Introduction and summary
Superconformal indices play a central role in the study of supersymmetric quantum field theories. They encode infinitely many protected quantities and provide powerful tools for studying dualities, renormalization-group flows, moduli spaces, and, through the AdS/CFT correspondence, black-hole entropy. In particular, they furnish stringent tests of infrared (IR) dualities: distinct ultraviolet descriptions that flow to the same IR fixed point are expected to yield identical superconformal indices. Such agreement provides strong evidence for a proposed duality by matching an infinite collection of protected quantities. When the indices admit contour-integral or infinite sum representations, these equalities also yield highly nontrivial identities of independent mathematical interest.
In this paper, we investigate a weaker notion of index matching: indices of theories that need not be IR dual can coincide after simple but nontrivial redefinitions and specializations of fugacities. Focusing on the Schur indices of four-dimensional N = 2 \mathcal{N}=2 N = 2 superconformal field theories (SCFTs), we systematically construct and prove infinitely many such identities. Through the SCFT/VOA correspondence, these results translate into identities among specialized characters of vertex operator algebras (VOAs).
Our primary examples are Argyres–Douglas (AD) theories with class S \mathcal{S} S realizations involving an irregular puncture and, in some cases, an additional regular puncture [1 ] . In suitable cases, the associated VOAs are affine vertex algebras or W W W -algebras at admissible levels [2 ] . These VOAs have a well-developed mathematical structure, including systematic constructions via Drinfeld–Sokolov (DS) reduction and uniform formulas for their characters [3 , 4 , 5 ] . They therefore provide a broad, mathematically tractable setting in which to identify and establish such relations.
A precursor to our construction is the identity I ( A 1 , A 6 n ) ( q ) = I ( A 1 , D 2 n + 1 ) ( q 3 ; q 1 / 2 ) \mathcal{I}_{(A_{1},A_{6n})}(q)=\mathcal{I}_{(A_{1},D_{2n+1})}\left(q^{3};q^{1/2}\right) I ( A 1 , A 6 n ) ( q ) = I ( A 1 , D 2 n + 1 ) ( q 3 ; q 1/2 ) for n ≥ 1 n\ge1 n ≥ 1 , established in [6 ] . The associated VOAs are Virasoro and affine vertex algebras, respectively. This identity was itself motivated by a relation between the indices of AD theories and N = 4 \mathcal{N}=4 N = 4 super-Yang–Mills theories [7 , 8 ] . Our construction extends these results to a broad range of SCFTs and VOAs.
For illustration, we present a few simple, representative examples:
I ( A N − 1 , A ( p − 1 ) ( N + 1 ) ) ( q ) = I D p ( S U ( N ) ) ( q N + 1 ; { q s − ( N + 1 ) / 2 } s = 1 N ) , gcd ( p , N ) = 1 , \mathcal{I}_{(A_{N-1},A_{(p-1)(N+1)})}(q)=\mathcal{I}_{D_{p}(SU(N))}\left(q^{N+1};\left\{q^{s-(N+1)/2}\right\}_{s=1}^{N}\right),\qquad\gcd(p,N)=1, I ( A N − 1 , A ( p − 1 ) ( N + 1 ) ) ( q ) = I D p ( S U ( N )) ( q N + 1 ; { q s − ( N + 1 ) /2 } s = 1 N ) , g cd( p , N ) = 1 ,
I ( A 1 , A 4 p − 3 ) ( q ; w ) = I ( A 1 , D 2 p ) ( q 2 ; y = q 1 / 2 , w ) , p ≥ 2 , \mathcal{I}_{(A_{1},A_{4p-3})}(q;w)=\mathcal{I}_{(A_{1},D_{2p})}\left(q^{2};y=q^{1/2},w\right),\qquad p\ge2, I ( A 1 , A 4 p − 3 ) ( q ; w ) = I ( A 1 , D 2 p ) ( q 2 ; y = q 1/2 , w ) , p ≥ 2 ,
I ( A 2 , A 4 ) ( q ) = I ( A 1 , A 5 ) ( q 2 ; 1 ) . \mathcal{I}_{(A_{2},A_{4})}(q)=\mathcal{I}_{(A_{1},A_{5})}(q^{2};1). I ( A 2 , A 4 ) ( q ) = I ( A 1 , A 5 ) ( q 2 ; 1 ) .
These examples illustrate different features of our construction. The first generalizes the identity of [6 ] to higher rank, relating affine vertex algebras to principal W W W -algebras. The second retains nontrivial flavor dependence on both sides, whereas the third relates unflavored indices. Combining the second and third examples further shows that these identities can be composed into chains connecting more than two SCFTs.
The index identities also imply relations among the anomaly coefficients of the SCFTs, including their a a a and c c c central charges and flavor levels. These relations follow by comparing the high-temperature asymptotics of the indices. In particular, the effective central charges obey a universal rescaling relation, with the rescaling fixed by the power of q q q appearing in the index identity.
These exact index identities also suggest structural connections between the underlying VOAs. More precisely, they imply graded vector space isomorphisms (GVISOs), with gradings defined by suitable combinations of conformal weights and flavor charges.1 This was first studied in [7 ] . Such isomorphisms are weaker than VOA isomorphisms, which must also preserve operator product expansions (OPEs) and normally ordered products. We seek a direct realization of these GVISOs in terms of VOA states and their relations. For a particular pair of VOAs, we explicitly construct the map and prove that it defines a GVISO. We also examine the obstacles to a general construction. For the universal affine and principal W W W -algebras considered here, the correspondence follows from a one-to-one matching of modes. Passing to the simple quotients, however, requires compatibility with null relations, making the construction substantially more subtle. Nevertheless, we explicitly demonstrate the matching of the lowest-degree null states on the two sides.
We further extend our discussion to non-vacuum VOA modules and their interpretation in terms of surface-defect insertions in the SCFTs. Analogous character and index identities also arise in this setting. The problem is more intricate than for vacuum modules: VOA modules generally have complicated structures, simple character formulas are not always available, and the numbers of modules on the two sides can differ substantially. In an example involving three AD theories, we analyze these issues in detail and establish character identities relating selected modules of the corresponding VOAs. We conjecture a broader extension to suitable non-vacuum modules and outline its form for affine and principal W W W -algebras at boundary admissible levels. In this setting, we express the module characters in plethystic exponential (PE) form, obtaining formulas useful for studying SCFTs with surface defects. A detailed treatment of the general extension, including its proof, is left to future work.
Finally, we extend the discussion beyond the SCFT/VOA correspondence. We establish a GVISO between a level-one affine VOA and the Ising VOA, together with explicit character identities that extend to non-vacuum modules. A free-field realization in terms of Majorana fermions provides a direct explanation of this correspondence. Generalizing further, this example leads us to character relations between pairs of affine VOAs whose ranks and levels are exchanged. We refer to these relations as spectral level-rank dualities , emphasizing the correspondence between graded state spaces under an exchange of rank and level. This terminology distinguishes them from the conventional topological level-rank dualities in Wess–Zumino–Witten (WZW) models, or Chern–Simons (CS) theories via the CS/WZW correspondence.
The paper is organized as follows. § 2 reviews the SCFT/VOA correspondence and the high-temperature asymptotics of Schur indices, and explains how index identities constrain anomaly coefficients. § 3 presents the basic family of index identities relating ( A N − 1 , A ( p − 1 ) ( N + 1 ) ) (A_{N-1}, A_{(p-1)(N+1)}) ( A N − 1 , A ( p − 1 ) ( N + 1 ) ) and D p ( S U ( N ) ) D_p(SU(N)) D p ( S U ( N )) . § 4 further extends this construction to identities relating ( A ( p − 1 ) ( h + 1 ) , G ) (A_{(p-1)(h+1)}, G) ( A ( p − 1 ) ( h + 1 ) , G ) and D p ( G ) D_p(G) D p ( G ) . § 5 develops the general construction and establishes further families of index identities for a broader class of AD theories. § 6 investigates explicit GVISOs between the underlying VOAs, going beyond character matching. § 7 explores extensions to non-vacuum modules. § 8 extends the analysis beyond the SCFT/VOA correspondence. § 9 concludes with a discussion of open questions. The appendices collect further technical details.
SCFT/VOA correspondence and index identity
Review of the SCFT/VOA correspondence
Every 4d N = 2 \mathcal{N}=2 N = 2 SCFT can be associated with a vertex operator algebra or chiral algebra [9 ]
V : 4 d N = 2 S C F T → V O A . (1) \mathbb{V}: 4\mathrm{d}\ \mathcal{N}=2\ \mathrm{SCFT} \to\mathrm{VOA}. \tag*{(1)} V : 4 d N = 2 SCFT → VOA . (1)
The SCFT/VOA correspondence captures the Schur sector of the SCFTs.
Two important entries in the SCFT/VOA dictionary are
c 2 d = − 12 c 4 d , k 2 d = − 1 2 k 4 d . (2) c_{2d}=-12c_{4d}, \qquad k_{2d}=-\frac{1}{2}k_{4d}. \tag*{(2)} c 2 d = − 12 c 4 d , k 2 d = − 2 1 k 4 d . (2)
The Schur index is defined as
I = Tr ( − 1 ) F q E − R x x J , q = e 2 π i τ , x = e 2 π i z , (3) \mathcal{I}=\operatorname{Tr}(-1)^Fq^{E-R_x}x^J,\qquad q=e^{2\pi i\tau},\qquad x=e^{2\pi iz}, \tag*{(3)} I = Tr ( − 1 ) F q E − R x x J , q = e 2 π i τ , x = e 2 π i z , (3)
where the trace is taken over all Schur operators, J J J is the flavor symmetry generator and x x x is its fugacity. A generalization to multiple flavor fugacities is straightforward.
In particular, the SCFT/VOA correspondence implies that the Schur index of a 4d N = 2 \mathcal{N}=2 N = 2 SCFT T \mathcal{T} T is identified with the vacuum (super)character of the associated VOA [9 ]
I T = χ V [ T ] . (4) \mathcal{I}_{\mathcal{T}}=\chi_{\mathcal{V}[\mathcal{T}]}. \tag*{(4)} I T = χ V [ T ] . (4)
In this paper, χ \chi χ denotes the normalized character, and χ ^ = q − c / 24 χ \widehat{\chi}=q^{-c/24}\chi χ = q − c /24 χ denotes the full character. The former has the nice feature that it starts with 1 for the unique vacuum and coincides with the index, while the latter transforms covariantly under modular transformations.
Cardy limit of the Schur index
In this subsection, we would like to make full use of the modular property in order to study the high temperature or Cardy limit of the Schur index. See [10 , 11 ] for previous discussions.
The VOAs associated with 4d SCFTs are expected to be quasi-lisse in general. For a quasi-lisse VOA, its full vacuum character satisfies an MLDE [10 , 12 ] . The solution space of MLDE gives a vector valued modular form, which includes the vacuum and non-vacuum module characters. Note that the modular vector may also need logarithmic or pseudotrace solutions in addition to ordinary module characters.
As a result, under the S S S transformation, the character vectors formed by the solutions of MLDEs transform as 2
χ ^ ( − 1 / τ ) = S χ ^ ( τ ) . (5) \widehat{\chi}(-1/\tau)=S\widehat{\chi}(\tau). \tag*{(5)} χ ( − 1/ τ ) = S χ ( τ ) . (5)
Taking into account possible logarithmic solutions, each character can be chosen to admit an expansion of the following form 3
χ ^ i ( τ ) = q h i − c / 24 A i ( − log q / ( 2 π ) ) d i [ 1 + O ( q ϵ ) ] , d i ∈ N , A i ≠ 0 , ϵ > 0. (6) \widehat{\chi}_i(\tau)=q^{h_i-c/24}A_i\left(-\log q/(2\pi)\right)^{d_i}\left[1+O\left(q^\epsilon\right)\right],\qquad d_i\in\mathbb{N},\qquad A_i\ne0,\qquad\epsilon>0. \tag*{(6)} χ i ( τ ) = q h i − c /24 A i ( − log q / ( 2 π ) ) d i [ 1 + O ( q ϵ ) ] , d i ∈ N , A i = 0 , ϵ > 0. (6)
Taking the limit τ → + ∞ i \tau\to+\infty i τ → + ∞ i , we learn from the S S S transformation (5) that
χ ^ 0 ( − 1 / τ ) = ∑ i S 0 i e 2 π i τ ( h i − c / 24 ) [ A i ( − log q / ( 2 π ) ) d i + O ( e 2 π i ϵ τ ) ] . (7) \widehat{\chi}_0(-1/\tau)=\sum_i S_{0i}e^{2\pi i\tau(h_i-c/24)}\left[A_i\left(-\log q/(2\pi)\right)^{d_i}+O\left(e^{2\pi i\epsilon\tau}\right)\right]. \tag*{(7)} χ 0 ( − 1/ τ ) = i ∑ S 0 i e 2 π i τ ( h i − c /24 ) [ A i ( − log q / ( 2 π ) ) d i + O ( e 2 π i ϵ τ ) ] . (7)
The dominant contribution comes from the one with the smallest h i h_i h i and largest exponent d i d_i d i 4
h min = min { h i ∣ S 0 i ≠ 0 } , d max = max { d i ∣ h i = h min , S 0 i ≠ 0 } . (8) h_{\min}=\min\{h_i\mid S_{0i}\ne0\},\qquad d_{\max}=\max\{d_i\mid h_i=h_{\min},S_{0i}\ne0\}. \tag*{(8)} h m i n = min { h i ∣ S 0 i = 0 } , d m a x = max { d i ∣ h i = h m i n , S 0 i = 0 } . (8)
We can then rewrite (7) and get the following asymptotic growth in the limit τ → i 0 + \tau\to i0^{+} τ → i 0 +
χ ( τ ) = e 2 π i τ c 2 d / 24 e − 2 π i ( h min − c 2 d / 24 ) / τ ( − i τ ) − d max [ ∑ i : h i = h min A i S 0 i ( − i τ ) d max − d i + O ( e − # / Im τ ) ] (9) \chi(\tau) = e^{2\pi i\tau c_{2d}/24}e^{-2\pi i(h_{\min}-c_{2d}/24)/\tau}(-i\tau)^{-d_{\max}}\left[\sum_{i:h_i=h_{\min}} A_iS_{0i}(-i\tau)^{d_{\max}-d_i}+O(e^{-\#/ \operatorname{Im}\tau})\right] \tag*{(9)} χ ( τ ) = e 2 π i τ c 2 d /24 e − 2 π i ( h m i n − c 2 d /24 ) / τ ( − i τ ) − d m a x [ i : h i = h m i n ∑ A i S 0 i ( − i τ ) d m a x − d i + O ( e − #/ Im τ ) ] (9)
where we omit the subscript 0 for the vacuum.
The universal expansion of the index is given by
log I = log χ = π i 12 τ c e f f + γ log ( − i τ ) + log S + π i τ C 12 + ⋯ , (10) \log\mathcal{I} = \log\chi= \frac{\pi i}{12\tau}c^{\mathrm{eff}}+\gamma\log(-i\tau)+\log S+\frac{\pi i\tau C}{12}+\cdots, \tag*{(10)} log I = log χ = 12 τ π i c eff + γ log ( − i τ ) + log S + 12 π i τ C + ⋯ , (10)
where the dots represent the exponentially small terms. By comparison, we find
c e f f = c 2 d − 24 h min , γ = − d max , S ( τ ) = ∑ i : h i = h min A i S 0 i ( − i τ ) d max − d i , C = c 2 d , (11) c^{\mathrm{eff}}=c_{2d}-24h_{\min},\qquad\gamma=-d_{\max},\qquad S(\tau)=\sum_{i:h_i=h_{\min}}A_iS_{0i}(-i\tau)^{d_{\max}-d_i},\qquad C=c_{2d}, \tag*{(11)} c eff = c 2 d − 24 h m i n , γ = − d m a x , S ( τ ) = i : h i = h m i n ∑ A i S 0 i ( − i τ ) d m a x − d i , C = c 2 d , (11)
where S S S is a finite polynomial of τ \tau τ , and only the minimal-exponent sector belongs in this polynomial.5
The logarithmic term γ log ( − i τ ) \gamma\log(-i\tau) γ log ( − i τ ) is absent if the MLDE has no logarithmic solution and the T T T matrix is diagonalizable, which holds for rational VOAs. In such a case d i = d max = 0 d_i=d_{\max}=0 d i = d m a x = 0 , and S S S is just a constant, which is related to the modular S S S -matrix.6
Note that the above equations apply to the asymptotic growth of the unflavored index in the high temperature limit τ → 0 \tau\to0 τ → 0 . If we also turn on the flavor fugacity and scale its additive coordinate as z = α τ → 0 z=\alpha\tau\to0 z = α τ → 0 , C C C also receives a contribution from the flavor symmetry, which is proportional to the corresponding level. In this case, we also need to include the flavor anomaly contribution exp ( π i k J z 2 / τ ) \exp(\pi i k_J z^2/\tau) exp ( π i k J z 2 / τ ) to the S S S transformation in (5) . Then the general relation is
C = c 2 d − 12 k J α 2 , (12) C=c_{2d}-12k_J\alpha^2, \tag*{(12)} C = c 2 d − 12 k J α 2 , (12)
where k J k_J k J is the level for the effective U(1) symmetry, namely J ( z ) J ( w ) ∼ k J / ( z − w ) 2 J(z)J(w)\sim k_J/(z-w)^2 J ( z ) J ( w ) ∼ k J / ( z − w ) 2 . For several currents we need to replace k J α 2 k_J\alpha^2 k J α 2 by their level quadratic form.
In principle, the effective central charge could also be corrected by the flavor contribution, but the extra flavor contribution turns out to be absent for the specializations considered here. The effective central charge is also related to the 4d central charges [10 ] 7
c e f f = 48 ( c − a ) . (13) c^{\mathrm{eff}}=48(c-a). \tag*{(13)} c eff = 48 ( c − a ) . (13)
This makes the effective central charge particularly useful.
AD theory and associated VOA
Our main examples of SCFTs are Argyres–Douglas (AD) theories [13 ] . We focus on theories admitting a class S S S realization by compactifying six-dimensional N = ( 2 , 0 ) \mathcal{N}=(2,0) N = ( 2 , 0 ) SCFTs on a sphere with one irregular puncture and, possibly, one regular puncture [1 ] . This construction can be extended by introducing outer-automorphism twists [14 ] . We mainly focus on the untwisted case, in which G G G is of ADE type. We denote its complexified Lie algebra by g \mathfrak{g} g and its Coxeter and dual Coxeter numbers by h G h_G h G and h G ∨ h_G^\vee h G ∨ , respectively; for simply-laced G G G , these coincide.
One family of interest is D p ( G ) D_p(G) D p ( G ) , with p ≥ 2 p \ge2 p ≥ 2 . These theories arise from the six-dimensional N = ( 2 , 0 ) \mathcal{N}=(2,0) N = ( 2 , 0 ) theory of type G G G compactified on a sphere with one full regular puncture and an irregular puncture of type G h G [ p − h G ] G^{h_G}[p-h_G] G h G [ p − h G ] [1 ] . The full puncture contributes a G G G flavor symmetry, while the irregular puncture may contribute additional flavor symmetry; the combined symmetry may also exhibit enhancement. When gcd ( p , h G ) = 1 \gcd(p,h_G)=1 g cd( p , h G ) = 1 , the irregular puncture contributes no additional flavor symmetry, and the flavor algebra of D p ( G ) D_p(G) D p ( G ) is precisely g \mathfrak{g} g .
More general regular punctures can be obtained by partially closing the full puncture through nilpotent Higgsing, whenever the corresponding Higgsing is allowed. The resulting theory is denoted by D p ( G , Y ) D_p(G,Y) D p ( G , Y ) , where Y Y Y labels the Nahm nilpotent orbit in g \mathfrak{g} g associated with the regular puncture.
For G = S U ( N ) G=SU(N) G = S U ( N ) , the two commonly used notations are related by
D p ( S U ( N ) , Y ) = ( A N − 1 N [ p − N ] , Y ) . (14) D_p(SU(N),Y)=\left(A^{N}_{N-1}[p-N],Y\right). \tag*{(14)} D p ( S U ( N ) , Y ) = ( A N − 1 N [ p − N ] , Y ) . (14)
The puncture Y Y Y is labeled by a partition [ λ 1 , … , λ t ] [\lambda_1,\ldots,\lambda_t] [ λ 1 , … , λ t ] of N N N , with λ 1 ≥ ⋯ ≥ λ t > 0 \lambda_1\ge\cdots\ge\lambda_t>0 λ 1 ≥ ⋯ ≥ λ t > 0 and ∑ i = 1 t λ i = N \sum_{i=1}^{t}\lambda_i=N ∑ i = 1 t λ i = N . Equivalently, grouping equal parts, we may write Y = [ μ 1 s 1 , … , μ k s k ] Y=[\mu_1^{s_1},\ldots,\mu_k^{s_k}] Y = [ μ 1 s 1 , … , μ k s k ] subject to μ 1 > ⋯ > μ k > 0 \mu_1>\cdots>\mu_k>0 μ 1 > ⋯ > μ k > 0 , s I > 0 s_I>0 s I > 0 and ∑ I = 1 k μ I s I = N \sum_{I=1}^{k}\mu_Is_I=N ∑ I = 1 k μ I s I = N , where s I s_I s I is the multiplicity of the part μ I \mu_I μ I and t = ∑ I = 1 k s I t=\sum_{I=1}^{k}s_I t = ∑ I = 1 k s I . The continuous flavor symmetry associated with the regular puncture is conventionally denoted by 8 9
F Y = S [ U ( s 1 ) × ⋯ × U ( s k ) ] , rank F Y = ∑ I = 1 k s I − 1 = t − 1. (15) F_Y=S[U(s_1)\times\cdots\times U(s_k)],\qquad\operatorname{rank}F_Y=\sum_{I=1}^{k}s_I-1=t-1. \tag*{(15)} F Y = S [ U ( s 1 ) × ⋯ × U ( s k )] , rank F Y = I = 1 ∑ k s I − 1 = t − 1. (15)
Some frequently encountered punctures are
full puncture: Y = [ 1 N ] ⟶ S U ( N ) , (16) \text{full puncture:}\quad Y=[1^N]\ \longrightarrow\ SU(N), \tag*{(16)} full puncture: Y = [ 1 N ] ⟶ S U ( N ) , (16)
simple (subregular) puncture: Y = [ N − 1 , 1 ] ⟶ U ( 1 ) , N ≥ 3 , (17) \text{simple (subregular) puncture:}\quad Y=[N-1,1]\ \longrightarrow\ U(1),\qquad N\ge3, \tag*{(17)} simple (subregular) puncture: Y = [ N − 1 , 1 ] ⟶ U ( 1 ) , N ≥ 3 , (17)
completely closed puncture: Y = [ N ] ⟶ none , (18) \text{completely closed puncture:}\quad Y=[N]\ \longrightarrow\ \text{none}, \tag*{(18)} completely closed puncture: Y = [ N ] ⟶ none , (18)
where we also indicate the flavor symmetry contributed by the regular puncture. In the Nahm labeling, the full and completely closed punctures correspond to the trivial and principal nilpotent orbits, respectively.
Another large class of AD theories also admits a geometric engineering description in type IIB string theory on noncompact Calabi–Yau threefolds with suitable isolated quasi-homogeneous hypersurface singularities. A distinguished family constructed in this way is denoted by ( G , G ′ ) (G,G') ( G , G ′ ) , where G G G and G ′ G' G ′ label ADE types and the defining polynomial is the sum of the corresponding two ADE singularity polynomials [15 ] .
The class S \mathcal{S} S and type IIB constructions overlap and can yield the same SCFT. In particular,
( A N − 1 , A k − 1 ) = D N + k ( S U ( N ) , [ N ] ) , N , k ≥ 2 , (19) (A_{N-1},A_{k-1})=D_{N+k}(SU(N),[N]), \qquad N,k\geq2, \tag*{(19)} ( A N − 1 , A k − 1 ) = D N + k ( S U ( N ) , [ N ]) , N , k ≥ 2 , (19)
( A 1 , D k ) = D k ( S U ( 2 ) ) = D k ( S U ( 2 ) , [ 1 , 1 ] ) , k ≥ 3 , (20) (A_1,D_k)=D_k(SU(2))=D_k(SU(2),[1,1]), \qquad k\geq3, \tag*{(20)} ( A 1 , D k ) = D k ( S U ( 2 )) = D k ( S U ( 2 ) , [ 1 , 1 ]) , k ≥ 3 , (20)
D 3 ( S U ( 2 ) ) = ( A 1 , D 3 ) = ( A 1 , A 3 ) . (21) D_3(SU(2))=(A_1,D_3)=(A_1,A_3). \tag*{(21)} D 3 ( S U ( 2 )) = ( A 1 , D 3 ) = ( A 1 , A 3 ) . (21)
The class S \mathcal{S} S description is particularly useful for identifying the associated VOAs. Under the coprimality assumption gcd ( p , h G ) = 1 \gcd(p,h_G)=1 g cd( p , h G ) = 1 , the VOA associated with D p ( G ) D_p(G) D p ( G ) is identified with the simple affine vertex algebra [2 , 16 ]
V [ D p ( G ) ] = L k ( g ) , k = − h G ∨ + h G ∨ p . (22) V[D_p(G)]=L_k(\mathfrak{g}), \qquad k=-h_G^\vee+\frac{h_G^\vee}{p}. \tag*{(22)} V [ D p ( G )] = L k ( g ) , k = − h G ∨ + p h G ∨ . (22)
Here k k k is the affine level, which is boundary admissible. The theory D p ( G ) D_p(G) D p ( G ) corresponds to the special case of D p ( G , Y ) D_p(G,Y) D p ( G , Y ) in which Y Y Y is the full puncture.
More generally, for a regular puncture Y Y Y compatible with the irregular data, the associated VOA is [2 , 16 ]
V [ D p ( G , Y ) ] = W k ( g , f Y ) , k = − h G ∨ + h G ∨ p , (23) V[D_p(G,Y)]=\mathcal{W}_k(\mathfrak{g},f_Y), \qquad k=-h_G^\vee+\frac{h_G^\vee}{p}, \tag*{(23)} V [ D p ( G , Y )] = W k ( g , f Y ) , k = − h G ∨ + p h G ∨ , (23)
where f Y f_Y f Y is a nilpotent element in the Nahm orbit determined by the puncture Y Y Y , and W k ( g , f Y ) \mathcal{W}_k(\mathfrak{g},f_Y) W k ( g , f Y ) denotes the simple W W W -algebra. The construction uses quantum Drinfeld–Sokolov (DS) reduction associated with f Y f_Y f Y . Thus, partial closure of the regular puncture is reflected by DS reduction on the VOA side, with the relevant reduction understood to be nonzero. For the full puncture, f Y = 0 f_Y=0 f Y = 0 , and the description reduces to the affine vertex algebra above. We will discuss affine and W W W -algebras in more detail below. See § 5 .
The construction can also be extended to twisted class S \mathcal{S} S setups [14 ] , leading to non-simply-laced flavor symmetries and associated affine and W W W -algebras. Such extensions require specifying the outer-automorphism twist, the relevant Lie algebra, and the level normalization. We will not discuss these additional subtleties here.
A recap on known index identities
The prototypical example of an index identity is given by [7 ]
I ( A 2 , D 4 ) ( q ) = I N = 4 S U ( 2 ) S Y M ( q 3 , q 1 / 2 ) , (24) \mathcal{I}_{(A_2,D_4)}(q)=\mathcal{I}_{\mathcal{N}=4\;SU(2)\;\mathrm{SYM}}(q^3,q^{1/2}), \tag*{(24)} I ( A 2 , D 4 ) ( q ) = I N = 4 S U ( 2 ) SYM ( q 3 , q 1/2 ) , (24)
where the second argument for the S U ( 2 ) SU(2) S U ( 2 ) flavor fugacity is specialized.
More generally, there is a class of AD theories which are obtained by the conformal gauging of a set of D p i ( G ) D_{p_i}(G) D p i ( G ) theories, denoted as T ( p , N ) \mathcal{T}_{(p,N)} T ( p , N ) with p ∈ { 2 , 3 , 4 , 6 } p\in\{2,3,4,6\} p ∈ { 2 , 3 , 4 , 6 } , N ≥ 2 N\geq2 N ≥ 2 and gcd ( p , N ) = 1 \gcd(p,N)=1 g cd( p , N ) = 1 [7 , 8 ] . Their Schur indices are also related to those of N = 4 \mathcal{N}=4 N = 4 SYM
I T ( p , N ) ( q ) = I N = 4 S U ( N ) S Y M ( q p , q p 2 − 1 ) . (25) \mathcal{I}_{\mathcal{T}_{(p,N)}}(q)=\mathcal{I}_{\mathcal{N}=4\ \mathrm{SU}(N)\ \mathrm{SYM}}\left(q^{p},q^{\frac{p}{2}-1}\right). \tag*{(25)} I T ( p , N ) ( q ) = I N = 4 SU ( N ) SYM ( q p , q 2 p − 1 ) . (25)
Note that ( A 2 , D 4 ) = T ( 3 , 2 ) (A_{2},D_{4})=\mathcal{T}_{(3,2)} ( A 2 , D 4 ) = T ( 3 , 2 ) . Another similarity between the two SCFTs above is that they both have the feature a = c a=c a = c .
As shown in [17 , 18 ] , there is another kind of character identity relating three theories
I ( A 2 , D 3 m + 1 ) ( q ) = I ( A 1 , A 2 m − 2 ) ( q ) I N = 4 S U ( 2 ) S Y M ( q 2 m + 1 , q 1 2 ) , m ≥ 1 , (26) \mathcal{I}_{(A_{2},D_{3m+1})}(q)=\mathcal{I}_{(A_{1},A_{2m-2})}(q)\mathcal{I}_{\mathcal{N}=4\ \mathrm{SU}(2)\ \mathrm{SYM}}\left(q^{2m+1},q^{\frac{1}{2}}\right),\qquad m\geq1, \tag*{(26)} I ( A 2 , D 3 m + 1 ) ( q ) = I ( A 1 , A 2 m − 2 ) ( q ) I N = 4 SU ( 2 ) SYM ( q 2 m + 1 , q 2 1 ) , m ≥ 1 , (26)
where the prefactor is trivial if m = 1 m=1 m = 1 , which reduces to (24) .
