Conformal current algebra in two dimensions

Conformal Current Algebra in Two Dimensions Sujay K. Ashoka,b , Raphael Benichouc and Jan Troostc a Institute arXiv:0903.4277v2 [hep-th] 10 Jun 2009 of Mathematical Sciences C.I.T Campus, Taramani Chennai, India 600113 b Perimeter Institute for Theoretical Physics Waterloo, Ontario, ON N2L2Y5, Canada c Laboratoire de Physique Théorique Unité Mixte du CRNS et de l’École Normale Supérieure associée à l’Université Pierre et Marie Curie 6 UMR 8549 1 École Normale Supérieure 24 Rue Lhomond Paris 75005, France Abstract We construct a non-chiral current algebra in two dimensions consistent with conformal invariance. We show that the conformal current algebra is realized in non-linear sigmamodels on supergroup manifolds with vanishing Killing form, with or without a WessZumino term. The current algebra is computed using two distinct methods. First we exploit special algebraic properties of supergroups to compute the exact two- and threepoint functions of the currents and from them we infer the current algebra. The algebra is also calculated by using conformal perturbation theory about the Wess-Zumino-Witten point and resumming the perturbation series. We also prove that these models realize a non-chiral Kac-Moody algebra and construct an infinite set of commuting operators that is closed under the action of the Kac-Moody generators. The supergroup models that we consider include models with applications to statistical mechanics, condensed matter and string theory. In particular, our results may help to systematically solve and clarify the quantum integrability of P SU (n|n) models and their cosets, which appear prominently in string worldsheet models on anti-deSitter spaces. 1 Preprint LPTENS-09/16. 1 Contents 1 Introduction 2 2 Current algebras in two dimensions 2.1 Locality, Lorentz invariance and PT-invariance 2.2 Current conservation . . . . . . . . . . . . . . . 2.3 The Maurer-Cartan equation . . . . . . . . . . 2.4 The Euclidean current algebra . . . . . . . . . 2.5 Conformal current algebra . . . . . . . . . . . . . . . . . 3 4 5 7 8 9 correlators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 11 13 19 4 Conformal perturbation theory 4.1 The current algebra in the Wess-Zumino-Witten model . . . . . . . . . . . . . 4.2 Perturbation of the kinetic term: classical analysis . . . . . . . . . . . . . . . 4.3 The current-current operator product expansions . . . . . . . . . . . . . . . . 19 20 20 21 5 The current algebra on the cylinder 23 6 Conclusions 25 A Perturbed operator product expansions 27 B Detailed operator product expansions 28 C Useful integrals 32 3 Current algebra from supergroup current 3.1 The model . . . . . . . . . . . . . . . . . . 3.2 Exact perturbation theory . . . . . . . . . 3.3 Summary of the current algebra . . . . . . 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Introduction Two-dimensional sigma-models on supergroups have applications to a wide range of topics such as the integer quantum hall effect, quenched disorder systems, polymers, string theory, as well as other domains in physics (see e.g. [1, 2, 3, 4, 5]). The principal chiral model is perturbatively conformal on various supergroups [6, 7] with or without the addition of a Wess-Zumino term, and at least to two loop order on their cosets with respect to a maximal regular subalgebra [7]. Sigma models on graded supercosets are also believed to be conformal [8]. Thus, these models have an infinite dimensional symmetry algebra that should tie in with their supergroup symmetry. An extended non-linear symmetry algebra was identified in [6] but the representation theory of the algebra seems difficult to establish. Steps towards solving these models were made using various techniques [9, 10, 11]. In this paper, we exhibit a conformal current algebra in these models. The algebra of currents is non-chiral and implies conformal symmetry, hence the name. The models under discussion enter as the key building blocks in worldsheet sigma-models on supersymmetric AdS backgrounds in string theory. The supergroup P SU (1, 1|2) principal chiral model corresponds to a supersymmetric AdS3 × S 3 background with Ramond-Ramond 2 flux [6, 12]. Since a theory of quantum gravity on asymptotically AdS3 space-times, supplemented with appropriate boundary conditions, exhibits an infinite dimensional conformal symmetry algebra [13], we should be able to construct those generators from the worldsheet theory. Indeed, for AdS3 string theory with only Neveu-Schwarz-Neveu-Schwarz flux, it has been shown how to construct the space-time Virasoro algebra in terms of the worldsheet current algebra [14, 15, 16]. To perform a similar construction in Ramond-Ramond backgrounds, one needs to understand the worldsheet current algebra for two-dimensional models with supergroup targets. A second application within this context is the extension of our analyis to supercoset manifolds, which includes the AdS5 ×S 5 background of string theory. The worldsheet current algebra is tied in with the integrability of the worldsheet theory [17]. Our work may help in systematically exploiting the integrability of the worldsheet model at the quantum level, with applications to the solution of the spectrum of planar four-dimensional gauge theories (see e.g. [18]) via the AdS/CF T correspondence. The plan of the paper is as follows. In section 2 we kick off with a general analysis of twodimensional current algebra operator product expansions that are consistent with locality, Lorentz invariance and parity-time reversal. We exhibit a particular current algebra that is non-chiral and consistent with conformal invariance. We then move on to exhibit a realization of the algebra in a conformally invariant model on a supergroup manifold. We calculate perturbatively the current two- and three-point functions of these models in section 3. From these correlators we infer the operator product expansions of the currents, which are shown to fall into the conformal current algebra class discussed in section 2. In section 4 we analyze deformed Wess-Zumino-Witten models on supergroups using conformal perturbation theory. We compute the operator product expansions of the currents to all orders and resum the series, thereby demonstrating that the resulting algebra matches the results obtained using purely algebraic properties of the supergroups. We analyze the current algebra on the cylinder in terms of a Fourier decomposition in section 5 and show the existence of a Kac-Moody subalgebra and an infinite set of commuting operators that transform amongst each other under the Kac-Moody subalgebra. In section 6 we discuss some applications of the conformal current algebra that we have found and possible future directions of work. 2 Current algebras in two dimensions Before we compute the current algebra operator product expansion [19] for supergroup sigmamodels, it is interesting to analyze the generic operator product expansions (OPEs) involving the currents for a two-dimensional model which is local, Lorentz invariant and which respects parity-time reversal. Previously, the generic two-dimensional current algebra was analyzed in [20] where it was applied to an asymptotically free sigma-model – the O(n) sigma-model. Parity symmetry was assumed to be valid in the analysis. Later, a similar analysis was performed [21, 22] for massive models with Wess-Zumino-Witten ultraviolet fixed points. In both cases, the study was applied to argue for the integrability of the two-dimensional sigma-model in the quantum theory. The models that we will study have the distinctive feature that they are conformal. Moreover, the Wess-Zumino term breaks parity invariance. Therefore, we start by analyzing the generic current operator product expansions consistent with locality, Lorentz invariance 3 and PT-invariance only. In the following, we generalize the methodology of [20]. 2.1 Locality, Lorentz invariance and PT-invariance In the absence of parity invariance, the vector representation of the two-dimensional Lorentz group splits into two one-dimensional irreducible representations. A current jµ can therefore be split into two irreducible representations j+ and j− of the two-dimensional Lorentz group2 . We write the OPEs in terms of these irreducible components, leading to three independent OPEs, between the pairs of current components (j+ , j+ ), (j+ , j− ), and (j− , j− ). We first analyze the OPE between the components j+ and j+ . We take the currents to be a t where t spans in the adjoint representation of a symmetry group G of the model: j+ = j+ a a the Lie algebra of G. We write down the generic OPE in terms of Lorentz invariant coefficient functions. Moreover, we assume that the only low-dimensional operators that appear in the operator product expansion are the identity operator, the currents and their derivatives. We also assume that the currents have conformal dimension one. Thus the previous list of allowed operators in the OPE should account for all the terms up to regular ones. The j+ j+ OPE is then given by:  a b c c j+ (x)j+ (0) ∼ αab (x− )2 d1 (x2 ) + β ab c (x− )2 d2 (x2 )x+ j+ (0) + d3 (x2 )x− j− (0) c c +e1 (x2 )x+ x− ∂+ j− (0) + e2 (x2 )x+ x− ∂− j+ (0)  c c +e3 (x2 )(x+ )2 ∂+ j+ (0) + e4 (x2 )(x− )2 ∂− j− (0) + . . . (2.1) where the functions di , ei are functions of the Lorentz invariant x2 = x+ x− . The tensor αab is an invariant two-tensor in the product of adjoint representations, and β ab c represents an adjoint representation in the product of two adjoints. We will assume that αab corresponds to a non-degenerate bi-invariant metric κab and that the tensor β ab c is equal to the structure constants f ab c of the Lie algebra of G. We take the structure constants to be anti-symmetric in its indices3 . We use the fact that we can interchange operators4 to determine that the above OPE should be equal to: b a c c j+ (0)j+ (x) ∼ κba (x− )2 d1 − f ba c (x− )2 (d2 x+ j+ (x) + d3 x− j− (x) c c c c − e1 x+ x− ∂+ j− (x) − e2 x+ x− ∂− j+ (x) − e3 (x+ )2 ∂+ j+ (x) − e4 (x− )2 ∂− j− (x)) c c ∼ κba x− x− d1 − f ba c x− x− (d2 x+ j+ (0) + d3 x− j− (0) c c + (d3 − e1 )x+ x− ∂+ j− (x) + (d2 − e2 )x+ x− ∂− j+ (x) c c + (d2 − e3 )x+ x+ ∂+ j+ (x) + (d3 − e4 )x− x− ∂− j− (x)) (2.2) from which we derive the equations: e1 = d3 /2 e2 = d2 /2 e3 = d2 /2 e4 = d3 /2, (2.3) We work in Lorentzian signature in this section. We define x± = x ± t and ds2 = −dt2 + dx2 = dx+ dx− . We denote x2 = x+ x− . 