Modular Invariance of Characters for Affine Lie Algebras at Subprincipal Admissible Levels

Victor G. Kac, Minoru Wakimoto · 2025

Read Modular Invariance of Characters for Affine Lie Algebras at Subprincipal Admissible Levels by Victor G. Kac, Minoru Wakimoto (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-17159.txt.

Introduction

Victor G Kac and Minoru Wakimoto present a 2025 arXiv paper in pure mathematics. It studies characters of modules over an affine Lie algebra. The work sits at a special class of levels called subprincipal admissible levels. The authors prove that the span of these normalized characters stays fixed under the modular group. They also give the explicit formula for how the characters transform. Such results anchor representation theory and its links to conformal field theory.

Catalog record

CREATORVictor G. Kac, Minoru Wakimoto
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE58 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-17159

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Modular Invariance of Characters for Affine Lie Algebras at Subprincipal Admissible Levels Victor G. Kac, Minoru Wakimoto Abstract We prove that the span of normalized characters of subprincipal admissible modules over an affine Lie algebra of subprincipal admissible level k k is S ​ L 2 ​ ( ℤ ) SL_{2}(\mathbb{Z}) -invariant and find the explicit modular transformation formula. 0 Introduction Let \mathfrak{g} be an affine Lie algebra over \mathbb{C} , and let L(\Lambda)=L(\Lambda,\mathfrak{g}) denote the irreducible highest weight module over \mathfrak{g} with highest weight \Lambda . Recall [ KW88 ] that the weight \Lambda and the corresponding \mathfrak{g} -module L(\Lambda) are called admissible if the following two properties hold: \displaystyle\left\langle\Lambda+\rho,\alpha\right\rangle\not\in\mathbb{Z}_{\leq 0}\quad\text{for all $\alpha\in\Delta_{+}^{\mathrm{re}\,\vee}$}, (0.1) where \Delta_{+}^{\mathrm{re}\,\vee} is the set of positive real coroots of \mathfrak{g} ; \displaystyle\mathbb{Q}\left\{\alpha\in\Delta_{+}^{\mathrm{re}\,\vee}\mid\left\langle\Lambda+\rho,\alpha\right\rangle\in\mathbb{Z}\right\}\supset\Delta_{+}^{\mathrm{re}\,\vee}. (0.2) These \mathfrak{g} -modules are important since we proved in [ KW88 ] the following formula for their characters \mathrm{tr}_{L(\Lambda)}e^{h},\ h\in\mathfrak{h} : \displaystyle\prod_{\alpha\in\Delta_{+}}\left(1-e^{-\alpha(h)}\right)^{\mathrm{mult}\ \alpha}\mathrm{tr}_{L(\Lambda)}e^{h}=\sum_{w\in W^{\Lambda}}\varepsilon(w)e^{(w.\Lambda)(h)}. (0.3) Here \mathfrak{h} is the Cartan subalgebra of \mathfrak{g} , W^{\Lambda} is the subgroup of the Weyl group W of \mathfrak{g} , generated by reflection in \alpha\in\Delta_{+}^{\mathrm{re}\,\vee} , integral with respect to \Lambda , i.e. such that \left\langle\Lambda+\rho,\alpha\right\rangle\in\mathbb{Z} , \displaystyle w.\Lambda=w(\Lambda+\rho)-\rho,\quad w\in W, (0.4) is the ‘shifted’ action of W , and \displaystyle\varepsilon(w)={\det}_{\mathfrak{h}}w. Using this formula, we showed that the characters, normalized by a power of q=e^{-\delta} factor, are meromorphic modular functions in the domain \left\{h\in\mathfrak{h}\mid\mathrm{Re}\ \delta(h)>0\right\} . We also conjectured that these are all irreducible highest weight \mathfrak{g} -modules L(\Lambda) with this property. From the viewpoint of the theory of vertex algebras the most interesting case is that of a non-twisted \mathfrak{g} , associated to a simple Lie algebra \overline{\mathfrak{g}} , and \Lambda=k\Lambda_{0} , where k\in\mathbb{C} , when L(k\Lambda_{0}) admits a structure of simple affine vertex algebra V_{k}(\overline{\mathfrak{g}}) of level k . If k\Lambda_{0} is an admissible weight, k is called an admissible level . It is easy to deduce from the classification of admissible weights in [ KW89 ] that k is an admissible level only when it is a rational number, and there are only the following two possibilities [ KW08 ] : (principal k ) k=-h^{\vee}+\frac{p}{u} , where p and u are coprime positive integers, u is coprime with the lacety r^{\vee} of \overline{\mathfrak{g}} , and p\geq h^{\vee} , the dual Coxeter number of \mathfrak{g} ; (subprincipal k ) k=-h^{\vee}+\frac{p}{u} , where p and u are coprime positve integers, u is divisible by r^{\vee}>1 , and p\geq h , the Coxeter number of \mathfrak{g} . The above mentioned conjecture leads to Conjecture 0.1. The series \mathrm{tr}_{V_{k}(\mathfrak{g})}e^{2\pi i\left(L_{0}-\frac{C_{k}}{24}\right)\tau} converges to a holomorphic modular function in the upper half-plane Im \tau>0 if and only if k is either principal or subprincipal admissible level. Here C_{k}=\frac{k\dim\overline{\mathfrak{g}}}{k+h^{\vee}} . Note that the ‘if’ part holds due to [ KW25 , formula (52)] if k is a principal admissible level, and due to the following similar formula for k=-h^{\vee}+\frac{p}{u} subprincipal admissible level: \displaystyle\mathrm{tr}_{L(\Lambda)}q^{L_{0}-\frac{C_{k}}{24}}=\frac{\sum_{\gamma\in\overline{Q}}d(\overline{\Lambda}+p\gamma)q^{\frac{u}{2p}\left|\overline{\Lambda}+\overline{\rho}+p\gamma\right|^{2}}}{\eta(\tau)^{\dim\overline{\mathfrak{g}}}}, (0.5) if \Lambda is a subprincipal admissible weight of level k whose restriction to \overline{\mathfrak{h}} is dominant integral for \overline{\mathfrak{g}} where \displaystyle d(\beta)=\prod_{\alpha\in\overline{\Delta}_{+}}\frac{\left(\beta+\overline{\rho}\mid\alpha\right)}{\left(\overline{\rho}\mid\alpha\right)},\quad\beta\in\overline{\mathfrak{h}}^{*}. In particular, (0.5) holds for \Lambda=k\Lambda_{0} . Note that (0.5) is a modular function since it is a specialization of the normalized character of L(\Lambda) . The most important among the admissible weights \Lambda of principal admissible level k is the set P_{+,\mathrm{pr}}^{k} of principal admissible weights . Such \Lambda are defined by the property that the set of real coroots, integral with respect to \Lambda , is isometric to that with respect to k\Lambda_{0} , or, equivalently, its set of simple roots is isometric to the set \displaystyle S_{(u)}^{\vee}=\left\{uK-\theta^{\vee},\alpha_{1}^{\vee},\cdots,\alpha_{l}^{\vee}\right\}, (0.6) where K is the canonical central element of \mathfrak{g} , \theta is the highest root of \overline{\mathfrak{g}} , and \alpha_{1}^{\vee},\cdots,\alpha_{l}^{\vee} are simple coroots of \overline{\mathfrak{g}} . (If \mathfrak{g} is of type A_{\ell}^{(1)} , then P_{+,\mathrm{pr}}^{k} is the set of all admissible weights, but not for other types.) It is shown in [ KW89 ] that P_{+,\mathrm{pr}}^{k} consists of weights of the form \displaystyle\Lambda=\left(t_{\beta}\overline{y}\right).