This kind of relation is further generalized to a pair of N = 2 \mathcal{N}=2 N = 2 AD theories. In [6 ] , it was found that10
I ( A 1 , D 2 n + 1 ) ( q 3 , q 1 2 ) = I ( A 1 , A 6 n ) ( q ) , (27) \mathcal{I}_{(A_{1},D_{2n+1})}\left(q^{3},q^{\frac{1}{2}}\right)=\mathcal{I}_{(A_{1},A_{6n})}(q), \tag*{(27)} I ( A 1 , D 2 n + 1 ) ( q 3 , q 2 1 ) = I ( A 1 , A 6 n ) ( q ) , (27)
where the ( A 1 , D 2 n + 1 ) (A_{1},D_{2n+1}) ( A 1 , D 2 n + 1 ) AD theory has flavor symmetry S U ( 2 ) \mathrm{SU}(2) SU ( 2 ) whose fugacity is specialized. This translates to a vacuum character identity. More non-trivially, it was also shown in [6 ] that similar identities can be generalized to the non-vacuum modules of VOAs.
These index and character identities motivate us to ask whether this is just a coincidence, or a general feature. Also, it is natural to wonder what they imply beyond the matching of indices.
This is the goal of this paper. We systematically generalize these index identities, which occur widely among AD theories and W W W -algebras. We also show that they imply simple relations between characteristic anomaly coefficients.
Anomaly relations from index identity
The general index identities take the form of
∏ i = 1 N L I L i ( Q i , X i ) = ∏ j = 1 N R I R j ( Q j ′ , X j ′ ) , (28) \prod_{i=1}^{N_{L}}\mathcal{I}_{\mathcal{L}_{i}}(Q_{i},X_{i})=\prod_{j=1}^{N_{R}}\mathcal{I}_{\mathcal{R}_{j}}(Q'_{j},X'_{j}), \tag*{(28)} i = 1 ∏ N L I L i ( Q i , X i ) = j = 1 ∏ N R I R j ( Q j ′ , X j ′ ) , (28)
where the fugacities on the LHS and RHS are all related by a specific map. This includes all the identities we reviewed before.
Our main goal is to study in detail the simplest case of N L = N R = 1 N_{L}=N_{R}=1 N L = N R = 1 :
I L ( q , x ) = I R ( q ℓ , X ∗ ( q , x ) ) , (29) \mathcal{I}_{\mathcal{L}}(q,x)=\mathcal{I}_{\mathcal{R}}\left(q^{\ell},X_{*}(q,x)\right), \tag*{(29)} I L ( q , x ) = I R ( q ℓ , X ∗ ( q , x ) ) , (29)
where X ∗ X_{*} X ∗ is specialized to a specific homogeneous function of q , x q,x q , x . Such identities can relate AD theories whose associated VOAs are not isomorphic.
As discussed before, the leading asymptotic growth of the index is controlled by the effective central charge (10) . The index identity in (29) thus implies that11
c L e f f = c R e f f / ℓ , C L = ℓ C R . (30) c_{\mathcal{L}}^{\mathrm{eff}} = c_{\mathcal{R}}^{\mathrm{eff}}/\ell,\qquad C_{\mathcal{L}} = \ell C_{\mathcal{R}}. \tag*{(30)} c L eff = c R eff / ℓ , C L = ℓ C R . (30)
This is particularly useful because these central charges are characteristic quantities, computable from the 4d central charges and levels in a simple way (13) . It also provides a natural way to understand why both theories in (25) feature a = c a=c a = c .
In case the flavor fugacities on the two sides can be set to one simultaneously, we have the simpler relation I L ( q ) = I R ( q ℓ ) \mathcal{I}_{\mathcal{L}}(q)=\mathcal{I}_{\mathcal{R}}(q^\ell) I L ( q ) = I R ( q ℓ ) . The asymptotic expansion (10) gives the stronger and simpler relations
c L e f f = c R e f f / ℓ , c L = ℓ c R . (31) c_{\mathcal{L}}^{\mathrm{eff}} = c_{\mathcal{R}}^{\mathrm{eff}}/\ell,\qquad c_{\mathcal{L}} = \ell c_{\mathcal{R}}. \tag*{(31)} c L eff = c R eff / ℓ , c L = ℓ c R . (31)
Note that the two central charges above scale oppositely, which is related to the factor τ \tau τ appearing in the denominator and numerator, respectively (10) .
For the more general case (29) , the τ \tau τ -linear coefficient C C C receives a correction from the flavor fugacity, as in (12) , and the c c c central charges on the two sides are therefore related in a more complicated but controlled way. One can also try to match other coefficients in (10) , in particular the constant term S S S . For γ \gamma γ , it actually always vanishes in our boundary admissible class of AD theories below.12
As a side remark, we also find many examples with index identities of the form
I L 1 ( Q 1 , X 1 ) I L 2 ( Q 2 , X 2 ) = I R 3 ( Q ′ , X ′ ) , (32) \mathcal{I}_{\mathcal{L}_1}(Q_1,\mathbf{X}_1)\mathcal{I}_{\mathcal{L}_2}(Q_2,\mathbf{X}_2)=\mathcal{I}_{\mathcal{R}_3}(Q',\mathbf{X}'), \tag*{(32)} I L 1 ( Q 1 , X 1 ) I L 2 ( Q 2 , X 2 ) = I R 3 ( Q ′ , X ′ ) , (32)
where the fugacities are related and specialized as well.
Index identity between ( A N − 1 , A ( N + 1 ) ( p − 1 ) ) (A_{N-1},A_{(N+1)(p-1)}) ( A N − 1 , A ( N + 1 ) ( p − 1 ) ) and D p ( S U ( N ) ) D_p(SU(N)) D p ( S U ( N ))
In this section, we generalize the result in [6 ] and establish that
I ( A N − 1 , A ( N + 1 ) ( p − 1 ) ) ( q ) = I D p ( S U ( N ) ) ( Q = q N + 1 , x = q ρ ) , gcd ( p , N ) = 1 , (33) \mathcal{I}_{(A_{N-1},A_{(N+1)(p-1)})}(q)=\mathcal{I}_{D_p(SU(N))}(Q=q^{N+1},x=q^\rho),\qquad\gcd(p,N)=1, \tag*{(33)} I ( A N − 1 , A ( N + 1 ) ( p − 1 ) ) ( q ) = I D p ( S U ( N )) ( Q = q N + 1 , x = q ρ ) , g cd( p , N ) = 1 , (33)
where we use Q , x Q,x Q , x to represent the fugacities of the flavored theory on the RHS. In particular, the flavor fugacities are set to the principal specialization
x i = q ρ i = q N + 1 − 2 i 2 , i = 1 , … , N , ∏ i = 1 N x i = 1 , (34) x_i=q^{\rho_i}=q^{\frac{N+1-2i}{2}},\qquad i=1,\ldots,N,\qquad\prod_{i=1}^{N}x_i=1, \tag*{(34)} x i = q ρ i = q 2 N + 1 − 2 i , i = 1 , … , N , i = 1 ∏ N x i = 1 , (34)
where ρ \rho ρ is the Weyl vector.
The Schur index of D p ( S U ( N ) ) D_p(SU(N)) D p ( S U ( N )) is given by
I D p ( S U ( N ) ) ( Q , x ) = PE [ Q − Q p ( 1 − Q ) ( 1 − Q p ) χ a d j S U ( N ) ( x ) ] . (35) \mathcal{I}_{D_p(SU(N))}(Q,x)=\operatorname{PE}\left[\frac{Q-Q^p}{(1-Q)(1-Q^p)}\chi_{\mathrm{adj}}^{SU(N)}(x)\right]. \tag*{(35)} I D p ( S U ( N )) ( Q , x ) = PE [ ( 1 − Q ) ( 1 − Q p ) Q − Q p χ adj S U ( N ) ( x ) ] . (35)
where the plethystic exponential is defined as
PE [ f ( q , x ) ] = exp [ ∑ n ≥ 1 f ( q n , x n ) n ] . (36) \operatorname{PE}[f(q,x)] = \exp\left[\sum_{n\geq1}\frac{f(q^n,x^n)}{n}\right]. \tag*{(36)} PE [ f ( q , x )] = exp [ n ≥ 1 ∑ n f ( q n , x n ) ] . (36)
The inverse is denoted as PL.
Under (34) , the adjoint character becomes
χ a d j S U ( N ) ( q ρ ) = ∑ 1 ≤ i , j ≤ N x i x j − 1 = ∑ 1 ≤ i , j ≤ N q j q i − 1 = ( 1 − q N − 1 ) ( 1 − q N + 1 ) q N − 1 ( 1 − q ) 2 . (37) \chi_{\mathrm{adj}}^{SU(N)}(q^\rho)=\sum_{1\leq i,j\leq N}\frac{x_i}{x_j}-1=\sum_{1\leq i,j\leq N}\frac{q^j}{q^i}-1=\frac{(1-q^{N-1})(1-q^{N+1})}{q^{N-1}(1-q)^2}. \tag*{(37)} χ adj S U ( N ) ( q ρ ) = 1 ≤ i , j ≤ N ∑ x j x i − 1 = 1 ≤ i , j ≤ N ∑ q i q j − 1 = q N − 1 ( 1 − q ) 2 ( 1 − q N − 1 ) ( 1 − q N + 1 ) . (37)
Setting Q = q N + 1 Q=q^{N+1} Q = q N + 1 gives
I D p ( q N + 1 , q ρ ) = PE [ q N + 1 ( 1 − q ( N + 1 ) ( p − 1 ) ) ( 1 − q N + 1 ) ( 1 − q ( N + 1 ) p ) χ a d j ( q ρ ) ] = PE [ q N + 1 ( 1 − q ( N + 1 ) ( p − 1 ) ) ( 1 − q N + 1 ) ( 1 − q ( N + 1 ) p ) ( 1 − q N − 1 ) ( 1 − q N + 1 ) q N − 1 ( 1 − q ) 2 ] = PE [ q 2 ( 1 − q ( N + 1 ) ( p − 1 ) ) ( 1 − q N − 1 ) ( 1 − q ) 2 ( 1 − q ( N + 1 ) p ) ] . \begin{aligned}
\mathcal{I}_{D_p}(q^{N+1},q^\rho)=\operatorname{PE}\left[\frac{q^{N+1}(1-q^{(N+1)(p-1)})}{(1-q^{N+1})(1-q^{(N+1)p})}\chi_{\mathrm{adj}}(q^\rho)\right] \\
&=\operatorname{PE}\left[\frac{q^{N+1}(1-q^{(N+1)(p-1)})}{(1-q^{N+1})(1-q^{(N+1)p})}\frac{(1-q^{N-1})(1-q^{N+1})}{q^{N-1}(1-q)^2}\right] \\
&=\operatorname{PE}\left[\frac{q^2(1-q^{(N+1)(p-1)})(1-q^{N-1})}{(1-q)^2(1-q^{(N+1)p})}\right].
\end{aligned} I D p ( q N + 1 , q ρ ) = PE [ ( 1 − q N + 1 ) ( 1 − q ( N + 1 ) p ) q N + 1 ( 1 − q ( N + 1 ) ( p − 1 ) ) χ adj ( q ρ ) ] = PE [ ( 1 − q N + 1 ) ( 1 − q ( N + 1 ) p ) q N + 1 ( 1 − q ( N + 1 ) ( p − 1 ) ) q N − 1 ( 1 − q ) 2 ( 1 − q N − 1 ) ( 1 − q N + 1 ) ] = PE [ ( 1 − q ) 2 ( 1 − q ( N + 1 ) p ) q 2 ( 1 − q ( N + 1 ) ( p − 1 ) ) ( 1 − q N − 1 ) ] .
Meanwhile, the Schur index of ( A N − 1 , A k − 1 ) (A_{N-1},A_{k-1}) ( A N − 1 , A k − 1 ) is given by
I ( A N − 1 , A k − 1 ) = PE [ q 2 ( 1 − q k − 1 ) ( 1 − q N − 1 ) ( 1 − q ) 2 ( 1 − q k + N ) ] , gcd ( k , N ) = 1. (38) \mathcal{I}_{(A_{N-1},A_{k-1})}=\operatorname{PE}\left[\frac{q^2(1-q^{k-1})(1-q^{N-1})}{(1-q)^2(1-q^{k+N})}\right],\qquad\gcd(k,N)=1. \tag*{(38)} I ( A N − 1 , A k − 1 ) = PE [ ( 1 − q ) 2 ( 1 − q k + N ) q 2 ( 1 − q k − 1 ) ( 1 − q N − 1 ) ] , g cd( k , N ) = 1. (38)
By comparing the two expressions above, it is easy to see that they coincide if we set k = ( N + 1 ) ( p − 1 ) + 1 = p N + p − N k=(N+1)(p-1)+1=pN+p-N k = ( N + 1 ) ( p − 1 ) + 1 = pN + p − N and gcd ( k , N ) = gcd ( p , N ) = 1 \gcd(k,N)=\gcd(p,N)=1 g cd( k , N ) = g cd( p , N ) = 1 as required. Therefore
I ( A N − 1 , A ( N + 1 ) ( p − 1 ) ) ( q ) = I D p ( S U ( N ) ) ( Q = q N + 1 , x = q ρ ) . (39) \mathcal{I}_{(A_{N-1},A_{(N+1)(p-1)})}(q)=\mathcal{I}_{D_p(SU(N))}(Q=q^{N+1},x=q^\rho). \tag*{(39)} I ( A N − 1 , A ( N + 1 ) ( p − 1 ) ) ( q ) = I D p ( S U ( N )) ( Q = q N + 1 , x = q ρ ) . (39)
Using (13) , one can easily compute the effective central charges of the two theories
c D p ( S U ( N ) ) e f f = ( N 2 − 1 ) ( p − 1 ) p , c ( A N − 1 , A k − 1 ) e f f = ( N − 1 ) ( k − 1 ) N + k . (40) c^{\mathrm{eff}}_{D_p(SU(N))}=\frac{(N^2-1)(p-1)}{p},\qquad c^{\mathrm{eff}}_{(A_{N-1},A_{k-1})}=\frac{(N-1)(k-1)}{N+k}. \tag*{(40)} c D p ( S U ( N )) eff = p ( N 2 − 1 ) ( p − 1 ) , c ( A N − 1 , A k − 1 ) eff = N + k ( N − 1 ) ( k − 1 ) . (40)
As a result, one can verify
c ( A N − 1 , A ( N + 1 ) ( p − 1 ) ) e f f = ( N − 1 ) ( p − 1 ) p = 1 N + 1 c D p ( S U ( N ) ) e f f . (41) c^{\mathrm{eff}}_{(A_{N-1},A_{(N+1)(p-1)})}=\frac{(N-1)(p-1)}{p}=\frac{1}{N+1}c^{\mathrm{eff}}_{D_p(SU(N))}. \tag*{(41)} c ( A N − 1 , A ( N + 1 ) ( p − 1 ) ) eff = p ( N − 1 ) ( p − 1 ) = N + 1 1 c D p ( S U ( N )) eff . (41)
Index identity between ( A ( h + 1 ) ( p − 1 ) , G ) (A_{(h+1)(p-1)},G) ( A ( h + 1 ) ( p − 1 ) , G ) and D p ( G ) D_p(G) D p ( G )
In this section, we generalize the identity in the previous section and prove that
I D p ( G ) ( q h + 1 , q ρ ) = I ( A ( h + 1 ) ( p − 1 ) , G ) ( q ) , gcd ( h , p ) = 1. (42) \mathcal{I}_{D_p(G)}(q^{h+1},q^\rho)=\mathcal{I}_{(A_{(h+1)(p-1)},G)}(q),\qquad\gcd(h,p)=1. \tag*{(42)} I D p ( G ) ( q h + 1 , q ρ ) = I ( A ( h + 1 ) ( p − 1 ) , G ) ( q ) , g cd( h , p ) = 1. (42)
for any simply laced G G G with Coxeter number h h h , which coincides with the dual Coxeter number h ∨ h^\vee h ∨ . Throughout the section, we always assume that p ≥ 2 p \ge2 p ≥ 2 is an integer and gcd ( p , h ) = 1 \gcd(p,h)=1 g cd( p , h ) = 1 .
For D p ( G ) D_p(G) D p ( G ) with no additional flavor symmetry from the irregular puncture, the flavored Schur index is [2 ]
I D p ( G ) ( Q , x ) = PE [ Q − Q p ( 1 − Q ) ( 1 − Q p ) χ a d j G ( x ) ] . (43) \mathcal{I}_{D_p(G)}(Q,x)=\operatorname{PE}\left[\frac{Q-Q^p}{(1-Q)(1-Q^p)}\chi_{\mathrm{adj}}^G(x)\right]. \tag*{(43)} I D p ( G ) ( Q , x ) = PE [ ( 1 − Q ) ( 1 − Q p ) Q − Q p χ adj G ( x ) ] . (43)
This is precisely the boundary-admissible affine vacuum character.
Let G G G be simply laced, with rank r r r , Coxeter number h = h ∨ h=h^\vee h = h ∨ , exponents e 1 , … , e r e_1,\ldots,e_r e 1 , … , e r , and Casimir degrees d i = e i + 1 d_i=e_i+1 d i = e i + 1 . Under the principal specialization x = q ρ x=q^\rho x = q ρ , we have
χ a d j ( q ρ ) = ∑ i = 1 r ∑ m = − e i e i q m = ∑ i = 1 r ∑ m = 1 − d i d i − 1 q m . (44) \chi_{\mathrm{adj}}(q^\rho)=\sum_{i=1}^{r}\sum_{m=-e_i}^{e_i}q^m=\sum_{i=1}^{r}\sum_{m=1-d_i}^{d_i-1}q^m. \tag*{(44)} χ adj ( q ρ ) = i = 1 ∑ r m = − e i ∑ e i q m = i = 1 ∑ r m = 1 − d i ∑ d i − 1 q m . (44)
Using e i + e r + 1 − i = h e_i+e_{r+1-i}=h e i + e r + 1 − i = h , one gets the following crucial identity
q h + 1 1 − q h + 1 χ a d j ( q ρ ) = ∑ i = 1 r q d i 1 − q . (45) \frac{q^{h+1}}{1-q^{h+1}}\chi_{\mathrm{adj}}(q^\rho)=\frac{\sum_{i=1}^{r}q^{d_i}}{1-q}. \tag*{(45)} 1 − q h + 1 q h + 1 χ adj ( q ρ ) = 1 − q ∑ i = 1 r q d i . (45)
Therefore setting Q = q h + 1 Q=q^{h+1} Q = q h + 1 in (43) gives
I D p ( G ) ( q h + 1 , q ρ ) = PE [ ( ∑ i = 1 r q d i ) ( 1 − q ( h + 1 ) ( p − 1 ) ) ( 1 − q ) ( 1 − q ( h + 1 ) p ) ] . (46) \mathcal{I}_{D_p(G)}(q^{h+1},q^\rho)=\operatorname{PE}\left[\frac{\left(\sum_{i=1}^{r}q^{d_i}\right)\left(1-q^{(h+1)(p-1)}\right)}{(1-q)\left(1-q^{(h+1)p}\right)}\right]. \tag*{(46)} I D p ( G ) ( q h + 1 , q ρ ) = PE [ ( 1 − q ) ( 1 − q ( h + 1 ) p ) ( ∑ i = 1 r q d i ) ( 1 − q ( h + 1 ) ( p − 1 ) ) ] . (46)
This immediately implies that
I D p ( G ) ( q h + 1 , q ρ ) = I ( A ( h + 1 ) ( p − 1 ) , G ) ( q ) , (47) \mathcal{I}_{D_p(G)}(q^{h+1},q^\rho)=\mathcal{I}_{(A_{(h+1)(p-1)},G)}(q), \tag*{(47)} I D p ( G ) ( q h + 1 , q ρ ) = I ( A ( h + 1 ) ( p − 1 ) , G ) ( q ) , (47)
by recalling the following Schur-index formula for ( A k − 1 , G ) (A_{k-1},G) ( A k − 1 , G ) with k ≥ 2 k\ge2 k ≥ 2 and gcd ( k , h ) = 1 \gcd(k,h)=1 g cd( k , h ) = 1
I ( A k − 1 , G ) ( q ) = PE [ ∑ i = 1 r ( q d i − q h + k + 1 − d i ) ( 1 − q ) ( 1 − q h + k ) ] . (48) \mathcal{I}_{(A_{k-1},G)}(q)=\operatorname{PE}\left[\frac{\sum_{i=1}^{r}\left(q^{d_i}-q^{h+k+1-d_i}\right)}{(1-q)(1-q^{h+k})}\right]. \tag*{(48)} I ( A k − 1 , G ) ( q ) = PE [ ( 1 − q ) ( 1 − q h + k ) ∑ i = 1 r ( q d i − q h + k + 1 − d i ) ] . (48)
More explicitly, the two related theories are
Table 4.8.
The associated VOAs of the two SCFTs are
V [ D p ( G ) ] = L − h + h / p ( g ) , \mathbb{V}[D_p(G)] = L_{-h+h/p}(\mathfrak{g}), V [ D p ( G )] = L − h + h / p ( g ) ,
V [ ( A ( h + 1 ) ( p − 1 ) , G ) ] = W − h + h / ( ( h + 1 ) p ) ( g , f p r i n ) = W G ( h , ( h + 1 ) p ) , (49) \mathbb{V}[(A_{(h+1)(p-1)},G)] = \mathcal{W}_{-h+h/((h+1)p)}(\mathfrak{g},f_{\mathrm{prin}}) = W_G(h,(h+1)p), \tag*{(49)} V [( A ( h + 1 ) ( p − 1 ) , G )] = W − h + h / (( h + 1 ) p ) ( g , f prin ) = W G ( h , ( h + 1 ) p ) , (49)
where W G ( h , h + k ) W_G(h,h+k) W G ( h , h + k ) is the simple principal W W W -algebra minimal model.
Since the Schur index of an SCFT equals the vacuum (super)character of the associated VOA, the index relation (47) of SCFTs translates to the character relation of VOAs
χ W − h + h / ( ( h + 1 ) p ) ( g , f p r i n ) ( q ) = χ L − h + h / p ( g ) ( q h + 1 , q ρ ) . (50) \chi_{\mathcal{W}_{-h+h/((h+1)p)}(\mathfrak{g},f_{\mathrm{prin}})}(q) = \chi_{L_{-h+h/p}(\mathfrak{g})}\left(q^{h+1},q^\rho\right). \tag*{(50)} χ W − h + h / (( h + 1 ) p ) ( g , f prin ) ( q ) = χ L − h + h / p ( g ) ( q h + 1 , q ρ ) . (50)
So the rescaling factor is generally given by ℓ = h G + 1 \ell=h_G+1 ℓ = h G + 1 . We next check the anomaly relations.
For gcd ( p , h ) = 1 \gcd(p,h)=1 g cd( p , h ) = 1 , the 4d central charges of D p ( G ) D_p(G) D p ( G ) are given by [19 ]
a D p ( G ) = ( 4 p − 1 ) ( p − 1 ) 48 p dim G , c D p ( G ) = p − 1 12 dim G , (51) a_{D_p(G)} = \frac{(4p-1)(p-1)}{48p}\dim G,\qquad c_{D_p(G)} = \frac{p-1}{12}\dim G, \tag*{(51)} a D p ( G ) = 48 p ( 4 p − 1 ) ( p − 1 ) dim G , c D p ( G ) = 12 p − 1 dim G , (51)
and the effective central charge is (13)
c D p ( G ) e f f = p − 1 p dim G = r ( h + 1 ) ( p − 1 ) p , (52) c^{\mathrm{eff}}_{D_p(G)} = \frac{p-1}{p}\dim G = \frac{r(h+1)(p-1)}{p}, \tag*{(52)} c D p ( G ) eff = p p − 1 dim G = p r ( h + 1 ) ( p − 1 ) , (52)
where we used dim G = r ( h + 1 ) \dim G=r(h+1) dim G = r ( h + 1 ) with r r r the rank of G G G .
Meanwhile, the central charges of ( A k − 1 , G ) (A_{k-1},G) ( A k − 1 , G ) in the coprime case gcd ( k , h ) = 1 \gcd(k,h)=1 g cd( k , h ) = 1 are
c ( A k − 1 , G ) = r ( k − 1 ) ( k h + k + h ) 12 ( k + h ) , a ( A k − 1 , G ) = r ( k − 1 ) ( 4 k h + 4 k + 4 h − 1 ) 48 ( k + h ) , (53) c_{(A_{k-1},G)} = \frac{r(k-1)(kh+k+h)}{12(k+h)},\qquad a_{(A_{k-1},G)} = \frac{r(k-1)(4kh+4k+4h-1)}{48(k+h)}, \tag*{(53)} c ( A k − 1 , G ) = 12 ( k + h ) r ( k − 1 ) ( k h + k + h ) , a ( A k − 1 , G ) = 48 ( k + h ) r ( k − 1 ) ( 4 k h + 4 k + 4 h − 1 ) , (53)
which can be computed from the type IIB realization. These expressions imply the following effective central charge
c ( A k − 1 , G ) e f f = r ( k − 1 ) k + h . (54) c^{\mathrm{eff}}_{(A_{k-1},G)} = \frac{r(k-1)}{k+h}. \tag*{(54)} c ( A k − 1 , G ) eff = k + h r ( k − 1 ) . (54)
This can also be confirmed from the direct high temperature limit of the index (48)
I ( A k − 1 , G ) ( e − β ) = S ( A k − 1 , G ) exp ( π 2 r ( k − 1 ) 6 ( h + k ) β − β c 2 d 24 ) [ 1 + O ( e − # / β ) ] , (55) \mathcal{I}_{(A_{k-1},G)}(e^{-\beta}) = S_{(A_{k-1},G)}\exp\left(\frac{\pi^2r(k-1)}{6(h+k)\beta}-\frac{\beta c_{2d}}{24}\right)\left[1+O(e^{-\#/\beta})\right], \tag*{(55)} I ( A k − 1 , G ) ( e − β ) = S ( A k − 1 , G ) exp ( 6 ( h + k ) β π 2 r ( k − 1 ) − 24 β c 2 d ) [ 1 + O ( e − #/ β ) ] , (55)
where
S ( A k − 1 , G ) = ( 2 π ) − r ( k − 1 ) / 2 ∏ i = 1 r ∏ s = 0 k − 2 Γ ( d i + s h + k ) . (56) S_{(A_{k-1},G)}=(2\pi)^{-r(k-1)/2}\prod_{i=1}^{r}\prod_{s=0}^{k-2}\Gamma\left(\frac{d_i+s}{h+k}\right). \tag*{(56)} S ( A k − 1 , G ) = ( 2 π ) − r ( k − 1 ) /2 i = 1 ∏ r s = 0 ∏ k − 2 Γ ( h + k d i + s ) . (56)
By comparing (52) and (54) , it is easy to see that
c ( A ( h + 1 ) ( p − 1 ) , G ) e f f = r ( p − 1 ) p = 1 h + 1 c D p ( G ) e f f . (57) c^{\mathrm{eff}}_{(A_{(h+1)(p-1)},G)}=\frac{r(p-1)}{p}=\frac{1}{h+1}c^{\mathrm{eff}}_{D_p(G)}. \tag*{(57)} c ( A ( h + 1 ) ( p − 1 ) , G ) eff = p r ( p − 1 ) = h + 1 1 c D p ( G ) eff . (57)
We can also match the τ \tau τ -linear term C C C in (10) once the flavor contribution is taken into account. On the affine side, we have
τ a f f = ℓ τ , z a f f = τ ρ = τ a f f ρ ℓ , k J ρ = k a f f ( ρ , ρ ) , ( ρ , ρ ) = h dim g 12 (58) \tau_{\mathrm{aff}}=\ell\tau,\qquad z_{\mathrm{aff}}=\tau\rho=\tau_{\mathrm{aff}}\frac{\rho}{\ell},\qquad k_{J\rho}=k_{\mathrm{aff}}(\rho,\rho),\qquad(\rho,\rho)=\frac{h\dim\mathfrak{g}}{12} \tag*{(58)} τ aff = ℓ τ , z aff = τ ρ = τ aff ℓ ρ , k J ρ = k aff ( ρ , ρ ) , ( ρ , ρ ) = 12 h dim g (58)
where ℓ = h + 1 \ell=h+1 ℓ = h + 1 . Therefore, (12) gives
C D p ( G ) = c 2 d , D p ( G ) − 12 k a f f ( ρ , ρ ) ℓ 2 = ( 1 − p ) dim g ( 1 − h 2 p ℓ 2 ) = − r ( p − 1 ) ℓ p [ p ( h + 1 ) 2 − h 2 ] (59) C_{D_p(G)}=c_{2d,D_p(G)}-\frac{12k_{\mathrm{aff}}(\rho,\rho)}{\ell^2}=(1-p)\dim\mathfrak{g}\left(1-\frac{h^2}{p\ell^2}\right)=-\frac{r(p-1)}{\ell p}\left[p(h+1)^2-h^2\right] \tag*{(59)} C D p ( G ) = c 2 d , D p ( G ) − ℓ 2 12 k aff ( ρ , ρ ) = ( 1 − p ) dim g ( 1 − p ℓ 2 h 2 ) = − ℓ p r ( p − 1 ) [ p ( h + 1 ) 2 − h 2 ] (59)
where we used (51) and k a f f = − h + h / p k_{\mathrm{aff}}=-h+h/p k aff = − h + h / p .
On the principal W W W side which has no flavor symmetry, we have
C A ( h + 1 ) ( p − 1 ) , G = c 2 d , A ( h + 1 ) ( p − 1 ) , G = − 12 c A ( h + 1 ) ( p − 1 ) , G = − r ( p − 1 ) [ p ( h + 1 ) 2 − h 2 ] p (60) C_{A_{(h+1)(p-1)},G}=c_{2d,A_{(h+1)(p-1)},G}=-12c_{A_{(h+1)(p-1)},G}=-\frac{r(p-1)\left[p(h+1)^2-h^2\right]}{p} \tag*{(60)} C A ( h + 1 ) ( p − 1 ) , G = c 2 d , A ( h + 1 ) ( p − 1 ) , G = − 12 c A ( h + 1 ) ( p − 1 ) , G = − p r ( p − 1 ) [ p ( h + 1 ) 2 − h 2 ] (60)
where we used (53) . By comparison, we obtain
C A ( h + 1 ) ( p − 1 ) , G = ℓ C D p ( G ) . (61) C_{A_{(h+1)(p-1)},G}=\ell\,C_{D_p(G)}. \tag*{(61)} C A ( h + 1 ) ( p − 1 ) , G = ℓ C D p ( G ) . (61)
This confirms (30) .