3 More precisely, we take the structure constants to be graded anti-symmetric when G is a supergroup. Although the grading is crucial, it will not affect our formulas, except for a plethora of minus signs when interchanging operators – these will not influence our final results much. We maintain consistency with the grading throughout section 2, but not necessarily through the rest of the paper. 4 Up to a minus sign for fermionic operators. We see, for example, that the interchange of fermionic operators will cancel a minus sign from the grading of the superalgebra when G is a supergroup. 2 4 which gives rise to the simplified operator product expansion:  d3 c a b ab − 2 ab − 2 c c j+ (x)j+ (0) ∼ κ (x ) d1 + f c (x ) d2 x+ j+ (0) (0) + d3 x− j− (0) + x+ x− ∂+ j− 2  d2 d3 d2 c c c + x+ x− ∂− j+ (0) + (x+ )2 ∂+ j+ (0) + (x− )2 ∂− j− (0) + . . . 2 2 2 (2.4) The OPE has one free coefficient d1 (function of x2 ) at leading order, and two at subleading order. We similarly obtain:  d5 c a b ab + 2 ab + 2 c c (0) j− (x)j− (0) ∼ κ (x ) d4 + f c (x ) d5 x− j− (0) + d6 x+ j+ (0) + x+ x− ∂+ j− 2  d5 − 2 d6 + − d6 + 2 c c c + (x ) ∂− j− (0) + x x ∂− j+ (0) + (x ) ∂+ j+ (0) + . . . (2.5) 2 2 2 For the OPE between j+ and j− we don’t get as many constraints. We find 7 more free functions:  a b c c c c j+ (x)j− (0) ∼ κab d7 + f ab c d8 x+ j+ (0) + d9 x− j− (0) + d13 x+ x− ∂+ j− (0) + d12 x+ x− ∂− j+ (0)  c c +d10 (x+ )2 ∂+ j+ (0) + d11 (x− )2 ∂− j− (0) + . . . (2.6) We have a total of 13 free coefficient functions. 2.2 Current conservation We now impose consistency of the operator product expansions of the currents with current conservation. We choose the relative normalization of the two components of the currents such that the equation of current conservation reads 5 : a a + ∂+ j− = 0. ∂ µ jµa = ∂− j+ (2.7) Current conservation implies that one of the coefficients in the j+ j− OPE (namely d12 + d13 ) becomes redundant. We check that the OPEs of the current conservation equation (2.7) with 5 We expect current conservation to only hold up to delta-function contact terms. We therefore do not keep track of contact terms in the following. 5 the currents j+ and j− vanish. That leads to the set of equations 6 : 1 x2 1 d8 x2 ′ d1 + d′7 = 0 , d2 + d′2 + d′8 + 2 = 0 , 2 2 2 2 2x 1 ′ 3 x2 ′ 1 ′ x2 ′ d6 + d6 + d8 = 0 , d4 + d4 + d7 = 0 , 2 2 2 2 2 3 x2 ′ 1 ′ x2 ′ d3 + d3 + d9 = 0 , 2d3 + d3 + (d′9 − d′11 ) = 0 2 2 2 2 2 x 2 x2 ′ d5 + d′5 + d′11 + 2 d11 = 0 , 2d6 + d6 + d′10 = 0 , 2 2 x 1 ′ 1 x2 ′ d5 + d5 + d9 + 2 d9 = 0 , 2 2 2x 2 x2 ′ d2 + d2 + (d′8 − d′10 ) + 2 (d8 − d10 ) = 0 2 x x2 ′ 1 3 (d6 − d5 ) + (d6 − d′5 ) + (d′12 − d′13 ) + 2 (d12 − d13 ) = 0 , 2 2 x 3 x2 ′ 1 (d2 − d3 ) + (d2 − d′3 ) + (d′8 − d′12 − (d′9 − d′13 )) + 2 (d8 − d12 − (d9 − d13 )) = 0 .(2.8) 2 2 x d1 + We get a system of twelve first-order differential equations for twelve functions. We will not try to solve them in full generality. Instead, we assume that the leading and subleading singularities in the OPEs are powerlike. This leads to the following ansatz: d1 (x2 ) = c1 /x4 , d2 (x2 ) = c2 /x4 , d3 (x2 ) = f1 /x4 , d4 (x2 ) = c3 /x4 , d5 (x2 ) = c4 /x4 , d6 (x2 ) = f7 /x4 , d7 (x2 ) = f2 /x2 , d8 (x2 ) = f3 /x2 , d9 (x2 ) = f4 /x2 . (2.9) The coefficient ci ’s and fi ’s are now constant coefficients. Plugging this ansatz into the equations (2.8), we get the following relations between the coefficients: f2 = 0, f4 = f1 , f3 = f7 = c4 − c2 + f1 . (2.10) The remaining equations in (2.8) then allow to solve for the subsubleading coefficient functions in the OPEs: d10 (x2 ) = f7 , x2 (d′12 − d′13 ) + d11 (x2 ) = 0, 1 c2 − f1 (d12 − d13 ) = . 2 x 2x4 (2.11) So it only remains to solve for d12 (x2 )−d13 (x2 ) in terms of the ci and f1 . Denoting c2 −f1 = g, the general solution to the differential equation (2.11) reads: d12 (x2 ) − d13 (x2 ) = 2c5 g + 2 log µ2 x2 2 x 2x (2.12) where c5 is a constant coefficient and µ is an arbitrary mass scale. We note that when the coefficient g is non-zero, we can absorb the coefficient c5 in a redefinition of the mass scale 6 There is no equation corresponding to the operator ∂µ j µ since the coefficient multiplies zero. 6 µ. So the OPEs are given in terms of five dimensionless coefficients. They read: 
κab c1 c2 c (c2 − g)x− c g x− a b ab c c j+ (x)j+ (0) ∼ + 2 + f c + j+ (0) + j (0) − ∂ j (0) − ∂ j (0) + − − − + (x ) x (x+ )2 4 x+  − c2 − g (x )2 c2 c c ∂− j− (0) + . . . + ∂+ j+ (0) + 2 2 (x+ )2 
κab c3 c4 c (c4 − g) x+ c g x+ a b ab c c j− (x)j− (0) ∼ − 2 + f c − j− (0) + j (0) + ∂+ j− (0) − ∂− j+ (0) + − 2 − (x ) x (x ) 4x  + c4 c4 − g (x )2 c c + ∂− j− (0) + ∂+ j+ (0) + . . . 2 2 (x− )2  (c2 − g) c (c4 − g)x+ c4 − g c c a b j+ (0) + j− (0) + ∂+ j+ (0) j+ (x)j− (0) ∼ f ab c − + x x x−  
i g c c (0) − ∂− j+ (0) + . . . − c5 + log µ2 x2 ∂+ j− 4 (2.13) It would be interesting to search for more general solutions to the set of differential equations. 2.3 The Maurer-Cartan equation In this subsection we show that under certain circumstances, we can obtain further constraints on the current algebra. Consider a field g taking values in a Lie group. The one-form dgg −1 satisfies the Maurer-Cartan equation d(dgg−1 ) = dgg−1 ∧ dgg−1 . (2.14) We will get further constraints if we suppose that the components of the current are related to the field g in the following way: j+ = c+ ∂+ gg−1 , j− = c− ∂− gg−1 (2.15) where c+ and c− are constant coefficients. The generators of the Lie (super-)algebra satisfy: [ta , tb ] = if c ab tc . Then the Maurer-Cartan equation takes the form c b a a j− = 0. − if a bc j+ − c+ ∂+ j− c− ∂− j+ (2.16) We want to ensure that this equation is also valid in the quantum theory. However the operator defined as the product of two currents needs to be regularized. For this reason the validity of the Maurer-Cartan equation in the quantum theory requires more discussion. Normal Ordering In the quantum theory, we introduce a normal ordering for composite operators based on a point-splitting procedure. The normal ordered product : O1 O2 : (y) of two operators O1 and O2 evaluated at the point y is defined to be the product of the operator O1 at the point x with the operator O2 at the point y, in the limit as x approaches y. The regularization amounts to dropping the terms that are singular in this limit. For this procedure to be well-defined, it is important that the resulting operator is evaluated at the point y. We will denote this procedure by : O1 O2 : (y) = lim O1 (x)O2 (y) . (2.17) :x→y: 7 We note that the operators within the normal ordered product : O1 O2 : do not commute7 . We will later confirm that a natural choice for the normal ordered Maurer-Cartan equation in this scheme is8 : i c b b c a a j− : +(−)bc : j− j+ :) = 0. c− ∂− j+ − c+ ∂+ j− − f a bc (: j+ 2 (2.18) Additional constraints from the Maurer-Cartan equation As for the current conservation equation, we ask for the OPE between the quantum MaurerCartan equation (2.18) and the current to vanish. The first non-trivial constraint is obtained for the subleading terms. This leads to a relation between the coefficient of the current algebra c1 , c2 and g, and the coefficients c+ and c− : (c+ + c− )(c2 − g) + ic1 = 0 . (2.19) We similarly find a constraint linking c3 , c4 and g to c+ and c− : (c+ + c− )(c4 − g) + ic3 = 0 . (2.20) When we consider concrete models realizing the current algebra, the coefficients c+ and c− can be derived from the Lagrangian. The Maurer-Cartan equation then reduces the number of free constant coefficients from five to three. 2.4 The Euclidean current algebra For future purposes, we wish to translate the result we obtained for the current operator algebra into euclidean signature. We perform the Wick rotation t → −iτ , and define the complex coordinates z = x − iτ , z = x + iτ . The current algebra OPEs become  c2 c z a b ab c1 ab jz (z)jz (0) ∼ κ 2 + f c jz (0) + (c2 − g) 2 jzc (0) z z z  c2 gz c2 − g z 2 c j − (∂z jzc (0) − ∂z jzc (0)) + ∂z jzc (0) + (0) + ... ∂ z z 4z 2 2 z2  c4 c 1 (c4 − g)z c ab a b ab jz (0) jz (z)jz (0) ∼ κ c3 2 + f c jz (0) + z z z2  gz (c4 − g) z 2 c4 c c c c + (∂z jz (0) − ∂z jz (0)) + ∂z jz (0) + ∂z jz (0) + . . . 4z 2 2 z2  (c4 − g) c (c2 − g) c (c4 − g)z jza (z)jzb (0) ∼ f ab c jz (0) + jz (0) + ∂z jzc (0) z z z i g −(c5 + log µ2 |z|2 )(∂z jzc (0) − ∂z jzc (0)) + . . . (2.21) 4 Later on, we will compare the current algebra operator product expansions in equations (2.21) to those of a supergroup non-linear sigma-model with Wess-Zumino term. We will find specific expressions for the coefficients ci and g in terms of the parameters in the Lagrangian. 