into definitions of wave functions. Since dµ ( a, ¯ a ) is invariant under a j → a j + m j + n j τ , the expression (27) shows explicit gauge invariance of the measure whichis incorporated with the zero modes. In this section, we construct a wave function corresponding to the vacuum state of (2+1)-dimensional Yang-Mills theory on S × S × R . Abelian Case
We first focus on abelian zero-mode dynamics and then move to a nonabelian case later.As in the previous section, we shall denote a as a complex variable for the moment. For anabalian case, we can substitute M † = e iθ ( z, ¯ z ) , where θ ( z, ¯ z ) is a function of z , ¯ z , into (23) toobtain e A ¯ z = i∂ ¯ z θ + π ¯ ω Im τ a . Thus, under a certain gauge, physical variables of e A ¯ z can be givensolely by a which satisfies the periodicity a → a + m + nτ .9e now consider geometric quantization of the U (1) Chern-Simons theory, following theline of [22, 35] in a slightly different manner. From (9), (13) and (22)-(24), we can expressthe K¨ahler form of zero modes asΩ = k a ¯ a π da ∧ d ¯ a Z z, ¯ z (cid:18) π ¯ ω Im τ (cid:19) ∧ (cid:18) πω Im τ (cid:19) = i πk a ¯ a Im τ da ∧ d ¯ a (28)where the integral is taken over dzd ¯ z and k a ¯ a is the level number associated to the abelianChern-Simons theory. The corresponding zero-mode K¨ahler potentials can generally beexpressed as W ( a, ¯ a ) = πk a ¯ a Im τ a ¯ a + g ( a ) + ¯ g (¯ a ) (29)where g ( a ) and ¯ g (¯ a ) are purely a -dependent and ¯ a -dependent functions, respectively.From (13)-(17) we have a following relation da ∧ d ¯ a Z z, ¯ z ¯ ω ∧ ω = da ∧ da Z z, ¯ z ω Im τ ∧ ω Im τ (30)where we use ω = ω − τ ω Im τ , ¯ ω = ω − ¯ τ ω Im τ . (31)In terms of a and a , the zero-mode K¨ahler form can then be expressed asΩ = k a ¯ a π (cid:18) π Im τ (cid:19) da ∧ d ¯ a Z z, ¯ z ¯ ω ∧ ω = k a ¯ a π (cid:18) π Im τ (cid:19) (2 i Im τ ) da ∧ d ¯ a = k a ¯ a π (cid:18) π Im τ (cid:19) da ∧ da Z z, ¯ z ω Im τ ∧ ω Im τ = − k a ¯ a π (cid:18) π Im τ (cid:19) da ∧ da . (32)A K¨ahler potential corresponding to the second line in (32) may be given by K ( a, ¯ a ) = iπk a ¯ a τ ) (¯ a − a )( τ ¯ a − ¯ τ a ) . (33) K ( a, ¯ a ) appears to differ from the general expression W ( a, ¯ a ) in (29). But this does notcause a problem since both W ( a, ¯ a ) and K ( a, ¯ a ) are derived from the same K¨ahler form Ωwith different choices of frames, i.e. , the two K¨ahler potentials describe the same physics ofzero modes. We shall choose K ( a, ¯ a ) as our zero-mode K¨ahler potential in the following.The symplectic potential for the zero modes can be expressed as A = πk a ¯ a τ ) Z z, ¯ z (cid:18) ω a Im τ ∧ ω Im τ da − ω a Im τ ∧ ω Im τ da (cid:19) = − πk a ¯ a τ ) ( a da + a da ) . (34)Note that we take account of the couplings of a i to ω i ( i = 1 ,
2) in K ( a, ¯ a ). A naive calculationof d A with respect to a i does not lead to Ω of (32) but this is not a discrepancy since Ω is10lso defined with a i coupled to ω i . From (18), a variation of A under a → a + m + nτ isgiven by A → A + d Λ m,n whereΛ m,n = − i πk a ¯ a τ ( ma − na ) . (35)A holomorphic wavefunction which satisfies the polarization condition can be expressedas Ψ[ e A ¯ z ] ≡ Ψ[ a ] = e − K ( a, ¯ a )2 f ( a ) . (36)We require zero-mode ‘gauge’ invariance on Ψ[ a ] under a → a + m + nτ by imposing e i Λ m,n Ψ[ a ] = Ψ[ a + m + nτ ] . (37)This leads to the following relation f ( a ) = e − iπk a ¯ a mn f ( a + m + nτ ) . (38)Since the periodicity property f ( a ) = f ( a + m + nτ ) is a natural requirement for any