Let us conclude this section by comparing the Coulomb-branch spectra of the two theories, which probe a protected sector different from the Schur sector.
The Coulomb spectra of the two theories are given by [19 ]
C B D p ( G ) = ⋃ i = 1 r { d i − h ℓ p | ℓ ∈ Z , 1 ≤ ℓ < p ( d i − 1 ) h } , (62) \mathrm{CB}_{D_p(G)}=\bigcup_{i=1}^{r}\left\{d_i-\frac{h\ell}{p}\ \middle|\ \ell\in\mathbb{Z},\quad1\leq\ell<\frac{p(d_i-1)}{h}\right\}, \tag*{(62)} CB D p ( G ) = i = 1 ⋃ r { d i − p h ℓ ℓ ∈ Z , 1 ≤ ℓ < h p ( d i − 1 ) } , (62)
C B A ( h + 1 ) ( p − 1 ) , G = ⋃ i = 1 r { d i − h j ( h + 1 ) p | j ∈ Z , d i ≤ j < ( h + 1 ) p ( d i − 1 ) h } , (63) \mathrm{CB}_{A_{(h+1)(p-1)},G}=\bigcup_{i=1}^{r}\left\{d_i-\frac{hj}{(h+1)p}\ \middle|\ j\in\mathbb{Z},\quad d_i\leq j<\frac{(h+1)p(d_i-1)}{h}\right\}, \tag*{(63)} CB A ( h + 1 ) ( p − 1 ) , G = i = 1 ⋃ r { d i − ( h + 1 ) p hj j ∈ Z , d i ≤ j < h ( h + 1 ) p ( d i − 1 ) } , (63)
where the second formula can be derived from the type IIB construction [20 ] . Here d i d_i d i ’s are the degrees of Casimirs of g \mathfrak{g} g , and all elements in the set above should be counted with multiplicity. In particular, the upper bounds on ℓ \ell ℓ and j j j select Coulomb-branch operators with Δ > 1 \Delta>1 Δ > 1 .
Consequently, one can verify that
C B D p ( G ) = { Δ ∈ C B A ( h + 1 ) ( p − 1 ) , G | p Δ ∈ Z } ⊊ C B A ( h + 1 ) ( p − 1 ) , G . (64) \mathrm{CB}_{D_p(G)}=\left\{\Delta\in\mathrm{CB}_{A_{(h+1)(p-1)},G}\ \middle|\ p\Delta\in\mathbb{Z}\right\}\subsetneq\mathrm{CB}_{A_{(h+1)(p-1)},G}. \tag*{(64)} CB D p ( G ) = { Δ ∈ CB A ( h + 1 ) ( p − 1 ) , G p Δ ∈ Z } ⊊ CB A ( h + 1 ) ( p − 1 ) , G . (64)
Indeed, starting from the spectrum in (63) , we have
p Δ = p d i − h j h + 1 ∈ Z ⟺ ℓ = j h + 1 ∈ Z ⟹ 1 ≤ ℓ < p ( d i − 1 ) h , (65) p\Delta=pd_i-\frac{hj}{h+1}\in\mathbb{Z}\quad\Longleftrightarrow\quad\ell=\frac{j}{h+1}\in\mathbb{Z}\quad\Longrightarrow\quad1\leq\ell<\frac{p(d_i-1)}{h}, \tag*{(65)} p Δ = p d i − h + 1 hj ∈ Z ⟺ ℓ = h + 1 j ∈ Z ⟹ 1 ≤ ℓ < h p ( d i − 1 ) , (65)
where we have used the fact that 2 ≤ d i ≤ h 2\leq d_i\leq h 2 ≤ d i ≤ h , and the range of j j j in (63) . The resulting ℓ \ell ℓ agrees exactly with that in (62) .
The ranks, which are dimensions of the Coulomb branches, of the two theories are also related:
dim C B D p ( G ) = r ( p − 1 ) 2 , dim C B ( A ( h + 1 ) ( p − 1 ) , G ) = r ( h + 1 ) ( p − 1 ) 2 = ℓ dim C B D p ( G ) . (66) \dim\mathrm{CB}_{D_p(G)}=\frac{r(p-1)}{2},\qquad\dim\mathrm{CB}_{(A_{(h+1)(p-1)},G)}=\frac{r(h+1)(p-1)}{2}=\ell\dim\mathrm{CB}_{D_p(G)}. \tag*{(66)} dim CB D p ( G ) = 2 r ( p − 1 ) , dim CB ( A ( h + 1 ) ( p − 1 ) , G ) = 2 r ( h + 1 ) ( p − 1 ) = ℓ dim CB D p ( G ) . (66)
Index identities for general AD theories
We now turn to more general AD theories and further extend the identities established previously. The resulting character identities involve W W W -algebras exhibiting a range of new features, possibly involving non-simply laced algebras, non-principal nilpotent elements, non-boundary admissible and even non-admissible levels.
W algebra at boundary admissible level
Let g \mathfrak{g} g be a finite-dimensional simple complex Lie algebra. Write h h h , h ∨ h^{\vee} h ∨ , r r r , and r ∨ r^{\vee} r ∨ for its Coxeter number, dual Coxeter number, rank, and lacing number, respectively. See Table 1 for details.
Table 1. Lie algebra data. dim g \dim\mathfrak{g} dim g , r r r , r ∨ r^{\vee} r ∨ , h h h , h ∨ h^{\vee} h ∨ denote the dimension, rank, lacing number, Coxeter number, and dual Coxeter number, respectively. Note that dim g = ( h + 1 ) r \dim\mathfrak{g} = (h+1)r dim g = ( h + 1 ) r . We restrict to A N − 1 : N ≥ 2 A_{N-1}: N \ge2 A N − 1 : N ≥ 2 , B n B_n B n , C n C_n C n : n ≥ 2 n \ge2 n ≥ 2 , D n : n ≥ 4 D_n: n \ge4 D n : n ≥ 4 .
In this paper, we mainly consider the principal boundary-admissible family, whose level is given by 13
k ( u ) = k = − h ∨ + h ∨ u , u = κ + h ∨ , (67) k(u)=k=-h^{\vee}+\frac{h^{\vee}}{u},\qquad u=\kappa+h^{\vee}, \tag*{(67)} k ( u ) = k = − h ∨ + u h ∨ , u = κ + h ∨ , (67)
subject to the conditions
u ∈ Z ≥ 1 , gcd ( u , h ∨ ) = gcd ( u , r ∨ ) = 1. (68) u\in\mathbb{Z}_{\geq1},\qquad\gcd(u,h^{\vee})=\gcd(u,r^{\vee})=1. \tag*{(68)} u ∈ Z ≥ 1 , g cd( u , h ∨ ) = g cd( u , r ∨ ) = 1. (68)
The corresponding simple affine vertex algebra will be denoted as L k ( g ) L_k(\mathfrak{g}) L k ( g ) , which is the simple quotient of the universal affine vertex algebra V k ( g ) V^k(\mathfrak{g}) V k ( g ) . The quotient removes all the null states. On the 4d SCFT side, this is the VOA associated with D u ( G ) D_u(G) D u ( G ) , constructed by compactifying a six-dimensional ( 2 , 0 ) (2,0) ( 2 , 0 ) theory on a Riemann sphere with an irregular puncture and a full puncture. 14
Performing a further DS reduction on the affine vertex algebra L k ( g ) L_k(\mathfrak{g}) L k ( g ) gives a W W W -algebra. On the 4d SCFT side, this corresponds to partially closing the full puncture. To specify the DS reduction, one also needs to choose a nilpotent element. The nilpotent element is denoted as f f f , and can be completed to an s l 2 \mathfrak{sl}_2 sl 2 -triple ( e f , h f , f ) (e_f,h_f,f) ( e f , h f , f ) . We also set h f = h f / 2 \mathbf{h}_f=h_f/2 h f = h f /2 . In particular, [ h f , e f ] = e f [\mathbf{h}_f,e_f]=e_f [ h f , e f ] = e f and [ h f , f ] = − f [\mathbf{h}_f,f]=-f [ h f , f ] = − f . Let g ♮ \mathfrak{g}^{\natural} g ♮ denote the centralizer of the triple. Under the commuting action of s l 2 ⊕ g ♮ ⊂ g \mathfrak{sl}_2\oplus\mathfrak{g}^{\natural}\subset\mathfrak{g} sl 2 ⊕ g ♮ ⊂ g , the adjoint representation of G G G , which is identified
k + h g ∨ = n p with n , p ∈ Z ≥ 1 , ( p , n ) = 1 , n ≥ { h g ∨ if ( r ∨ , p ) = 1 , h g if ( r ∨ , p ) = r ∨ , k+h_g^{\vee}=\frac{n}{p}\qquad\text{with }n,p\in\mathbb{Z}_{\geq1},\qquad(p,n)=1,\qquad n\geq
\begin{cases}
h_g^{\vee} & \text{if }(r^{\vee},p)=1,\\
h_g & \text{if }(r^{\vee},p)=r^{\vee},
\end{cases} k + h g ∨ = p n with n , p ∈ Z ≥ 1 , ( p , n ) = 1 , n ≥ { h g ∨ h g if ( r ∨ , p ) = 1 , if ( r ∨ , p ) = r ∨ ,
where r ∨ r^{\vee} r ∨ is the lacing number. If k k k is admissible with ( r ∨ , p ) = 1 (r^{\vee},p)=1 ( r ∨ , p ) = 1 , we say that k k k is principal. If k k k is admissible with ( r ∨ , p ) = r ∨ (r^{\vee},p)=r^{\vee} ( r ∨ , p ) = r ∨ , we say that k k k is coprincipal. We say k k k is boundary admissible if k k k is admissible and n = h g ∨ n=h_g^{\vee} n = h g ∨ . with g \mathfrak{g} g , can be decomposed as
g = ⨁ j ∈ 1 2 Z ≥ 0 V j ⊗ R j , χ V j ( q ) = ∑ ℓ = − j j q ℓ = q − j − q j + 1 1 − q . (69) \mathfrak{g}=\bigoplus_{j\in\frac{1}{2}\mathbb{Z}_{\geq0}} V_j\otimes R_j,\qquad\chi_{V_j}(q)=\sum_{\ell=-j}^{j}q^\ell=\frac{q^{-j}-q^{j+1}}{1-q}. \tag*{(69)} g = j ∈ 2 1 Z ≥ 0 ⨁ V j ⊗ R j , χ V j ( q ) = ℓ = − j ∑ j q ℓ = 1 − q q − j − q j + 1 . (69)
Here V j V_j V j is the spin-j j j representation of s u ( 2 ) \mathfrak{su}(2) su ( 2 ) , namely dim V j = 2 j + 1 \dim V_j=2j+1 dim V j = 2 j + 1 , and R j R_j R j is a representation of g \mathfrak{g} g . We will use x \boldsymbol{x} x to denote the flavor fugacity of g \mathfrak{g} g .
The corresponding simple W W W -algebra after DS reduction specified by nilpotent element f f f is denoted as W k ( g , f ) \mathcal{W}_{k}(\mathfrak{g},f) W k ( g , f ) ,15 which should be distinguished from the universal algebra W k ( g , f ) \mathcal{W}^{k}(\mathfrak{g},f) W k ( g , f ) .
The character and index formula
We now present the vacuum-character formula for the W W W -algebra above. At boundary-admissible levels, these formulas are particularly simple.
The character is defined as16
χ W k ( g , f ) ( q , x ) = Tr W k ( g , f ) q L 0 x J 0 , (70) \chi_{\mathcal{W}_{k}(\mathfrak{g},f)}(q,\boldsymbol{x})=\operatorname{Tr}_{\mathcal{W}_{k}(\mathfrak{g},f)}q^{L_0}\boldsymbol{x}^{J_0}, \tag*{(70)} χ W k ( g , f ) ( q , x ) = Tr W k ( g , f ) q L 0 x J 0 , (70)
where we omit the factor q − c / 24 q^{-c/24} q − c /24 , and normalize the character such that it starts with 1 1 1 for the vacuum.
Let us introduce
A f ( q , x ) = ∑ j q j + 1 χ R j ( x ) , (71) A_f(q,\boldsymbol{x})=\sum_j q^{j+1}\chi_{R_j}(\boldsymbol{x}), \tag*{(71)} A f ( q , x ) = j ∑ q j + 1 χ R j ( x ) , (71)
B f ( q , x ) = χ a d j ( x q h f ) = ∑ j χ V j ( q ) χ R j ( x ) . (72) B_f(q,\boldsymbol{x})=\chi_{\mathrm{adj}}(\boldsymbol{x}q^{h_f})=\sum_j\chi_{V_j}(q)\chi_{R_j}(\boldsymbol{x}). \tag*{(72)} B f ( q , x ) = χ adj ( x q h f ) = j ∑ χ V j ( q ) χ R j ( x ) . (72)
The multiplicative notation x q h f \boldsymbol{x}q^{h_f} x q h f means that a weight λ \lambda λ contributes x λ q λ ( h f ) \boldsymbol{x}^{\lambda}q^{\lambda(h_f)} x λ q λ ( h f ) .
Then the character of the W W W -algebra is given by [23 eq. (36)]
χ W k ( u ) ( g , f ) ( q , x ) = PE [ A f ( q , x ) 1 − q − q u 1 − q u B f ( q , x ) ] , (73) \chi_{\mathcal{W}_{k(u)}(\mathfrak{g},f)}(q,\boldsymbol{x})=\operatorname{PE}\left[\frac{A_f(q,\boldsymbol{x})}{1-q}-\frac{q^u}{1-q^u}B_f(q,\boldsymbol{x})\right], \tag*{(73)} χ W k ( u ) ( g , f ) ( q , x ) = PE [ 1 − q A f ( q , x ) − 1 − q u q u B f ( q , x ) ] , (73)
where A f , B f A_f,B_f A f , B f are given by (71) and (72) .
Using the formula in (69) , this can be simplified further [23 eq. (39)]
χ W k ( u ) ( g , f ) ( q , x ) = PE [ ∑ j ( q j + 1 − q u − j ) χ R j ( x ) ( 1 − q ) ( 1 − q u ) ] = PE [ A f ( q , x ) − q u + 1 A f ( q − 1 , x ) ( 1 − q ) ( 1 − q u ) ] . (74) \chi_{W_{k}(u)}^{(\mathfrak{g},f)}(q,x)=\operatorname{PE}\left[\frac{\sum_{j}\left(q^{j+1}-q^{u-j}\right)\chi_{R_j}(x)}{(1-q)(1-q^u)}\right]=\operatorname{PE}\left[\frac{A_f(q,x)-q^{u+1}A_f(q^{-1},x)}{(1-q)(1-q^u)}\right]. \tag*{(74)} χ W k ( u ) ( g , f ) ( q , x ) = PE [ ( 1 − q ) ( 1 − q u ) ∑ j ( q j + 1 − q u − j ) χ R j ( x ) ] = PE [ ( 1 − q ) ( 1 − q u ) A f ( q , x ) − q u + 1 A f ( q − 1 , x ) ] . (74)
These character formulas allow us to derive the effective central charge17
c W k ( u ) e f f ( g , f ) = dim g f − dim g u , dim g f = ∑ j dim R j , (75) c^{\mathrm{eff}}_{W_{k}(u)}(\mathfrak{g},f)=\dim\mathfrak{g}^{f}-\frac{\dim\mathfrak{g}}{u},\qquad\dim\mathfrak{g}^{f}=\sum_j\dim R_j, \tag*{(75)} c W k ( u ) eff ( g , f ) = dim g f − u dim g , dim g f = j ∑ dim R j , (75)
where g f \mathfrak{g}^{f} g f is the centralizer of f f f , whereas g ♮ ≅ R 0 ⊂ g f \mathfrak{g}^{\natural}\cong R_0\subset\mathfrak{g}^{f} g ♮ ≅ R 0 ⊂ g f is the centralizer of the full s l 2 \mathfrak{sl}_2 sl 2 -triple and has dimension dim R 0 \dim R_0 dim R 0 .
If f f f is the principal nilpotent element, then g ♮ \mathfrak{g}^{\natural} g ♮ is trivial and we have the decomposition
g = ⨁ i = 1 r V d i − 1 , (76) \mathfrak{g}=\bigoplus_{i=1}^{r}V_{d_i-1}, \tag*{(76)} g = i = 1 ⨁ r V d i − 1 , (76)
where the d i d_i d i are the degrees of the Casimirs of g \mathfrak{g} g . See Table 1 for explicit values. In this case the character simplifies to
χ W k ( u ) ( g , f p r i n ) ( q ) = PE [ ∑ i = 1 r ( q d i − q u + 1 − d i ) ( 1 − q ) ( 1 − q u ) ] . (77) \chi_{W_{k}(u)}^{(\mathfrak{g},f_{\mathrm{prin}})}(q)=\operatorname{PE}\left[\frac{\sum_{i=1}^{r}\left(q^{d_i}-q^{u+1-d_i}\right)}{(1-q)(1-q^u)}\right]. \tag*{(77)} χ W k ( u ) ( g , f prin ) ( q ) = PE [ ( 1 − q ) ( 1 − q u ) ∑ i = 1 r ( q d i − q u + 1 − d i ) ] . (77)
If f f f is a trivial nilpotent element, we have g = g ♮ \mathfrak{g}=\mathfrak{g}^{\natural} g = g ♮ and only V 0 V_0 V 0 survives in (69) . In this case, we get the original affine vertex algebra, whose character is given by
χ L k ( u ) ( g ) ( Q , x ) = PE [ Q − Q u ( 1 − Q ) ( 1 − Q u ) χ a d ( x ) ] , (78) \chi_{L_{k}(u)}(\mathfrak{g})(Q,x)=\operatorname{PE}\left[\frac{Q-Q^u}{(1-Q)(1-Q^u)}\chi_{\mathrm{ad}}(x)\right], \tag*{(78)} χ L k ( u ) ( g ) ( Q , x ) = PE [ ( 1 − Q ) ( 1 − Q u ) Q − Q u χ ad ( x ) ] , (78)
where we use the symbol Q Q Q in place of q q q for later convenience.
In the two cases above, the effective central charges are
c L k ( u ) e f f ( g ) = ( 1 − 1 u ) dim g , c W k ( u ) e f f ( g , f p r i n ) = r − dim g u = dim g h + 1 − dim g u , (79) c^{\mathrm{eff}}_{L_{k}(u)}(\mathfrak{g})=\left(1-\frac{1}{u}\right)\dim\mathfrak{g},\qquad c^{\mathrm{eff}}_{W_{k}(u)}(\mathfrak{g},f_{\mathrm{prin}})=r-\frac{\dim\mathfrak{g}}{u}=\frac{\dim\mathfrak{g}}{h+1}-\frac{\dim\mathfrak{g}}{u}, \tag*{(79)} c L k ( u ) eff ( g ) = ( 1 − u 1 ) dim g , c W k ( u ) eff ( g , f prin ) = r − u dim g = h + 1 dim g − u dim g , (79)
where we used the formula dim g = r ( h + 1 ) \dim\mathfrak{g}=r(h+1) dim g = r ( h + 1 ) . As a consistency check, if u = h + 1 u=h+1 u = h + 1 , then the second formula gives vanishing effective central charge, which is expected as the corresponding principal W W W algebra is trivial.
Master equation
Now we show a very useful identity:
χ W − h ∨ + h ∨ ℓ ( g , f ) ( q , x ) = χ W − h ∨ + h ∨ ℓ ( g , f ) ( q , x ) χ L − h ∨ + h ∨ p ( g ) ( q ℓ , x q h f ) . (80) \chi_{W_{-h^\vee+\frac{h^\vee}{\ell}}}^{(\mathfrak{g},f)}(q,x)=\chi_{W_{-h^\vee+\frac{h^\vee}{\ell}}}^{(\mathfrak{g},f)}(q,x)\chi_{L_{-h^\vee+\frac{h^\vee}{p}}}(\mathfrak{g})(q^\ell,xq^{h_f}). \tag*{(80)} χ W − h ∨ + ℓ h ∨ ( g , f ) ( q , x ) = χ W − h ∨ + ℓ h ∨ ( g , f ) ( q , x ) χ L − h ∨ + p h ∨ ( g ) ( q ℓ , x q h f ) . (80)
To ensure the admissibility and nonvanishing of DS reduction, we require
ℓ , p ∈ Z ≥ 1 , gcd ( ℓ p , h ∨ r ∨ ) = 1 , ℓ > j max ≡ max { j : R j ≠ 0 } . (81) \ell,p\in\mathbb{Z}_{\geq1},\qquad\operatorname{gcd}(\ell p,h^\vee r^\vee)=1,\qquad\ell>j_{\max}\equiv\max\{j:R_j\ne0\}. \tag*{(81)} ℓ , p ∈ Z ≥ 1 , gcd ( ℓ p , h ∨ r ∨ ) = 1 , ℓ > j m a x ≡ max { j : R j = 0 } . (81)
The proof of (80) is simple once we use the explicit character formula. We consider the quotient of the characters of W − h ∨ + h ∨ ℓ p ( g , f ) \mathcal{W}_{-h^\vee+\frac{h^\vee}{\ell p}}(\mathfrak{g},f) W − h ∨ + ℓ p h ∨ ( g , f ) and W − h ∨ + h ∨ ℓ ( g , f ) \mathcal{W}_{-h^\vee+\frac{h^\vee}{\ell}}(\mathfrak{g},f) W − h ∨ + ℓ h ∨ ( g , f ) , and then take plethystic logarithms:
PL χ W − h ∨ + h ∨ ℓ p ( g , f ) χ W − h ∨ + h ∨ ℓ ( g , f ) = ( q ℓ 1 − q ℓ − q ℓ p 1 − q ℓ p ) B f = q ℓ − q ℓ p ( 1 − q ℓ ) ( 1 − q ℓ p ) χ a d j ( x q h f ) . (82) \operatorname{PL}\frac{\chi_{\mathcal{W}_{-h^\vee+\frac{h^\vee}{\ell p}}(\mathfrak{g},f)}}{\chi_{\mathcal{W}_{-h^\vee+\frac{h^\vee}{\ell}}(\mathfrak{g},f)}}=\left(\frac{q^\ell}{1-q^\ell}-\frac{q^{\ell p}}{1-q^{\ell p}}\right)B_f=\frac{q^\ell-q^{\ell p}}{(1-q^\ell)(1-q^{\ell p})}\chi_{\mathrm{adj}}(xq^{h_f}). \tag*{(82)} PL χ W − h ∨ + ℓ h ∨ ( g , f ) χ W − h ∨ + ℓ p h ∨ ( g , f ) = ( 1 − q ℓ q ℓ − 1 − q ℓ p q ℓ p ) B f = ( 1 − q ℓ ) ( 1 − q ℓ p ) q ℓ − q ℓ p χ adj ( x q h f ) . (82)
Note that the A f / ( 1 − q ) A_f/(1-q) A f / ( 1 − q ) terms cancel. The surviving term has exactly the structure of the affine character in (78) , after the substitutions Q → q ℓ Q\to q^\ell Q → q ℓ and x → x q h f x\to xq^{h_f} x → x q h f . This gives the proof of (80) .
The character identity (80) translates to the following index identity among three AD theories:
I D ℓ p ( G , Y ) ( q ; x ) = I D ℓ ( G , Y ) ( q ; x ) I D p ( G ) ( q ℓ ; x q h f Y ) , (83) \mathcal{I}_{D_{\ell p}(G,Y)}(q;x)=\mathcal{I}_{D_\ell(G,Y)}(q;x)\mathcal{I}_{D_p(G)}(q^\ell;xq^{h_f^Y}), \tag*{(83)} I D ℓ p ( G , Y ) ( q ; x ) = I D ℓ ( G , Y ) ( q ; x ) I D p ( G ) ( q ℓ ; x q h f Y ) , (83)
where certain conditions on the parameters are imposed implicitly as explained before.
Let us see a simple example of (80) by taking g = s l 2 \mathfrak{g}=\mathfrak{sl}_2 g = sl 2 and odd ℓ ≥ 5 \ell\geq5 ℓ ≥ 5 and odd p ≥ 3 p\geq3 p ≥ 3 , f = [ 2 ] f=[2] f = [ 2 ] . Then we have
I ( A 1 , A ℓ p − 3 ) ( q ) = I ( A 1 , D p ) ( q ℓ ; x = q 1 / 2 ) I ( A 1 , A ℓ − 3 ) ( q ) , (84) \mathcal{I}_{(A_1,A_{\ell p-3})}(q)=\mathcal{I}_{(A_1,D_p)}(q^\ell;x=q^{1/2})\mathcal{I}_{(A_1,A_{\ell-3})}(q), \tag*{(84)} I ( A 1 , A ℓ p − 3 ) ( q ) = I ( A 1 , D p ) ( q ℓ ; x = q 1/2 ) I ( A 1 , A ℓ − 3 ) ( q ) , (84)
where D p ( S U ( 2 ) ; [ 2 ] ) = ( A 1 , A p − 3 ) D_p(SU(2);[2])=(A_1,A_{p-3}) D p ( S U ( 2 ) ; [ 2 ]) = ( A 1 , A p − 3 ) and D p ( S U ( 2 ) ) = ( A 1 , D p ) D_p(SU(2))=(A_1,D_p) D p ( S U ( 2 )) = ( A 1 , D p ) .
The character identity in (80) involves three VOAs. It is natural to ask when we can simplify and establish identities between a pair of VOAs or SCFTs.
For this purpose, we use the following lemma. The following identity holds
I D ℓ p ( G , f ) ( q , x ) = I D p ( G ) ( q ℓ , x q h f ) , ℓ , p ≥ 2 , gcd ( ℓ p , h ∨ r ∨ ) = 1 , (85) \mathcal{I}_{D_{\ell p}(G,f)}(q,x)=\mathcal{I}_{D_p(G)}(q^\ell,xq^{h_f}),\qquad\ell,p\geq2,\quad\operatorname{gcd}(\ell p,h^\vee r^\vee)=1, \tag*{(85)} I D ℓ p ( G , f ) ( q , x ) = I D p ( G ) ( q ℓ , x q h f ) , ℓ , p ≥ 2 , gcd ( ℓ p , h ∨ r ∨ ) = 1 , (85)
with generic surviving flavor fugacities, if and only if
R j ≃ R ℓ − 1 − j , ∀ j . (86) R_j\simeq R_{\ell-1-j},\qquad\forall j. \tag*{(86)} R j ≃ R ℓ − 1 − j , ∀ j . (86)
Here R j R_j R j is given in the decomposition of adjoint of g \mathfrak{g} g in (69) . An absent R j R_j R j is understood to be zero.
The relation in (86) is equivalent to
A f ( q , x ) = q ℓ + 1 A f ( q − 1 , x ) , (87) A_f(q,x)=q^{\ell+1}A_f(q^{-1},x), \tag*{(87)} A f ( q , x ) = q ℓ + 1 A f ( q − 1 , x ) , (87)
where A f A_f A f is defined in (71) .
To prove (85) , we first note that (74) can be rewritten as
χ W − h ∨ + h ∨ ℓ ( g , f ) ( q , x ) = PE [ A f ( q , x ) − q ℓ + 1 A f ( q − 1 , x ) ( 1 − q ) ( 1 − q ℓ ) ] . (88) \chi\mathcal{W}_{-h^\vee+\frac{h^\vee}{\ell}}^{(g,f)}(q,x)=\operatorname{PE}\left[\frac{A_f(q,x)-q^{\ell+1}A_f(q^{-1},x)}{(1-q)(1-q^\ell)}\right]. \tag*{(88)} χ W − h ∨ + ℓ h ∨ ( g , f ) ( q , x ) = PE [ ( 1 − q ) ( 1 − q ℓ ) A f ( q , x ) − q ℓ + 1 A f ( q − 1 , x ) ] . (88)
which equals 1 if (87) holds. Further applying (80) gives
χ W − h ∨ + h ∨ ℓ p ( g , f ) ( q , x ) = χ L − h ∨ + h ∨ p ( g ) ( q ℓ , x q h f ) . (89) \chi\mathcal{W}_{-h^\vee+\frac{h^\vee}{\ell p}}^{(g,f)}(q,x)=\chi\mathcal{L}_{-h^\vee+\frac{h^\vee}{p}}^{(g)}(q^\ell,xq^{h^f}). \tag*{(89)} χ W − h ∨ + ℓ p h ∨ ( g , f ) ( q , x ) = χ L − h ∨ + p h ∨ ( g ) ( q ℓ , x q h f ) . (89)
This then proves the sufficient direction. The necessary condition can be shown similarly. This establishes the lemma
I D ℓ p ( G , f ) ( q , x ) = I D p ( G ) ( q ℓ , x q h f ) ⟺ R j ≅ R ℓ − 1 − j ∀ j . (90) \mathcal{I}D_{\ell p}^{(G,f)}(q,x)=\mathcal{I}D_p^{(G)}(q^\ell,xq^{h^f})\quad\Longleftrightarrow\quad R_j\cong R_{\ell-1-j}\quad\forall j. \tag*{(90)} I D ℓ p ( G , f ) ( q , x ) = I D p ( G ) ( q ℓ , x q h f ) ⟺ R j ≅ R ℓ − 1 − j ∀ j . (90)
Principal case: trivial puncture
For a principal nilpotent element f f f , corresponding to the completely closed trivial regular puncture, the character is given by (77)
χ W − h ∨ + h ∨ ℓ ( g , f p r i n ) ( q ) = PE [ ∑ i = 1 r ( q d i − q ℓ + 1 − d i ) ( 1 − q ) ( 1 − q ℓ ) ] . (91) \chi\mathcal{W}_{-h^\vee+\frac{h^\vee}{\ell}}(g,f_{\mathrm{prin}})(q)=\operatorname{PE}\left[\frac{\sum_{i=1}^{r}\left(q^{d_i}-q^{\ell+1-d_i}\right)}{(1-q)(1-q^\ell)}\right]. \tag*{(91)} χ W − h ∨ + ℓ h ∨ ( g , f prin ) ( q ) = PE [ ( 1 − q ) ( 1 − q ℓ ) ∑ i = 1 r ( q d i − q ℓ + 1 − d i ) ] . (91)
For ordered Casimir degrees d i d_i d i , they satisfy d i + d r + 1 − i = h + 2 d_i+d_{r+1-i}=h+2 d i + d r + 1 − i = h + 2 . This means that if we choose ℓ = h + 1 \ell=h+1 ℓ = h + 1 , then the numerator above vanishes. Therefore
χ W − h ∨ + h ∨ h + 1 ( g , f p r i n ) ( q ) = 1 , W − h ∨ + h ∨ h + 1 ( g , f p r i n ) ≅ C . (92) \chi\mathcal{W}_{-h^\vee+\frac{h^\vee}{h+1}}(g,f_{\mathrm{prin}})(q)=1,\qquad\mathcal{W}_{-h^\vee+\frac{h^\vee}{h+1}}(g,f_{\mathrm{prin}})\cong\mathbb{C}. \tag*{(92)} χ W − h ∨ + h + 1 h ∨ ( g , f prin ) ( q ) = 1 , W − h ∨ + h + 1 h ∨ ( g , f prin ) ≅ C . (92)
This comes from the fact that the trivial VOA has only one state given by the vacuum. Combining (92) with (80) gives the identity
χ W − h ∨ + h ∨ ( h + 1 ) p ( g , f p r i n ) ( q ) = χ L − h ∨ + h ∨ p ( g ) ( q h + 1 , q ρ ∨ ) , gcd ( p , h ∨ r ∨ ) = 1. (93) \chi\mathcal{W}_{-h^\vee+\frac{h^\vee}{(h+1)p}}(g,f_{\mathrm{prin}})(q)=\chi\mathcal{L}_{-h^\vee+\frac{h^\vee}{p}}^{(g)}(q^{h+1},q^{\rho^\vee}),\qquad\gcd(p,h^\vee r^\vee)=1. \tag*{(93)} χ W − h ∨ + ( h + 1 ) p h ∨ ( g , f prin ) ( q ) = χ L − h ∨ + p h ∨ ( g ) ( q h + 1 , q ρ ∨ ) , g cd( p , h ∨ r ∨ ) = 1. (93)
In particular, the condition in (81) is simplified as indicated above, because gcd ( h + 1 , h ∨ r ∨ ) = 1 \gcd(h+1,h^\vee r^\vee)=1 g cd( h + 1 , h ∨ r ∨ ) = 1 , which can be easily checked using the data in Table 1 .