7 8 See e.g. [23] for a discussion of this fact in the context of chiral conformal field theories. We use the notation (−)a = +1 if a is a bosonic index, and −1 if a is a fermionic index. 8 2.5 Conformal current algebra It turns out that the above current algebra can become the building unit for a conformal algebra when the Killing form of the (super-)group vanishes9 . In that special case the current algebra is promoted to a conformal current algebra, namely, the Sugawara stress-energy tensor built from the currents satisfies the canonical conformal operator product expansion. The terms that in other circumstances spoil conformality are eliminated through the fact that the Killing form is zero. The holomorphic component of the stress-energy tensor is T (w) = 1 : jbz jzb : (w), 2c1 (2.22) as we will demonstrate10 . The current as a conformal primary First we compute the OPE between the current jza and its bilinear combination : jbz jzb :, using a point-splitting procedure: h i jza (z) : jbz jzb : (w) = lim jza (z) jbz (x)jzb (w) (2.23) :x→w:    c1 δba c2 c (c2 − g)(z − x) c a = lim + f j (x) + j (x) jzb (w) bc z :x→w: (z − x)2 z−x z (z − x)2    c2 c (c2 − g)(z − w) c c1 κab ab d ab +jz (x)(−1) κdb +f c j (w) + jz (w) . (z − w)2 z−w z (z − w)2 At this point we have to perform the OPE between the operators evaluated at the point x and the operators evaluated at the point w. Then we take the limit where x goes to w and discard the singular terms. We notice already that only the regular terms in the OPEs of the second line will contribute to the final result. We get   c2 f a bc c1 κcb jza (w) a b + jz (z) : jbz jz : (w) = lim c1 :x→w: (z − x)2 z − x (x − w)2    d x−w d c2 jz (w) c b cb + (c2 − g) j (w) + : jz jz : (w) +f d x−w (x − w)2 z   (c4 − g)jzd (w) (c2 − g)jzd (x) (c2 − g)f a bc (z − x) cb f + + d (z − x)2 x−w x−w    i g − c5 + log µ2 |x − w|2 (∂z jzd (w) − ∂z jzd (w)) + : jzc jzb : (w) 4 jza (w) c2 + c1 : j b j c : (w) + (−1)bc f a bc (z − w)2 z−w z z (c2 − g)(z − w) b c : jz jz : (w) . + (−1)bc f a bc (z − w)2 (2.24) 9 For simple super Lie algebras the vanishing of the Killing form is equivalent to the vanishing of the dual Coxeter number. 10 We defined ja = j b κba and κac κcb = δ a b . In the following we also use the convention fabc = fab d κdc . 9 Among the remaining terms, many cancel: every contraction of the invariant metric with a structure constant gives zero by symmetry, and the double contractions of two structure constants are proportional to the Killing form and thus also vanish. We are left with  f a bc  jza (w) bc b c c b (−1) : j j : (w)+ : j j : (w) + c 2 z z z z (z − w)2 z−w   z−w c b bc b c : (w)+ : j j : (w) (2.25) . (−1) : j j +(c2 − g)f a bc z z z z (z − w)2 jza (z) : jbz jzb : (w) = 2c1 The second term vanishes because of the anti-(super)symmetry of the structure constants. We can simplify the third term using the Maurer-Cartan identity: jza (z) : jbz jzb : (w) = 2c1
z−w jza (w) + 2i(c2 − g) c− ∂jza (w) − c+ ∂jza (w) . 2 2 (z − w) (z − w) (2.26) By current conservation this can be rewritten as: jza (z) : jbz jzb : (w) = 2c1 z−w jza (w) + 2i(c2 − g) (c− + c+ )∂jza (w). 2 (z − w) (z − w)2 (2.27) We can now show that the current jza is a primary field of conformal weight one. We deduce from the previous computation the OPE between the stress-energy tensor and the current jza , by expanding the operators on the right-hand side in the neigbourhood of the point z: 2c1 T (w)jza (z) = 2c1 z−w jza (z) ∂jza (z) ∂j a (z) (2.28) +2c +2(−c1 +i(c2 −g)(c− +c+ )) 1 2 (w − z) w−z (z − w)2 z Using the relation obtained in equation (2.19), we finally have T (w)jza (z) = jza (z) ∂jza (z) + , (w − z)2 w−z (2.29) which shows that the current jz is a primary field of conformal dimension one. It can similarly be checked that jz is a conformal primary of dimension zero. The stress-energy tensor We now want to compute the OPE between T (z) and T (w). This calculation relies on the preceeding calculation and on the double pole in the current-current operator product expansion. We get: T (z)T (w) = = 1 lim T (z) [jza (x)jza (w)] 2c1 :x→w:   1 ∂jza (x) a jza (x) lim + j (w) 2c1 :x→w:  (z − x)2 z − x  z ∂jza (w) jza (w) + +jza (x) (z − w)2 z−w (2.30) In the second line, only the regular terms in the remaining OPE’s will contribute to the final result. In the first line, all the terms proportional to the structure constants disappear once 10 again. We get: 1 T (z)T (w) = lim 2c1 :x→w:  c1 κba κba : jza jza : (w) 2c1 κba κba + − (z − x)2 (x − w)2 (z − x)2 (z − x)(x − w)3   a a : jza jz : (w) : jza (∂jza ) : (w) : (∂jza )jz : (w) + + (2.31) + z−x (z − w)2 z−w To take the limit, we expand all the functions of x in the neighbourhood of the point w and keep only the regular term: T (z)T (w) = dim G 2T (w) ∂T (w) + + , 2(z − w)4 (z − w)2 z−w (2.32) which proves that we have indeed a conformal algebra of central charge c = dim G. For supergroups, the relevant dimension is the superdimension (which is the self-contraction of the invariant metric). Summary We have shown that a fairly generic current algebra leads to a conformal theory when the Killing form is zero. The corresponding stress-energy tensor is given by the Sugawara construction. The holomorphic current component is a conformal primary with respect to the holomorphic Virasoro algebra. The Sugawara energy-momentum tensor has a central charge equal to the superdimension of the supergroup. We note that a supergroup with zero Coxeter number shares some features with a free theory. The central charge takes its naive value. The composite part of the Maurer-Cartan equation does not need to be renormalized. We will see other simplifications for these models further on. 3 Current algebra from supergroup current correlators We now switch gears and consider a concrete model in which the generic analysis of twodimensional current algebras of section 2 can be applied. We consider a conformal supergroup sigma-model from the list given in [7]. Though we believe our analysis applies to the whole list, some facts that we use below have been proven explicitly only for the P SL(n|n) models. We will calculate two-, three- and four-point functions of currents. Later we will infer the operator algebra of the currents from those correlation functions. 3.1 The model We consider a supergroup non-linear sigma-model with standard kinetic term based on a bi-invariant metric on the supergroup. It is the principal chiral model on the supergroup. In addition we allow for a Wess-Zumino term. Therefore, we have two coupling constants, namely the coefficient of the kinetic term 1/f 2 and the coupling constant k preceding the 11 Wess-Zumino term. The action is 11 : S = Skin + SW Z Z 1 d2 xT r ′ [−∂ µ g−1 ∂µ g] Skin = 16πf 2 Z ik SW Z = − d3 yǫαβγ T r ′ (g−1 ∂α gg−1 ∂β gg−1 ∂γ g) 24π B Using complex coordinates, and after taking the trace, the kinetic term becomes: Z 1 d2 z(∂gg−1 )c (∂gg−1 )c . Skin = − 4πf 2 (3.1) (3.2) The field g takes values in a supergroup. From the action we can calculate the classical currents associated to the invariance of the theory under left multiplication of the field g by a group element in GL and right multiplication by a group element in GR . The classical equations of motion for the model read:

(3.3) kf 2 + 1 ∂J a + kf 2 − 1 ∂(gJ g−1 )a = 0, where we have used the standard expressions for the left- and right-current at the WessZumino-Witten point12 : J(z, z) = −k∂gg−1 and J(z, z) = kg−1 ∂g . The classical GL currents are given by:   1 1 (1 + kf 2 ) −1 jz = − + k ∂gg = J 2 f2 2kf 2   1 1 (1 − kf 2 ) −1 ∂gg = − jz = − − k (gJ g−1 ). 2 f2 2kf 2 (3.4) (3.5) At the Wess-Zumino-Witten point f 2 = 1/k the z-component of the left-moving current becomes zero. As a consequence, the z-component J becomes holomorphic. A similar phenomenon happens at the other Wess-Zumino-Witten point (f 2 = −1/k) for the antiholomorphic component gJ g−1 . From now on, we will concentrate on the current j associated to the left action of the group. For future reference we note that the coefficients that relate the left current components to the derivative of the group element are (see section 2): c+ = − (1 + kf 2 ) 2f 2 and 11 c− = − (1 − kf 2 ) . 2f 2 (3.6) Our normalizations and conventions are mostly as in [23]. In particular we define the primed trace as T r ′ (ta tb ) = 2κab where κab is normalized to be the Kronecker delta-function for a compact subgroup. The action is written in terms of real euclidean coordinates. We soon switch to complex coordinates via z = x1 +ix2 . See [23] for further details. Starting in this section, we will no longer be careful in keeping track of the signs due to the bosonic or fermionic nature of the super Lie algebra generators. They can consistently be restored. 