functionsdefined on torus, the relation (38) means that Ψ[ a ] can be ‘gauge’ invariant given that f ( a )satisfies a Dirac-like quantization condition for k a ¯ a , i.e. , k a ¯ a ∈ Z . This is another indicationof level quantization for the Chern-Simons theory on torus.For choices of arbitrary winding numbers ( m, n ), one may make n be absorbed into k a ¯ a .This arrows us to identify 2 n with the level number of abelian Chern-Simons theory encodingthe zero mode dynamics. We shall later choose mn to be a large integer such that m ≫ n = 1as mentioned in Fig.1.An inner product of the holomorphic wavefunctions can be expressed as h | i = Z dµ ( a, ¯ a ) e − K ( a, ¯ a ) f ( a ) f ( a ) (39)where f ( a ) denotes the complex conjugate of the function f ( a ). Note that, as a requirementfor a holomorphic function, the factor of πk a ¯ a Im τ ¯ a is realized by an operation of ∂∂a on f ( a ). Weshall discuss this point later below equation (60). Nonabelian Case
Let us now turn to the main part of the present paper. From (8)-(10) and Narashimhan-Seshadri theorem, we can express a wave functional for vacuum states of (2 + 1)-dimensionalYang-Mills theory on torus asΨ[ e A ¯ z ] ≡ Ψ[ f M † ] = e − e K e ˜ k S WZW ( e M † ) Υ( a ) (40)where a now has an algebraic structure as in (20) and f K is a toric version of (9), i.e. , f K = − ˜ kπ Z Σ Tr( e A ¯ z e A z ) (41)11ith ˜ k being a toric version of the level number k defined in (9). Note that Υ( a ) in (40) issome functions of a which does not depend on f H = f M † f M as Ψ[ f M † ] being a wave functionalfor the vacuum states. The inner product can be given by h | i = Z dµ ( e C )Ψ ∗ [ f M † ]Ψ [ f M † ]= Z dµ ( f H ) e (2 c A +˜ k ) S WZW ( e H ) Υ ( a )Υ ( a ) (42)where Υ ( a ) is a complex conjugate of Υ ( a ) and Ψ ∗ [ f M † ] is defined byΨ ∗ [ f M † ] ≡ Ψ[ f M ] = e − e K e ˜ k S WZW ( e M ) Υ( a ) (43)We may naively follow the lines of discussion in the planar case to assume ˜ k → k can essentially be given by the zero-mode level number k a ¯ a in the previous subsection.To circumvent the problem, we can make a dimensional analysis. One of the interestingfeatures of Yang-Mills theory in three dimensions is that a coupling constant e has massdimension of . Gauge potentials also have mass dimension for the three-dimensionaltheory. Conventionally, this is realized by absorption of e into dimensionless A ¯ z in the formof (1). Similarly, we regard Im τ in e A ¯ z of (22) as dimensionless. Mass dimension of Im τ ,however, can be expressed as [(Im τ ) − ] = [ e ] = 1. We can then consider π Im τ as a unit ofthe level number ˜ k , where the factor of π for Im τ − arises from periodicity conditions (19)or from the matrix parametrization (22), (23). In terms of dimensionful parameters, we mayexplicitly write the level number ˜ k as π Im τe ˜ k . This vanishes as Im τ → ∞ , which correspondsto the planar case. For finite Im τ with nonzero k a ¯ a , however, the toric level number doesnot vanish and we have a critical value of τ at which the factor of 2 c A + ˜ k in (42) vanishes.In the KKN Hamiltonian approach, this factor, or the coefficient of S W ZW ( H ), involves incalculations of string tension and mass gap. We can therefore read off a deconfinementtemperature from the critical value of τ , which we shall discuss in the next section.To understand the relation between ˜ k and k a ¯ a , we now consider decomposition of a , ¯ a out of f H with an imposition of (26). The K¨ahler potential (41) can be expressed as f K decom. = − ˜ kπ Z Σ Tr( A ¯ z A z ) − ˜ k Im τ Z Σ Tr( A ¯ z M ω ¯ aM − + M †− ¯ ωaM † A z )+ N − Y j π ˜ k τ a j ¯ a j (44)The last term can be written as Q j πk a ¯ a Im τ a j ¯ a j with an identification of˜ k = 2 k a ¯ a . (45)As we have discussed below (38), we can express k a ¯ a = 2 n for any integer n . So theidentification (45) means ˜ k = 4 n . Note that we may choose a sign of ˜ k since the order ofvariables does not matter in K¨ahler potentials of zero modes; this implies a change of sign12or (positive) k a ¯ a in the definition of zero-mode K¨ahler form in (28). As discussed earlier,by a change of frames we may replace πk a ¯ a Im τ a j ¯ a j with K ( a j , ¯ a j ) of (33). Upon decomposition,the wave functional (40) can then be written asΨ[ f M † ] decom. = e − Q j K ( a j , ¯ a j ) Υ( a ) e ˜ k π R Σ Tr( A ¯ z A z ) e ˜ k S WZW ( M † ) × exp " ˜ k τ Z Σ Tr( A ¯ z M ω ¯ aM − + M †− ¯ ωaM † A z ) × exp " − ˜ k Im τ Z Σ Tr(¯ ωa∂ z M † M †− ) (46)where we use the Polyakov-Wiegmann identity (11) for S W ZW ( g M † ) = S W ZW ( e γ ¯ z M † ) and ∂ ¯ z e γ ¯ z = π ¯ ω Im τ a e γ ¯ z , ∂ z e γ ¯ z = 0 for e γ ¯ z defined in (24). In analogy with the abelian case, we cannow regard Υ( a ) as a general function satisfying invariance under a j → a j + m j + τ n j for a = a j t diag j and m j , n j ∈ Z . We can have a similar expression for Ψ[ f M ] decom. and its productwith Ψ[ f M † ] decom. can be written asΨ ∗ Ψ ] decom. = e − Q j K ( a j , ¯ a j ) Υ ( a )Υ ( a ) e ˜ k S WZW ( H ) × exp " ˜ k Im τ Z Σ Tr( H − ∂ ¯ z Hω ¯ a − ¯ ωa∂ z HH − ) . (47)Earlier we have obtained a gauge invariant wave function for an abelian case but fornonabelian cases it is not possible to do so. This is because a corresponding vacuum wavefunctional has properties arising from a WZW action. Note that neither of Ψ[ f M † ] or Ψ[ f M ]is gauge invariant in terms of the zero mode variables. This is related to the fact that thereis no gauge invariant WZW action for S W ZW ( g M † ) = S W ZW ( e γ ¯ z M † ). Technically speaking, S W ZW ( g M † ) does not satisfy the so-called anomaly-free condition [26], which is a sort ofchirality condition in terms of transformations from M † to g M † . On the other hand, S W ZW ( ˜ H )satisfies the anomaly condition. So we may obtain a gauge invariant value for the productin (47). Indeed, we can introduce the following gauged WZW action [24, 25, 26]: I ( H, a ) = S W ZW ( H ) + 1 π Z Σ Tr (cid:18) H − ∂ ¯ z H πω Im τ ¯ a − π ¯ ω Im τ a∂ z HH − + H − π ¯ ω Im τ aH πω Im τ ¯ a − π ¯ ω Im τ a πω Im τ ¯ a (cid:19) . (48)The inner product is then given by h | i decom. = Z dµ ( H ) dµ ( a, ¯ a ) e (2 c A +˜ k ) I ( H,a ) Ψ ( a )Ψ ( a ) (cid:21) [ a,H ]=0 (49)where Ψ( a ) is now defined asΨ( a ) = exp − N − Y j K ( a j , ¯ a j )2 Υ( a ) . (50)13( a ) is a complex conjugate of Ψ( a ). Υ( a ) is a general function on torus with a Cartansubalgebra structure for a . Equation (49) shows manifest gauge invariance of the innerproduct in terms of H and ( a, ¯ a ), including the measure.What we have done here is to obtain a lower class of the vacuum wave functional (40)and a corresponding inner product by imposition of the decomposition assumption (26).The decomposition condition leads to expressions in terms of H and ( a, ¯ a ) by ‘gauging away’( a, ¯ a )-dependence in f H = e γ ¯ z H e γ z so that we can extract the gauge invariant matrix H . In(47) we find a coupling between the current of S W ZW ( H ), ∂ z HH − , and the zero modevariable ¯ ωa . This coupling implies an identification