The effective central charge provides a further consistency check via (79) .
c W − h ∨ + h ∨ ( h + 1 ) p ( g , f p r i n ) e f f = ( 1 h + 1 − 1 p ( h + 1 ) ) dim g (94) c^{\mathrm{eff}}_{\mathcal{W}_{-h^\vee+\frac{h^\vee}{(h+1)p}}(g,f_{\mathrm{prin}})}=\left(\frac{1}{h+1}-\frac{1}{p(h+1)}\right)\dim\mathfrak{g} \tag*{(94)} c W − h ∨ + ( h + 1 ) p h ∨ ( g , f prin ) eff = ( h + 1 1 − p ( h + 1 ) 1 ) dim g (94)
= 1 h + 1 ( 1 − 1 p ) dim g = 1 h + 1 c L − h ∨ + h ∨ p ( g ) e f f . (95) =\frac{1}{h+1}\left(1-\frac{1}{p}\right)\dim\mathfrak{g}=\frac{1}{h+1}c^{\mathrm{eff}}_{\mathcal{L}_{-h^\vee+\frac{h^\vee}{p}}^{(g)}}. \tag*{(95)} = h + 1 1 ( 1 − p 1 ) dim g = h + 1 1 c L − h ∨ + p h ∨ ( g ) eff . (95)
For simply laced G G G , ρ ∨ = ρ \rho^\vee=\rho ρ ∨ = ρ , r ∨ = 1 r^\vee=1 r ∨ = 1 , 18 the character identity translates into an index identity between a pair of untwisted AD theories
I D p ( G ) ( q h + 1 , q ρ ) = I ( G , A ( h + 1 ) ( p − 1 ) ) ( q ) , gcd ( p , h ) = 1. (96) \mathcal{I}_{D_p(G)}(q^{h+1},q^{\rho})=\mathcal{I}_{(G,A_{(h+1)(p-1)} )}(q),\qquad\gcd(p,h)=1. \tag*{(96)} I D p ( G ) ( q h + 1 , q ρ ) = I ( G , A ( h + 1 ) ( p − 1 ) ) ( q ) , g cd( p , h ) = 1. (96)
For non-simply-laced types, (93) is the unambiguous VOA statement. A four-dimensional interpretation also requires the twisted puncture conventions to be specified.
Non-principal case: non-trivial puncture
We now switch to the non-principal nilpotent element case, and try to relate its character with that of an affine vertex algebra. In the 4d SCFT, this corresponds to a non-trivial regular puncture, which can contribute to the flavor symmetry.
slN
We start with the case of g = s l N \mathfrak{g}=\mathfrak{sl}_N g = sl N . This case is particularly simple, and we can represent the regular puncture using the partition of N N N . For example, [ N ] [N] [ N ] corresponds to the principal nilpotent element that we considered in the previous subsection. Now, we want to consider more general partitions. We will not perform a systematic classification, but only focus on selected families.
Our strategy is to search for cases where one of the factors on the RHS of (80) is trivial, such that the resulting character identity only involves two VOAs.
For this purpose, we use the result in [24 Theorem 8.7] : if ℓ ≥ 2 \ell\ge2 ℓ ≥ 2 , s = 1 s=1 s = 1 or ℓ − 1 \ell-1 ℓ − 1 , then
W − N + N / ℓ ( s l N , [ ℓ m , s ] ) ≅ C , χ W − N + N / ℓ ( s l N , f ) = 1 , (97) \mathcal{W}_{-N+N/\ell}(\mathfrak{sl}_N,[\ell^m,s])\cong\mathbb{C},\qquad\chi_{\mathcal{W}_{-N+N/\ell}(\mathfrak{sl}_N,f)}=1, \tag*{(97)} W − N + N / ℓ ( sl N , [ ℓ m , s ]) ≅ C , χ W − N + N / ℓ ( sl N , f ) = 1 , (97)
where N = ℓ m + s N=\ell m+s N = ℓ m + s . Here N ≥ 2 N\ge2 N ≥ 2 , m ≥ 0 m\ge0 m ≥ 0 , and we have N ≡ ± 1 ( m o d ℓ ) N\equiv\pm1\pmod{\ell} N ≡ ± 1 ( mod ℓ ) . The triviality of character can be verified using (86) .
Combining (97) with (80) gives
χ W − N + N / ℓ ( s l N , f = [ ℓ m , s ] ) ( q , x ) = χ L − N + N / p ( s l N ) ( q ℓ , x q h f ) . (98) \chi_{\mathcal{W}_{-N+N/\ell}(\mathfrak{sl}_N,f=[\ell^m,s])}(q,x)=\chi_{\mathcal{L}_{-N+N/p}(\mathfrak{sl}_N)}(q^{\ell},xq^{h_f}). \tag*{(98)} χ W − N + N / ℓ ( sl N , f = [ ℓ m , s ]) ( q , x ) = χ L − N + N / p ( sl N ) ( q ℓ , x q h f ) . (98)
This gives the following identity between the flavored indices of the two SCFTs
I D ℓ p ( S U ( N ) , f = [ ℓ m , s ] ) ( q ; x ) = I D p ( S U ( N ) ) ( q ℓ ; x q h f ) . (99) \mathcal{I}_{D_{\ell p}(SU(N),f=[\ell^{m},s])}(q;\boldsymbol{x})=\mathcal{I}_{D_{p}(SU(N))}(q^{\ell};\boldsymbol{x}q^{h_f}). \tag*{(99)} I D ℓ p ( S U ( N ) , f = [ ℓ m , s ]) ( q ; x ) = I D p ( S U ( N )) ( q ℓ ; x q h f ) . (99)
Note that the theory D ℓ p ( S U ( N ) , [ ℓ m , s ] ) D_{\ell p}(SU(N),[\ell^{m},s]) D ℓ p ( S U ( N ) , [ ℓ m , s ]) on the LHS has symmetry S U ( m ) × U ( 1 ) SU(m)\times U(1) S U ( m ) × U ( 1 ) for m ≥ 1 m\geq1 m ≥ 1 , which may get further enhancement.
The effective central charge provides a further consistency check. In particular, for s = 1 s=1 s = 1 , we find that
c D ℓ p ( S U ( N ) , [ ℓ m , 1 ] ) = p − 1 12 [ ℓ ( N 2 − 1 ) − N m ( ℓ 2 − 1 ) p ] , (100) c_{D_{\ell p}(SU(N),[\ell^{m},1])}=\frac{p-1}{12}\left[\ell(N^{2}-1)-\frac{Nm(\ell^{2}-1)}{p}\right], \tag*{(100)} c D ℓ p ( S U ( N ) , [ ℓ m , 1 ]) = 12 p − 1 [ ℓ ( N 2 − 1 ) − p N m ( ℓ 2 − 1 ) ] , (100)
a D ℓ p ( S U ( N ) , [ ℓ m , 1 ] ) = c D ℓ p ( S U ( N ) , [ ℓ m , 1 ] ) − ( p − 1 ) m ( N + 1 ) 48 p . (101) a_{D_{\ell p}(SU(N),[\ell^{m},1])}=c_{D_{\ell p}(SU(N),[\ell^{m},1])}-\frac{(p-1)m(N+1)}{48p}. \tag*{(101)} a D ℓ p ( S U ( N ) , [ ℓ m , 1 ]) = c D ℓ p ( S U ( N ) , [ ℓ m , 1 ]) − 48 p ( p − 1 ) m ( N + 1 ) . (101)
Therefore the effective central charge is
c D ℓ p ( S U ( N ) , [ ℓ m , 1 ] ) e f f = ( p − 1 ) ( N 2 − 1 ) ℓ p = 1 ℓ c D p ( S U ( N ) ) e f f , (102) c^{\mathrm{eff}}_{D_{\ell p}(SU(N),[\ell^{m},1])}=\frac{(p-1)(N^{2}-1)}{\ell p}=\frac{1}{\ell}c^{\mathrm{eff}}_{D_p(SU(N))}, \tag*{(102)} c D ℓ p ( S U ( N ) , [ ℓ m , 1 ]) eff = ℓ p ( p − 1 ) ( N 2 − 1 ) = ℓ 1 c D p ( S U ( N )) eff , (102)
where we used (52) . For s = ℓ − 1 s=\ell-1 s = ℓ − 1 , one can show that the effective central charge is the same, which gives the same relation with the parent theory.
If we choose ℓ = N + 1 , m = 0 , s = ℓ − 1 = N \ell=N+1,m=0,s=\ell-1=N ℓ = N + 1 , m = 0 , s = ℓ − 1 = N , the nilpotent element on the LHS becomes the principal one, and we recover the previous identity (42) between principal W W W algebra and affine algebra.
Since ℓ m + s = N \ell m+s=N ℓ m + s = N and s = 1 s=1 s = 1 or ℓ − 1 \ell-1 ℓ − 1 , it is easy to see that ℓ ∣ ( N − 1 ) \ell\mid(N-1) ℓ ∣ ( N − 1 ) or ℓ ∣ ( N + 1 ) \ell\mid(N+1) ℓ ∣ ( N + 1 ) , depending on the value of s s s . Therefore, there are several values of ℓ \ell ℓ , and we could establish the index identity for the family of theories D ℓ p ( S U ( N ) , f = [ ℓ m , s ] ) D_{\ell p}(SU(N),f=[\ell^{m},s]) D ℓ p ( S U ( N ) , f = [ ℓ m , s ]) with s ∈ { 1 , ℓ − 1 } s\in\{1,\ell-1\} s ∈ { 1 , ℓ − 1 } , labelled by ℓ \ell ℓ . Two members in the family are related by
I D ℓ p ( S U ( N ) , [ ℓ m , s ] ) ( q ; x ) = I D ℓ ^ p ^ ( S U ( N ) , [ ℓ ^ m ^ , s ^ ] ) ( q ℓ / ℓ ^ ; x ^ ) , ℓ m + s = ℓ ^ m ^ + s ^ , (103) \mathcal{I}_{D_{\ell p}(SU(N),[\ell^{m},s])}(q;\boldsymbol{x})=\mathcal{I}_{D_{\hat{\ell}\hat{p}(SU(N),[\hat{\ell}^{\hat{m}},\hat{s}])}}(q^{\ell/\hat{\ell}};\hat{\boldsymbol{x}}),\qquad\ell m+s=\hat{\ell}\hat{m}+\hat{s}, \tag*{(103)} I D ℓ p ( S U ( N ) , [ ℓ m , s ]) ( q ; x ) = I D ℓ ^ p ^ ( S U ( N ) , [ ℓ ^ m ^ , s ^ ]) ( q ℓ / ℓ ^ ; x ^ ) , ℓ m + s = ℓ ^ m ^ + s ^ , (103)
where the flavor fugacities x ^ \hat{\boldsymbol{x}} x ^ and x \boldsymbol{x} x are related in a specific way, and the identity holds when the two sides share a compatible specialization of flavor symmetry.
In particular, in (103) we can choose ℓ = N + 1 , m = 0 , s = ℓ − 1 = N \ell=N+1,m=0,s=\ell-1=N ℓ = N + 1 , m = 0 , s = ℓ − 1 = N , which implies that the nilpotent element is principal on the LHS. This gives the following index identity
I D n p ( S U ( N ) , f p r i n ) ( q ) = I D p ( S U ( N ) , [ t n − 1 , t − 1 ] ) ( q n ; { q j − n / 2 } j = 1 n − 1 ; 1 ) , (104) \mathcal{I}_{D_{np}(SU(N),f_{\mathrm{prin}})}(q)=\mathcal{I}_{D_p(SU(N),[t^{n-1},t-1])}\left(q^{n};\{q^{j-n/2}\}_{j=1}^{n-1};1\right), \tag*{(104)} I D n p ( S U ( N ) , f prin ) ( q ) = I D p ( S U ( N ) , [ t n − 1 , t − 1 ]) ( q n ; { q j − n /2 } j = 1 n − 1 ; 1 ) , (104)
where p ≥ t p\geq t p ≥ t , gcd ( p , N ) = 1 \gcd(p,N)=1 g cd( p , N ) = 1 , t n = N + 1 tn=N+1 t n = N + 1 and we have also provided the explicit flavor fugacity map. Further setting t = 1 t=1 t = 1 yields a trivial nilpotent element on the RHS, and we again recover the identity in (42) .
If we set n = 2 n=2 n = 2 in (104) , then
I D 2 p ( S U ( N ) , f p r i n ) ( q ) = I D p ( S U ( N ) , [ N + 1 2 , N − 1 2 ] ) ( q 2 ; 1 ) . (105) \mathcal{I}_{D_{2p}(SU(N),f_{\mathrm{prin}})}(q)=\mathcal{I}_{D_p(SU(N),[\frac{N+1}{2},\frac{N-1}{2}])}(q^{2};1). \tag*{(105)} I D 2 p ( S U ( N ) , f prin ) ( q ) = I D p ( S U ( N ) , [ 2 N + 1 , 2 N − 1 ]) ( q 2 ; 1 ) . (105)
In a more standard notation, this is just
I ( A 2 m , A 2 n ) ( q ) = I D m + n + 1 ( s l 2 m + 1 , [ m + 1 , m ] ) ( q 2 ; 1 ) = I D ( m + n + 1 ) / ( m + 1 ) ( s l 2 m + 1 ) ( q 2 m + 2 ; q i − ( m + 1 ) ) . (106) \mathcal{I}_{(A_{2m},A_{2n})}(q)=\mathcal{I}_{D_{m+n+1}(\mathfrak{sl}_{2m+1},[m+1,m])}(q^2;1)=\mathcal{I}_{D_{(m+n+1)/(m+1)}(\mathfrak{sl}_{2m+1})}(q^{2m+2};q^{i-(m+1)}). \tag*{(106)} I ( A 2 m , A 2 n ) ( q ) = I D m + n + 1 ( sl 2 m + 1 , [ m + 1 , m ]) ( q 2 ; 1 ) = I D ( m + n + 1 ) / ( m + 1 ) ( sl 2 m + 1 ) ( q 2 m + 2 ; q i − ( m + 1 ) ) . (106)
for m , n ∈ Z + m,n\in\mathbb{Z}_{+} m , n ∈ Z + and gcd ( 2 m + 1 , 2 n + 1 ) = 1 \operatorname{gcd}(2m+1,2n+1)=1 gcd ( 2 m + 1 , 2 n + 1 ) = 1 , where we also include the identity in (42) with the extra condition ( m + 1 ) ∣ n (m+1)\mid n ( m + 1 ) ∣ n .
Since the LHS is invariant under the exchange of m , n m,n m , n , the first expression on the right should share this symmetry. In fact, using (74) , one can show that the symmetry persists with the flavor fugacity turned on:
I D m + n + 1 ( s l 2 m + 1 , [ m + 1 , m ] ) ( q ; w ) = I D m + n + 1 ( s l 2 n + 1 , [ n + 1 , n ] ) ( q ; w ) (107) \mathcal{I}_{D_{m+n+1}(\mathfrak{sl}_{2m+1},[m+1,m])}(q;w)=\mathcal{I}_{D_{m+n+1}(\mathfrak{sl}_{2n+1},[n+1,n])}(q;w) \tag*{(107)} I D m + n + 1 ( sl 2 m + 1 , [ m + 1 , m ]) ( q ; w ) = I D m + n + 1 ( sl 2 n + 1 , [ n + 1 , n ]) ( q ; w ) (107)
= PE [ q ( 1 − q m ) ( 1 − q n ) [ 1 + q + ( w + w − 1 ) q 1 / 2 ] ( 1 − q ) 2 ( 1 − q m + n + 1 ) ] , (108) =\operatorname{PE}\left[\frac{q(1-q^m)(1-q^n)\left[1+q+(w+w^{-1})q^{1/2}\right]}{(1-q)^2(1-q^{m+n+1})}\right], \tag*{(108)} = PE [ ( 1 − q ) 2 ( 1 − q m + n + 1 ) q ( 1 − q m ) ( 1 − q n ) [ 1 + q + ( w + w − 1 ) q 1/2 ] ] , (108)
where w w w is the U ( 1 ) U(1) U ( 1 ) fugacity and gcd ( 2 m + 1 , 2 n + 1 ) = 1 \operatorname{gcd}(2m+1,2n+1)=1 gcd ( 2 m + 1 , 2 n + 1 ) = 1 .
Similarly, if we consider s = 1 s=1 s = 1 in (99) , we can also derive
I D n p ( S U ( N ) , [ N − 1 , 1 ] ) ( q , w N ) = I D p ( S U ( N ) , [ t n , 1 ] ) ( q n ; { y j = w q j − ( n + 1 ) / 2 } j = 1 n ; w − ( N − 1 ) ) , (109) \mathcal{I}_{D_{np}(SU(N),[N-1,1])}(q,w^N)=\mathcal{I}_{D_p(SU(N),[tn,1])}\left(q^n;\{y_j=wq^{j-(n+1)/2}\}_{j=1}^{n};w^{-(N-1)}\right), \tag*{(109)} I D n p ( S U ( N ) , [ N − 1 , 1 ]) ( q , w N ) = I D p ( S U ( N ) , [ t n , 1 ]) ( q n ; { y j = w q j − ( n + 1 ) /2 } j = 1 n ; w − ( N − 1 ) ) , (109)
where t n = N − 1 tn=N-1 t n = N − 1 , p ≥ t p\geq t p ≥ t and gcd ( p , N ) = 1 \operatorname{gcd}(p,N)=1 gcd ( p , N ) = 1 . If n = 2 n=2 n = 2 , then N = 2 t + 1 N=2t+1 N = 2 t + 1 and
I D 2 p ( S U ( N ) , f s u b r e g = [ N − 1 , 1 ] ) ( q , w N ) = I D p ( S U ( N ) , [ t 2 , 1 ] ) ( q 2 ; w q − 1 / 2 , w q 1 / 2 ; w − ( N − 1 ) ) . (110) \mathcal{I}_{D_{2p}(SU(N),f_{\mathrm{subreg}=[N-1,1])}}(q,w^N)=\mathcal{I}_{D_p(SU(N),[t^2,1])}\left(q^2;wq^{-1/2},wq^{1/2};w^{-(N-1)}\right). \tag*{(110)} I D 2 p ( S U ( N ) , f subreg = [ N − 1 , 1 ]) ( q , w N ) = I D p ( S U ( N ) , [ t 2 , 1 ]) ( q 2 ; w q − 1/2 , w q 1/2 ; w − ( N − 1 ) ) . (110)
Let us be more concrete and take N = 3 N=3 N = 3 , p = 2 p=2 p = 2 , n = 2 n=2 n = 2 , t = 1 t=1 t = 1 , f = [ 2 , 1 ] f=[2,1] f = [ 2 , 1 ] . This then gives
I ( A 1 , A 5 ) ( q , w 3 ) = I ( A 1 , D 4 ) ( q 2 ; w q 1 / 2 , w q − 1 / 2 ; w − 2 ) , (111) \mathcal{I}_{(A_1,A_5)}(q,w^3)=\mathcal{I}_{(A_1,D_4)}(q^2;wq^{1/2},wq^{-1/2};w^{-2}), \tag*{(111)} I ( A 1 , A 5 ) ( q , w 3 ) = I ( A 1 , D 4 ) ( q 2 ; w q 1/2 , w q − 1/2 ; w − 2 ) , (111)
where we have used the identifications
D 4 ( S U ( 3 ) , [ 2 , 1 ] ) = ( A 1 , A 5 ) , D 2 ( S U ( 3 ) , [ 1 , 1 , 1 ] ) = ( A 1 , D 4 ) . (112) D_4(SU(3),[2,1])=(A_1,A_5),\qquad D_2(SU(3),[1,1,1])=(A_1,D_4). \tag*{(112)} D 4 ( S U ( 3 ) , [ 2 , 1 ]) = ( A 1 , A 5 ) , D 2 ( S U ( 3 ) , [ 1 , 1 , 1 ]) = ( A 1 , D 4 ) . (112)
It turns out this can be generalized further, as we will discuss in subsection 5.7.
Orthogonal s o N \mathfrak{so}_N so N
There is another useful result proved in [24 Theorem 9.14(1)] : The following VOA is trivial
W − h ∨ + h ∨ ℓ ( s o N , f ) = C , (113) \mathcal{W}_{-h^\vee+\frac{h^\vee}{\ell}}(\mathfrak{so}_N,f)=\mathbb{C}, \tag*{(113)} W − h ∨ + ℓ h ∨ ( so N , f ) = C , (113)
for an odd integer ℓ ∈ 2 Z + 1 \ell\in2\mathbb{Z}+1 ℓ ∈ 2 Z + 1 coprime to h ∨ h^\vee h ∨ and a nilpotent element f f f corresponding to partition [ ℓ m , ϵ ] [\ell^m,\epsilon] [ ℓ m , ϵ ] with ϵ ∈ { 0 , 1 } \epsilon\in\{0,1\} ϵ ∈ { 0 , 1 } . Here N = ℓ m + ϵ N=\ell m+\epsilon N = ℓ m + ϵ , ℓ ≥ 3 \ell\geq3 ℓ ≥ 3 , and a zero part is omitted.
This gives χ W − h ∨ + h ∨ ℓ ( s o N , f ) = 1 \chi_{\mathcal{W}_{-h^\vee+\frac{h^\vee}{\ell}}(\mathfrak{so}_N,f)}=1 χ W − h ∨ + ℓ h ∨ ( so N , f ) = 1 . Combining this with (80) gives
χ W − h ∨ + h ∨ ℓ p ( s o N , f ) ( q , x ) = χ L − h ∨ + h ∨ p ( s o N ) ( q ℓ , x q h f ) . (114) \chi_{\mathcal{W}_{-h^\vee+\frac{h^\vee}{\ell p}}(\mathfrak{so}_N,f)}(q,x)=\chi_{\mathcal{L}_{-h^\vee+\frac{h^\vee}{p}}(\mathfrak{so}_N)}(q^\ell,xq^{h_f}). \tag*{(114)} χ W − h ∨ + ℓ p h ∨ ( so N , f ) ( q , x ) = χ L − h ∨ + p h ∨ ( so N ) ( q ℓ , x q h f ) . (114)
For even N N N , this gives untwisted type-D AD index identities:
I D ℓ p ( S O ( N ) , f ) ( q , x ) = I D p ( S O ( N ) ) ( q ℓ , x q h f ∨ ) , (115) \mathcal{I}_{D\ell p}(SO(N),f)(q,\mathbf{x})=\mathcal{I}_{Dp}(SO(N))\left(q^{\ell},xq^{h^{\vee}_{f}}\right), \tag*{(115)} I D ℓ p ( S O ( N ) , f ) ( q , x ) = I D p ( S O ( N )) ( q ℓ , x q h f ∨ ) , (115)
for f = [ ℓ m , ε ] f=[\ell^{m},\varepsilon] f = [ ℓ m , ε ] with ε ∈ { 0 , 1 } \varepsilon\in\{0,1\} ε ∈ { 0 , 1 } , ℓ ∈ 2 N + 1 \ell\in2\mathbb{N}+1 ℓ ∈ 2 N + 1 .
Symplectic s p N \mathfrak{sp}_N sp N
There is another useful result proved in [24 Theorem 9.7(1)] : The following VOA is trivial
W − h ∨ + h ∨ ℓ ( s p N , f ) = C , h s p N ∨ = N 2 + 1 , (116) \mathcal{W}_{-h^{\vee}+\frac{h^{\vee}}{\ell}}(\mathfrak{sp}_N,f)=\mathbb{C},\qquad h^{\vee}_{\mathfrak{sp}_N}=\frac{N}{2}+1, \tag*{(116)} W − h ∨ + ℓ h ∨ ( sp N , f ) = C , h sp N ∨ = 2 N + 1 , (116)
for an odd integer ℓ \ell ℓ coprime to h s p N ∨ h^{\vee}_{\mathfrak{sp}_N} h sp N ∨ and nilpotent element f = [ ℓ m , ε ] f=[\ell^{m},\varepsilon] f = [ ℓ m , ε ] with ε = 0 , ℓ − 1 \varepsilon=0,\ell-1 ε = 0 , ℓ − 1 . Note that both N = ℓ m + ε N=\ell m+\varepsilon N = ℓ m + ε and m m m are even.19
This gives χ W − h ∨ + h ∨ ℓ ( s p N , f ) = 1 \chi_{\mathcal{W}_{-h^{\vee}+\frac{h^{\vee}}{\ell}}(\mathfrak{sp}_N,f)}=1 χ W − h ∨ + ℓ h ∨ ( sp N , f ) = 1 . Combining this with (80) gives
χ W − h ∨ + h ∨ ℓ p ( s p N , f ) ( q , x ) = χ L − h ∨ + h ∨ p ( s p N ) ( q ℓ , x q h f ∨ ) , (117) \chi_{\mathcal{W}_{-h^{\vee}+\frac{h^{\vee}}{\ell p}}(\mathfrak{sp}_N,f)}(q,\mathbf{x})=\chi_{\mathcal{L}_{-h^{\vee}+\frac{h^{\vee}}{p}}(\mathfrak{sp}_N)}\left(q^{\ell},\mathbf{x}q^{h_f^{\vee}}\right), \tag*{(117)} χ W − h ∨ + ℓ p h ∨ ( sp N , f ) ( q , x ) = χ L − h ∨ + p h ∨ ( sp N ) ( q ℓ , x q h f ∨ ) , (117)
for f = [ ℓ m , ε ] f=[\ell^{m},\varepsilon] f = [ ℓ m , ε ] with ε = 0 , ℓ − 1 \varepsilon=0,\ell-1 ε = 0 , ℓ − 1 .
Exceptional E 7 E_7 E 7
We also study the exceptional case, and use [24 Theorem 10.6] :20
W − 18 + 18 7 ( E 7 , A 6 ) ≅ C , χ W − 18 + 18 7 ( E 7 , A 6 ) = 1 , (118) \mathcal{W}_{-18+\frac{18}{7}}(E_7,A_6)\cong\mathbb{C},\qquad\chi_{\mathcal{W}_{-18+\frac{18}{7}}(E_7,A_6)}=1, \tag*{(118)} W − 18 + 7 18 ( E 7 , A 6 ) ≅ C , χ W − 18 + 7 18 ( E 7 , A 6 ) = 1 , (118)
where A 6 A_6 A 6 denotes the Bala–Carter label of the nilpotent orbit in E 7 E_7 E 7 .