12 At the Wess-Zumino-Witten point, the parameters satisfy the equation: 1/f 2 = k. 12 3.2 Exact perturbation theory Elegant arguments were given [6] for the exactness of low-order perturbation theory for the calculation of various observables in the supergroup model on P SL(n|n). In particular, we will use these arguments to compute the (left) current-current two-point function exactly to all orders in perturbation theory using the free theory. Similarly we also compute the current three-point functions to all orders using perturbation theory up to first order in the structure constants. Below, we summarize some important facts that lead to these results [6]. The argument is essentially based on the special feature of the Lagrangian that all interaction vertices are proportional to (powers of) the structure constants as well as certain properties of the Lie superalgebra, which we now list: • If structure constants are doubly contracted, the result is proportional to the Killing form, which is zero for the supergroups under consideration. • The only invariant three-tensor is proportional to the structure constant and the only invariant two-tensor is the invariant metric13 . • Traceless invariant 4-tensors made of structure constants and the invariant metric give zero when contracted with the structure constants over two indices [6]. Using all these facts, and by using a pictorial representation of the correlation functions, one can show that vacuum diagrams with at least one interaction vertex all vanish and that group invariant correlation functions can be computed in the free theory. However, in order to compute the OPEs in the theory of interest, we have to calculate the 2 and 3-point functions of the right-invariant currents J a and (gJ g−1 )a , which are not fully invariant under the group action. In [6], it is shown that a correlation function that is invariant under only the right group action, and which is a two-tensor under the left group action (or vice versa), can be computed by setting all structure constants to zero. Similarly, a correlation function that is invariant under the right group action, and a three-tensor under the left group action can be computed by taking into account only contributions with at most one structure constant. We will present the argument for the simplicity of the 3-point function in the next section. In the following we also find it instructive to compute a four-point function to second order in the structure constants. In order to perform these calculations it is useful to expand the various terms in the action as well as the currents to second order in the structure constants. 13 These and the following are statements taken from [6]. A detailed proof is lacking. The condition of the uniqueness of the three-tensor can be relaxed to the condition that any invariant three-tensor contracted with the structure constants vanishes, which is a statement that has been proven in detail in an appendix to [11] for the particular case of the psl(2|2) Lie superalgebra. 13 Ingredients of perturbation theory We gather all our ingredients expanded to second order in the structure constants. We use the conventions g = eA and A = iAa ta . For the left current components we obtain: J gJ g−1 1 1 = −k∂gg−1 = −k(∂A + [A, ∂A] + [A, [A, ∂A]] + O(f 3 )) 2 3! 1 a c f bc a Ab ∂Ac + f bc f de Ab Ad ∂Ae + O(f 3 ))ta = −ki(∂Aa − 2 6 1 1 −1 = k∂gg = k(∂A + [A, ∂A] + [A, [A, ∂A]] + O(f 3 )) 2 3! f bc a 1 Ab ∂Ac + f a bc f c de Ab Ad ∂Ae + O(f 3 ))ta , = +ki(∂Aa − 2 6 (3.7) where O(f 3 ) indicates terms of third order or higher in the structure constants. The kinetic term and the Wess-Zumino term become: Z 1 1 d2 z(∂Aa ∂Aa − f a bc faij Ab ∂Ac Ai ∂Aj + O(f 4 )) Skin = 4πf 2 12 Z k d2 zf abc Aa ∂Ab ∂Ac + O(f 3 ). SW Z = − (3.8) 12π C The quadratic terms in the action give rise to the free propagators: Aa (z, z)Ab (w, w) = −f 2 κab log µ2 |z − w|2 , (3.9) where µ is an infrared regulator. Two-point functions Consider the Feynman diagrams with two external lines and pull out a structure constant where the external line enters. The rest of the diagram is now a blob with three external lines, with two of them contracted with the structure constant (the interaction strength). Now, from a group theoretic perspective, the three-spoked blob is also an invariant 3-tensor, which, by the properties itemized at the beginning of the section, is proportional to the structure constant. As a result, the whole graph is proportional to the metric times the Killing form: fabc fdbc = 2ȟgad , which vanishes for the supergroups under consideration. The two-point functions are therefore perturbatively exact when computed by setting all structure constants to zero and follows directly from the free propagator (3.9): hJ a (z, z)J b (w, w)i = h(gJ g−1 )a (z, z)(gJ g−1 )b (w, w)i = f 2 k2 κab (z − w)2 f 2 k2 κab (z − w)2 hJ a (z, z)(gJ g−1 )b (w, w)i = 2πf 2 k2 κab δ(2) (z − w). Three-point functions Consider the Feynman diagrams that contribute to hJ a (z, z)J b (w, w)J c (x, x)i . 14 (3.10) In their evaluation, there are strucutre constants coming both from the expansion of the currents in (3.7) and also from the interaction vertices. We would like to argue that only those diagrams which contain a single structure constant contribute. In order to show this, consider pulling out one structure constant out of the vertex where the external line enters. The rest of the diagram can be thought of as a blob with four external lines and which has the group structure of a rank 4 invariant tensor. Contracting two of its indices, the resulting graph contains a structure constant inside and vanishes, following the same argument that allows to compute the two-point functions by setting all structure constants to zero. We have now shown that the group structure of the four-spoked blob is that of a traceless rank 4 tensor. The full Feynman graph is evaluated by contracting a structure constant with this traceless rank-4 tensor. Using the special properties of the Lie superalgebra of P SL(n|n) we listed earlier in the section, it is clear that such a term evaluates to zero. Thus, the three-point functions are perturbatively exact at first order in the structure constants. There are two non-trivial contributions to this calculation. We have one contribution coming from the term proportional to the structure constants in the expansion (3.7) of the current components, and one from the first order Wess-Zumino interaction (3.8). Let us compute the first contribution for the JJJ three-point function in some detail: 1 hJ a (z, z)J b (w, w)J c (x, x)i1 = +ik3 h(∂Aa (z, z) − f dea Ad ∂Ae ) 2 1 1 (∂Ab − f deb Ad ∂Ae )(∂Ac − f dec Ad ∂Ae )i 2 2 abc 1 f f bac = −ik3 f 4 ((+) + (+) + cyclic) 2 (z − x)(w − x)2 (w − x)(z − x)2 z−w 1 + cyclic = −i k3 f 4 f abc 2 (z − x)2 (w − x)2 3 1 = −i k3 f 4 f abc . (3.11) 2 (z − w)(w − x)(x − z) The Wess-Zumino contribution is14 : E D D J a (z, z)J b (w, w)J c (x, x) = +ik4 ∂Aa (z, z)∂Ab (w, w)∂Ac (x, x)× 2  Z 1 2 deg d yf Ag ∂Ad ∂Ae (y, y) 12π C 1 1 . (3.12) = +ik4 f 6 f abc 2 (z − w)(w − x)(x − z) Adding the two contributions we get the three-point function: 1 1 hJ a (z, z)J b (w, w)J c (x, x)i = −i k3 f 4 (3 − kf 2 )f abc . (3.13) 2 (z − w)(w − x)(x − z) A quick check on the calculation is that it matches the known three-point function at the Wess-Zumino-Witten point, where it can be evaluated using the holomorphy of the currents. All other left current three-point functions can be computed analogously. They are (up to 14 A minus sign arises from expanding e−S to first order. 15 contact terms):  3 hJ (z, z)J (w, w)J (x, x)i = 2  1 hJ a (z, z)J b (w, w)(gJ g−1 )c (x, x)i = 2  1 h(gJ g−1 )a (z, z)(gJ g−1 )b (w, w)J c (x, x)i = 2  3 h(gJ g−1 )a (z, z)(gJ g−1 )b (w, w)(gJ g−1 )c (x, x)i = 2 a b c − − + +  kf 2 −ik3 f 4 f abc 2 (z − w)(w − x)(x − z)  2 kf −ik3 f 4 f abc (z − w) 2 (z − w)2 (x − w)(x − z)  kf 2 +ik3 f 4 f abc (z − w) 2 (z − w)2 (x − w)(x − z)  +ik3 f 4 f abc kf 2 . (3.14) 2 (z − w)(w − x)(x − z) Coincidence limit and operator product expansions When we take coincidence limits of the three-point functions, we expect to be able to replace the product of two operators by their operator product expansion. Using the general form of the current-current operator product expansions, and the exact two-point functions, we can infer from the above three-point functions a proposal for the current-current operator product expansions. Up to contact terms the two- and three-point functions can be reproduced in their coincidence limits by the OPEs 15 : k2 f 2 κab kf 2 ab J c (w) 2 3 + kf ( − )if c (z − w)2 2 2 z−w 2 z−w 1 kf )if ab c (gJ g−1 )c (w) + . . . + (−kf 2 )( − 2 2 (z − w)2 1 kf 2 ab (gJ g−1 )c (w) )if c J a (z, z)(gJ g−1 )b (w, w) ∼ 2πk2 f 2 κab δ(2) (z − w) + kf 2 ( + 2 2 z−w 2 1 1 kf )if ab c J c (w) + . . . − kf 2 ( − 2 2 z−w k2 f 2 κab kf 2 ab (gJ g−1 )c (w) 2 3 (gJ g−1 )a (z, z)(gJ g−1 )b (w, w) ∼ − kf ( + )if c (z − w)2 2 2 z−w 2 z−w c 1 kf )if ab c J (w) + . . . (3.15) + kf 2 ( + 2 2 (z − w)2 J a (z, z)J b (w, w) ∼ 15 From now on we will no longer always make explicit the fact that all the operators depend both on the holomorphic as well as the anti-holomorphic coordinate at a generic point in the moduli space. 16 When we normalize the currents as in (3.5) to agree with section 2, we find the following OPEs: c (1 + kf 2 )2 κab i 2 2 ab jw (w) (1 + kf )(3 − kf )f + c 4f 2 (z − w)2 4 z−w z − w i + (1 + kf 2 )2 f ab c j c (w) + . . . 4 (z − w)2 w ab c i (1 − kf 2 )2 κab 2 2 f c jw (w) b + (1 − kf )(3 + kf ) (w) ∼ jza (z)jw 4f 2 (z − w)2 4 z−w ab c (z − w) f c jw (w) i + ... + (1 − kf 2 )2 4 (z − w)2 c (w) i(1 + kf 2 )2 f ab c jw 2π b (w) ∼ − 2 (1 + kf 2 )(1 − kf 2 )κab δ(2) (z − w) + jza (z)jw 4f 4 z−w 2 2 ab c i(1 − kf ) f c jw (w) + + ... (3.16) 4 z−w b jza (z)jw (w) ∼ We can read from these formulas the coefficients of the generic current algebra (2.21): i (1 + kf 2 )2 c2 = (1 + kf 2 )(3 − kf 2 ) 2 4f 4 2 2 (1 − kf ) i c3 = c4 = (1 − kf 2 )(3 + kf 2 ) 4f 2 4 i and g = (1 + kf 2 )(1 − kf 2 ) . 