of the current as a non-perturbativegluon field as discussed in the Hamiltonian approach. This is an interesting point, however,the upshot of the decomposition analysis here lies in gauge invariance of the inner productwhich leads to the relation of level number ˜ k with k a ¯ a as in (45) and properties of Υ( a ) asa general non-abelian function on torus. Since a = a j t diag j is diagonal, the decompositioncondition (26) does not affect on the properties of Υ( a ); hence, that in (40) remains thesame as a Cartan subalgebraic function on torus throughout the present subsection.The expression (49) suggests that the effects of zero modes on torus can be interpretedas changes in level numbers of the WZW action. In this sense, (2 + 1)-dimensional Yang-Mills theory on S × S × R may be regarded as Yang-Mills-Chern-Simons theory. This is,however, not in contradiction to our consideration of pure Yang-Mills theory because of thefollowing. As mentioned in the introduction, physical states of (2 + 1)-dimensional Yang-Mills theory can be described by holomorphic wave functionals of Chern-Simons theory. Theapparent changes of level numbers in the WZW action arise from the zero-mode contributionsto the holomorphic wave functionals of Chern-Simons theory in the toric case. Thus weare considering pure Yang-Mills theory on S × S × R , with the zero-mode contributionsrepresented by ˜ k in the exponent of (49) along with an additional abelian measure dµ ( a, ¯ a ). Vacuum states and theta functions
So far we have chosen a K¨ahler potential of either abelian or nonabelian theory such thata holomorphic function, f ( a ) or Υ( a ), has a obvious periodic relation characterized by (38).We can however use different K¨ahler potentials to start with, as long as the potentials leadto the same K¨ahler form. For example, it is known a certain choice of K¨ahler potential givesrise to theta functions for holomorphic functions for an abelian case [22, 35]. We shall brieflyreview this fact in the rest of this section for better understanding of the above mentionedresults in the framework of geometric quantization.A K¨ahler potential we choose is W τ ( a, ¯ a ) = iπk a ¯ a τ ) (¯ a − a ) τ (51)which, along with an implicit assumption of Re τ = 0, can be expressed in the form of (29) andleads to the K¨ahler form (28). A polarization condition for a holomorphic function Ψ τ ( a )is given by D ¯ a Ψ τ ( a ) = ( ∂ ¯ a − i A ¯ a )Ψ τ ( a ) = 0 where A ¯ a = i ∂ ¯ a W τ ( a, ¯ a ). The holomorphic14unction is then given by Ψ τ ( a ) = exp " − iπk a ¯ a τ ) (¯ a − a ) τ f τ ( a ) . (52)This is a general expression for a holomorphic functions upon a choice of the K¨ahler potentialas we have seen earlier. Properties of a holomorphic function f τ ( a ) can similarly be obtainedby imposing gauge invariance on Ψ τ ( a ). With a choice of the corresponding symplecticone-form of the form [22, 23]: A τ = − πk a ¯ a τ ) (¯ a − a )( τ d ¯ a − ¯ τ da ) , (53)we have δ A τ = d Λ τ m,n for transformations of a → a + m + nτ withΛ τ m,n = iπk a ¯ a Im τ n ( τ ¯ a − ¯ τ a ) . (54)Gauge invariance on Ψ τ ( a ) is then given by e i Λ τ m,n Ψ τ ( a ) = Ψ τ ( a + m + nτ ). This leads tothe following relation f τ ( a ) = e i πk a ¯ a (cid:16) n τ + an (cid:17) f τ ( a + m + nτ ) . (55)This shows that f τ ( a ) is a Jacobi θ -function defined by θ ( a, τ ) = Θ " ( a, τ ) (56)where Θ " ab ( z, τ ) = X n ∈ Z e iπk a ¯ a τ ( n + a ) +2 iπk a ¯ a ( n + a )( z + b ) . (57)An operation of ∂∂a on f τ ( a ) corresponds πk a ¯ a Im