Applying this to (80) , we get the character identity
χ W − 18 + 18 7 p ( E 7 , A 6 ) ( q , x ) = χ L − 18 + 18 p ( E 7 ) ( q 7 , x q h A 6 ∨ ) . (119) \chi_{\mathcal{W}_{-18+\frac{18}{7p}}(E_7,A_6)}(q,\mathbf{x})=\chi_{\mathcal{L}_{-18+\frac{18}{p}}(E_7)}\left(q^7,\mathbf{x}q^{h_{A_6}^{\vee}}\right). \tag*{(119)} χ W − 18 + 7 p 18 ( E 7 , A 6 ) ( q , x ) = χ L − 18 + p 18 ( E 7 ) ( q 7 , x q h A 6 ∨ ) . (119)
This translates to the following index identity
I D 7 p ( E 7 , A 6 ) ( q ; y ) = I D p ( E 7 ) ( q 7 ; y q h A 6 ∨ ) , gcd ( p , 18 ) = 1. (120) \mathcal{I}_{D7p}(E_7,A_6)(q;y)=\mathcal{I}_{Dp}(E_7)\left(q^7;yq^{h_{A_6}^{\vee}}\right),\qquad\gcd(p,18)=1. \tag*{(120)} I D 7 p ( E 7 , A 6 ) ( q ; y ) = I D p ( E 7 ) ( q 7 ; y q h A 6 ∨ ) , g cd( p , 18 ) = 1. (120)
A non-boundary admissible family
We now consider the following pair of VOAs
V t , p = W − 3 + t p ( s l 3 , f [ 2 , 1 ] ) , W t , p = W − 3 + t 2 p ( s l 3 , f p r i n ) = W 3 ( t , 2 p ) , (121) V_{t,p}=\mathcal{W}_{-3+\frac{t}{p}}(\mathfrak{sl}_3,f_{[2,1]}),\qquad W_{t,p}=\mathcal{W}_{-3+\frac{t}{2p}}(\mathfrak{sl}_3,f_{\mathrm{prin}})=W_3(t,2p), \tag*{(121)} V t , p = W − 3 + p t ( sl 3 , f [ 2 , 1 ] ) , W t , p = W − 3 + 2 p t ( sl 3 , f prin ) = W 3 ( t , 2 p ) , (121)
which correspond to the Bershadsky–Polyakov algebra and W 3 W_3 W 3 minimal model, respectively. We also impose the following constraints on the parameters
3 ≤ t ∈ 2 Z + 1 , p ≥ 2 , gcd ( t , p ) = 1. (122) 3 \le t \in2\mathbb{Z}+1,\qquad p \ge2,\qquad\gcd(t,p)=1. \tag*{(122)} 3 ≤ t ∈ 2 Z + 1 , p ≥ 2 , g cd( t , p ) = 1. (122)
Since t t t can differ from 3, these W algebras are admissible, but not necessarily boundary admissible. Their central charges are
c V t , p = 25 − 6 t p − 24 p t , c W t , p = 50 − 12 t p − 48 p t . (123) c_{V_{t,p}}=25-\frac{6t}{p}-\frac{24p}{t},\qquad c_{W_{t,p}}=50-\frac{12t}{p}-\frac{48p}{t}. \tag*{(123)} c V t , p = 25 − p 6 t − t 24 p , c W t , p = 50 − p 12 t − t 48 p . (123)
Obviously, we have the relation
c W t , p = 2 c V t , p . (124) c_{W_{t,p}}=2c_{V_{t,p}}. \tag*{(124)} c W t , p = 2 c V t , p . (124)
This naturally suggests a possible character identity relating the two following (31) . We claim that the following identity holds
χ V t , p ( q 2 , 1 ) = χ W t , p ( q ) . (125) \chi_{V_{t,p}}(q^2,1)=\chi_{W_{t,p}}(q). \tag*{(125)} χ V t , p ( q 2 , 1 ) = χ W t , p ( q ) . (125)
This can be proved if we use the following character formulae21
χ V t , p ( q , w ) = N t , p ( q 1 / 2 , w ) ( q ; q ) ∞ 2 ( w q 1 / 2 ; q ) ∞ ( w − 1 q 1 / 2 ; q ) ∞ , χ W 3 ( t , 2 p ) ( q ) = N t , p ( q , 1 ) ( q ; q ) ∞ . (126) \chi_{V_{t,p}}(q,w)=\frac{\mathcal{N}_{t,p}(q^{1/2},w)}{(q;q)_\infty^2(wq^{1/2};q)_\infty(w^{-1}q^{1/2};q)_\infty},\qquad
\chi_{W_3(t,2p)}(q)=\frac{\mathcal{N}_{t,p}(q,1)}{(q;q)_\infty}. \tag*{(126)} χ V t , p ( q , w ) = ( q ; q ) ∞ 2 ( w q 1/2 ; q ) ∞ ( w − 1 q 1/2 ; q ) ∞ N t , p ( q 1/2 , w ) , χ W 3 ( t , 2 p ) ( q ) = ( q ; q ) ∞ N t , p ( q , 1 ) . (126)
where
N t , p ( ξ , w ) = ∑ m , n ∈ Z ξ 2 t p ( m 2 + n 2 − m n ) − t ( m + n ) w t ( n − m ) [ ξ 2 p ( m + n ) − ξ 4 − 2 p ( m + n ) − w ξ 1 + 2 p ( − m + 2 n ) − w − 1 ξ 1 + 2 p ( 2 m − n ) + w ξ 3 + 2 p ( − 2 m + n ) + w − 1 ξ 3 + 2 p ( m − 2 n ) ] . (127) \mathcal{N}_{t,p}(\xi,w)=\sum_{m,n\in\mathbb{Z}}\xi^{2tp(m^2+n^2-mn)-t(m+n)}w^{t(n-m)}
\left[\xi^{2p(m+n)}-\xi^{4-2p(m+n)}-w\xi^{1+2p(-m+2n)}-w^{-1}\xi^{1+2p(2m-n)}+w\xi^{3+2p(-2m+n)}+w^{-1}\xi^{3+2p(m-2n)}\right]. \tag*{(127)} N t , p ( ξ , w ) = m , n ∈ Z ∑ ξ 2 tp ( m 2 + n 2 − mn ) − t ( m + n ) w t ( n − m ) [ ξ 2 p ( m + n ) − ξ 4 − 2 p ( m + n ) − w ξ 1 + 2 p ( − m + 2 n ) − w − 1 ξ 1 + 2 p ( 2 m − n ) + w ξ 3 + 2 p ( − 2 m + n ) + w − 1 ξ 3 + 2 p ( m − 2 n ) ] . (127)
Under the conditions (122) , the identity therefore provides a family of character identities beyond the boundary admissible case.
To make contact with 4d AD theories, we specialize to the case t = 3 t=3 t = 3 and gcd ( 3 , p ) = 1 \gcd(3,p)=1 g cd( 3 , p ) = 1 , which corresponds to the boundary admissible case. We also take p > 2 p>2 p > 2 in order to get a negative 2d and hence a positive 4d central charge.22
Then, the two AD theories are given by
T p = D p ( S U ( 3 ) , [ 2 , 1 ] ) = ( A 2 3 [ p − 3 ] , [ 2 , 1 ] ) ↔ ( A 2 , A 2 p − 4 ) . (128) T_p=D_p(SU(3),[2,1])=(A_2^3[p-3],[2,1])\ \leftrightarrow\ (A_2,A_{2p-4}). \tag*{(128)} T p = D p ( S U ( 3 ) , [ 2 , 1 ]) = ( A 2 3 [ p − 3 ] , [ 2 , 1 ]) ↔ ( A 2 , A 2 p − 4 ) . (128)
The first theory is obtained from D p ( S U ( 3 ) ) D_p(SU(3)) D p ( S U ( 3 )) by partially closing the full puncture to [2 , 1 ] . Their associated VOAs are
V [ T p ] = W − 3 + 3 / p ( s l 3 , f [ 2 , 1 ] ) = V 3 , p , (129) \mathbb{V}[\mathcal{T}_p] = \mathcal{W}_{-3+3/p}(\mathfrak{sl}_3,f_{[2,1]}) = V_{3,p}, \tag*{(129)} V [ T p ] = W − 3 + 3/ p ( sl 3 , f [ 2 , 1 ] ) = V 3 , p , (129)
and
V [ ( A 2 , A 2 p − 4 ) ] = W − 3 + 3 / ( 2 p ) ( s l 3 , f p r i n ) = W 3 ( 3 , 2 p ) . (130) \mathbb{V}[(A_2,A_{2p-4})] = \mathcal{W}_{-3+3/(2p)}(\mathfrak{sl}_3,f_{\mathrm{prin}}) = W_3(3,2p). \tag*{(130)} V [( A 2 , A 2 p − 4 )] = W − 3 + 3/ ( 2 p ) ( sl 3 , f prin ) = W 3 ( 3 , 2 p ) . (130)
Under the [2 , 1 ] embedding, the adjoint decomposes as
s l 3 ≅ ( V 1 ) 0 ⊕ ( V 0 ) 0 ⊕ ( V 1 / 2 ) 1 ⊕ ( V 1 / 2 ) − 1 , (131) \mathfrak{sl}_3 \cong(V_1)_0 \oplus(V_0)_0 \oplus(V_{1/2})_1 \oplus(V_{1/2})_{-1}, \tag*{(131)} sl 3 ≅ ( V 1 ) 0 ⊕ ( V 0 ) 0 ⊕ ( V 1/2 ) 1 ⊕ ( V 1/2 ) − 1 , (131)
where the outer subscripts denote the U(1) charges. The boundary-admissible index formula then gives
I T p ( q , w ) = PE [ q + q 2 + ( w + w − 1 ) q 3 / 2 − q p − 1 − q p − ( w + w − 1 ) q p − 1 / 2 ( 1 − q ) ( 1 − q p ) ] , (132) \mathcal{I}_{\mathcal{T}_p}(q,w) = \operatorname{PE}\left[\frac{q+q^2+(w+w^{-1})q^{3/2}-q^{p-1}-q^p-(w+w^{-1})q^{p-1/2}}{(1-q)(1-q^p)}\right], \tag*{(132)} I T p ( q , w ) = PE [ ( 1 − q ) ( 1 − q p ) q + q 2 + ( w + w − 1 ) q 3/2 − q p − 1 − q p − ( w + w − 1 ) q p − 1/2 ] , (132)
which can also be written as
I T p ( q , w ) = PE [ ( q + ( w + w − 1 ) q 3 / 2 + q 2 ) ( 1 − q p − 2 ) ( 1 − q ) ( 1 − q p ) ] = PE [ ( q + ( w + w − 1 ) q 3 / 2 + q 2 ) ∑ j = 0 p − 3 q j 1 − q p ] . (133) \begin{aligned}
\mathcal{I}_{\mathcal{T}_p}(q,w) &= \operatorname{PE}\left[\frac{(q+(w+w^{-1})q^{3/2}+q^2)(1-q^{p-2})}{(1-q)(1-q^p)}\right]\\
&= \operatorname{PE}\left[\frac{(q+(w+w^{-1})q^{3/2}+q^2)\sum_{j=0}^{p-3}q^j}{1-q^p}\right]. \tag*{(133)}
\end{aligned} I T p ( q , w ) = PE [ ( 1 − q ) ( 1 − q p ) ( q + ( w + w − 1 ) q 3/2 + q 2 ) ( 1 − q p − 2 ) ] = PE [ 1 − q p ( q + ( w + w − 1 ) q 3/2 + q 2 ) ∑ j = 0 p − 3 q j ] . (133)
Comparing with (38) , we obtain
I D p ( s l 3 , [ 2 , 1 ] ) ( q 2 , 1 ) = PE [ ( q 2 + q 3 ) ( 1 − q 2 p − 4 ) ( 1 − q ) ( 1 − q 2 p ) ] = I ( A 2 , A 2 p − 4 ) ( q ) . (134) \mathcal{I}_{D_p(\mathfrak{sl}_3,[2,1])}(q^2,1)=\operatorname{PE}\left[\frac{(q^2+q^3)(1-q^{2p-4})}{(1-q)(1-q^{2p})}\right]=\mathcal{I}_{(A_2,A_{2p-4})}(q). \tag*{(134)} I D p ( sl 3 , [ 2 , 1 ]) ( q 2 , 1 ) = PE [ ( 1 − q ) ( 1 − q 2 p ) ( q 2 + q 3 ) ( 1 − q 2 p − 4 ) ] = I ( A 2 , A 2 p − 4 ) ( q ) . (134)
This actually corresponds to the identity (105) with N = 3 N=3 N = 3 .
A non-admissible family
We now consider another family whose natural s l 2 \mathfrak{sl}_2 sl 2 levels are non-admissible. The pair of AD theories is given by
( A 1 , D 2 p ) = D 2 p ( S U ( 2 ) , [ 1 , 1 ] ) ⟷ D 4 p ( S U ( 2 ) , [ 2 ] ) = ( A 1 , A 4 p − 3 ) , p ≥ 2. (135) (A_1,D_{2p})=D_{2p}(SU(2),[1,1])\longleftrightarrow D_{4p}(SU(2),[2])=(A_1,A_{4p-3}),\qquad p\ge2. \tag*{(135)} ( A 1 , D 2 p ) = D 2 p ( S U ( 2 ) , [ 1 , 1 ]) ⟷ D 4 p ( S U ( 2 ) , [ 2 ]) = ( A 1 , A 4 p − 3 ) , p ≥ 2. (135)
Since D 2 p ( S U ( 2 ) ) D_{2p}(SU(2)) D 2 p ( S U ( 2 )) and D 4 p ( S U ( 2 ) ) D_{4p}(SU(2)) D 4 p ( S U ( 2 )) violate the coprimality condition, their associated VOAs are not given by the boundary-admissible s l 2 \mathfrak{sl}_2 sl 2 formulas.
The associated VOAs are identified in [27 , 28 ]
V [ ( A 1 , D 2 p ) ] = R ( p ) , V [ ( A 1 , A 4 p − 3 ) ] = B ( 2 p ) . (136) \mathbb{V}[(A_1,D_{2p})]=\mathcal{R}^{(p)},\qquad\mathbb{V}[(A_1,A_{4p-3})]=\mathcal{B}^{(2p)}. \tag*{(136)} V [( A 1 , D 2 p )] = R ( p ) , V [( A 1 , A 4 p − 3 )] = B ( 2 p ) . (136)
Interestingly, they can nevertheless be realized as boundary admissible W W W -algebras:
R ( p ) ≅ W − ( p + 1 ) + p + 1 p ( s l p + 1 , [ p − 1 , 1 , 1 ] ) , (137) \mathcal{R}^{(p)} \cong\mathcal{W}_{-(p+1)+\frac{p+1}{p}}\left(\mathfrak{sl}_{p+1},[p-1,1,1]\right), \tag*{(137)} R ( p ) ≅ W − ( p + 1 ) + p p + 1 ( sl p + 1 , [ p − 1 , 1 , 1 ] ) , (137)
and
B ( 2 p ) ≅ W − ( 2 p − 1 ) + 2 p − 1 2 p ( s l 2 p − 1 , [ 2 p − 2 , 1 ] ) . (138) \mathcal{B}^{(2p)} \cong\mathcal{W}_{-(2p-1)+\frac{2p-1}{2p}}\left(\mathfrak{sl}_{2p-1},[2p-2,1]\right). \tag*{(138)} B ( 2 p ) ≅ W − ( 2 p − 1 ) + 2 p 2 p − 1 ( sl 2 p − 1 , [ 2 p − 2 , 1 ] ) . (138)
Note the two VOAs arise from DS reduction of different parent affine vertex algebras.
The 4d central charges of the two AD theories are
a ( A 1 , D 2 p ) = 6 p − 5 12 , c ( A 1 , D 2 p ) = 3 p − 2 6 , (139) a_{(A_1,D_{2p})}=\frac{6p-5}{12},\qquad c_{(A_1,D_{2p})}=\frac{3p-2}{6}, \tag*{(139)} a ( A 1 , D 2 p ) = 12 6 p − 5 , c ( A 1 , D 2 p ) = 6 3 p − 2 , (139)
a ( A 1 , A 4 p − 3 ) = p − 29 24 + 1 4 p , c ( A 1 , A 4 p − 3 ) = p − 7 6 + 1 4 p . (140) a_{(A_1,A_{4p-3})}=p-\frac{29}{24}+\frac{1}{4p},\qquad c_{(A_1,A_{4p-3})}=p-\frac{7}{6}+\frac{1}{4p}. \tag*{(140)} a ( A 1 , A 4 p − 3 ) = p − 24 29 + 4 p 1 , c ( A 1 , A 4 p − 3 ) = p − 6 7 + 4 p 1 . (140)
The effective central charges are
c ( A 1 , D 2 p ) e f f = 4 , c ( A 1 , A 4 p − 3 ) e f f = 2. (141) c^{\mathrm{eff}}_{(A_1,D_{2p})}=4,\qquad c^{\mathrm{eff}}_{(A_1,A_{4p-3})}=2. \tag*{(141)} c ( A 1 , D 2 p ) eff = 4 , c ( A 1 , A 4 p − 3 ) eff = 2. (141)
The flavor symmetry of ( A 1 , A 4 p − 3 ) (A_1,A_{4p-3}) ( A 1 , A 4 p − 3 ) is U ( 1 ) U(1) U ( 1 ) for p ≥ 2 p \ge2 p ≥ 2 . The flavor symmetry of ( A 1 , D 2 p ) (A_1,D_{2p}) ( A 1 , D 2 p ) is S U ( 2 ) × U ( 1 ) SU(2)\times U(1) S U ( 2 ) × U ( 1 ) for p > 2 p > 2 p > 2 , and is further enhanced to S U ( 3 ) SU(3) S U ( 3 ) for p = 2 p=2 p = 2 , which is more manifest in terms of the equivalence D 4 ( S U ( 2 ) ) = D 2 ( S U ( 3 ) ) = ( A 1 , D 4 ) D_4(SU(2))=D_2(SU(3))=(A_1,D_4) D 4 ( S U ( 2 )) = D 2 ( S U ( 3 )) = ( A 1 , D 4 ) . Note that the p = 2 p=2 p = 2 case has been discussed in (111) .
The vacuum characters of the two VOAs are given by
I ( A 1 , D 2 p ) ( Q ; y , w ) = PE [ ( Q 1 − Q − Q p 1 − Q p ) ( y 2 + 2 + y − 2 ) + Q p / 2 1 − Q p ( w + w − 1 ) ( y + y − 1 ) ] , (142) \mathcal{I}_{(A_1,D_{2p})}(Q;y,w)=\operatorname{PE}\left[\left(\frac{Q}{1-Q}-\frac{Q^p}{1-Q^p}\right)(y^2+2+y^{-2})+\frac{Q^{p/2}}{1-Q^p}(w+w^{-1})(y+y^{-1})\right], \tag*{(142)} I ( A 1 , D 2 p ) ( Q ; y , w ) = PE [ ( 1 − Q Q − 1 − Q p Q p ) ( y 2 + 2 + y − 2 ) + 1 − Q p Q p /2 ( w + w − 1 ) ( y + y − 1 ) ] , (142)
and
I ( A 1 , A 4 p − 3 ) ( q ; w ) = PE [ 1 + q 1 − q 2 p ( q − q 2 p − 1 1 − q + q p − 1 2 ( w + w − 1 ) ) ] , (143) \mathcal{I}_{(A_1,A_{4p-3})}(q;w)=\operatorname{PE}\left[\frac{1+q}{1-q^{2p}}\left(\frac{q-q^{2p-1}}{1-q}+q^{p-\frac{1}{2}}(w+w^{-1})\right)\right], \tag*{(143)} I ( A 1 , A 4 p − 3 ) ( q ; w ) = PE [ 1 − q 2 p 1 + q ( 1 − q q − q 2 p − 1 + q p − 2 1 ( w + w − 1 ) ) ] , (143)
where w w w is the fugacity for u ( 1 ) \mathfrak{u}(1) u ( 1 ) and y y y is the fugacity for s u ( 2 ) \mathfrak{su}(2) su ( 2 ) . They can be derived from the general formula (74) [23 ] .
It is easy to check that
I ( A 1 , A 4 p − 3 ) ( q ; w ) = I ( A 1 , D 2 p ) ( q 2 ; y = q 1 / 2 , w ) , p ≥ 2. (144) \mathcal{I}_{(A_1,A_{4p-3})}(q;w)=\mathcal{I}_{(A_1,D_{2p})}\left(q^2;y=q^{1/2},w\right),\qquad p\ge2. \tag*{(144)} I ( A 1 , A 4 p − 3 ) ( q ; w ) = I ( A 1 , D 2 p ) ( q 2 ; y = q 1/2 , w ) , p ≥ 2. (144)
This is also consistent with the relation of effective central charges (141) .
Graded vector space isomorphism for VOAs
In the previous sections, we established numerous index identities relating pairs of AD theories with specialized fugacities. On the VOA side, these yield various character identities between affine and W W W -algebras. Since our character identities involve a non-trivial fugacity map, one should not expect an isomorphism between the two VOAs. Instead, they are related by a graded vector space isomorphism.
In this section, we first formalize the general aspects of graded vector space isomorphisms. As a proof of concept, we construct such a graded isomorphism for a concrete pair of AD theories. We then discuss the graded vector space isomorphism between the universal algebras of affine and principal W W W -algebras, as well as their simple quotients up to the lowest null state.
General aspects
Let Γ \Gamma Γ be an additive commutative monoid. A Γ \Gamma Γ -graded vector space is a complex vector space V V V equipped with a direct-sum decomposition
V = ⨁ γ ∈ Γ V γ . (145) V = \bigoplus_{\gamma\in\Gamma} V_\gamma. \tag*{(145)} V = γ ∈ Γ ⨁ V γ . (145)
When Γ ⊂ R \Gamma\subset\mathbb{R} Γ ⊂ R , this grading can be encoded by a diagonalizable grading operator K K K satisfying K v = γ v Kv = \gamma v K v = γ v , v ∈ V γ v \in V_\gamma v ∈ V γ . In physics, the operator K K K corresponds to a symmetry generator or charge. Assuming that every graded component is finite-dimensional, we define the formal graded character by
χ V , K ( q ) = Tr V q K = ∑ γ ∈ Γ dim ( V γ ) q γ . (146) \chi_{V,K}(q) = \operatorname{Tr}_{V} q^K = \sum_{\gamma\in\Gamma} \dim(V_\gamma) q^\gamma. \tag*{(146)} χ V , K ( q ) = Tr V q K = γ ∈ Γ ∑ dim ( V γ ) q γ . (146)
A graded vector-space isomorphism (GVISO) between V V V and W W W is a linear bijection Φ : V → ∼ W \Phi: V \xrightarrow{\sim} W Φ : V ∼ W such that
Φ ( V γ ) = W γ for every γ ∈ Γ . (147) \Phi(V_\gamma) = W_\gamma\qquad\text{for every } \gamma\in\Gamma. \tag*{(147)} Φ ( V γ ) = W γ for every γ ∈ Γ. (147)
When the gradings are encoded by K V K_V K V and K W K_W K W , this is equivalent to
Φ K V = K W Φ . (148) \Phi K_V = K_W \Phi. \tag*{(148)} Φ K V = K W Φ. (148)
Therefore Φ \Phi Φ acts as an intertwiner for the grading generators. In particular, this implies the following identity
χ V , K V ( y ) = χ W , K W ( y ) , (149) \chi_{V,K_V}(y) = \chi_{W,K_W}(y), \tag*{(149)} χ V , K V ( y ) = χ W , K W ( y ) , (149)
or equivalently
Tr V y K V = Tr W y K W . (150) \operatorname{Tr}_{V} y^{K_V} = \operatorname{Tr}_{W} y^{K_W}. \tag*{(150)} Tr V y K V = Tr W y K W . (150)
This kind of relation will be referred to as a graded vector space isomorphism (GVISO) and will be denoted as
Φ : ( V , K V ) ≅ g r ( W , K W ) . (151) \Phi: (V,K_V) \cong_{\mathrm{gr}} (W,K_W). \tag*{(151)} Φ : ( V , K V ) ≅ gr ( W , K W ) . (151)
Our character identities in the previous sections are exactly of the form in (149) , where the two vector spaces are now given by two VOAs graded by a linear combination of conformal and flavor symmetry generators. This naturally motivates us to find an explicit map Φ \Phi Φ between two VOAs which further satisfies (148) .
Note that a GVISO is weaker than an isomorphism of VOAs, which must preserve additional algebraic structures. A GVISO identifies graded vector spaces and their characters, but generally does not preserve OPEs or normally ordered products.
In the rest of this section, we need to translate between operators and modes. The convention is as follows. For an operator O O O with conformal weight Δ \Delta Δ , we will use the mode expansion
O ( z ) = ∑ n + Δ ∈ Z O n z − n − Δ . (152) O(z)=\sum_{n+\Delta\in\mathbb{Z}}O_n z^{-n-\Delta}. \tag*{(152)} O ( z ) = n + Δ ∈ Z ∑ O n z − n − Δ . (152)
Note that the dimension Δ \Delta Δ can be either integer or half-integer, independent of the statistics.
(A1,A4p-3) and (A1,D2p)
In this subsection, we prove the GVISO between ( A 1 , A 4 p − 3 ) (A_1,A_{4p-3}) ( A 1 , A 4 p − 3 ) and ( A 1 , D 2 p ) (A_1,D_{2p}) ( A 1 , D 2 p ) explicitly.
Decomposition of two VOAs
We consider the pair of AD theories ( A 1 , A 4 p − 3 ) (A_1,A_{4p-3}) ( A 1 , A 4 p − 3 ) and ( A 1 , D 2 p ) (A_1,D_{2p}) ( A 1 , D 2 p ) with p ≥ 2 p\geq2 p ≥ 2 . Their associated VOAs are given by [27 , 28 ]
V = V [ ( A 1 , D 2 p ) ] = R ( p ) , W = V [ ( A 1 , A 4 p − 3 ) ] = B ( 2 p ) . (153) V=\mathbb{V}[(A_1,D_{2p})]=\mathcal{R}^{(p)},\qquad W=\mathbb{V}[(A_1,A_{4p-3})]=\mathcal{B}^{(2p)}. \tag*{(153)} V = V [( A 1 , D 2 p )] = R ( p ) , W = V [( A 1 , A 4 p − 3 )] = B ( 2 p ) . (153)
Note that ( A 1 , A 4 p − 3 ) (A_1,A_{4p-3}) ( A 1 , A 4 p − 3 ) has U ( 1 ) U(1) U ( 1 ) flavor symmetry, and ( A 1 , D 2 p ) (A_1,D_{2p}) ( A 1 , D 2 p ) has flavor symmetry S U ( 2 ) × U ( 1 ) SU(2)\times U(1) S U ( 2 ) × U ( 1 ) for p ≥ 3 p\geq3 p ≥ 3 , but gets enhanced to S U ( 3 ) SU(3) S U ( 3 ) at p = 2 p=2 p = 2 .
The central charges of the two VOAs are
c V = 4 − 6 p , c W = 14 − 12 p − 3 p . (154) c_V=4-6p,\qquad c_W=14-12p-\frac{3}{p}. \tag*{(154)} c V = 4 − 6 p , c W = 14 − 12 p − p 3 . (154)
Their strong generators have conformal weights
V : 1 , 1 , 1 , 1 , p 2 , p 2 , p 2 , p 2 , W : 1 , 2 , p − 1 2 , p − 1 2 . (155) V:1,1,1,1,\frac{p}{2},\frac{p}{2},\frac{p}{2},\frac{p}{2},\qquad W:1,2,p-\frac{1}{2},p-\frac{1}{2}. \tag*{(155)} V : 1 , 1 , 1 , 1 , 2 p , 2 p , 2 p , 2 p , W : 1 , 2 , p − 2 1 , p − 2 1 . (155)
For p = 2 p=2 p = 2 , the VOAs are given by the affine vertex algebra and BP algebra, reviewed in Appendix § 11 :
V = R ( 2 ) ≅ L − 3 / 2 ( s l 3 ) , W = B ( 4 ) ≅ W − 9 / 4 ( s l 3 , f s u b r e g ) = B P − 9 / 4 . (156) V=\mathcal{R}^{(2)}\cong L_{-3/2}(\mathfrak{sl}_3),\qquad W=\mathcal{B}^{(4)}\cong\mathcal{W}_{-9/4}(\mathfrak{sl}_3,f_{\mathrm{subreg}})=\mathrm{BP}_{-9/4}. \tag*{(156)} V = R ( 2 ) ≅ L − 3/2 ( sl 3 ) , W = B ( 4 ) ≅ W − 9/4 ( sl 3 , f subreg ) = BP − 9/4 . (156)
We have established the index identity (144) for (153) . This naturally motivates us to find its interpretation at the level of VOAs. Their different central charges already show that they cannot be isomorphic as conformal VOAs. We thus want to find a weaker isomorphism, which can be used to explain (144) .
Our goal in this subsection is to show that this is indeed possible, and it is realized by the GVISO below
Φ : ( V , H , a 0 ) ≅ g r ( W , L 0 W , J 0 ) . (157) \Phi:(V,H,a_0)\cong_{\mathrm{gr}}(W,L_0^W,J_0). \tag*{(157)} Φ : ( V , H , a 0 ) ≅ gr ( W , L 0 W , J 0 ) . (157)
In the rest of this subsection, we construct Φ \Phi Φ and prove the GVISO (157) . We sketch our procedure below and discuss each step in more detail in the following subsections.
First, we note that the two VOAs have some useful module decompositions. They admit the following module decompositions [28 Eqs. (27)–(28)]
V = ⨁ m ≥ 0 ν = − m , − m + 2 , … , m V m , ν , V m , ν = V m ( p ) ⊗ F ν a , (158) V = \bigoplus_{\substack{m \ge0 \\ \nu=-m,-m+2,\ldots,m}} V_{m,\nu}, \qquad V_{m,\nu} = V_m^{(p)} \otimes\mathcal{F}_{\nu}^{a}, \tag*{(158)} V = m ≥ 0 ν = − m , − m + 2 , … , m ⨁ V m , ν , V m , ν = V m ( p ) ⊗ F ν a , (158)
W = ⨁ m ≥ 0 ν = − m , − m + 2 , … , m W m , ν , W m , ν = M m ( 2 p ) ⊗ F ν J , (159) W = \bigoplus_{\substack{m \ge0 \\ \nu=-m,-m+2,\ldots,m}} W_{m,\nu}, \qquad W_{m,\nu} = M_m^{(2p)} \otimes\mathcal{F}_{\nu}^{J}, \tag*{(159)} W = m ≥ 0 ν = − m , − m + 2 , … , m ⨁ W m , ν , W m , ν = M m ( 2 p ) ⊗ F ν J , (159)
where V , M , F \mathsf{V},\mathsf{M},\mathsf{F} V , M , F are affine, Virasoro and Heisenberg modules, respectively, defined below. The module decompositions (158) and (159) allow us to decompose Φ = ⨁ m , ν Φ m , ν \Phi=\bigoplus_{m,\nu}\Phi_{m,\nu} Φ = ⨁ m , ν Φ m , ν above as
Φ m , ν : ( V m , ν , H , a 0 ) ≅ g r ( W m , ν , L 0 W , J 0 ) . (160) \Phi_{m,\nu}:(V_{m,\nu},H,a_0)\cong_{\mathrm{gr}}(W_{m,\nu},L_0^W,J_0). \tag*{(160)} Φ m , ν : ( V m , ν , H , a 0 ) ≅ gr ( W m , ν , L 0 W , J 0 ) . (160)
Φ m , ν a 0 = J 0 Φ m , ν \Phi_{m,\nu}a_0=J_0\Phi_{m,\nu} Φ m , ν a 0 = J 0 Φ m , ν . Φ m , ν H = L 0 W Φ m , ν \Phi_{m,\nu}H=L_0^W\Phi_{m,\nu} Φ m , ν H = L 0 W Φ m , ν .