2 c1 = (3.17) We note that the coefficients automatically satisfy the extra constraints (2.19) and (2.20) one gets by requiring consistency of the current algebra with the Maurer-Cartan equation. A four-point function From the three-point functions, we conclude that the coefficient g is non-zero. We have argued in section 2 that that is associated to the appearance of logarithms in the regular term of the jz jz operator product expansion. We would like to check the coefficient of the logarithm more directly in a perturbative calculation. For that purpose it is sufficient to study a four-point function at second order in the structure constants. The computation will be exact at that order. In particular we want to compute the four-point function: hJ [a (z)(gJ g−1 )b] (w)J c (x)(gJ g−1 )d (y)iO(f 2 ) , (3.18) at second order in the structure constants, and in the z → w limit. We anti-symmetrized in the a and b indices (and weighted each term with a factor 1/2). In the coincidence limit, logarithms appear in the regular terms in the OPE between J and (gJ g−1 ) and they will give non-zero contribution to the four-point functions. In our calculation we focus on the terms proportional to log |z − w|2 (and which are not contact terms). We distinghuish the following contributions at this order. We can expand a single current to second order, and compute in the free theory. We can expand two currents to first order and compute in the free theory. We can add one linear Wess-Zumino interaction term and expand one current to first order. Or we can add two linear Wess-Zumino interaction terms 17 and take only the leading terms in the currents. Finally, we can add one quadratic principal chiral model interaction term, and treat the currents at zeroth order. We found the following results. The second order term in a current cannot give rise to logarithmic contributions. Two currents at first order can be contracted to give a logarithm. It is easy to see that there are few contributions to the terms of interest, and they give: − 1 k4 f 6 abe cd f f e log µ2 |z − w|2 8 (z − x)2 (w − y)2 (3.19) The term linear in the Wess-Zumino interaction term gives no contribution. The term arising from the quartic interaction term in the principal chiral model doubles the previous non-zero term. The Wess-Zumino term squared gives a contribution of a different type equal to: + k6 f 10 1 log µ2 |z − w|2 f abe f cde . 2 4 (z − x) (w − y)2 (3.20) The latter contribution is the hardest to calculate. It consists of the order of 216 free field contractions, which exhibit a lot of symmetries. Some logarithmic terms in the double integrals over the points of interaction need to be evaluated, but all of these integrals are straightforwardly performed using partial integrations and other elementary techniques. The tedious but elementary calculation leads to the above result. In total we get, at second order in the structure constants, and only regarding the logarithmic contribution in z, w: hJ [a (z)(gJ g−1 )b] (w)J c (x)(gJ g−1 )d (y)iO(f 2 ),log = 1 1 − k4 f 6 (1 − k2 f 4 )f abe f cd e log µ2 |z − w|2 . (3.21) 2 4 (z − x) (w − y)2 Using the relative normalization between the J’s and the j’s, we find that b] hjz[a (z)jw (w)jxc (x)jyd (y)iO(f 2 ),log = 1 1 − (1 + kf 2 )2 (1 − kf 2 )(1 − k2 f 4 )f abe f cd e log µ2 |z − w|2 . (3.22) 64f 2 (z − x)2 (w − y)2 Let us see how to use this result to check the coefficient g in the operator product expansion. From the expressions for the operator product algebra (2.21) and from the exact three-point functions (3.14), we find that the logarithmic term in the normalized four-point function in the coincidence limit z → w is: g e c d e b c d )jx jy i jx jy i ≈ − f ab e log µ2 |z − w|2 h(∂jw − ∂jw hjza jw 4 g 1 (1 + kf 2 )(1 − kf 2 )f ab e log µ2 |z − w|2 ≈ + 4 8k3 f 6 1 w−y (−∂w (1 − kf 2 )ik3 f 4 f edc (1 + kf 2 ) 2 (w − y)2 (x − w)(x − y) w−x 1 ) −∂ w (1 + kf 2 )f ecd(−)ik3 f 4 (1 − kf 2 ) 2 2 (w − x) (y − w)(y − x) g (1 + kf 2 )2 (1 − kf 2 )2 f ab e f ecd ≈ +i 32f 2 1 log µ2 |z − w|2 . (3.23) 2 (w − x) (w − y)2 18 We recall the value for g: i (1 + kf 2 )(1 − kf 2 ), 2 so the operator algebra and the three-point functions predict: g= 1 (1 + kf 2 )2 (1 − kf 2 )2 (1 − k2 f 4 )f ab e f ecd 64f 2 1 log µ2 |z − w|2 . 2 (w − x) (w − y)2 (3.24) b c d jx jy i ≈ − hjza jw (3.25) The prediction is matched by our perturbative calculation of the four-point function. Moreover, since the coefficient g is fixed to all orders by the calculation of the three-point function, our result at second order in the structure constants is exact. The calculation is a good consistency check on the correlators and operator product expansions. The full four-point function is a function of the cross ratio of the four insertion points, in which the regulator µ drops out. We note also that the appearance of logarithms in four-point functions of operators that differ by an integer in their conformal dimension is generic. In our case, the scale µ must appear in the operator product expansion because we lifted space-time fermionic zeromodes [6]. These in turn are linked to the non-diagonalizable nature of the scaling operator in sigma-models on supergroups [24]. 3.3 Summary of the current algebra We summarize the current algebra for the left group action: ab c i (1 + kf 2 )2 κab 2 2 f c jz (w) + (1 + kf )(3 − kf ) 4f 2 (z − w)2 4 z−w z − w i + (1 + kf 2 )2 f ab c jzc (w)+ : jza jzb : (w) 4 (z − w)2 π i(1 + kf 2 )2 f ab c jzc (w) jza (z)jzb (w) = − 2 (1 + kf 2 )(1 − kf 2 )κab δ(2) (z − w) + 4f 4 z−w ab c 2 2
i i(1 − kf ) f c jz (w) − (1 − k2 f 4 )f ab c log |z − w|2 ∂jzc (w) − ∂jzc (w) + 4 z−w 8 a b + : jz jz : (w) jza (z)jzb (w) = (1 − kf 2 )2 κab i f ab j c (w) + (1 − kf 2 )(3 + kf 2 ) c w 2 2 4f (z − w) 4 z−w 2 2 i(1 − kf ) z − w ab c + f c j (w)+ : jza jzb : (w) . 4 (z − w)2 jza (z)jzb (w) = (3.26) The current algebra for the right group action can be obtained through the combined operation g → g−1 and worldsheet parity P , which is a symmetry of the model 4 Conformal perturbation theory In this section we study Wess-Zumino-Witten models with a perturbed kinetic term, both for its intrinsic interest and as a tractable example of conformal perturbation theory. For a general group manifold, the deformed model becomes non-conformal. For supergroup manifolds with vanishing Killing form, the models remain conformal. In the earlier sections, we 19 have computed, using the exact two- and three-point functions, the current algebra of the deformed theory. In this section, we re-derive these results using conventional conformal perturbation theory. This will be a consistency check of the deformed conformal current algebra we have obtained in equation (3.26). 4.1 The current algebra in the Wess-Zumino-Witten model We first review the chiral current algebra of the Wess-Zumino-Witten model. We recall the action Z k SW ZW = d2 xT r ′ [−∂ µ g−1 ∂µ g] + kΓ (4.1) 16π where Γ is the Wess-Zumino term, and the field g(z, z) takes values in a (super)group G. The model has a global GL × GR invariance by left and right multiplication of the group element. The currents associated to these symmetries are (in complex coordinates): J(z) = −k∂gg−1 and J(z) = kg−1 ∂g . (4.2) The right-invariant current J(z) is holomorphic and generates the left-action of the group GL . The left-invariant anti-holomorphic current J generates the right translation by a group element. The components of the current J satisfy the OPE: J a (z)J b (w) ∼ c kκab ab J (w) + if , c (z − w)2 z−w (4.3) and the components of the current J (z) satisfy the same OPE, with anti-holomorphic coordinates instead of holomorphic ones. In particular, in our conventions the pole term keeps the same sign. These currents generate a large chiral affine current algebra whose existence is useful in solving the model via the Knizhnik-Zamolodchikov equations. 4.2 Perturbation of the kinetic term: classical analysis We are interested in the following marginal deformation of the Wess-Zumino-Witten model: Z λ S = SW ZW + d2 z Φ(z, z) . (4.4) 4πk where 1 Φ = (: J c (gJ g−1 )c : + : (gJ g−1 )c J c :) . (4.5) 2 In other words, we perturb the kinetic term by multiplying it with a factor 1 + λ. Comparing the action with the action in the earlier section, we find that λ is related to the kinetic coefficient f defined in the previous section by the relation 1 = k(1 + λ) . f2 (4.6) We note that, analogous to the composite operator that appeared in the Maurer-Cartan equation, we have chosen a symmetric combination of the product of J and gJ g−1 operators to represent the marginal operator in the quantum theory. 