τ (¯ a − a ). acting on f τ ( a ). This is in consistentwith the K¨ahler form written by Ω = − i πk a ¯ a Im τ d (¯ a − a ) ∧ da . In terms of f τ ( a ), the inner productfor the holomorphic functions is expressed by h | i = Z dµ ( a, ¯ a ) e − W τ ( a, ¯ a ) f τ ( a ) f τ ( a ) . (58)Expanding the K¨ahler potential, we can rewrite this as h | i = Z dµ ( a, ¯ a ) e − πka ¯ a Im τ a ¯ a g τ ( a ) g τ ( a ) (59)where we introduce g τ ( a ) = exp " πk a ¯ a a τ f τ ( a ) . (60)We now find the operation of ∂∂a on g τ ( a ) is realized by πk a ¯ a Im τ ¯ a . We further find that g τ ( a ) = g τ ( a + m + nτ ) ( m = 0 , Re τ = 0) (61)15egardless a choice of k a ¯ a . This periodic relation is a subsector of the relation for more generalholomorphic function f ( a ) in (38) since relation (61) is realized when m = 0 and Re τ = 0are satisfied. The choice of f ( a ) = g τ ( a ) is then a concrete realization of the relation ∂∂a f ( a ) = πk a ¯ a Im τ ¯ a f ( a ) . (62)What is essential in construction of wave functions of zero modes and a corresponding innerproduct is the K¨ahler form to start with. Different K¨ahler potentials may lead to differentexpressions, e.g. , (58) and (59), yet physical consequences should be unaltered. With such aprinciple, we may require the relation (62) for the previously discussed holomorphic function f ( a ) in (36).Extension to a nonabelian case is straightforward with a knowledge of SU ( N ) algebra.The theta function is to be replaced by a higher dimensional theta function, more precisely,the Weyl-Kac character for SU ( N ) algebra with level number k a ¯ a [36]: ch ˆ λ ( a, τ ) = Tr ˆ λ e πik a ¯ a τh − πik a ¯ a ( a h + a h + ··· + a N − h N − ) (63)where h = h + h + · · · + h N − and the trace meansTr ˆ λ = X h ∈ Z + λka ¯ a . (64)Note that for vacuum wave functions, we should take the ground state for ˆ λ , i.e. , λ = 0. Anonabelian version of a vacuum wave functional Ψ [ e A ¯ z ] can be constructed by use of (63).Such a wave functional has been studied before and is explicitly given in [23]. In this section, we return to a physical part of the present paper. We consider contributionsof the zero modes to the planar case by taking the winding numbers in the limit of ( m, n ) =( ∞ , k , the effect of zero modes can be evaluated with a replacement of 2 c A with (2 c A + ˜ k ) as seen in equation (42) (see also [30], for rigorous discussion). From earlierdiscussion below equation (43), we can express ˜ k as − πk a ¯ a Im τe with k a ¯ a = 2 n . The criticaltemperature is therefore given by (cid:16) τ (cid:17) c = e N π where we use c A = N and n = 1. Thedeconfinement temperature is then expressed as T c = e N π (65)which is the same as a mass for non-perturbative gluons predicted in the KKN Hamiltonianapproach. Thus T c in (65) is a natural result and it is what we seek for in the present study.In what follows, we shall briefly review the calculation of string tension in the Hamiltonianapproach for completion of our discussion. 