This turns out to be decomposable further into the two factors, depending on m m m and ν \nu ν , respectively:
φ ν : ( F ν a , a 0 ) ≅ g r ( F ν , e v e n J , J 0 ) , φ ν a n = J 2 n φ ν , (161) \varphi_\nu:(\mathcal{F}_\nu^a,a_0)\cong_{\mathrm{gr}}(\mathcal{F}_{\nu,\mathrm{even}}^J,J_0), \qquad\varphi_\nu a_n=J_{2n}\varphi_\nu, \tag*{(161)} φ ν : ( F ν a , a 0 ) ≅ gr ( F ν , even J , J 0 ) , φ ν a n = J 2 n φ ν , (161)
and
ψ m : ( V m ( p ) , H s l 2 ) ≅ g r ( M m ( 2 p ) ⊗ F 0 , o d d J , S 0 + N o d d ) . (162) \psi_m:(V_m^{(p)},H_{\mathfrak{sl}_2})\cong_{\mathrm{gr}}(M_m^{(2p)}\otimes\mathcal{F}_{0,\mathrm{odd}}^J,S_0+N_{\mathrm{odd}}). \tag*{(162)} ψ m : ( V m ( p ) , H sl 2 ) ≅ gr ( M m ( 2 p ) ⊗ F 0 , odd J , S 0 + N odd ) . (162)
The relevant charges are related as follows and will be described in detail later:
H = 2 L 0 V + 1 2 h 0 = H s l 2 − p 2 a 0 2 − p ∑ n > 0 a − n a n . (163) H=2L_0^V+\frac{1}{2}h_0=H_{\mathfrak{sl}_2}-\frac{p}{2}a_0^2-p\sum_{n>0}a_{-n}a_n. \tag*{(163)} H = 2 L 0 V + 2 1 h 0 = H sl 2 − 2 p a 0 2 − p n > 0 ∑ a − n a n . (163)
L 0 W = S 0 − p 2 J 0 2 − p ∑ r > 0 J − r J r = S 0 + N o d d − p 2 J 0 2 − p ∑ r ∈ 2 Z > 0 J − r J r . (164) L_0^W=S_0-\frac{p}{2}J_0^2-p\sum_{r>0}J_{-r}J_r=S_0+N_{\mathrm{odd}}-\frac{p}{2}J_0^2-p\sum_{r\in2\mathbb{Z}_{>0}}J_{-r}J_r. \tag*{(164)} L 0 W = S 0 − 2 p J 0 2 − p r > 0 ∑ J − r J r = S 0 + N odd − 2 p J 0 2 − p r ∈ 2 Z > 0 ∑ J − r J r . (164)
These decompositions make our strategy for establishing GVISO transparent. We will discuss φ ν \varphi_\nu φ ν (161) in the next subsection 6.2.2, and then ψ m \psi_m ψ m (162) in subsection 6.2.3. Their combination to Φ m , ν \Phi_{m,\nu} Φ m , ν (160) will be discussed in subsection 6.2.4. The actual GVISO Φ \Phi Φ (157) and its relation to index identity will be discussed in subsection 6.2.5.
The graded isomorphism φ ν \varphi_\nu φ ν
We establish (161) in this subsection. The two VOAs in (153) contain two U ( 1 ) U(1) U ( 1 ) currents a ( z ) a(z) a ( z ) and J ( z ) J(z) J ( z ) with levels − 2 / p -2/p − 2/ p and − 1 / p -1/p − 1/ p , respectively:
a ( z ) ∈ V = R ( p ) , a ( z ) a ( w ) ∼ − 2 / p ( z − w ) 2 , a(z)\in V=\mathcal{R}^{(p)}, \qquad a(z)a(w)\sim-\frac{2/p}{(z-w)^2}, a ( z ) ∈ V = R ( p ) , a ( z ) a ( w ) ∼ − ( z − w ) 2 2/ p ,
J ( z ) ∈ W = B ( 2 p ) , J ( z ) J ( w ) ∼ − 1 / p ( z − w ) 2 . J(z)\in W=\mathcal{B}^{(2p)}, \qquad J(z)J(w)\sim-\frac{1/p}{(z-w)^2}. J ( z ) ∈ W = B ( 2 p ) , J ( z ) J ( w ) ∼ − ( z − w ) 2 1/ p .
Note that R ( p ) \mathcal{R}^{(p)} R ( p ) has flavor symmetry s l 2 ⊕ u 1 \mathfrak{sl}_2 \oplus\mathfrak{u}_1 sl 2 ⊕ u 1 , which gets enhanced to s l 3 \mathfrak{sl}_3 sl 3 at p = 2 p = 2 p = 2 , and contains affine factors [28 ]
L − 2 + 1 / p ( s l 2 ) ⊗ H ⊂ R ( p ) . (165) L_{-2+1/p}(\mathfrak{sl}_2) \otimes\mathcal{H} \subset\mathcal{R}^{(p)}. \tag*{(165)} L − 2 + 1/ p ( sl 2 ) ⊗ H ⊂ R ( p ) . (165)
Therefore a a a commutes with the s l 2 \mathfrak{sl}_2 sl 2 currents, and generates the Heisenberg algebra H \mathcal{H} H . The modes of the two Heisenberg algebras satisfy the commutation relations
[ a r , a s ] = − 2 r p δ r + s , 0 , [ J r , J s ] = − r p δ r + s , 0 . (166) [a_r,a_s] = -\frac{2r}{p}\delta_{r+s,0}, \qquad[J_r,J_s] = -\frac{r}{p}\delta_{r+s,0}. \tag*{(166)} [ a r , a s ] = − p 2 r δ r + s , 0 , [ J r , J s ] = − p r δ r + s , 0 . (166)
For a fixed ν \nu ν , we define the ground states as
a 0 ∣ ν ⟩ a = ν ∣ ν ⟩ a , J 0 ∣ ν ⟩ J = ν ∣ ν ⟩ J , (167) a_0|\nu\rangle_a = \nu|\nu\rangle_a, \qquad J_0|\nu\rangle_J = \nu|\nu\rangle_J, \tag*{(167)} a 0 ∣ ν ⟩ a = ν ∣ ν ⟩ a , J 0 ∣ ν ⟩ J = ν ∣ ν ⟩ J , (167)
which are also annihilated by all positive modes
a r ∣ ν ⟩ a = 0 , J r ∣ ν ⟩ J = 0 , r > 0. (168) a_r|\nu\rangle_a = 0, \qquad J_r|\nu\rangle_J = 0, \qquad r > 0. \tag*{(168)} a r ∣ ν ⟩ a = 0 , J r ∣ ν ⟩ J = 0 , r > 0. (168)
The two Fock spaces are then given by
F ν a = C [ a − 1 , a − 2 , … ] ∣ ν ⟩ a , F ν , e v e n J = C [ J − 2 , J − 4 , … ] ∣ ν ⟩ J , (169) \mathcal{F}^{a}_{\nu} = \mathbb{C}[a_{-1},a_{-2},\ldots]|\nu\rangle_a, \qquad\mathcal{F}^{J}_{\nu,\mathrm{even}} = \mathbb{C}[J_{-2},J_{-4},\ldots]|\nu\rangle_J, \tag*{(169)} F ν a = C [ a − 1 , a − 2 , … ] ∣ ν ⟩ a , F ν , even J = C [ J − 2 , J − 4 , … ] ∣ ν ⟩ J , (169)
We can construct a map between the two spaces
φ ν ( ∏ n ≥ 1 a − n b n ∣ ν ⟩ a ) = ∏ n ≥ 1 J − 2 n b n ∣ ν ⟩ J . (170) \varphi_\nu\left(\prod_{n\geq1}a_{-n}^{b_n}|\nu\rangle_a\right) = \prod_{n\geq1}J_{-2n}^{b_n}|\nu\rangle_J. \tag*{(170)} φ ν ( n ≥ 1 ∏ a − n b n ∣ ν ⟩ a ) = n ≥ 1 ∏ J − 2 n b n ∣ ν ⟩ J . (170)
This is a basis bijection, with inverse J − 2 n ↦ a − n J_{-2n} \mapsto a_{-n} J − 2 n ↦ a − n . The ground charges agree, and
[ a n , a − r ] = − 2 n p δ n , r = [ J 2 n , J − 2 r ] . (171) [a_n,a_{-r}] = -\frac{2n}{p}\delta_{n,r} = [J_{2n},J_{-2r}]. \tag*{(171)} [ a n , a − r ] = − p 2 n δ n , r = [ J 2 n , J − 2 r ] . (171)
Together with (168) , we thus have
φ ν a n = J 2 n φ ν n ∈ Z . (172) \varphi_\nu a_n = J_{2n}\varphi_\nu\qquad n \in\mathbb{Z}. \tag*{(172)} φ ν a n = J 2 n φ ν n ∈ Z . (172)
The corresponding oscillators both have degree 2 n 2n 2 n in the full gradings defined below. Therefore we have
φ ν : ( F ν a , a 0 ) ≅ g r ( F ν , e v e n J , J 0 ) , φ ν a n = J 2 n φ ν . (173) \varphi_\nu: \left(\mathcal{F}^{a}_{\nu},a_0\right) \cong_{\mathrm{gr}} \left(\mathcal{F}^{J}_{\nu,\mathrm{even}},J_0\right), \qquad\varphi_\nu a_n = J_{2n}\varphi_\nu. \tag*{(173)} φ ν : ( F ν a , a 0 ) ≅ gr ( F ν , even J , J 0 ) , φ ν a n = J 2 n φ ν . (173)
The graded isomorphism ψ m \psi_m ψ m
In this subsection, we establish the GVISO
ψ m : ( V m ( p ) , H s l 2 ) ≅ g r ( M m ( 2 p ) ⊗ F 0 , o d d J , S 0 + N o d d ) . (174) \psi_m : \left(\mathcal{V}^{(p)}_m,\mathcal{H}_{\mathfrak{sl}_2}\right) \cong_{\mathrm{gr}} \left(\mathcal{M}^{(2p)}_m \otimes\mathcal{F}^{J}_{0,\mathrm{odd}}, S_0 + N_{\mathrm{odd}}\right). \tag*{(174)} ψ m : ( V m ( p ) , H sl 2 ) ≅ gr ( M m ( 2 p ) ⊗ F 0 , odd J , S 0 + N odd ) . (174)
Affine module V m ( p ) V_m^{(p)} V m ( p )
We introduce V s l 2 = L k s l 2 ( s l 2 ) ⊂ V V_{\mathfrak{sl}_2}=L_{k_{\mathfrak{sl}_2}}(\mathfrak{sl}_2)\subset V V sl 2 = L k sl 2 ( sl 2 ) ⊂ V with level k s l 2 = − 2 + 1 / p k_{\mathfrak{sl}_2}=-2+1/p k sl 2 = − 2 + 1/ p , which is the affine subsector of R ( p ) \mathcal{R}^{(p)} R ( p ) in (153) . The three affine currents are denoted as e , f , h e,f,h e , f , h .
We use the conventions for s l 2 \mathfrak{sl}_2 sl 2 :
[ h , e ] = 2 e , [ h , f ] = − 2 f , [ e , f ] = h , ( h , h ) = 2 , ( e , f ) = 1. (175) [h,e]=2e,\qquad[h,f]=-2f,\qquad[e,f]=h,\qquad(h,h)=2,\qquad(e,f)=1. \tag*{(175)} [ h , e ] = 2 e , [ h , f ] = − 2 f , [ e , f ] = h , ( h , h ) = 2 , ( e , f ) = 1. (175)
Let E m E_m E m be the irreducible ( m + 1 ) (m+1) ( m + 1 ) -dimensional s l 2 \mathfrak{sl}_2 sl 2 -module. Choose its lowest-weight vector ∣ u m ⟩ |u_m\rangle ∣ u m ⟩ :
h 0 ∣ u m ⟩ = − m ∣ u m ⟩ , f 0 ∣ u m ⟩ = 0 , E m = Span { e 0 s ∣ u m ⟩ : 0 ≤ s ≤ m } , (176) h_0|u_m\rangle=-m|u_m\rangle,\qquad f_0|u_m\rangle=0,\qquad E_m=\operatorname{Span}\{e_0^s|u_m\rangle:0\le s\le m\}, \tag*{(176)} h 0 ∣ u m ⟩ = − m ∣ u m ⟩ , f 0 ∣ u m ⟩ = 0 , E m = Span { e 0 s ∣ u m ⟩ : 0 ≤ s ≤ m } , (176)
subject to
e 0 m + 1 ∣ u m ⟩ = 0. (177) e_0^{m+1}|u_m\rangle=0. \tag*{(177)} e 0 m + 1 ∣ u m ⟩ = 0. (177)
Next, we extend the s l 2 \mathfrak{sl}_2 sl 2 -module to an affine module of V s l 2 V_{\mathfrak{sl}_2} V sl 2 by inducing from E m E_m E m , on which all strictly positive current modes act trivially. The resulting module will be denoted as V m ( p ) V_m^{(p)} V m ( p ) ,23 and it admits a basis
U F e 0 s ∣ u m ⟩ , 0 ≤ s ≤ m , (178) UF e_0^s|u_m\rangle,\qquad0\le s\le m, \tag*{(178)} U F e 0 s ∣ u m ⟩ , 0 ≤ s ≤ m , (178)
where U U U is an ordered product of h − n , e − n h_{-n},e_{-n} h − n , e − n and F F F is an ordered product of f − n f_{-n} f − n , with n ≥ 1 n\ge1 n ≥ 1 .
The s l 2 \mathfrak{sl}_2 sl 2 affine Sugawara field and its central charge are
T s l 2 = : e f : + : f e : + 1 2 : h h : 2 ( k s l 2 + 2 ) , c s l 2 = 3 k s l 2 k s l 2 + 2 = 3 − 6 p . (179) T_{\mathfrak{sl}_2}=\frac{:{ef}:+:{fe}:+\frac{1}{2}:{hh}:}{2(k_{\mathfrak{sl}_2}+2)},\qquad c_{\mathfrak{sl}_2}=\frac{3k_{\mathfrak{sl}_2}}{k_{\mathfrak{sl}_2}+2}=3-6p. \tag*{(179)} T sl 2 = 2 ( k sl 2 + 2 ) : e f : + : f e : + 2 1 : hh : , c sl 2 = k sl 2 + 2 3 k sl 2 = 3 − 6 p . (179)
We introduce
H s l 2 = 2 L 0 s l 2 + 1 2 h 0 , (180) H_{\mathfrak{sl}_2}=2L_0^{\mathfrak{sl}_2}+\frac{1}{2}h_0, \tag*{(180)} H sl 2 = 2 L 0 sl 2 + 2 1 h 0 , (180)
where L 0 s l 2 L_0^{\mathfrak{sl}_2} L 0 sl 2 is the Virasoro mode of T s l 2 T_{\mathfrak{sl}_2} T sl 2 .
Using (176) , one can show
H s l 2 ∣ u m ⟩ = λ m ∣ u m ⟩ , λ m = p m ( m + 2 ) − m 2 , (181) H_{\mathfrak{sl}_2}|u_m\rangle=\lambda_m|u_m\rangle,\qquad\lambda_m=\frac{pm(m+2)-m}{2}, \tag*{(181)} H sl 2 ∣ u m ⟩ = λ m ∣ u m ⟩ , λ m = 2 p m ( m + 2 ) − m , (181)
and
[ H s l 2 , e − n ] = ( 2 n + 1 ) e − n , [ H s l 2 , f − n ] = ( 2 n − 1 ) f − n , [ H s l 2 , h − n ] = 2 n h − n , n ≥ 1 , (182) [H_{\mathfrak{sl}_2},e_{-n}]=(2n+1)e_{-n},\qquad[H_{\mathfrak{sl}_2},f_{-n}]=(2n-1)f_{-n},\qquad[H_{\mathfrak{sl}_2},h_{-n}]=2nh_{-n},\qquad n\ge1, \tag*{(182)} [ H sl 2 , e − n ] = ( 2 n + 1 ) e − n , [ H sl 2 , f − n ] = ( 2 n − 1 ) f − n , [ H sl 2 , h − n ] = 2 n h − n , n ≥ 1 , (182)
[ H s l 2 , e 0 ] = e 0 . (183) [H_{\mathfrak{sl}_2},e_0]=e_0. \tag*{(183)} [ H sl 2 , e 0 ] = e 0 . (183)
Virasoro module M m ( 2 p ) M_m^{(2p)} M m ( 2 p )
The generalized BP algebra W W W in (153) contains a stress tensor T T T with central charge c W = 14 − 12 p − 3 / p c_W = 14 - 12p - 3/p c W = 14 − 12 p − 3/ p , together with a primary current J J J :
J ( z ) J ( w ) ∼ − 1 / p ( z − w ) 2 , T ( z ) J ( w ) ∼ J ( w ) ( z − w ) 2 + ∂ J ( w ) z − w . (184) J(z)J(w) \sim-\frac{1/p}{(z-w)^2}, \qquad T(z)J(w) \sim\frac{J(w)}{(z-w)^2}+\frac{\partial J(w)}{z-w}. \tag*{(184)} J ( z ) J ( w ) ∼ − ( z − w ) 2 1/ p , T ( z ) J ( w ) ∼ ( z − w ) 2 J ( w ) + z − w ∂ J ( w ) . (184)
The current J J J generates a Heisenberg subalgebra, whose stress tensor T J = − p 2 : J J : T_J=-\frac{p}{2}:JJ: T J = − 2 p : J J : has central charge 1. One can then construct
S = T − T J = T + p 2 : J J : . (185) S=T-T_J=T+\frac{p}{2}:JJ:. \tag*{(185)} S = T − T J = T + 2 p : J J : . (185)
which satisfies
S ( z ) J ( w ) ∼ 0 , [ S m , J n ] = 0. (186) S(z)J(w)\sim0,\qquad[S_m,J_n]=0. \tag*{(186)} S ( z ) J ( w ) ∼ 0 , [ S m , J n ] = 0. (186)
Since S S S also commutes with T J T_J T J , it is a Virasoro field with central charge
c S = c W − 1 = 13 − 12 p − 3 p = 1 − 6 ( 2 p − 1 ) 2 2 p = c 1 , 2 p , (187) c_S=c_W-1=13-12p-\frac{3}{p}=1-\frac{6(2p-1)^2}{2p}=c_{1,2p}, \tag*{(187)} c S = c W − 1 = 13 − 12 p − p 3 = 1 − 2 p 6 ( 2 p − 1 ) 2 = c 1 , 2 p , (187)
which coincides with the central charge of the nonrational ( 1 , 2 p ) (1,2p) ( 1 , 2 p ) Virasoro algebra. From (185) , we have the relation
S n = L n W + p 2 ∑ r ∈ Z : J n − r J r : . (188) S_n=L_n^W+\frac{p}{2}\sum_{r\in\mathbb{Z}}:J_{n-r}J_r:. \tag*{(188)} S n = L n W + 2 p r ∈ Z ∑ : J n − r J r : . (188)
where S n S_n S n , L n W L_n^W L n W are the Virasoro modes of S S S and T T T . In particular, the zero mode of T T T is
L 0 W = S 0 − p 2 ∑ r ∈ Z : J − r J r : = S 0 − p 2 J 0 2 − p ∑ r > 0 J − r J r . (189) L_0^W=S_0-\frac{p}{2}\sum_{r\in\mathbb{Z}}:J_{-r}J_r:=S_0-\frac{p}{2}J_0^2-p\sum_{r>0}J_{-r}J_r. \tag*{(189)} L 0 W = S 0 − 2 p r ∈ Z ∑ : J − r J r := S 0 − 2 p J 0 2 − p r > 0 ∑ J − r J r . (189)
The field S S S generates the simple Virasoro algebra at central charge (187) ; its universal vacuum algebra is already simple. We study its modules generated by highest-weight states [6 ]
S n ∣ w m ⟩ = 0 ( n > 0 ) , S 0 ∣ w m ⟩ = λ m ∣ w m ⟩ , (190) S_n|w_m\rangle=0\ (n>0),\qquad S_0|w_m\rangle=\lambda_m|w_m\rangle, \tag*{(190)} S n ∣ w m ⟩ = 0 ( n > 0 ) , S 0 ∣ w m ⟩ = λ m ∣ w m ⟩ , (190)
with conformal weight
λ m = ( 2 p ( m + 1 ) − 1 ) 2 − ( 2 p − 1 ) 2 8 p = p m ( m + 2 ) − m 2 , (191) \lambda_m=\frac{(2p(m+1)-1)^2-(2p-1)^2}{8p}=\frac{pm(m+2)-m}{2}, \tag*{(191)} λ m = 8 p ( 2 p ( m + 1 ) − 1 ) 2 − ( 2 p − 1 ) 2 = 2 p m ( m + 2 ) − m , (191)
which coincides with (181) .
Acting with all ordered products of negative modes S − n S_{-n} S − n with n ≥ 1 n\geq1 n ≥ 1 on ∣ w m ⟩ |w_m\rangle ∣ w m ⟩ generates the Verma module. Its maximal proper submodule is generated by the singular vector P m + 1 ( 2 p ) ∣ w m ⟩ P_{m+1}(2p)\lvert w_m\rangle P m + 1 ( 2 p ) ∣ w m ⟩ for the singular polynomial [6 ] 24
P r ( 2 p ) = S − 1 r + terms with fewer than r Virasoro factors. (192) P_r(2p)=S_{-1}^{r}+\text{terms with fewer than }r\text{ Virasoro factors.} \tag*{(192)} P r ( 2 p ) = S − 1 r + terms with fewer than r Virasoro factors. (192)
After quotienting out the corresponding null submodules, we obtain the irreducible Virasoro module M m ( 2 p ) \mathrm{M}_{m}^{(2p)} M m ( 2 p ) at central charge c 1 , 2 p c_{1,2p} c 1 , 2 p [28 ] .
We can choose the following basis for M m ( 2 p ) \mathrm{M}_{m}^{(2p)} M m ( 2 p )
S − μ S − 1 s ∣ w m ⟩ , μ i ≥ 2 , 0 ≤ s ≤ m , (193) S_{-\mu}S_{-1}^{s}\lvert w_m\rangle,\qquad\mu_i\ge2,\quad0\le s\le m, \tag*{(193)} S − μ S − 1 s ∣ w m ⟩ , μ i ≥ 2 , 0 ≤ s ≤ m , (193)
where S − μ = S − μ 1 ⋯ S − μ t S_{-\mu}=S_{-\mu_1}\cdots S_{-\mu_t} S − μ = S − μ 1 ⋯ S − μ t for the partition μ \mu μ subject to μ 1 ≥ ⋯ ≥ μ t ≥ 2 \mu_1\ge\cdots\ge\mu_t\ge2 μ 1 ≥ ⋯ ≥ μ t ≥ 2 , allowing the empty partition. Indeed, if s = m + 1 s=m+1 s = m + 1 in (193) , we can use the null relation P m + 1 ( 2 p ) ∣ w m ⟩ = 0 P_{m+1}(2p)\lvert w_m\rangle=0 P m + 1 ( 2 p ) ∣ w m ⟩ = 0 to rewrite S − 1 m + 1 ∣ w m ⟩ S_{-1}^{m+1}\lvert w_m\rangle S − 1 m + 1 ∣ w m ⟩ in terms of ordered products S − ν S_{-\nu} S − ν containing fewer than m + 1 m+1 m + 1 factors of S − 1 S_{-1} S − 1 . Repeating this procedure removes all powers S − 1 s S_{-1}^{s} S − 1 s with s > m s>m s > m and proves that (193) spans the simple quotient. One can further justify that the basis is independent and complete by matching the simple module character and the generating function of the spanning set (193) .
Neutral odd-mode Fock space F 0 , o d d J \mathcal{F}^{J}_{0,\mathrm{odd}} F 0 , odd J
For the Heisenberg algebra (166) , we constructed the even charged sector F ν , e v e n J \mathcal{F}^{J}_{\nu,\mathrm{even}} F ν , even J using the even modes of J J J . Similarly, we can also construct the neutral odd-mode space and its degree operator
F 0 , o d d J = C [ J − 1 , J − 3 , … ] ∣ 0 ⟩ J , N o d d = − p ∑ n ≥ 1 J − ( 2 n − 1 ) J 2 n − 1 , (194) \mathcal{F}^{J}_{0,\mathrm{odd}}=\mathbb{C}[J_{-1},J_{-3},\ldots]\lvert0\rangle_J,\qquad N_{\mathrm{odd}}=-p\sum_{n\ge1}J_{-(2n-1)}J_{2n-1}, \tag*{(194)} F 0 , odd J = C [ J − 1 , J − 3 , … ] ∣ 0 ⟩ J , N odd = − p n ≥ 1 ∑ J − ( 2 n − 1 ) J 2 n − 1 , (194)
where ∣ 0 ⟩ J \lvert0\rangle_J ∣ 0 ⟩ J is annihilated by all the non-negative modes J r ∣ 0 ⟩ J = 0 J_r\lvert0\rangle_J=0 J r ∣ 0 ⟩ J = 0 , r ≥ 0 r\ge0 r ≥ 0 . As a result, J − ( 2 n − 1 ) J_{-(2n-1)} J − ( 2 n − 1 ) has degree 2 n − 1 2n-1 2 n − 1 , namely N o d d J − ( 2 n − 1 ) ∣ 0 ⟩ J = ( 2 n − 1 ) J − ( 2 n − 1 ) ∣ 0 ⟩ J N_{\mathrm{odd}}J_{-(2n-1)}\lvert0\rangle_J=(2n-1)J_{-(2n-1)}\lvert0\rangle_J N odd J − ( 2 n − 1 ) ∣ 0 ⟩ J = ( 2 n − 1 ) J − ( 2 n − 1 ) ∣ 0 ⟩ J .
The m m m -dependent bijection ψ m \psi_m ψ m
On the affine side, the module V m ( p ) \mathcal{V}^{(p)}_m V m ( p ) has a basis U F e 0 s ∣ u m ⟩ UF e_0^s\lvert u_m\rangle U F e 0 s ∣ u m ⟩ given in (178) , where we order the h , e h,e h , e -modes in U U U as ⋯ , e − 2 , h − 2 , e − 1 , h − 1 \cdots,e_{-2},h_{-2},e_{-1},h_{-1} ⋯ , e − 2 , h − 2 , e − 1 , h − 1 , followed by the ordered f f f -modes in F F F , then e 0 s e_0^s e 0 s with 0 ≤ s ≤ m 0\le s\le m 0 ≤ s ≤ m .
We then construct the following mode dictionary
Table 3.
where n ≥ 1 n \ge1 n ≥ 1 , 0 ≤ s ≤ m 0 \le s \le m 0 ≤ s ≤ m . Using the replacement rule in the dictionary (6.51), we can map ordered products U , F U,F U , F from the affine side to ordered products on the target side, which will be denoted as U S , F J U^{S},F^{J} U S , F J .
We can further define the map between the affine basis and the tensor-product basis as follows:
ψ m ( U F e 0 s ∣ u m ⟩ ) = U S S − 1 s ∣ w m ⟩ ⊗ F J ∣ 0 ⟩ J . (195) \psi_m\left(UF e_0^s \lvert u_m\rangle\right)=U^{S}S_{-1}^{s}\lvert w_m\rangle\otimes F^{J}\lvert0\rangle_J . \tag*{(195)} ψ m ( U F e 0 s ∣ u m ⟩ ) = U S S − 1 s ∣ w m ⟩ ⊗ F J ∣ 0 ⟩ J . (195)
First, it is easy to check that the dictionary (6.51) sends the affine basis (178) bijectively to the Virasoro basis (193) tensored with the odd Fock basis (194) . Its inverse replaces S − 2 n , S − ( 2 n + 1 ) , J − ( 2 n − 1 ) S_{-2n},S_{-(2n+1)},J_{-(2n-1)} S − 2 n , S − ( 2 n + 1 ) , J − ( 2 n − 1 ) by h − n , e − n , f − n h_{-n},e_{-n},f_{-n} h − n , e − n , f − n and S − 1 S S_{-1}^{S} S − 1 S by e 0 s e_0^s e 0 s . So ψ m \psi_m ψ m (195) is a bijection of bases. Second, using (182) (183) , we see that the degree increment in the last column of (6.51) on the affine side coincides with the increment of H s l 2 H_{\mathfrak{sl}_2} H sl 2 . But the degree increment in the last column of (6.51) also equals the sum of the conformal weight for S S S and the degree N o d d N_{\mathrm{odd}} N odd (194) for J J J . Consequently, this implies that
ψ m H s l 2 = ( S 0 + N o d d ) ψ m . (196) \psi_m H_{\mathfrak{sl}_2}=(S_0+N_{\mathrm{odd}})\psi_m . \tag*{(196)} ψ m H sl 2 = ( S 0 + N odd ) ψ m . (196)
Therefore, the map ψ m \psi_m ψ m is a graded vector-space isomorphism
( V m ( p ) , H s l 2 ) ≅ g r ( M m ( 2 p ) ⊗ F 0 , o d d J , S 0 + N o d d ) . (197) \left(\mathbb{V}^{(p)}_m,H_{\mathfrak{sl}_2}\right)\cong_{\mathrm{gr}}\left(\mathbb{M}^{(2p)}_m\otimes\mathcal{F}^{J}_{0,\mathrm{odd}},S_0+N_{\mathrm{odd}}\right). \tag*{(197)} ( V m ( p ) , H sl 2 ) ≅ gr ( M m ( 2 p ) ⊗ F 0 , odd J , S 0 + N odd ) . (197)
The graded isomorphism Φ m , ν \Phi_{m,\nu} Φ m , ν
We have just established (197) , where F 0 , o d d J \mathcal{F}^{J}_{0,\mathrm{odd}} F 0 , odd J is the neutral odd sector for the Heisenberg algebra, constructed in (194) , while V m ( p ) \mathbb{V}^{(p)}_m V m ( p ) and M m ( 2 p ) \mathbb{M}^{(2p)}_m M m ( 2 p ) are the affine and Virasoro modules, respectively.