20 4.3 The current-current operator product expansions In this subsection we compute the correction to the holomorphic current-current operator product expansion induced by the perturbation of the kinetic term of the Wess-ZuminoWitten model for a simple (super) Lie algebra. In order to perform the calculation we require the OPEs between the currents J and gJ g−1 at the WZW point. These are obtained by requiring that the Maurer-Cartan equation holds in the quantum WZW model, as shown in Appendix B: we compute the OPE of the current J with the Maurer-Cartan equation for a generic value of the dual Coxeter number and demand that it vanish. This constraint leads to the operator product expansion (gJ g−1 )c (w, w) + : J a (gJ g−1 )b : (w, w) . z−w (4.7) −1 A similar demand on contact terms and the most singular terms in the OPE of (gJ g ) with the Maurer-Cartan equation leads to the OPE J a (z)(gJ g−1 )b (w, w) = 2πkκab δ(2) (z − w) + if ab c (gJ g−1 )a (z, z)(gJ g−1 )b (w, w) = kκab J c (w)(z − w) (gJ g−1 )c (w, w) + if ab c − 2if ab c 2 2 (z − w) (z − w) z−w + : (gJ g−1 )a (gJ g−1 )b : (w, w) . (4.8) A general discussion of higher order corrections to operator product expansions is given in Appendix A. Here we focus on applying the discussion to the case of a supergroup with vanishing Killing form. We compute the corrections induced by the exactly marginal perturbation to the J a J b OPE. The nth order correction is denoted by # " n Z n Y (−λ) a (4.9) (JJ)ab d2 xi Φ(xi , xi ) J b (w, w) n (z − w, w) = J (z, z) (4πk)n n! i=1 where the square bracket means that we have to contract J a (z, z) with all the integrated operators before we contract it with J b (w, w). We define Hna (z, z) to be this complete contraction: " # n Z Y 1 2 a a d xi Φ(xi , xi ) . Hn (z, z) = J (z, z) (4.10) (4πk)n n! i=1 One can similarly define another contraction, with J replaced by the current (gJ g−1 )a : " # n Z Y 1 a −1 a 2 H n (z, z) = (gJ g ) (z, z) (4.11) d xi Φ(xi , xi ) . (4πk)n n! i=1 The basic building blocks we need to carry out this computation are the OPE of the currents J a and (gJ g−1 )a with the marginal operator Φ. As we will see, once these OPEs are obtained, the nth order correction can be obtained by a process of iteration. Here, we list the two OPEs of interest and refer the reader to Appendix B for details. Z Z (gJ g−1 )a (w, w) a 2 + 2πkJ a (w) J (w) d xΦ(x, x) ∼ k d2 x (w − x)2 Z Z J a (x, x) −1 a 2 −1 a . (4.12) (gJ g ) (w, w) d xΦ(x, x) ∼ 6πk(gJ g ) (w, w) − k d2 x (w − x)2 21 With these basic OPEs, let us contract the current with one of the integrated marginal operators: Hna (z, z) n = (4πk)n n! Z 1 = d2 x 4π Z  n−1 YZ k(gJ g−1 )a (x, x) (2) a d2 xi Φ(xi , xi ) + 2πkδ (z − x)J (x, x) d x (z − x)2 i=1 ! a H n−1 (x, x) 1 a (z, z) + Hn−1 2 (z − x) 2  2 (4.13) a One can do a similar operation on H n and we get  n−1  Z a YZ n −1 a 2 kJ (x, x) d2 xi Φ(xi , xi ) 6πk(gJ g ) (z, z) − d x (4πk)n n! (z − x)2 i=1   a Z (x, H x) 3 a 1 n−1 . = H n−1 (z, z) − d2 x 2 4π (z − x)2 a H n (z, z) = (4.14) a These are coupled recursion relations for Hna and H n subject to the initial conditions: H0a (z, z) = J a (z, z) a H 0 (z, z) = (gJ g−1 )a (z, z) These recursion relations have the following solutions: Z  (gJ g−1 )a (x, x) n a n 1 a J (z, z) + Hn (z, z) = 1 − d2 x 2 2 2π (z − x)2 Z   J a (x, x) n n 1 a (gJ g−1 )a (z, z) − . d2 x H n (z, z) = 1 + 2 2 2π (z − x)2 (4.15) (4.16) In particular we deduce that   Z −1 a n a n 1 n b 2 (gJ g ) (x, x) (JJ)ab (z − w, w) = (−λ) J (w, 1 − w) J (z, z) + d x n 2 2 2π (z − x)2     ab c kκ J (w) n + if abc = (−λ)n 1 − 2 2 (z − w) z−w   Z −1 c 1 n 1 ab (2) abc (gJ g ) (w, w) 2 2πkκ δ (z − x) + if d x + 2 2π (z − x)2 x−w    c ab n J (w) kκ + if abc 1 − = (−λ)n 2 (z − w) 2 z−w  n z−w +i f abc (gJ g−1 )c (w, w) + . . . . (4.17) 2 (z − w)2 We can now sum the perturbative series in λ. We get the OPE in the perturbed theory: J a (z, z)J b (w, w) = ∞ X (−λ)n (JJ)ab n (z − w, w) n=0 = kκab c 1 2 + 3λ λ z−w abc J (w) + if − if abc (gJ g−1 )c (w, w) 2 2 2 1 + λ (z − w) 2(1 + λ) z−w 2(1 + λ) (z − w)2 22 Using the map (kf 2 )−1 = 1 + λ, one can check that this coincides with the OPE in equation (3.15). With the same techniques we can also compute the corrections to the J a (gJ g−1 )b and (gJ g−1 )a (gJ g−1 )b OPEs. We get the results J a (z, z)(gJ g−1 )b (w, w) = 1 2 + λ (gJ g−1 )c (w, w) 2πkκab δ(2) (z − w) + if abc 1+λ 2(1 + λ)2 z−w J c (w) λ + . . . (4.18) − if abc 2(1 + λ)2 z − w (gJ g−1 )a (z, z)(gJ g−1 )b (w, w) = kκab 4 + 3λ if abc (gJ g−1 )c (w, w 1 − 1 + λ (z − w)2 2(1 + λ)2 z−w 2 + λ (z − w)if abc J c (w) + + ... (4.19) 2(1 + λ)2 (z − w)2 Again this matches with the OPE obtained in equation (3.15). Summary In this section, we have shown by resumming conformal perturbation theory that the deformed current algebra we obtain this way matches the algebra obtained in section 3 through the calculation of 2- and 3-point functions to all orders in perturbation theory. 5 The current algebra on the cylinder In this section we consider the sigma-model on a cylinder and Fourier decompose the current algebra. The representation in terms of Fourier modes is often more conventient. To put the algebra on a cylinder, we first compute the operator algebra on the plane, and then compactify the plane. We consider the complex plane z = σ − iτ and consider τ as time and σ as the spatial coordinate. Denoting the currents as jµ (σ, τ ), the commutator of equal-time operators is defined to be the limit of the difference of time-ordered operators (evaluated at τ = 0) [jµa (σ, 0), jνb (0, 0)] = lim (jµa (σ, iǫ)jνb (0, 0) − jνb (0, iǫ)jµa (σ, 0)) . (5.1) ǫ→0 Using this definition, let us compute the commutators for the holomorphic component of the current (we suppress the τ = 0 argument within the currents in what follows):  c2 c σ + iǫ c c1 κab a b + f ab c ( jz (0) + (c2 − g) j (0)) [jz (σ), jz (0)] = lim 2 ǫ→0 (σ − iǫ) σ − iǫ (σ − iǫ)2 z  c1 κab −σ + iǫ c c2 c ab − j (σ) + (c2 − g) − f c( j (σ)) + . . . (σ + iǫ)2 −σ − iǫ z (σ + iǫ)2 z = −2πic1 δ′ (σ)κab + 2πic2 δ(σ)f ab c jzc (0) + 2πi(c2 − g)δ(σ)f ab c jzc (0) + . . . (5.2) where we used 1 1 − = 2πiδ(σ) ǫ→0 σ − iǫ σ + iǫ 1 1 lim − = −2πiδ′ (σ) ǫ→0 (σ − iǫ)2 (σ + iǫ)2 σ − iǫ σ + iǫ − = 2πiδ(σ). lim 2 ǫ→0 (σ − iǫ) (σ + iǫ)2 lim 23 (5.3) For other components we find: [jza (σ), jzb (0)] = +2πic3 δ′ (σ)κab − 2πic4 δ(σ)f ab c jzc (0) − 2πi(c4 − g)δ(σ)f ab c jzc (0) [jza (σ), jzb (0)] = −2πi(c4 − g)δ(σ)f ab c jzc (0) + 2πi(c2 − g)δ(σ)f ab c jzc (0) . (5.4) It is now straightforward to compactify σ ≡ σ + 2π and Fourier decompose the operator algebra on the cylinder using: X jz = +i e−inσ jz,n n∈Z jz = −i X e−inσ jz,n n∈Z δ(σ) = 1 X inσ e . 2π (5.5) n∈Z We find: a b c c [jz,n , jz,m ] = c1 κab nδn+m,0 + c2 f ab c jz,n+m − (c2 − g)f ab c jz,n+m a b c c [jz,n , jz,m ] = −c3 κab nδn+m,0 + c4 f ab c jz,n+m − (c4 − g)f ab c jz,n+m c c a b ] = (c4 − g)f ab c jz,n+m . + (c2 − g)f ab c jz,n+m [jz,n , jz,m (5.6) We can check the validity of the Jacobi identity, which follows from the validity of the Maurer-Cartan equation along with the Jacobi identity for the Lie algebra of G. Conserved charges We note that the current one-form satisfies the conservation equation d ∗ j = 0, and that therefore the integral of the time-component of the current over the spatial circle is conserved in time. The corresponding charges are easily determined to be the sum of the zero-modes of the current algebra. They generate the Lie algebra of G. We recall that the group action generated by these charges corresponds to the left group action GL , and that there is an analogous right group action GR . Kac-Moody subalgebra Let us consider the combination of the currents jza −jza and compute the commutation relations of its modes with themselves. Using the above basic commutation relations, we find b b c c a a ), (jz,m + jz,m )] = (c1 − c3 )κab nδm+n,0 + (c2 + c4 − g)f ab c (jz,n+m ), + jz,n+m [(jz,n + jz,n (5.7) which is a Kac-Moody algebra at level k+ = − c1 − c3 (c2 + c4 − g)2 (5.8) as becomes manifest in terms of the rescaled currents J a = −i jza − jza . c2 + c4 − g 24 (5.9) We observe that for the case of the supergroup considered in the earlier section, substituting the values of the ci in (3.17), we obtain a Kac-Moody algebra at level k+ = k, with the currents taking the simple form J a = (jza − jza ) . (5.10) When we choose a real form of the supergroup that has a compact subgroup, the level k will be integer. We also observe that the current associated to the σ-component of the canonical right-invariant one-form dgg−1 is: J ′a = c− jz + c+ jz . (5.11) In term of these currents, we find the mode algebra: c1 − c3 c κab nδm+n,0 + if ab c Jn+m (c2 + c4 − g)2 c1 c− + c3 c+ ab ′c κ nδm+n,0 − if ab c Jn+m [Jn′a , Jmb ] = −i c2 + c4 − g ′c [Jn′a , Jm′b ] = (c2− c1 − c2+ c3 )κab nδm+n,0 + f ab c (2c2 c− − 2c4 c+ + g(c+ − c− ))Jn+m [Jna , Jmb ] = − c . −if ab c (c2 + c4 − g)(c2− c2 + c4 c2+ − g(c2+ + c+ c− + c2− ))Jn+m (5.12) For the specific case of the supergroup model, we find b c [Jna , Jm ] = kκab nδm+n,0 + if ab c Jn+m (kf 2 − 1)(kf 2 + 1) ab ′c κ nδm+n,0 − if ab c Jn+m 4f 4 [Jn′a , Jm′b ] = 0 . b [Jn′a , Jm ]= (5.13) We identified a Kac-Moody subalgebra J and an infinite set of modes J ′ that commute amongst themselves. The latter modes transform into the identity and themselves under the Kac-Moody algebra. We also note that we can obtain a second Virasoro algebra by applying the Sugawara construction to the Kac-Moody algebra J . The corresponding energy-momentum tensor generates a Virasoro algebra at central charge sdim G. It is not holomorphic. The difference of these energy momentum tensors for the left and right group is proportional to the difference of the holomorphic and anti-holomorphic energy momentum tensors. That indicates the existence of a non-chiral analogue of the Knizhnik-Zamolodchikov equation. 