16he vacuum expectation value of the Wilson loop operator h W ( C ) i = h Ψ | W ( C ) | Ψ i can be calculated as h W ( C ) i = Z dµ ( f H ) e (2 c A +˜ k ) S WZW ( e H ) e − S ( e H ) Υ( a )Υ( a ) W ( C ) (66)where the Wilson loop operator is given by W ( C ) = TrP exp (cid:16) − H e A (cid:17) = TrP exp (cid:16) πc A H e J (cid:17) with e J = c A π ∂ z f H f H − . The function S ( f H ) denotes a contribution from the potential energyof the Yang-Mills theory. For modes of low momenta, or for a (continuum) strong couplinglimit, this function can be evaluated. Using an analog of two-dimensional Yang-Mills theoryand setting Υ( a ) = 1, we can evaluate the vacuum expectation as h W ( C ) i ≈ exp [ − ˜ σ A C ] (67)where A C is the area of the loop C and ˜ σ is the string tension on torus given by˜ σ = e π (cid:18) c A + 12 ˜ k (cid:19) c F . (68)Here c F = ( N − N + 1) / N is the quadratic Casimir for SU ( N ) in the fundamentalrepresentation. Substituting ˜ k = − πk a ¯ a Im τe ( k a ¯ a = 2 n , n = 1), we find vanishing of the stringtension at T c .Temperatures corresponding to n > n = 1 as long as m ≫ n ≥ τ is scaledto n Im τ such that ˜ k remains the same for any n . Notice that for the case of n = 0, whichmay be possible as we consider n as the winding number of the beta cycle of torus, we havevanishing of n Im τ but yet ˜ k remains as ˜ k = − π Im τe . In terms of the picture in Fig.1, thechoice of n = 0 means that the torus of our interest is dimensionally reduced to a circle(times a point). Thus, in our setting the choice of n = 0 may be ruled out for a dimensionalreasoning.In the planer case, the string tension σ is given by σ = e N − π ! . (69)Comparisons with numerical data can be made for dimensionless parameter T c / √ σ . Ourprediction for this value is T c √ σ = s π s N N −
1= 0 . s N N − . (70)Lattice simulations show 0.865, 0.903 and 0.86(7) for this value at N → ∞ . These values aretaken from references [14], [15] and [17], respectively. Corresponding error percentages are8.40%, 13.2% and 8.65%. Among the lattice data, the one given by [15] actually providesthe most updated and reliable result. Obviously, we need to make further investigation tofigure out the relatively large deviation between the lattice data and the value (70).17 Concluding remarks
In the present paper, we consider (2 + 1)-dimensional Yang-Mills theory on S × S × R inthe framework of the so-called Karabali-Kim-Nair (KKN) Hamiltonian approach. A physicalmotivation to consider the toric theory is clear since we may regard it as the planar theoryat a finite temperature in the limit of a large radius for one of the S ’s of torus ( S × S ),and, hence, we can discuss deconfinement transition in terms of the other radius. In orderto execute a calculation of a deconfinement temperature, however, we need to understandsome mathematical aspects of the KKN Hamiltonian approach. In section 2, we reviewfew features of the Hamiltonian approach which are pertinent to our discussion. Detailedanalysis on dynamics or geometry of zero modes of torus is given in section 3 for both abelianand nonabelian cases. For a nonabelian case, we construct vacuum-state wave functionalsfor (2 + 1)-dimensional Yang-Mills theory on S × S × R by use of Narashimhan-Seshadritheorem. We further consider a subsector of the vacuum wave functionals by imposing acertain condition (26) to discuss gauge invariance of an inner product of the vacuum wavefunctionals. Along the way, we also find zero-mode contributions to the planar Yang-Millstheory. In section 4, we compute a string tension of pure Yang-Mills theory on S × S × R inthe Hamiltonian framework, namely, in the so-called continuous strong coupling limit, andfind a deconfinement temperature (65). This value agrees with numerical data from latticesimulation quite roughly in 10%. We shall leave the explanation of this rather large errorfor future studies.Now we would like to comment on subtle points in the arguments which we have used inthe present paper. We argue that we take a cylindrical