Now we can combine the odd and even sectors of F J \mathcal{F}^{J} F J together by defining a map
Υ ν : F 0 , o d d J ⊗ F ν , e v e n J ⟶ F ν J , (198) \Upsilon_\nu:\mathcal{F}^{J}_{0,\mathrm{odd}}\otimes\mathcal{F}^{J}_{\nu,\mathrm{even}}\longrightarrow\mathcal{F}^{J}_\nu, \tag*{(198)} Υ ν : F 0 , odd J ⊗ F ν , even J ⟶ F ν J , (198)
such that
Υ ν ( P ∣ 0 ⟩ J ⊗ Q ∣ ν ⟩ J ) = P Q ∣ ν ⟩ J , (199) \Upsilon_\nu(P\lvert0\rangle_J\otimes Q\lvert\nu\rangle_J)=PQ\lvert\nu\rangle_J , \tag*{(199)} Υ ν ( P ∣ 0 ⟩ J ⊗ Q ∣ ν ⟩ J ) = P Q ∣ ν ⟩ J , (199)
where P P P uses odd negative J J J -modes and Q Q Q uses even ones. Its inverse separates these two sets of modes unambiguously. The charge ν \nu ν occurs once; after gluing together in the way above, the odd modes act on ∣ ν ⟩ J \lvert\nu\rangle_J ∣ ν ⟩ J as
F ν , o d d J = C [ J − 1 , J − 3 , … ] ∣ ν ⟩ J . (200) \mathcal{F}^{J}_{\nu,\mathrm{odd}}=\mathbb{C}[J_{-1},J_{-3},\ldots]\lvert\nu\rangle_J . \tag*{(200)} F ν , odd J = C [ J − 1 , J − 3 , … ] ∣ ν ⟩ J . (200)
We can finally combine φ ν \varphi_\nu φ ν in (161) and ψ m \psi_m ψ m (162) together by considering:
Φ m , ν = ( id M m ( 2 p ) ⊗ Υ ν ) ∘ ( ψ m ⊗ φ ν ) . (201) \Phi_{m,\nu}=\left(\operatorname{id}_{\mathbb{M}^{(2p)}_m}\otimes\Upsilon_\nu\right)\circ\left(\psi_m\otimes\varphi_\nu\right). \tag*{(201)} Φ m , ν = ( id M m ( 2 p ) ⊗ Υ ν ) ∘ ( ψ m ⊗ φ ν ) . (201)
Explicitly, for 0 ≤ s ≤ m 0\le s\le m 0 ≤ s ≤ m , the affine basis (178) and the Virasoro basis (193) are related explicitly using the dictionary in (6.51):
Φ m , ν ( U F e 0 s ∣ u m ⟩ ⊗ Q ( a − 1 , a − 2 , … ) ∣ ν ⟩ a ) = U S S − 1 s ∣ w m ⟩ ⊗ F J Q ( J − 2 , J − 4 , … ) ∣ ν ⟩ J (202) \Phi_{m,\nu}\left(UF e_0^s|u_m\rangle\otimes Q(a_{-1},a_{-2},\ldots)|\nu\rangle_a\right)=U^S S_{-1}^s|w_m\rangle\otimes F^J Q(J_{-2},J_{-4},\ldots)|\nu\rangle_J \tag*{(202)} Φ m , ν ( U F e 0 s ∣ u m ⟩ ⊗ Q ( a − 1 , a − 2 , … ) ∣ ν ⟩ a ) = U S S − 1 s ∣ w m ⟩ ⊗ F J Q ( J − 2 , J − 4 , … ) ∣ ν ⟩ J (202)
where 0 ≤ s ≤ m 0 \le s \le m 0 ≤ s ≤ m and Q Q Q is an arbitrary polynomial. Since each factor in (201) is bijective, the map Φ m , ν : V m , ν → W m , ν \Phi_{m,\nu}:V_{m,\nu}\to W_{m,\nu} Φ m , ν : V m , ν → W m , ν is also bijective. To establish that Φ m , ν \Phi_{m,\nu} Φ m , ν is a GVISO, we also need to verify it acts as an intertwiner.
The full stress tensor of R ( p ) \mathcal{R}^{(p)} R ( p ) is
T V = T s l 2 + T a = T s l 2 − p 4 : a a : , c V = 4 − 6 p . (203) T_V=T_{\mathfrak{sl}_2}+T_a=T_{\mathfrak{sl}_2}-\frac{p}{4}:aa:,\qquad c_V=4-6p. \tag*{(203)} T V = T sl 2 + T a = T sl 2 − 4 p : aa : , c V = 4 − 6 p . (203)
This follows because (165) is a conformal embedding, so the stress tensor is a composite of the affine and Heisenberg currents. With this we have
H = 2 L 0 V + 1 2 h 0 = H s l 2 − p 2 a 0 2 − p ∑ n > 0 a − n a n , (204) H=2L_0^V+\frac{1}{2}h_0=H_{\mathfrak{sl}_2}-\frac{p}{2}a_0^2-p\sum_{n>0}a_{-n}a_n, \tag*{(204)} H = 2 L 0 V + 2 1 h 0 = H sl 2 − 2 p a 0 2 − p n > 0 ∑ a − n a n , (204)
where H s l 2 H_{\mathfrak{sl}_2} H sl 2 is given in (180) , and L 0 V L_0^V L 0 V is the zero mode of T V T_V T V .
Combining (196) and (173) gives
Φ m , ν H = Φ m , ν ( H s l 2 − p 2 a 0 2 − p ∑ n > 0 a − n a n ) (205) \Phi_{m,\nu}H=\Phi_{m,\nu}\left(H_{\mathfrak{sl}_2}-\frac{p}{2}a_0^2-p\sum_{n>0}a_{-n}a_n\right) \tag*{(205)} Φ m , ν H = Φ m , ν ( H sl 2 − 2 p a 0 2 − p n > 0 ∑ a − n a n ) (205)
= ( S 0 + N o d d − p 2 J 0 2 − p ∑ n > 0 J − 2 n J 2 n ) Φ m , ν (206) =\left(S_0+N_{\mathrm{odd}}-\frac{p}{2}J_0^2-p\sum_{n>0}J_{-2n}J_{2n}\right)\Phi_{m,\nu} \tag*{(206)} = ( S 0 + N odd − 2 p J 0 2 − p n > 0 ∑ J − 2 n J 2 n ) Φ m , ν (206)
= ( S 0 − p 2 J 0 2 − p ∑ r > 0 J − r J r ) Φ m , ν , (207) =\left(S_0-\frac{p}{2}J_0^2-p\sum_{r>0}J_{-r}J_r\right)\Phi_{m,\nu}, \tag*{(207)} = ( S 0 − 2 p J 0 2 − p r > 0 ∑ J − r J r ) Φ m , ν , (207)
where we used the definition of N o d d N_{\mathrm{odd}} N odd given in (194) . Using (189) , one realizes that the new grading is precisely L 0 W L_0^W L 0 W . As a result, we have
Φ m , ν H = L 0 W Φ m , ν . \Phi_{m,\nu}H=L_0^W\Phi_{m,\nu}. Φ m , ν H = L 0 W Φ m , ν .
Therefore, we have the following graded vector-space isomorphism for each allowed pair ( m , ν ) (m,\nu) ( m , ν ) :
Φ m , ν : ( V m , ν , H , a 0 ) ≅ g r ( W m , ν , L 0 W , J 0 ) , (208) \Phi_{m,\nu}:(V_{m,\nu},H,a_0)\cong_{\mathrm{gr}}(W_{m,\nu},L_0^W,J_0), \tag*{(208)} Φ m , ν : ( V m , ν , H , a 0 ) ≅ gr ( W m , ν , L 0 W , J 0 ) , (208)
which satisfies the further property Φ m , ν a n = J 2 n Φ m , ν \Phi_{m,\nu}a_n=J_{2n}\Phi_{m,\nu} Φ m , ν a n = J 2 n Φ m , ν for every n ∈ Z n\in\mathbb{Z} n ∈ Z .
Character identity from graded isomorphism
For the lowest-degree vectors v ∈ V m , ν v\in V_{m,\nu} v ∈ V m , ν and w ∈ W m , ν w\in W_{m,\nu} w ∈ W m , ν , we have H v = Δ m , ν v Hv=\Delta_{m,\nu}v H v = Δ m , ν v and L 0 W w = Δ m , ν w L_0^Ww=\Delta_{m,\nu}w L 0 W w = Δ m , ν w , where
Δ m , ν = λ m − p 2 ν 2 = p m ( m + 2 ) − m − p ν 2 2 . (209) \Delta_{m,\nu}=\lambda_m-\frac{p}{2}\nu^2=\frac{pm(m+2)-m-p\nu^2}{2}. \tag*{(209)} Δ m , ν = λ m − 2 p ν 2 = 2 p m ( m + 2 ) − m − p ν 2 . (209)
On the W W W side, this follows from (191) , (167) and (189) . On the V V V side, this follows from (181) , (167) and (204) . Thus ν \nu ν fixes the charge and ( m , ν ) (m,\nu) ( m , ν ) fixes the lowest weight, while descendants occupy higher weights in the same branch.
Taking the direct sum over the allowed labels gives
Φ = ⨁ m , ν Φ m , ν : ( V , H , a 0 ) ≅ g r ( W , L 0 W , J 0 ) . (210) \Phi= \bigoplus_{m,\nu} \Phi_{m,\nu} : (V,H,a_0) \cong_{\mathrm{gr}} (W,L_0^W,J_0). \tag*{(210)} Φ = m , ν ⨁ Φ m , ν : ( V , H , a 0 ) ≅ gr ( W , L 0 W , J 0 ) . (210)
Since ∣ ν ∣ ≤ m |\nu| \le m ∣ ν ∣ ≤ m , we have Δ m , ν ≥ ( p − 1 2 ) m \Delta_{m,\nu} \ge(p-\frac{1}{2})m Δ m , ν ≥ ( p − 2 1 ) m ; hence each total weight receives contributions from finitely many branches.
The GVISO gives the following identity of graded characters, which can be computed explicitly
Tr V m , ν q H w a 0 = Tr W m , ν q L 0 W w J 0 (211) \operatorname{Tr}_{V_{m,\nu}}q^{H}w^{a_0}=\operatorname{Tr}_{W_{m,\nu}}q^{L_0^W}w^{J_0} \tag*{(211)} Tr V m , ν q H w a 0 = Tr W m , ν q L 0 W w J 0 (211)
= w ν q Δ m , ν 1 + q + ⋯ + q m ( q 2 ; q 2 ) ∞ 2 ( q 3 ; q 2 ) ∞ ( q ; q 2 ) ∞ (212) = w^\nu q^{\Delta_{m,\nu}}\frac{1+q+\cdots+q^m}{(q^2;q^2)_\infty^2(q^3;q^2)_\infty(q;q^2)_\infty} \tag*{(212)} = w ν q Δ m , ν ( q 2 ; q 2 ) ∞ 2 ( q 3 ; q 2 ) ∞ ( q ; q 2 ) ∞ 1 + q + ⋯ + q m (212)
= w ν q Δ m , ν 1 − q m + 1 ( q ; q 2 ) ∞ 2 , (213) = w^\nu q^{\Delta_{m,\nu}}\frac{1-q^{m+1}}{(q;q^2)_\infty^2}, \tag*{(213)} = w ν q Δ m , ν ( q ; q 2 ) ∞ 2 1 − q m + 1 , (213)
where ( b ; q ) ∞ = ∏ n = 0 ∞ ( 1 − b q n ) (b;q)_\infty=\prod_{n=0}^{\infty}(1-bq^n) ( b ; q ) ∞ = ∏ n = 0 ∞ ( 1 − b q n ) . On the V V V side, V m , ν = V m ( p ) ⊗ F ν a V_{m,\nu}=V_m^{(p)}\otimes\mathcal{F}_\nu^a V m , ν = V m ( p ) ⊗ F ν a in (158) admits a basis (178) tensored with (169) . Therefore in (212) the numerator comes from e 0 s e_0^s e 0 s and the denominator accounts for the contribution of a − n , h − n , e − n , f − n a_{-n},h_{-n},e_{-n},f_{-n} a − n , h − n , e − n , f − n , respectively, after using their degrees in (6.51).
The flavored characters of the two VOAs are defined as
χ V ( Q , x , w ) = Tr V Q L 0 V x h 0 w a 0 , χ W ( q , w ) = Tr W q L 0 W w J 0 , (214) \chi_V(Q,x,w)=\operatorname{Tr}_V Q^{L_0^V}x^{h_0}w^{a_0},\qquad\chi_W(q,w)=\operatorname{Tr}_W q^{L_0^W}w^{J_0}, \tag*{(214)} χ V ( Q , x , w ) = Tr V Q L 0 V x h 0 w a 0 , χ W ( q , w ) = Tr W q L 0 W w J 0 , (214)
where x x x is the S U ( 2 ) SU(2) S U ( 2 ) flavor fugacity and w w w is the U ( 1 ) U(1) U ( 1 ) fugacity.
Summing over different sectors in (211) thus gives
χ V ( q 2 , q 1 / 2 , w ) = χ W ( q , w ) = ∑ m , ν Tr V m , ν q H w a 0 = ∑ m , ν Tr W m , ν q L 0 W w J 0 = PE [ q + q 2 − q 2 p − 1 − q 2 p + ( w + w − 1 ) ( q p − 1 / 2 − q p + 3 / 2 ) ( 1 − q ) ( 1 − q 2 p ) ] . (215) \begin{aligned}
\chi_V(q^2,q^{1/2},w)&=\chi_W(q,w)\\
&=\sum_{m,\nu}\operatorname{Tr}_{V_{m,\nu}}q^Hw^{a_0}
=\sum_{m,\nu}\operatorname{Tr}_{W_{m,\nu}}q^{L_0^W}w^{J_0}\\
&=\operatorname{PE}\left[\frac{q+q^2-q^{2p-1}-q^{2p}+(w+w^{-1})(q^{p-1/2}-q^{p+3/2})}{(1-q)(1-q^{2p})}\right]. \tag*{(215)}
\end{aligned} χ V ( q 2 , q 1/2 , w ) = χ W ( q , w ) = m , ν ∑ Tr V m , ν q H w a 0 = m , ν ∑ Tr W m , ν q L 0 W w J 0 = PE [ ( 1 − q ) ( 1 − q 2 p ) q + q 2 − q 2 p − 1 − q 2 p + ( w + w − 1 ) ( q p − 1/2 − q p + 3/2 ) ] . (215)
where the last step is derived in Appendix C. This agrees with the index prediction (144) and (143) .
( A ( h + 1 ) ( p − 1 ) , G ) (A_{(h+1)(p-1)},G) ( A ( h + 1 ) ( p − 1 ) , G ) and D p ( G ) D_p(G) D p ( G )
We want to study the GVISO between the affine VOA and the corresponding principal W-algebra, which are the associated VOAs of D p ( G ) D_p(G) D p ( G ) and ( A ( h + 1 ) ( p − 1 ) , G ) (A_{(h+1)(p-1)},G) ( A ( h + 1 ) ( p − 1 ) , G ) , respectively.
Consider the principal boundary admissible level
k V = − h ∨ + h ∨ p , k W = − h ∨ + h ∨ u , u = p ( h + 1 ) , (216) k_V=-h^\vee+\frac{h^\vee}{p},\qquad k_W=-h^\vee+\frac{h^\vee}{u},\qquad u=p(h+1), \tag*{(216)} k V = − h ∨ + p h ∨ , k W = − h ∨ + u h ∨ , u = p ( h + 1 ) , (216)
with
p ≥ 2 , gcd ( p , h ∨ ) = gcd ( p , r ∨ ) = 1 , (217) p\ge2,\qquad\gcd(p,h^\vee)=\gcd(p,r^\vee)=1, \tag*{(217)} p ≥ 2 , g cd( p , h ∨ ) = g cd( p , r ∨ ) = 1 , (217)
where r ∨ r^\vee r ∨ is the lacing number.
Let us use V ^ , W ^ \widehat{V},\widehat{W} V , W for the universal algebras, and V , W V,W V , W for their simple quotients, namely
V ^ = V k V ( g ) , V = L k V ( g ) = V ^ / N V , \widehat{V}=V^{k_V}(\mathfrak{g}),\qquad V=L_{k_V}(\mathfrak{g})=\widehat{V}/\mathcal{N}_V, V = V k V ( g ) , V = L k V ( g ) = V / N V ,
W ^ = W k W ( g , f p r i n ) , W = W k W ( g , f p r i n ) = W ^ / N W , \widehat{W}=W^{k_W}(\mathfrak{g},f_{\mathrm{prin}}),\qquad W=W_{k_W}(\mathfrak{g},f_{\mathrm{prin}})=\widehat{W}/\mathcal{N}_W, W = W k W ( g , f prin ) , W = W k W ( g , f prin ) = W / N W ,
where N V , N W \mathcal{N}_V,\mathcal{N}_W N V , N W are the maximal proper ideals of V ^ , W ^ \widehat{V},\widehat{W} V , W formed by all the null states.
We consider the mode expansion for affine currents and W W W generators:
J a ( z ) = ∑ n ∈ Z J n a z − n − 1 , W i ( z ) = ∑ n ∈ Z W n i z − n − d i . (218) J^a(z)=\sum_{n\in\mathbb{Z}}J^a_nz^{-n-1},\qquad W^i(z)=\sum_{n\in\mathbb{Z}}W^i_nz^{-n-d_i}. \tag*{(218)} J a ( z ) = n ∈ Z ∑ J n a z − n − 1 , W i ( z ) = n ∈ Z ∑ W n i z − n − d i . (218)
Here d i = e i + 1 d_i=e_i+1 d i = e i + 1 , where e i , d i e_i,d_i e i , d i are the exponents and degrees of the Casimirs of g \mathfrak{g} g , respectively. See Table 1 for explicit values.
The principal W W W algebra has stress tensor given by
W 1 ( z ) = T W ( z ) = ∑ m ∈ Z L m W z − m − 2 . (219) W^1(z)=T_W(z)=\sum_{m\in\mathbb{Z}}L^W_mz^{-m-2}. \tag*{(219)} W 1 ( z ) = T W ( z ) = m ∈ Z ∑ L m W z − m − 2 . (219)
Set
ℓ = h + 1 , H = ℓ L 0 V − J 0 ρ ∨ , (220) \ell=h+1,\qquad H=\ell L^V_0-J^{\rho^\vee}_0, \tag*{(220)} ℓ = h + 1 , H = ℓ L 0 V − J 0 ρ ∨ , (220)
where J 0 ρ ∨ J^{\rho^\vee}_0 J 0 ρ ∨ is the zero mode of the affine Cartan current associated with the dual Weyl vector ρ ∨ \rho^\vee ρ ∨ .
Universal algebra
The vacuum conditions imply that the Poincaré–Birkhoff–Witt (PBW) creation modes are
J − n a , n ≥ 1 , a = 1 , … , dim g , (221) J^a_{-n},\qquad n\ge1,\qquad a=1,\ldots,\dim\mathfrak{g}, \tag*{(221)} J − n a , n ≥ 1 , a = 1 , … , dim g , (221)
and
W − n i , n ≥ d i , i = 1 , … , r . (222) W^i_{-n},\qquad n\ge d_i,\qquad i=1,\ldots,r. \tag*{(222)} W − n i , n ≥ d i , i = 1 , … , r . (222)
We now organize the current labels of J a J^a J a into principal s l 2 \mathfrak{sl}_2 sl 2 multiplets:
a = ( i , t ) , − e i ≤ t ≤ e i , i = 1 , … , r , (223) a=(i,t),\qquad-e_i\le t\le e_i,\qquad i=1,\ldots,r, \tag*{(223)} a = ( i , t ) , − e i ≤ t ≤ e i , i = 1 , … , r , (223)
so that
[ J 0 ρ ∨ , J n i , t ] = t J n i , t . (224) [J^{\rho^\vee}_0,J^{i,t}_n]=tJ^{i,t}_n. \tag*{(224)} [ J 0 ρ ∨ , J n i , t ] = t J n i , t . (224)
Consequently,
[ H , J − n i , t ] = ( ℓ n − t ) J − n i , t , [ L 0 W , W − m i ] = m W − m i . (225) [H,J_{-n}^{i,t}] = (\ell n-t)J_{-n}^{i,t}, \qquad[L_0^W,W_{-m}^{i}] = mW_{-m}^{i}. \tag*{(225)} [ H , J − n i , t ] = ( ℓ n − t ) J − n i , t , [ L 0 W , W − m i ] = m W − m i . (225)
Every current creator has positive H H H -degree
ℓ n − t ≥ ℓ − ( h − 1 ) = 2 , (226) \ell n-t \ge\ell-(h-1)=2, \tag*{(226)} ℓ n − t ≥ ℓ − ( h − 1 ) = 2 , (226)
because n ≥ 1 n \ge1 n ≥ 1 and t ≤ e i ≤ h − 1 t \le e_i \le h-1 t ≤ e i ≤ h − 1 .
All the exponents can be grouped in pairs such that e i + e i ^ = h e_i+e_{\hat{i}}=h e i + e i ^ = h . Without loss of generality, we can assume that e i ≤ e i ^ e_i \le e_{\hat{i}} e i ≤ e i ^ . Then we can use the following mode map between affine and principal W W W algebra:
Φ ^ : J − n i , t ⟷ W − ( ℓ n − t ) i ^ , − e i ≤ t ≤ e i , (227) \widehat{\Phi}: \quad J_{-n}^{i,t} \longleftrightarrow W_{-(\ell n-t)}^{\hat{i}}, \qquad-e_i \le t \le e_i, \tag*{(227)} Φ : J − n i , t ⟷ W − ( ℓ n − t ) i ^ , − e i ≤ t ≤ e i , (227)
and
Φ ^ : J ^ − n i , t ⟷ { W − ( ℓ n − t ) i , − e i ≤ t ≤ e i ^ , W − ( ℓ n − t ) i ^ , − e i ^ ≤ t ≤ − e i − 1. (228) \widehat{\Phi}: \quad\widehat{J}_{-n}^{i,t} \longleftrightarrow
\begin{cases}
W_{-(\ell n-t)}^{i}, & -e_i \le t \le e_{\hat{i}}, \\
W_{-(\ell n-t)}^{\hat{i}}, & -e_{\hat{i}} \le t \le-e_i-1.
\end{cases} \tag*{(228)} Φ : J − n i , t ⟷ { W − ( ℓ n − t ) i , W − ( ℓ n − t ) i ^ , − e i ≤ t ≤ e i ^ , − e i ^ ≤ t ≤ − e i − 1. (228)
For an unpaired self-dual exponent e i = e i ^ = h / 2 e_i=e_{\hat{i}}=h/2 e i = e i ^ = h /2 with i = i ^ i=\hat{i} i = i ^ , the rule reduces to
Φ ^ : J − n i , r ⟷ W − ( ℓ n − r ) i , − e i ≤ r ≤ e i . (229) \widehat{\Phi}: \quad J_{-n}^{i,r} \longleftrightarrow W_{-(\ell n-r)}^{i}, \qquad-e_i \le r \le e_i. \tag*{(229)} Φ : J − n i , r ⟷ W − ( ℓ n − r ) i , − e i ≤ r ≤ e i . (229)
This is bijective for a simple reason. For fixed n n n and a pair ( i , i ^ ) (i,\hat{i}) ( i , i ^ ) , as t t t varies, the map in (228) sends J − n J_{-n} J − n to W − k i W_{-k}^{i} W − k i with mode numbers
k : ℓ n − e i ^ , … , ℓ n + e i . (230) k:\quad\ell n-e_{\hat{i}},\ldots,\ell n+e_i. \tag*{(230)} k : ℓ n − e i ^ , … , ℓ n + e i . (230)
This is a consecutive block of length ℓ \ell ℓ , because e i + e i ^ + 1 = h + 1 = ℓ e_i+e_{\hat{i}}+1=h+1=\ell e i + e i ^ + 1 = h + 1 = ℓ . At n = 1 n=1 n = 1 , it starts at ℓ − e i ^ = d i \ell-e_{\hat{i}}=d_i ℓ − e i ^ = d i . Increasing n n n by one gives the next consecutive block. Thus every allowed creator
W − d i i , W − ( d i + 1 ) i , … , (231) W_{-d_i}^{i},W_{-(d_i+1)}^{i},\ldots, \tag*{(231)} W − d i i , W − ( d i + 1 ) i , … , (231)
appears exactly once. For W i ^ W^{\hat{i}} W i ^ , the map in (227) (228) sends J − n J_{-n} J − n to W − k i ^ W_{-k}^{\hat{i}} W − k i ^ with mode numbers
k : ℓ n − e i , … , ℓ n + e i ^ . (232) k:\quad\ell n-e_i,\ldots,\ell n+e_{\hat{i}}. \tag*{(232)} k : ℓ n − e i , … , ℓ n + e i ^ . (232)
Again, this is a consecutive block of length ℓ \ell ℓ . For n = 1 n=1 n = 1 it starts at ℓ − e i = d i ^ \ell-e_i=d_{\hat{i}} ℓ − e i = d i ^ , and it is easy to see that every mode of W k i ^ W_k^{\hat{i}} W k i ^ appears exactly once for k ≥ d i ^ k \ge d_{\hat{i}} k ≥ d i ^ . Using the PBW basis, the map (227) (228) (229) further allows us to establish a bijection between the two vector spaces of the affine and W W W algebras.
Using the commutator in (225) , one can also see that (227) (228) (229) preserves the grading, namely
Φ ^ H = L 0 W Φ ^ . (233) \widehat{\Phi}H=L_0^W\widehat{\Phi}. \tag*{(233)} Φ H = L 0 W Φ . (233)
Therefore, we have proved the GVISO for the universal algebras of affine and principal $W algebras:
Φ ^ : ( V ^ , H ) ≅ g r ( W ^ , L 0 W ) . (234) \widehat{\Phi} : (\widehat{V}, H) \cong_{\mathrm{gr}} (\widehat{W}, L_0^W). \tag*{(234)} Φ : ( V , H ) ≅ gr ( W , L 0 W ) . (234)
This is consistent with the following character identity between affine vertex and principal W W W algebras:
Tr V ^ q H = PE [ ∑ i = 1 r ∑ t = − e i e i q ℓ − t 1 − q ℓ ] = PE [ ∑ i = 1 r q d i 1 − q ] = Tr W ^ q L 0 W , (235) \operatorname{Tr}_{\widehat{V}} q^H = \operatorname{PE}\left[\frac{\sum_{i=1}^{r}\sum_{t=-e_i}^{e_i}q^{\ell-t}}{1-q^\ell}\right] = \operatorname{PE}\left[\frac{\sum_{i=1}^{r}q^{d_i}}{1-q}\right] = \operatorname{Tr}_{\widehat{W}}q^{L_0^W}, \tag*{(235)} Tr V q H = PE [ 1 − q ℓ ∑ i = 1 r ∑ t = − e i e i q ℓ − t ] = PE [ 1 − q ∑ i = 1 r q d i ] = Tr W q L 0 W , (235)
where the H H H -grading on the left can be seen from (225) .
Simple quotient
Now we want to generalize and establish the GVISO for the simple quotients V , W V,W V , W
Φ : ( V , H ) ≅ g r ( W , L 0 W ) . (236) \Phi: (V,H) \cong_{\mathrm{gr}} (W,L_0^W). \tag*{(236)} Φ : ( V , H ) ≅ gr ( W , L 0 W ) . (236)
For this purpose, we need to account for the null states. The previous map Φ ^ \widehat{\Phi} Φ defined in (227) (228) (229) holds for the universal algebras where the ordered creation-mode monomials form PBW bases. Therefore, we need to modify and construct a new Φ ^ ′ \widehat{\Phi}' Φ ′ such that it preserves the null states, in the sense that
Φ ^ ′ : ( V ^ , H ) ≅ g r ( W ^ , L 0 W ) , Φ ^ ′ ( N V ) = N W . (237) \widehat{\Phi}':(\widehat{V},H) \cong_{\mathrm{gr}} (\widehat{W},L_0^W),\qquad\widehat{\Phi}'(\mathcal{N}_V)=\mathcal{N}_W. \tag*{(237)} Φ ′ : ( V , H ) ≅ gr ( W , L 0 W ) , Φ ′ ( N V ) = N W . (237)
Linearity implies that
Φ ^ ′ ( v + N V ) = Φ ^ ′ ( v ) + N W . (238) \widehat{\Phi}'(v+\mathcal{N}_V)=\widehat{\Phi}'(v)+\mathcal{N}_W. \tag*{(238)} Φ ′ ( v + N V ) = Φ ′ ( v ) + N W . (238)
Let us denote by [ v ] ∈ V [v]\in V [ v ] ∈ V the image of v ∈ V ^ v\in\widehat{V} v ∈ V under the quotient map, and similarly for W W W . Then we can define
Φ ( [ v ] ) = [ Φ ^ ′ ( v ) ] ∈ W . (239) \Phi([v])=[\widehat{\Phi}'(v)]\in W. \tag*{(239)} Φ ([ v ]) = [ Φ ′ ( v )] ∈ W . (239)
This map is injective because
Φ ( [ v ] ) = 0 ⇒ Φ ^ ′ ( v ) ∈ N W ⇒ v ∈ N V ⇒ [ v ] = 0. (240) \Phi([v])=0 \Rightarrow\widehat{\Phi}'(v)\in\mathcal{N}_W \Rightarrow v\in\mathcal{N}_V \Rightarrow[v]=0. \tag*{(240)} Φ ([ v ]) = 0 ⇒ Φ ′ ( v ) ∈ N W ⇒ v ∈ N V ⇒ [ v ] = 0. (240)
Surjectivity follows from that of Φ ^ ′ \widehat{\Phi}' Φ ′ , and grading preservation descends to the quotient. This gives the GVISO (236) . The key in this construction is thus the map Φ ^ ′ \widehat{\Phi}' Φ ′ in (237) for the universal algebras
Φ ^ ′ : N V → N W , dim ( N V ) d = dim ( N W ) d , ∀ d , (241) \widehat{\Phi}':\mathcal{N}_V\to\mathcal{N}_W,\qquad\dim(\mathcal{N}_V)_d=\dim(\mathcal{N}_W)_d,\qquad\forall d, \tag*{(241)} Φ ′ : N V → N W , dim ( N V ) d = dim ( N W ) d , ∀ d , (241)
where the subscript denotes the degree-d d d component of the respective null ideal under the H H H - and L 0 W L_0^W L 0 W -gradings.
The map Φ ^ ′ \widehat{\Phi}' Φ ′ is guaranteed to exist if the character identity holds, but constructing this explicitly is not straightforward, because a concrete construction requires control of the full null spaces. We are not able to construct the desired Φ ^ ′ \widehat{\Phi}' Φ ′ universally. Nevertheless, below we show this indeed holds for the lowest null state for all g \mathfrak{g} g .