6 Conclusions In this paper, we have performed a generic analysis of the conditions imposed on local Lorentz covariant and P T invariant current algebras. In particular we allowed for parity-breaking models and found a class of solutions to the conditions. In the case for which the algebra has vanishing Killing form, we showed that one can construct an energy momentum tensor in terms of a current component in a way similar to the Sugawara construction. The current component is then a conformal primary and the central charge is the (super)dimension of the group. This gives a constructive proof of conformality of the quantum model. We moreover computed exact two- and three-point functions for principal chiral models with Wess-Zumino term for supergroups with vanishing Killing form. Using these exact 25 results, we showed that the current algebra is realized in these models, and we calculated the coefficients in the current algebra. We performed a check on a logarithmic regular term by computing the relevant part of a four-point function. The algebra was independently derived by using the techniques of conformal perturbation theory about the Wess-ZuminoWitten point. We hope the existence of such current algebras will prove useful in furthering the solution of these models [9, 10, 11]. Another avenue to explore is to systematically analyze the exactness of low-order perturbation theory for various current and group valued correlators. One of the examples to which our discussion applies is the sigma model on the supergroup P SU (1, 1|2). This particular supergroup is useful to quantize string theory on AdS3 × S 3 [12, 6]. To quantize the string in the presence of Ramond-Ramond fluxes, we can, in this instance, use the six-dimensional hybrid formalism with eight [28] or sixteen [25] manifest supercharges. In the first case, the P SU (1, 1|2) sigma-model is at the core of the worldsheet theory [12]. It is possible to realize the AdS3 × S 3 spacetime as the near-horizon limit of a D5-NS5D1-F1 system. We can then write the parameters of the non-chiral current algebra in terms of the numbers of D5 and NS5 branes [12]. The integer parameter k that multiplies the WessZumino term in the action is equal to the number NN S5 of NS5 branes while the parameter 1/f is the radius of curvature of spacetime. When the number ND5 of D5-branes is equal to zero, the parameters satisfy kf 2 = 1 and the non-holomorphic component of the rightinvariant current vanishes : we have a chiral current algebra. When we turn on the RR fluxes, we obtain the generic current algebra given in equation (3.26). It is important to further investigate this algebra in the context of string theory on AdS3 . Exploring the integrability of these supergroup models will prove useful in understanding better the properties of the AdS3 /CF T2 correspondence. The presence of a Kac-Moody algebra at level k over the whole moduli space of the theory may also help in the construction of the string spectrum in AdS3 × S 3 with Ramond-Ramond fluxes. Likewise, another application of our analysis is to coset models G/H where G is a supergroup with zero Killing form. In [7] it was shown that a number of coset models where H is a maximal regular subalgebra are conformal to two loops. Graded supercosets based on supergroups with vanishing Killing form are also believed to be conformal [8]. These cosets occur in the worldsheet description of certain string theory backgrounds, for instance, they appear as the central building block of the AdS5 × S 5 background. Moreover, as symmetric spaces or right coset manifolds, they retain a left group action as a symmetry and we therefore expect that parts of our analysis still apply. It is certainly worth exploring the quantum integrability of these coset models per se, and how it ties in with the conformal current algebra that we have exhibited. Acknowledgements We would like to thank Costas Bachas, Zaara Benbadis, Denis Bernard, Christian Hagendorf, Christoph Keller, Andre LeClair, Giuseppe Policastro, Thomas Quella and Walter Troost for discussions. We are grateful to Matthias Gaberdiel, Anatoly Konechny, Thomas Quella and an anonymous referee for comments and corrections. 26 A Perturbed operator product expansions We consider the corrections to an OPE induced by an exactly marginal deformation of a conformal field theory. The deformation parameter is denoted by λ. In the deformed theory, we can write the OPE between two operators A and B as: X lim A(z)B(w) = C(z, w) = λn Cn (z − w, w) , (A.1) z→w n≥0 where it is implicit that the dependence on the variables does not have to be holomorphic. We expand the result in a basis of operators evaluated at the point w. The operator Cn (z − w, w) is usually written as a series in powers of z − w. It is not obvious that the operators A and B (and therefore C) are well-defined operators in the perturbed conformal field theory, but we will assume that that is the case for the model at hand. Let us see how to compute the operators Cn (z − w, w) at order n. By definition, we have:   X lim hA(z)B(w)φ1 (x1 )...φp (xp )iλ = h λn Cn (z − w, w) φ1 (y1 )...φp (yp )iλ (A.2) z→w n≥0 for any operators φ1 (y1 )...φp (yp ). If we want to perform the computation at the non-perturbed point, we write the previous equality as:  m Z X λm Y d2 xi Φ(xi )i0 lim hA(z)B(w)φ1 (y1 )...φp (yp )  z→w m! i=1 m≥0     m Z X X λm Y d2 xi Φ(xi )i0 (A.3) = h λn Cn (z − w, w) φ1 (y1 )...φp (yp )  m!  n≥0 m≥0 i=1 where Φ is the exactly marginal operator we use to deform the theory. We isolate the term proportional to λn : n 1 Y lim hA(z)B(w)φ1 (y1 )...φp (yp ) z→w n! i=1 =h n X l=0 Z d2 xi Φ(xi )i0 n−l Z Y 1 Cl (z − w, w) (n − l)! ! d2 xi Φ(xi ) φ1 (y1 )...φp (yp )i0 (A.4) i=1 This becomes an operator identity in the non-perturbed theory: n lim A(z)B(w) z→w 1 Y n! i=1 Z d2 xi Φ(xi ) = n X Cl (z − w, w) n−l Z Y 1 (n − l)! d2 xi Φ(xi ) . (A.5) i=1 l=0 The previous equation defines iteratively the operator Cn which appears in the operator product expansion at order n. We would like to give a prescription to compute the n-th order term in the OPE, Cn (z − w, w). At zeroth order, the definition is lim A(z)B(w) = C0 (z − w, w) . z→w 27 (A.6) As expected the zeroth-order OPE is the OPE in the non-deformed model. At order one, we have Z Z 2 lim A(z)B(w) d xΦ(x) = C0 (z − w, w) d2 xΦ(x) + C1 (z − w, w) (A.7) z→w Here is one proposal on how to deal with the left-hand side of this equation. First we let the operator A(z) approach B(w) and Φ(x) (separately):   Z Z Z 2 2 2 lim A(z)B(w) d xΦ(x) = lim (AB)(z − w, w) d xΦ + B(w) d x(AΦ)(z − x, x) z→w z→w (A.8) where (AB)(z − w, w) denotes the contraction of A(z) and B(w) (in the unperturbed theory), with the resulting operators evaluated at point w. It is clear that the first term on the R the 2 right-hand side is equal to C0 (z − w, w) d xΦ(x), so the OPE at first-order is given by the second term: Z C1 (z − w, w) = lim B(w) d2 x(AΦ)(z − x, x) . (A.9) z→w At higher order, the same structure appears. We can always recognize in the computation the lower-order contributions, and isolate the highest-order term. We use the definition n n Z n−l X YZ 1 1 Y Cl (z − w, w) d2 xi Φ(xi ) = d2 xi Φ(xi ) . (A.10) lim A(z)B(w) z→w n! (n − l)! i=1 i=1 l=0 To evaluate the left-hand side, we let the operator A(z) approach the other ones. As it approaches B(w), we generate the term with l = 0 on the right-hand side. As it approaches one of the copies of the marginal operator Φ, we get Z n−1 YZ 1 2 d2 xi Φ(xi ) (A.11) d x(AΦ)(z − x, x) lim B(w) z→w n! i=1 To carry on, we take the operators (AΦ)(z − x, x) that was just generated at the point x and let it approach the other operators in the expression. If it approaches B(w), then we generate the term with l = 1 in the right-hand side of the definition. Otherwise we generate a new expression on which we apply the same procedure. Finally, we understand how to obtain directly the order-n OPE Cn (z − w): it is the term that we get by first contracting A(z) with all the integrated operators, and then contracting with B(w) at the very end. We will denote it as: # " n Z 1 Y 2 d xi Φ(xi ) B(w) (A.12) Cn (z − w, w) = A(z) n! i=1 All the operators inside the square brackets have to be contracted, before performing the last contraction with the operator outside the square brackets. We should stress that the previous procedure is not always well-defined. In the computation described in the bulk of this paper, this prescription leads to an unambiguous result for the poles of the current-current OPEs. However, in a more general context, the integrals appearing in the above calculations need a more careful regularization. B Detailed operator product expansions In this appendix we show how to compute OPEs involving the operator (gJ g−1 ) in the WZW model. 28 The contact terms It is natural to postulate contact terms between the left- and right-invariant currents (see e.g. [29]). Indeed, even for a U (1) current algebra, contact terms can be derived from the representation of the current algebra in terms of a free boson and its logarithmic propagator. Since at large level k, the group manifold flattens and is equivalent to a set of free fields, we do expect contact terms to arise. We propose the following contact terms: J a (z, z)(gJ g−1 )b (w, w) ∼ 2πkκab δ(2) (z − w) + ... (B.1) The Maurer-Cartan equation in the quantum theory In the quantum theory, the composite operator in the Maurer-Cartan equation is ambiguous due to normal ordering. With our choice of normal ordering, it is natural to propose the quantum Maurer-Cartan equation ∂J c + ∂(gJ g−1 )c + i c f (: J d (gJ g−1 )e : + : (gJ g−1 )e J d :) = 0 . 2k de (B.2) One way to check this proposal is to compute the OPE between the current components J a and the operator on the left hand side of equation (B.2) which is classically zero due to the Maurer-Cartan equation. In the calculation, it is crucial to apply the normal ordering prescription we introduced in section 2. We not only confirm the above proposal for the quantum Maurer-Cartan equation, but also find that we need to fix the OPE between J a and (gJ g−1 )b to be (gJ g−1 )c (w, w) + : J a (gJ g−1 )b : (w, w) . z−w (B.3) Let us show this calculation in some detail in order to illustrate the techniques involved. Using the holomorphy of the current J and the knowledge of the naive conformal dimensions of the operators, we can make the ansatz J a (z)(gJ g−1 )b (w, w) ∼ 2πκab δ(2) (z − w) + if ab c (gJ g−1 )c (w, w) + : J a (gJ g−1 )b : (w, w) z−w (B.4) With the definition of the normal ordering above, let us compute the operator product expansion between J a (z) and the Maurer-Cartan equation. We distinguish two terms. The first term is J a (z)(gJ g−1 )b (w, w) = 2πkκab δ(2) (z − w) + αif abc