limit of the torus, i.e. , S × S × R → S × R , in the end of calculations. In our framework, one of the S directions of thetorus corresponds to the time coordinate. The Lorentz invariance however implies thatwe can interchange this temporal direction with the spatial direction R . Thus our resultssuggest that a change of topologies (from plane to cylinder) leads to an apparent changeof level numbers in a WZW model which is relevant to the gauge invariant measure in(2 + 1)-dimensional Yang-Mills theory. This is a nontrivial result and probably containssome subtleties because that a topology change causes a change of level numbers is simplycounter-intuitive. In this paper, we present one of the justifications of this issue by use ofgauge invariance. We have made the following argument.In the KKN Hamiltonian approach, the Gauss law constraint (or the integrability) of thegauge potentials should be satisfied regardless what Riemann surfaces we use in constructionof the WZW action which is relevant to the gauge invariant measure of (2 + 1)-dimensionalYang-Mills theory. Thus, the wave functionals in the toric theory can be written as (40).(This is why the use of Narashimhan-Seshadri theorem has been emphasized in the presentpaper.) The symplectic structure of the zero modes is essentially given by (32). This isnot exactly the symplectic structure of Chern-Simos theory. However, with an algebraicextension (20), we can relate the level number k a ¯ a of (32) to the level number ˜ k of (40) sothat we can encode the zero-mode contributions in the level number ˜ k . A detailed explanationof this relation is given by the argument of gauge invariance in section 4. The argument islimited to a particular case, where we can explicitly write down the gauge invariant measure18n terms of zero modes. Although this will be sufficient to show the relation for our purposes,it is desirable to confirm this relation in more general cases. We shall leave this task for futurestudies.Lastly, we would like to emphasize that the validity of our analyses and results is limitedin the framework of the KKN Hamiltonian approach. It is within this framework that we canproperly use the abelian Chern-Simons symplectic form to discuss zero-mode contributionsto (2 + 1)-dimensional Yang-Mills theory. We have not proven the use of the Chern-Simonssymplectic form in general. Furthermore there may be some subtleties to justify this analysisin the more standard approaches to Yang-Mills theory. Although this might be the case,what is significant in the present paper from a physical perspective is that the use of theChern-Simons symplectic form does seem to lead a reasonable estimate for the deconfine-ment temperature and that this fact itself suggests the usefulness of the KKN Hamiltonianapproach in the future investigations of (2 + 1)-dimensional Yang-Mills theory. Acknowledgments
The author would like to thank Professor V.P. Nair for introduction to the present work andhelpful discussions in the spring of 2006. Later correspondence with Professors D. Karabaliand V.P. Nair was of significant help in improving a manuscript. The author thanks theYukawa Institute for Theoretical Physics at Kyoto University. Discussions during the YITPworkshop YITP-W-07-05 on “String Theory and Quantum Field Theory” were useful forthe present work. The author also thanks Professor Holland for updated information onlattice results. Lastly the author is grateful to the referee of this paper for many critical andstimulating comments.