\left(\Lambda^{0}+\frac{1-u}{u}p\Lambda_{0}\right), (0.7) where \beta\in\overline{P} and y\in\overline{W} are such that \displaystyle t_{\beta}\overline{y}\left(S_{(u)}^{\vee}\right)\subset\Delta_{+}^{\mathrm{re}\,\vee}, and \Lambda^{0} is a dominant integral weight for \mathfrak{g} of level p-h^{\vee} . The importance of principal admissible weights is due to the fact, proved in [ KW89 ] , that the span of normalized characters of \mathfrak{g} -modules L(\Lambda) with \Lambda\in P_{+,\mathrm{pr}}^{k} is SL_{2}(\mathbb{Z}) -invariant, and the fact, proved in [ Ara16 ] that all \mathfrak{g} -modules L(\Lambda) , which descend to V_{k}(\mathfrak{g}) are of the form (0.7), with \beta=0,\overline{y}=e , i.e. \Lambda is quasidominant. The set of admissible weights of subprincipal admissible level k , denoted by P_{+}^{k} in the paper, is defined in a similar way. A weight of such level k is called subprincipal admissible if the set of coroots of \mathfrak{g} , which are integral with respect to \Lambda , is isometric to that for k\Lambda_{0} , i.e. to the set \displaystyle S_{(u)}^{\#\vee}=\left\{uK-\theta_{s}^{\vee},\alpha_{1}^{\vee},\cdots,\alpha_{l}^{\vee}\right\}, (0.8) where \theta_{s} is the highest short root of \overline{\mathfrak{g}} (cf. (0.6)). In this case the set of coroots of \mathfrak{g} , integral with respect to \Lambda , is isometric to the set of coroots of the twisted affine Lie algebra \mathfrak{g}^{\#}=\mathfrak{g}\left(A^{\#}\right) , whose Cartan matrix A^{\#} is A^{\top^{\prime}} , where A is the Cartan matrix of \mathfrak{g} , \top stands for transpose and ′ stands for taking the adjacent Cartan matrix [ Kac90 , §13.9] . Recall that A^{\prime}=A for all affine Cartan matrices, except that ′ switches A_{2\ell-1}^{(2)} with D_{\ell+1}^{(2)} . The description of the set of subprincipal admissible weights is similar to that of principal ones, see Theorem 2.1. There is an important difference, however. All k\in\mathbb{Z}_{\geq 0} are principal admissible levels and the corresponding principal admissible weights are just level k dominant integral weights, hence the span of the normalized characters of \mathfrak{g} -modules L(\Lambda) with principal admissible \Lambda of level k\in\mathbb{Z}_{\geq 0} is SL_{2}(\mathbb{Z}) -invariant [ Kac90 ] . This is not the case for subprincipal admissible levels, which cannot be integer numbers. It is related to the fact that the numerators of characters of subprincipal admissible L(\Lambda) are expressed in terms of the Weyl group W^{\#} of the twisted affine Lie algebra \mathfrak{g}^{\#} . But the span of normalized characters of modules L(\Lambda,\mathfrak{g}^{\#}) with \Lambda dominant integral is invariant only with respect to the action of the subgroup \Gamma_{0}(r^{\vee}) of SL_{2}(\mathbb{Z}) [ Kac90 , Theorem 13.9] . It is quite surprising that nevertheless the span of the normalized characters of subprincipal admissible \mathfrak{g} -modules L(\Lambda) , \Lambda\in P_{+}^{k} , is SL_{2}(\mathbb{Z}) -invariant. This is our main Theorem 5.1. Due to the character formula (3.6), it suffices to prove SL_{2}(\mathbb{Z}) -invariance of the span of the numerators A_{\Lambda+\rho} , since \mathbb{C}R_{\mathfrak{g}} , where R_{\mathfrak{g}} is the Weyl denominator, is SL_{2}(\mathbb{Z}) -invariant. The SL_{2}(\mathbb{Z}) -invariance of the span of \left\{A_{\Lambda+\rho}\right\}_{\Lambda\in P_{+}^{k}} is proved, using their expression in terms of Jacobi forms in Proposition 5.1. Note that Theorem 5.1 implies the same modular transformation formulas for quantum Hamiltonian reduction H_{f}(\Lambda) for subprincipal admissible \Lambda , as those, given by [ KW25 , Theorem 4] for the principal admissible ones. In the last section of the paper we show that the simple vertex algebra V_{k}(\overline{\mathfrak{g}}) of subprincipal admissible level k is modular invariant in the sense of [ KW25

Read the complete text