J a (z) ∂J c (w) + ∂(gJ g−1 )c (w, w) =     d −1 d kκac acd J (w) ac (2) acd (gJ g ) (w, w) + if + ∂w 2πkκ δ (z − w) + αif + ... ∂w (z − w)2 z−w z−w = −if acd J d (w)2πδ(2) (z − w) + αif acd + if acd (gJ g−1 )d (w, w) (z − w)2
α ∂(gJ g−1 )d (w, w) + ∂J d (w) + . . . (B.5) z−w From the last terms we see that we can obtain a pole term proportional to the Maurer-Cartan equation if we put α = 1. It can be shown that this is the only consistent possibility, and we 29 will freely put α = 1 from now on. The second term with an extra minus sign is given by c ifde J a (z) lim (J d (x)(gJ g−1 )e (w, w) + (gJ g−1 )e (w, w)J d (x)) . (B.6) :x→w: k where we have used the normal ordering prescription. Let us start with the first of the two c ): terms in (B.6) (suppressing the overall − ki fde # (" ad J g (x) ad f kκ g (gJ g−1 )e (w, w) +i J a (z) lim J d (x)(gJ g−1 )e (w, w) = lim :x→w: :x→w: (z − x)2 (z − x) #) " ae (gJ g −1 )g (w, w) f g . (B.7) +J d (x) 2πkκae δ(2) (z − w) + iα (z − w) − We perform successive contractions, and subtract singular terms according to the normal ordering procedure to obtain c f ad f ge )(gJ g −1 )h (w, w) i(fdca δhd − αk fde g h (z − w)2 f c (fgad : J g (gJ g−1 )e : (w, w) + αfgae : J d (gJ g−1 )g : (w, w)) . (B.8) + de k(z − w) − 2iπfdca δ(2) (z − w)J d (w) − When α = 1 and using the Jacobi identity, we can simplify further: ! d : J g (gJ g −1 )e : (w, w) fdca(gJ g−1 )d fdca fge 2 ĥ − . (B.9) − 2πifdca δ(2) (z − w)J d (w) − i 1 + k (z − w)2 k(z − w) Analogously, the second part of the second term becomes ! (fdca (gJ g−1 )d 2 ĥ − 2πifdca δ2 (z − w)J d (w) − i 1 − k (z − w)2 + c fde (f ae : (gJ g−1 )g J d : (w, w) + fgad : (gJ g−1 )e J g : (w, w)) . (B.10) k(z − w) g Combining the two parts of the second term we find 2πifdac δ2 (z − w)J d (w) + ifdac (gJ g−1 )d (z − w)2
d : (gJ g −1 )g J e : (w, w)+ : J e (gJ g −1 )g : (w, w) fdac feg . (B.11) + 2k(z − w) Comparing with the first term, we see that the contact term as well as the double pole term cancel exactly while the single pole term vanishes using the Maurer-Cartan equation itself, normal ordered as in our proposal. We note that the operator product expansion between J a and (gJ g−1 )b obtained this way matches the operator product expansion obtained in (3.15) at the Wess-Zumino-Witten point. We will also need the OPE of (gJ g−1 )a (z, z) with itself at the Wess-Zumino-Witten point: ifcab (z − w)J c (w) ifcab (gJ g−1 )c (w, w) kκab + − 2 . (z − w)2 (z − w)2 z−w (B.12) The coefficients can be argued for by analyzing the contact and most singular terms in the OPE of the current gJ g−1 with the Maurer-Cartan equation. (gJ g−1 )a (z, z)(gJ g−1 )b (w, w) ∼ 30 Operator product expansions of currents with marginal operator Since the computations are fairly similar, let us consider the more complicated OPE of the current (gJ g−1 )a (w, w) with the marginal operator Φ. The first part of the computation involves the OPE (gJ g−1 )b (w, w) lim J c (y)(gJ g−1 )c (x, x) ∼ :y→x:   ifdbc (gJ g−1 )d (y, y) (2) (gJ g−1 )c (x, x) lim 2πkδ (w − y) + :y→x w−y  b (w − x)J d (x) b (gJ g −1 )d (x, x)  ifcd 2ifcd kδcb c + − +J (y) (w − x)2 (w − x)2 w−x b 2ifcd kJ b (x) − : J c (gJ g−1 )d : (x, x) (w − x)2 (w − x) b ifcd if b (w − x) c d : (gJ g−1 )c (gJ g−1 )d : (x, x) . (B.13) : J J : (x) − + cd (w − x)2 w−x ∼ 2πkδ(2) (w − x)(gJ g−1 )b (x, x) + Similarly, exchanging the order of J and gJ g−1 , we find an identical OPE to the above one except that the terms in the last line have opposite sign. Combining these two, we therefore find that (gJ g−1 )b (w, w)Φ(x, x) ∼ 2πkδ(2) (w − x)(gJ g−1 )b (x, x) + − kJ b (x) (w − x)2 b 2ifcd (: (gJ g−1 )d J c : + : J c (gJ g−1 )d :)(x, x) . (B.14) w−x Now, the last term can be rewritten using the Maurer-Cartan identity and we obtain (gJ g−1 )b (w, w)Φ(x, x) ∼ 2πkδ(2) (w − x)(gJ g−1 )b (x, x) + + kJ b (x) (w − x)2 2k (∂J b + ∂(gJ g−1 )b )(x, x) . (B.15) w−x Integrating over the location of the marginal operator and using the identities Z J b (x) ∂J b = − d2 x w−x (w − x)2 Z ∂(gJ g−1 )b = 2π(gJ g−1 )a (w, w) , d2 x w−x Z d2 x we find the contraction Z Z J b (x) (gJ g−1 )b (w, w) d2 xΦ(x, x) = 6πk(gJ g−1 )b (w, w) − k d2 x . (w − x)2 31 (B.16) (B.17) (B.18) For the OPE of J a (w) with the marginal operator, it turns out that both orderings lead to the same answer, so we only exhibit the following OPE:   ifdbc J d (y) kκbc + J b (w) lim : J c (y)(gJ g−1 )c (x, x) : ∼ lim (gJ g−1 )c (x, x) :y→x: :y→x: (w − y)2 w−y  b (gJ g −1 )d (x, x)  ifcd c (2) b +J (y) 2πkδ (w − x)δc + w−y k(gJ g−1 )b (x, x) + 2πkδ(2) (w − x)J b (x) (w − y)2 b ifcd + (: J d (gJ g−1 )c : + : J c (gJ g−1 )d :)(x, x) w−y k(gJ g−1 )b (x, x) ∼ 2πkδ(2) (w − x)J b (x) + . (B.19) (w − y)2 ∼ C Useful integrals We tabulate a few useful integrals that have been used throughout the article (see e.g. [29]): Z d2 x = −2π log |z − w|2 (C.1) (x − w)(x − z) Z 1 d2 x = 2π (C.2) (x − w)2 (x − z) z−w Z d2 x 1 = −2π (C.3) 2 (x − w)(x − z) z−w Z d2 x = 4π 2 δ(2) (z − w) (C.4) (x − w)2 (x − z)2 Z d2 x z−w = −2π . (C.5) (z − x)2 (w − x) (z − w)2 References [1] G. Parisi, N. Sourlas, “Self avoiding walk and supersymmetry,” J. Phys. Lett. 41 (1980) 403. [2] K. B. Efetov, “Supersymmetry and theory of disordered metals,” Adv. Phys. 32 (1983) 53. [3] S. Sethi, “Supermanifolds, rigid manifolds and mirror symmetry,” Nucl. Phys. B 430, 31 (1994) [arXiv:hep-th/9404186]. [4] M. R. Zirnbauer, “Conformal field theory of the integer quantum Hall plateau transition,” arXiv:hep-th/9905054. [5] S. Guruswamy, A. LeClair and A. W. W. Ludwig, “gl(N—N) super-current algebras for disordered Dirac fermions in two dimensions,” Nucl. Phys. B 583, 475 (2000) [arXiv:cond-mat/9909143]. 32 [6] M. Bershadsky, S. Zhukov and A. Vaintrob, “PSL(n—n) sigma model as a conformal field theory,” Nucl. Phys. B 559 (1999) 205 [arXiv:hep-th/9902180]. [7] A. Babichenko, “Conformal invariance and quantum integrability of sigma models on symmetric superspaces,” Phys. Lett. B 648, 254 (2007) [arXiv:hep-th/0611214]. [8] D. Kagan and C. A. S. Young, “Conformal Sigma-Models on Supercoset Targets,” Nucl. Phys. B 745 (2006) 109 [arXiv:hep-th/0512250]. [9] N. Read and H. Saleur, “Exact spectra of conformal supersymmetric nonlinear sigma models in two dimensions,” Nucl. Phys. B 613, 409 (2001) [arXiv:hep-th/0106124]. [10] G. Gotz, T. Quella and V. Schomerus, “The WZNW model on PSU(1,1;2),” JHEP 0703, 003 (2007) [arXiv:hep-th/0610070]. [11] T. Quella, V. Schomerus and T. Creutzig, “Boundary Spectra in Superspace SigmaModels,” arXiv:0712.3549 [hep-th]. [12] N. Berkovits, C. Vafa and E. Witten, “Conformal field theory of AdS background with Ramond-Ramond flux,” JHEP 9903 (1999) 018 [arXiv:hep-th/9902098]. [13] J. D. Brown and M. Henneaux, “Central Charges in the Canonical Realization of Asymptotic Symmetries: An Example from Three-Dimensional Gravity,” Commun. Math. Phys. 104, 207 (1986). [14] A. Giveon, D. Kutasov and N. Seiberg, “Comments on string theory on AdS(3),” Adv. Theor. Math. Phys. 2, 733 (1998) [arXiv:hep-th/9806194]. [15] D. Kutasov and N. Seiberg, “More comments on string theory on AdS(3),” JHEP 9904 (1999) 008 [arXiv:hep-th/9903219]. [16] J. de Boer, H. Ooguri, H. Robins and J. Tannenhauser, “String theory on AdS(3),” JHEP 9812 (1998) 026 [arXiv:hep-th/9812046]. [17] I. Bena, J. Polchinski and R. Roiban, “Hidden symmetries of the AdS(5) x S**5 superstring,” Phys. Rev. D 69, 046002 (2004) [arXiv:hep-th/0305116]. [18] N. Gromov, V. Kazakov and P. Vieira, “Integrability for the Full Spectrum of Planar AdS/CFT,” arXiv:0901.3753 [hep-th]. [19] K. G. Wilson, “Nonlagrangian models of current algebra,” Phys. Rev. 179 (1969) 1499. [20] M. Luscher, “Quantum Nonlocal Charges And Absence Of Particle Production In The Two-Dimensional Nonlinear Sigma Model,” Nucl. Phys. B 135, 1 (1978). [21] D. Bernard, “Hidden Yangians in 2-D massive current algebras,” Commun. Math. Phys. 137, 191 (1991). [22] D. Bernard, “Quantum arXiv:hep-th/9109058. Symmetries In 2-D Massive Field Theories,” [23] P. Di Francesco, P. Mathieu, D. Senechal, “Conformal Field Theory,” Springer, 1997. [24] L. Rozansky and H. Saleur, “Quantum field theory for the multivariable AlexanderConway polynomial,” Nucl. Phys. B 376 (1992) 461. 33 [25] N. Berkovits, “Quantization of the type II superstring in a curved six-dimensional background,” Nucl. Phys. B 565, 333 (2000) [arXiv:hep-th/9908041]. [26] R. R. Metsaev and A. A. Tseytlin, “Type IIB superstring action in AdS(5) x S(5) background,” Nucl. Phys. B 533 (1998) 109 [arXiv:hep-th/9805028]. [27] N. Berkovits, M. Bershadsky, T. Hauer, S. Zhukov and B. Zwiebach, “Superstring theory on AdS(2) x S(2) as a coset supermanifold,” Nucl. Phys. B 567 (2000) 61 [arXiv:hep-th/9907200]. [28] N. Berkovits and C. Vafa, Nucl. Phys. B 433 (1995) 123 [arXiv:hep-th/9407190]. [29] D. Kutasov, “Geometry on the space of conformal field theories and contact terms,” Phys. Lett. B 220, 153 (1989). 34