References [1] D. Karabali and V. P. Nair, Nucl. Phys. B , 135 (1996) [arXiv:hep-th/9510157];Phys. Lett. B , 141 (1996) [arXiv:hep-th/9602155]; Int. J. Mod. Phys. A , 1161(1997) [arXiv:hep-th/9610002].[2] D. Karabali, C. j. Kim and V. P. Nair, Nucl. Phys. B , 661 (1998) [arXiv:hep-th/9705087].[3] D. Karabali, C. j. Kim and V. P. Nair, Phys. Lett. B , 103 (1998) [arXiv:hep-th/9804132].[4] B. Bringoltz and M. Teper, PoS LAT2006 , 041 (2006) [arXiv:hep-lat/0610035]; Phys.Lett. B , 383 (2007) [arXiv:hep-th/0611286].[5] R. G. Leigh, D. Minic and A. Yelnikov, Phys. Rev. Lett. , 222001 (2006) [arXiv:hep-th/0512111]; arXiv:hep-th/0604060.[6] L. Brits, JHEP , 012 (2007) [arXiv:hep-th/0702156].197] A. Agarwal, D. Karabali and V. P. Nair, arXiv:0705.0394 [hep-th].[8] D. Karabali and V. P. Nair, arXiv:0705.2898 [hep-th].[9] M. Fukuma, K. Katayama and T. Suyama, arXiv:0711.4191v1 [hep-th].[10] P. Orland, Phys. Rev. D , 101702 (2007) [arXiv:0704.0940 [hep-th]]; Phys. Rev. D , 025001 (2007) [arXiv:hep-th/0608067]; Phys. Rev. D , 085001 (2006) [arXiv:hep-th/0607013].[11] K. Papadodimas, H. H. Shieh and M. Van Raamsdonk, JHEP , 069 (2007)[arXiv:hep-th/0612066].[12] C. Feuchter and H. Reinhardt, arXiv:0711.2452 [hep-th].[13] P. Orland and G. W. Semenoff, Nucl. Phys. B , 627 (2000) [arXiv:hep-th/9912009].[14] J. Liddle and M. Teper, PoS LAT2005 , 188 (2006) [arXiv:hep-lat/0509082].[15] J. Liddle and M. Teper, arXiv:0803.2128 [hep-lat].[16] R. Narayanan and H. Neuberger, Phys. Rev. Lett. , 081601 (2003) [arXiv:hep-lat/0303023].[17] R. Narayanan, H. Neuberger and F. Reynoso, arXiv:0704.2591 [hep-lat].[18] K. Holland, M. Pepe and U. J. Wiese, JHEP , 041 (2008) [arXiv:0712.1216 [hep-lat]].[19] K. Holland, JHEP , 023 (2006) [arXiv:hep-lat/0509041].[20] M. S. Narasimhan and C. S. Seshadri, “Stable and unitary bundles on a compact Rie-mann surface,” Ann. Math. (1965) 540.[21] K. Gaw¸edzki and A. Kupiainen, Phys. Lett. B , 119 (1988); Nucl. Phys. B , 625(1989).[22] M. Bos and V. P. Nair, Phys. Lett. B , 61 (1989).[23] M. Bos and V. P. Nair, Int. J. Mod. Phys. A , 959 (1990).[24] D. Karabali, Q-H. Park, H. J. Schnitzer and Z. Yang, Phys. Lett. B , 307 (1989).[25] D. Karabali and H. J. Schnitzer, Nucl. Phys. B , 412 (1989).[26] E. Witten, Commun. Math. Phys. , 189 (1992).[27] E. Witten, Commun. Math. Phys. , 351 (1989).[28] S. Axelrod, S. Della Pietra and E. Witten, J. Diff. Geom. , 787 (1991).[29] A. M. Polyakov and P. B. Wiegmann, Phys. Lett. B , 121 (1983).2030] D. Karabali, C. j. Kim and V. P. Nair, Nucl. Phys. B , 331 (2000) [arXiv:hep-th/9907078].[31] M. Asorey, F. Falceto, J. L. Lopez and G. Luzon, Phys. Lett. B , 125 (1995)[arXiv:hep-th/9502024].[32] J. M. Cornwall, Phys. Rev. D , 1814 (1996) [arXiv:hep-th/9602157].[33] D. Bernard, Nucl. Phys. B , 77 (1988); Nucl. Phys. B , 145 (1988).[34] F. Falceto and K. Gaw¸edzki, Commun. Math. Phys. , 549 (1994) [arXiv:hep-th/9211003].[35] V. P. Nair, Quantum Field Theory: A Modern Perspective , New York, USA: Springer(2004), see pp.515-522.[36] P. Di Francesco, P. Mathieu and D. Senechal,