On the affine side, the first null state takes the following form
ψ V = ∑ a i , n i C n 1 ⋯ n s a 1 ⋯ a s J − n 1 a 1 ⋯ J − n s a s ∣ 0 ⟩ , n j > 0 , ∑ j n j = p , (242) \psi_V = \sum_{a_i,n_i} C^{a_1\cdots a_s}_{n_1\cdots n_s} J^{a_1}_{-n_1}\cdots J^{a_s}_{-n_s}|0\rangle,\qquad n_j>0,\qquad\sum_j n_j=p, \tag*{(242)} ψ V = a i , n i ∑ C n 1 ⋯ n s a 1 ⋯ a s J − n 1 a 1 ⋯ J − n s a s ∣0 ⟩ , n j > 0 , j ∑ n j = p , (242)
where p p p is the denominator in the boundary admissible level (216) . This state is a vacuum descendant and, as the generating affine singular vector at this boundary level, is also an affine primary whose horizontal highest weight is the highest root θ \theta θ [30 ] . It generates an adjoint representation under the affine current zero modes. Therefore 25
L 0 V ψ V = p ψ V , J 0 ρ ψ V = ( h − 1 ) ψ V , (243) L^V_0\psi_V=p\psi_V,\qquad J^\rho_0\psi_V=(h-1)\psi_V, \tag*{(243)} L 0 V ψ V = p ψ V , J 0 ρ ψ V = ( h − 1 ) ψ V , (243)
and
H ψ V = d 0 ψ V , d 0 = p ( h + 1 ) − ( h − 1 ) = u − h + 1 , (244) H\psi_V=d_0\psi_V,\qquad d_0=p(h+1)-(h-1)=u-h+1, \tag*{(244)} H ψ V = d 0 ψ V , d 0 = p ( h + 1 ) − ( h − 1 ) = u − h + 1 , (244)
where u u u is given in (216) .
Finite lowering-current zero modes J 0 − α J^{-\alpha}_0 J 0 − α generate the adjoint multiplet from ψ V \psi_V ψ V ; they preserve the L 0 V L^V_0 L 0 V weight and increase the H H H weight. Therefore ψ V \psi_V ψ V spans the null space of lowest H H H -degree, given by (244) .
On the W W W algebra side, the first null state has the structure 26
ψ W = ∑ i j , m j B m 1 ⋯ m s i 1 ⋯ i s W − m 1 i 1 ⋯ W − m s i s ∣ 0 ⟩ , ∑ j m j = d 0 . (245) \psi_W=\sum_{i_j,m_j}B^{i_1\cdots i_s}_{m_1\cdots m_s}W^{i_1}_{-m_1}\cdots W^{i_s}_{-m_s}|0\rangle,\qquad\sum_jm_j=d_0. \tag*{(245)} ψ W = i j , m j ∑ B m 1 ⋯ m s i 1 ⋯ i s W − m 1 i 1 ⋯ W − m s i s ∣0 ⟩ , j ∑ m j = d 0 . (245)
As a singular vector, this state is annihilated by all positive modes W m i ψ W = 0 W^i_m\psi_W=0 W m i ψ W = 0 , m > 0 m>0 m > 0 , which implies that ψ W \psi_W ψ W is also Virasoro primary. The state ψ W \psi_W ψ W has conformal weight d 0 d_0 d 0 and spans the null space at this weight
L 0 W ψ W = d 0 ψ W . (246) L^W_0\psi_W=d_0\psi_W. \tag*{(246)} L 0 W ψ W = d 0 ψ W . (246)
As a result, on both sides the lowest-degree null spaces are one-dimensional and occur in degree d 0 d_0 d 0
dim ( N V ) d 0 = dim ( N W ) d 0 = 1. (247) \dim(\mathcal{N}_V)_{d_0}=\dim(\mathcal{N}_W)_{d_0}=1. \tag*{(247)} dim ( N V ) d 0 = dim ( N W ) d 0 = 1. (247)
This is thus consistent with (241) , and gives a natural map up to this degree such that Φ ^ ′ : ψ V ↦ ψ W \widehat{\Phi}':\psi_V\mapsto\psi_W Φ ′ : ψ V ↦ ψ W .
The null spaces become increasingly complicated at higher degrees, so we will not try
L 0 V ψ V = h θ ψ V , J 0 ρ ψ V = ⟨ θ , ρ ∨ ⟩ ψ V . (248) L^V_0\psi_V=h_\theta\psi_V,\qquad J^\rho_0\psi_V=\langle\theta,\rho^\vee\rangle\psi_V. \tag*{(248)} L 0 V ψ V = h θ ψ V , J 0 ρ ψ V = ⟨ θ , ρ ∨ ⟩ ψ V . (248)
Here
h θ = ( θ , θ + 2 ρ ) 2 ( k V + h ∨ ) = h ∨ k V + h ∨ = p , ⟨ θ , ρ ∨ ⟩ = h − 1. h_\theta=\frac{(\theta,\theta+2\rho)}{2(k_V+h^\vee)}=\frac{h^\vee}{k_V+h^\vee}=p,\qquad\langle\theta,\rho^\vee\rangle=h-1. h θ = 2 ( k V + h ∨ ) ( θ , θ + 2 ρ ) = k V + h ∨ h ∨ = p , ⟨ θ , ρ ∨ ⟩ = h − 1.
where we used ( ρ , θ ) = ⟨ ρ , θ ∨ ⟩ = h ∨ − 1 (\rho,\theta)=\langle\rho,\theta^\vee\rangle=h^\vee-1 ( ρ , θ ) = ⟨ ρ , θ ∨ ⟩ = h ∨ − 1 , ( θ , θ ) = 2 (\theta,\theta)=2 ( θ , θ ) = 2 . to establish (241) in general.
To be more concrete, let us briefly look at the simplest example of g = s l 2 \mathfrak{g}=\mathfrak{sl}_2 g = sl 2 . The two VOAs are given by the L − 2 + 2 / p ( s l 2 ) L_{-2+2/p}(\mathfrak{sl}_2) L − 2 + 2/ p ( sl 2 ) affine vertex algebra and Virasoro minimal model Vir ( 2 , 3 p ) \operatorname{Vir}(2,3p) Vir ( 2 , 3 p ) . Our map in (227) (228) (229) then gives
Φ ^ : J − n e ⟷ L − ( 3 n − 1 ) W , J − n h ⟷ L − 3 n W , J − n f ⟷ L − ( 3 n + 1 ) W . (249) \widehat{\Phi}: \quad J^{e}_{-n} \longleftrightarrow L^{W}_{-(3n-1)}, \qquad J^{h}_{-n} \longleftrightarrow L^{W}_{-3n}, \qquad J^{f}_{-n} \longleftrightarrow L^{W}_{-(3n+1)}. \tag*{(249)} Φ : J − n e ⟷ L − ( 3 n − 1 ) W , J − n h ⟷ L − 3 n W , J − n f ⟷ L − ( 3 n + 1 ) W . (249)
This provides the GVISO for universal algebras. For the simple quotient, Φ ^ \widehat{\Phi} Φ requires modification to preserve the null ideals. 27 In [6 ] , the authors studied the GVISO of the simple quotients in detail and proved (236) using the character formula.
(A1,A5) and (A2,A4)
Our final example concerns
( A 1 , A 5 ) ⟷ ( A 2 , A 4 ) , (250) (A_1,A_5) \longleftrightarrow(A_2,A_4), \tag*{(250)} ( A 1 , A 5 ) ⟷ ( A 2 , A 4 ) , (250)
which is the special case of subsection 5.6 with p = 4 p=4 p = 4 . Their associated VOAs and central charges are
V = V [ ( A 1 , A 5 ) ] = B P − 9 / 4 = W − 9 / 4 ( s l 3 , f s u b r e g ) , c V = − 23 2 , (251) V=V[(A_1,A_5)]=\mathrm{BP}_{-9/4}=\mathcal{W}_{-9/4}(\mathfrak{sl}_3,f_{\mathrm{subreg}}), \qquad c_V=-\frac{23}{2}, \tag*{(251)} V = V [( A 1 , A 5 )] = BP − 9/4 = W − 9/4 ( sl 3 , f subreg ) , c V = − 2 23 , (251)
and
W = V [ ( A 2 , A 4 ) ] = W − 21 / 8 ( s l 3 , f p r i n ) = W 3 ( 3 , 8 ) , c W = − 23 , (252) W=V[(A_2,A_4)]=\mathcal{W}_{-21/8}(\mathfrak{sl}_3,f_{\mathrm{prin}})=\mathcal{W}_3(3,8), \qquad c_W=-23, \tag*{(252)} W = V [( A 2 , A 4 )] = W − 21/8 ( sl 3 , f prin ) = W 3 ( 3 , 8 ) , c W = − 23 , (252)
corresponding to the Bershadsky–Polyakov algebra and the W 3 \mathcal{W}_3 W 3 minimal model. The Bershadsky–Polyakov algebra V V V has four bosonic strong generators J J J , G + G^{+} G + , G − G^{-} G − , T T T of conformal weights 1 1 1 , 3 2 \frac{3}{2} 2 3 , 3 2 \frac{3}{2} 2 3 , 2 2 2 , whose OPEs are given in Appendix § 11 . The W 3 \mathcal{W}_3 W 3 minimal model has strong generators T W T_W T W , U U U of weight 2 , 3 2,3 2 , 3 , where T W T_W T W is the stress tensor. Their mode expansions follow the general equation (152) .
As shown in (134) , we have the following identity
χ V ( q 2 , 1 ) = I ( A 1 , A 5 ) ( q 2 , 1 ) = I ( A 2 , A 4 ) ( q ) = χ W ( q ) = PE [ q 2 + q 3 ( 1 − q ) ( 1 + q 4 ) ] . (253) \chi_V(q^2,1)=\mathcal{I}_{(A_1,A_5)}(q^2,1)=\mathcal{I}_{(A_2,A_4)}(q)=\chi_W(q)=\operatorname{PE}\left[\frac{q^2+q^3}{(1-q)(1+q^4)}\right]. \tag*{(253)} χ V ( q 2 , 1 ) = I ( A 1 , A 5 ) ( q 2 , 1 ) = I ( A 2 , A 4 ) ( q ) = χ W ( q ) = PE [ ( 1 − q ) ( 1 + q 4 ) q 2 + q 3 ] . (253)
This character identity suggests a graded vector-space isomorphism
Φ : ( V , 2 L 0 ) ≅ g r ( W , L 0 W ) . (254) \Phi:(V,2L_0)\cong_{\mathrm{gr}}(W,L_0^W). \tag*{(254)} Φ : ( V , 2 L 0 ) ≅ gr ( W , L 0 W ) . (254)
ψ V = [ ( J − 1 h ) 2 J − 1 e + 4 J − 1 f ( J − 1 e ) 2 + 22 3 J − 2 h J − 1 e − 16 3 J − 2 e J − 1 h − 88 9 J − 3 e ] 1 V , \psi_V=\left[(J^h_{-1})^2J^e_{-1}+4J^f_{-1}(J^e_{-1})^2+\frac{22}{3}J^h_{-2}J^e_{-1}-\frac{16}{3}J^e_{-2}J^h_{-1}-\frac{88}{9}J^e_{-3}\right]1_V, ψ V = [ ( J − 1 h ) 2 J − 1 e + 4 J − 1 f ( J − 1 e ) 2 + 3 22 J − 2 h J − 1 e − 3 16 J − 2 e J − 1 h − 9 88 J − 3 e ] 1 V ,
ψ W = [ ( L − 2 W ) 4 − 4 9 ( L − 3 W ) 2 L − 2 W − 26 9 L − 4 W ( L − 2 W ) 2 + 7 9 ( L − 2 W ) 4 + 88 27 L − 5 W L − 3 W − 278 27 L − 6 W L − 2 W − 81 81 L − 8 W ] 1 W . \psi_W=\left[(L^W_{-2})^4-\frac{4}{9}(L^W_{-3})^2L^W_{-2}-\frac{26}{9}L^W_{-4}(L^W_{-2})^2+\frac{7}{9}(L^W_{-2})^4+\frac{88}{27}L^W_{-5}L^W_{-3}-\frac{278}{27}L^W_{-6}L^W_{-2}-\frac{81}{81}L^W_{-8}\right]1_W. ψ W = [ ( L − 2 W ) 4 − 9 4 ( L − 3 W ) 2 L − 2 W − 9 26 L − 4 W ( L − 2 W ) 2 + 9 7 ( L − 2 W ) 4 + 27 88 L − 5 W L − 3 W − 27 278 L − 6 W L − 2 W − 81 81 L − 8 W ] 1 W .
There is an immediate obstruction to the literal PBW substitution (249) : ψ V \psi_V ψ V contains at most three current factors, whereas the unique target null contains the nonzero four-Virasoro term. Thus the ordered PBW replacement rule (249) would not map ψ V \psi_V ψ V to a nonzero multiple of ψ W \psi_W ψ W , and therefore fails to preserve the null ideals.
Let us first look at the universal algebras and forget about the null relations. Then the vacuum module of V ^ \widehat{V} V is generated by acting with J − n J_{-n} J − n , G − n − 1 2 + G^{+}_{-n-\frac{1}{2}} G − n − 2 1 + , G − n − 1 2 − G^{-}_{-n-\frac{1}{2}} G − n − 2 1 − , L − n L_{-n} L − n on the vacuum state 1 V \mathbf{1}_{V} 1 V . Similarly, the vacuum module of W ^ \widehat{W} W is generated by acting with L − m W L^{W}_{-m} L − m W , U − n U_{-n} U − n on the vacuum state 1 W \mathbf{1}_{W} 1 W . We can then establish a one-to-one correspondence Φ ^ \widehat{\Phi} Φ between them via
J − n ⟼ L − 2 n W , n ≥ 1 , G − n − 1 2 + ⟼ L − ( 2 n + 1 ) W , n ≥ 1 , G − n − 1 2 − ⟼ U − ( 2 n + 1 ) , n ≥ 1 , L − n ⟼ U − 2 n , n ≥ 2. (255) \begin{aligned}
J_{-n} &\longmapsto L^{W}_{-2n}, && n \ge1,\\
G^{+}_{-n-\frac{1}{2}} &\longmapsto L^{W}_{-(2n+1)}, && n \ge1,\\
G^{-}_{-n-\frac{1}{2}} &\longmapsto U_{-(2n+1)}, && n \ge1,\\
L_{-n} &\longmapsto U_{-2n}, && n \ge2. \tag*{(255)}
\end{aligned} J − n G − n − 2 1 + G − n − 2 1 − L − n ⟼ L − 2 n W , ⟼ L − ( 2 n + 1 ) W , ⟼ U − ( 2 n + 1 ) , ⟼ U − 2 n , n ≥ 1 , n ≥ 1 , n ≥ 1 , n ≥ 2. (255)
together with 1 V ↦ 1 W \mathbf{1}_{V} \mapsto\mathbf{1}_{W} 1 V ↦ 1 W . This can also be understood from the character identity between two universal algebras V ^ \widehat{V} V and W ^ \widehat{W} W
χ V ^ ( q 2 , 1 ) = PE [ q 2 + 2 q 3 + q 4 1 − q 2 ] = PE [ q 2 + q 3 1 − q ] = χ W ^ ( q ) . (256) \chi_{\widehat{V}}(q^{2},1)=\operatorname{PE}\left[\frac{q^{2}+2q^{3}+q^{4}}{1-q^{2}}\right]=\operatorname{PE}\left[\frac{q^{2}+q^{3}}{1-q}\right]=\chi_{\widehat{W}}(q). \tag*{(256)} χ V ( q 2 , 1 ) = PE [ 1 − q 2 q 2 + 2 q 3 + q 4 ] = PE [ 1 − q q 2 + q 3 ] = χ W ( q ) . (256)
To establish the correspondence between the simple quotients V V V and W W W , we need to account for the null relations.
On the BP side, the first vacuum null has conformal weight 3:
ψ V = ( G − 3 2 + G − 3 2 − + J − 1 3 − 3 J − 2 J − 1 + 5 4 J − 3 − 3 2 J − 1 L − 2 + 3 8 L − 3 ) 1 V . (257) \psi_{V}=\left(G^{+}_{-\frac{3}{2}}G^{-}_{-\frac{3}{2}}+J^{3}_{-1}-3J_{-2}J_{-1}+\frac{5}{4}J_{-3}-\frac{3}{2}J_{-1}L_{-2}+\frac{3}{8}L_{-3}\right)\mathbf{1}_{V}. \tag*{(257)} ψ V = ( G − 2 3 + G − 2 3 − + J − 1 3 − 3 J − 2 J − 1 + 4 5 J − 3 − 2 3 J − 1 L − 2 + 8 3 L − 3 ) 1 V . (257)
On the W 3 W_{3} W 3 side, one similarly finds a null state at weight 628
ψ W = 1 558 ( 558 U − 3 2 − 32 ( L − 2 W ) 3 + 51 ( L − 2 W ) 2 + 168 L − 4 W L − 2 W − 216 L − 6 W ) 1 W . (258) \psi_{W}=\frac{1}{558}\left(558U^{2}_{-3}-32(L^{W}_{-2})^{3}+51(L^{W}_{-2})^{2}+168L^{W}_{-4}L^{W}_{-2}-216L^{W}_{-6}\right)\mathbf{1}_{W}. \tag*{(258)} ψ W = 558 1 ( 558 U − 3 2 − 32 ( L − 2 W ) 3 + 51 ( L − 2 W ) 2 + 168 L − 4 W L − 2 W − 216 L − 6 W ) 1 W . (258)
We can then apply the map (255) to (257)
Φ ^ ( ψ V ) = ( L − 3 W U − 3 + ( L − 2 W ) 3 − 3 L − 4 W L − 2 W + 5 4 L − 6 W − 3 2 L − 2 W U − 4 + 3 8 U − 6 ) 1 W . (259) \widehat{\Phi}(\psi_{V})=\left(L^{W}_{-3}U_{-3}+(L^{W}_{-2})^{3}-3L^{W}_{-4}L^{W}_{-2}+\frac{5}{4}L^{W}_{-6}-\frac{3}{2}L^{W}_{-2}U_{-4}+\frac{3}{8}U_{-6}\right)\mathbf{1}_{W}. \tag*{(259)} Φ ( ψ V ) = ( L − 3 W U − 3 + ( L − 2 W ) 3 − 3 L − 4 W L − 2 W + 4 5 L − 6 W − 2 3 L − 2 W U − 4 + 8 3 U − 6 ) 1 W . (259)
Therefore the resulting state Φ ^ ( ψ V ) ∝̸ ψ W \widehat{\Phi}(\psi_{V})\not\propto\psi_{W} Φ ( ψ V ) ∝ ψ W , hence is not null. So we need to modify the map Φ ^ \widehat{\Phi} Φ to a new one Φ ^ ′ \widehat{\Phi}' Φ ′ at this order. One choice is given by
Φ ^ ′ ( v ) = Φ ^ ( v ) , v ∈ Span { J − 1 3 1 V , J − 2 J − 1 1 V , J − 3 1 V , J − 1 L − 2 1 V , L − 3 1 V } , (260) \widehat{\Phi}'(v)=\widehat{\Phi}(v),\qquad v\in\operatorname{Span}\{J^{3}_{-1}\mathbf{1}_{V},J_{-2}J_{-1}\mathbf{1}_{V},J_{-3}\mathbf{1}_{V},J_{-1}L_{-2}\mathbf{1}_{V},L_{-3}\mathbf{1}_{V}\}, \tag*{(260)} Φ ′ ( v ) = Φ ( v ) , v ∈ Span { J − 1 3 1 V , J − 2 J − 1 1 V , J − 3 1 V , J − 1 L − 2 1 V , L − 3 1 V } , (260)
Φ ^ ′ ( G − 3 2 + G − 3 2 − 1 V ) = ( L − 3 W ) 2 1 W , Φ ^ ′ ( G − 3 2 − G − 3 2 − 1 V ) = L − 3 W U − 3 1 W , (261) \widehat{\Phi}'(G^{+}_{-\frac{3}{2}}G^{-}_{-\frac{3}{2}}\mathbf{1}_{V})=(L^{W}_{-3})^{2}\mathbf{1}_{W},\qquad\widehat{\Phi}'(G^{-}_{-\frac{3}{2}}G^{-}_{-\frac{3}{2}}\mathbf{1}_{V})=L^{W}_{-3}U_{-3}\mathbf{1}_{W}, \tag*{(261)} Φ ′ ( G − 2 3 + G − 2 3 − 1 V ) = ( L − 3 W ) 2 1 W , Φ ′ ( G − 2 3 − G − 2 3 − 1 V ) = L − 3 W U − 3 1 W , (261)
Φ ^ ′ ( G − 3 2 + G − 3 2 − 1 V ) = ψ W − Φ ^ ( ψ V − G − 3 2 + G − 3 2 − 1 V ) . (262) \widehat{\Phi}'(G^{+}_{-\frac{3}{2}}G^{-}_{-\frac{3}{2}}\mathbf{1}_{V})=\psi_{W}-\widehat{\Phi}\left(\psi_{V}-G^{+}_{-\frac{3}{2}}G^{-}_{-\frac{3}{2}}\mathbf{1}_{V}\right). \tag*{(262)} Φ ′ ( G − 2 3 + G − 2 3 − 1 V ) = ψ W − Φ ( ψ V − G − 2 3 + G − 2 3 − 1 V ) . (262)
The corrected map satisfies Φ ^ ′ ( ψ V ) = ψ W \widehat{\Phi}'(\psi_{V})=\psi_{W} Φ ′ ( ψ V ) = ψ W . It is also easy to show that this map is invertible. Thus the map identifies the one-dimensional null spaces and induces the claimed GVISO through degree six, by (239) .
We do not give an explicit formula at higher degrees. Existence follows by choosing bases of the null spaces and their complements at each degree: equality of both universal and simple characters implies equality of the corresponding dimensions.
Character identities for modules
In this section, we generalize the previous discussion of vacuum character identities to non-vacuum modules. On the SCFT side, non-vacuum modules correspond to the insertion of certain BPS surface defects [32 ] . We will first discuss the module identities in a concrete example involving three AD theories in the next subsection, and then formulate a more general conjecture.
A detailed example: (A1,A5), (A1,D4), (A2,A4)
We start by considering the index identities previously established for a triple of AD theories. The three AD theories under consideration are ( A 1 , A 5 ) (A_1,A_5) ( A 1 , A 5 ) , ( A 1 , D 4 ) (A_1,D_4) ( A 1 , D 4 ) , ( A 2 , A 4 ) (A_2,A_4) ( A 2 , A 4 ) , and the corresponding index identities are: 29
I ( A 2 , A 4 ) ( q ) = I ( A 1 , A 5 ) ( q 2 ; 1 ) , (263) \mathcal{I}_{(A_2,A_4)}(q)=\mathcal{I}_{(A_1,A_5)}(q^2;1), \tag*{(263)} I ( A 2 , A 4 ) ( q ) = I ( A 1 , A 5 ) ( q 2 ; 1 ) , (263)
I ( A 1 , A 5 ) ( q ; w ) = I ( A 1 , D 4 ) ( q 2 ; w q − 1 / 2 , w − 1 q − 1 / 2 , q ) , (264) \mathcal{I}_{(A_1,A_5)}(q;w)=\mathcal{I}_{(A_1,D_4)}(q^2;wq^{-1/2},w^{-1}q^{-1/2},q), \tag*{(264)} I ( A 1 , A 5 ) ( q ; w ) = I ( A 1 , D 4 ) ( q 2 ; w q − 1/2 , w − 1 q − 1/2 , q ) , (264)
Together these identities imply
I ( A 2 , A 4 ) ( q ) = I ( A 1 , D 4 ) ( q 4 ; q − 1 , q − 1 , q 2 ) . (265) \mathcal{I}_{(A_2,A_4)}(q)=\mathcal{I}_{(A_1,D_4)}(q^4;q^{-1},q^{-1},q^2). \tag*{(265)} I ( A 2 , A 4 ) ( q ) = I ( A 1 , D 4 ) ( q 4 ; q − 1 , q − 1 , q 2 ) . (265)
We will use the shorthand notation s l 3 \mathfrak{sl}_3 sl 3 , BP, and W 3 W_3 W 3 to denote their associated VOAs:
s l 3 ≡ L − 3 / 2 ( s l 3 ) = V [ ( A 1 , D 4 ) ] , (266) \mathfrak{sl}_3\equiv L_{-3/2}(\mathfrak{sl}_3)=V[(A_1,D_4)], \tag*{(266)} sl 3 ≡ L − 3/2 ( sl 3 ) = V [( A 1 , D 4 )] , (266)
B P ≡ W − 9 / 4 ( s l 3 , f s u b r e g ) = V [ ( A 1 , A 5 ) ] , (267) \mathrm{BP}\equiv W_{-9/4}(\mathfrak{sl}_3,f_{\mathrm{subreg}})=V[(A_1,A_5)], \tag*{(267)} BP ≡ W − 9/4 ( sl 3 , f subreg ) = V [( A 1 , A 5 )] , (267)
W 3 ≡ W − 21 / 8 ( s l 3 , f p r i n ) = W 3 ( 3 , 8 ) = V [ ( A 2 , A 4 ) ] . (268) W_3\equiv W_{-21/8}(\mathfrak{sl}_3,f_{\mathrm{prin}})=W_3(3,8)=V[(A_2,A_4)]. \tag*{(268)} W 3 ≡ W − 21/8 ( sl 3 , f prin ) = W 3 ( 3 , 8 ) = V [( A 2 , A 4 )] . (268)
The central charges are given by
c W 3 = − 23 , c B P = − 23 / 2 , c s l 3 = − 8. (269) c_{W_3}=-23,\qquad c_{\mathrm{BP}}=-23/2,\qquad c_{\mathfrak{sl}_3}=-8. \tag*{(269)} c W 3 = − 23 , c BP = − 23/2 , c sl 3 = − 8. (269)
Note that the simple relation c W 3 = 2 c B P c_{W_3}=2c_{\mathrm{BP}} c W 3 = 2 c BP implies that (263) can be further promoted to an identity for the full characters with the factor of q − c / 24 q^{-c/24} q − c /24 : χ B P ( q 2 , 1 ) = χ W 3 ( q ) \chi^{\mathrm{BP}}(q^2,1)=\chi^{W_3}(q) χ BP ( q 2 , 1 ) = χ W 3 ( q ) . From (264) and using (12) , we find
C s l 3 = c s l 3 − 12 k a f f ( ξ , ξ ) = − 8 + 9 4 = − 23 4 , C B P = c B P = − 23 2 . (270) \mathcal{C}_{\mathfrak{sl}_3}=c_{\mathfrak{sl}_3}-12k_{\mathrm{aff}}(\xi,\xi)=-8+\frac{9}{4}=-\frac{23}{4},\qquad\mathcal{C}_{\mathrm{BP}}=c_{\mathrm{BP}}=-\frac{23}{2}. \tag*{(270)} C sl 3 = c sl 3 − 12 k aff ( ξ , ξ ) = − 8 + 4 9 = − 4 23 , C BP = c BP = − 2 23 . (270)
where we have used k a f f = − 3 2 k_{\mathrm{aff}}=-\frac{3}{2} k aff = − 2 3 and the following embedding of U ( 1 ) U(1) U ( 1 ) into s l 3 \mathfrak{sl}_3 sl 3
ξ = 1 2 × 1 2 ( e 1 − e 2 ) , ( ξ , ξ ) = 1 8 . (271) \xi=\frac{1}{2}\times\frac{1}{2}(e_1-e_2),\qquad(\xi,\xi)=\frac{1}{8}. \tag*{(271)} ξ = 2 1 × 2 1 ( e 1 − e 2 ) , ( ξ , ξ ) = 8 1 . (271)
Therefore we have the simple relation
c B P = 2 c s l 3 , (272) c_{\mathrm{BP}}=2c_{\mathfrak{sl}_3}, \tag*{(272)} c BP = 2 c sl 3 , (272)
in agreement with (30) .
The effective central charges are
c W 3 e f f = 1 , c B P e f f = 2 , c s l 3 e f f = 4 , (273) c^{\mathrm{eff}}_{W_3}=1,\qquad c^{\mathrm{eff}}_{\mathrm{BP}}=2,\qquad c^{\mathrm{eff}}_{\mathfrak{sl}_3}=4, \tag*{(273)} c W 3 eff = 1 , c BP eff = 2 , c sl 3 eff = 4 , (273)
which can be obtained from the 4d central charges using (13) . They are consistent with (30) , see (7.1) and (7.2).
sl3 affine modules
The simple affine vertex algebra L − 3 / 2 ( s l 3 ) L_{-3/2}(\mathfrak{sl}_3) L − 3/2 ( sl 3 ) has four admissible irreducible highest-weight modules. Their finite s l 3 \mathfrak{sl}_3 sl 3 highest weights are
0 , − 3 2 ω 1 , − 3 2 ω 2 , − 1 2 ρ , (274) 0,\qquad-\frac{3}{2}\omega_1,\qquad-\frac{3}{2}\omega_2,\qquad-\frac{1}{2}\rho, \tag*{(274)} 0 , − 2 3 ω 1 , − 2 3 ω 2 , − 2 1 ρ , (274)
where ω 1 , ω 2 \omega_1,\omega_2 ω 1 , ω 2 are the fundamental weights of s l 3 \mathfrak{sl}_3 sl 3 and ρ = ω 1 + ω 2 \rho=\omega_1+\omega_2 ρ = ω 1 + ω 2 . The conformal weights are 0 , − 1 2 , − 1 2 , − 1 2 0,-\frac{1}{2},-\frac{1}{2},-\frac{1}{2} 0 , − 2 1 , − 2 1 , − 2 1 , respectively.
It is convenient to introduce four vectors as follows:
K = { ( 0 , 0 , 0 ) , ( 1 , 0 , − 1 ) , ( 0 , 1 , − 1 ) , ( 1 , 1 , 0 ) } . (275) \mathcal{K}=\{(0,0,0),(1,0,-1),(0,1,-1),(1,1,0)\}. \tag*{(275)} K = {( 0 , 0 , 0 ) , ( 1 , 0 , − 1 ) , ( 0 , 1 , − 1 ) , ( 1 , 1 , 0 )} . (275)
Then their corresponding highest weights and conformal weights are given by
λ v = − 1 2 ∑ i = 1 3 v i α i , h v s l 3 = − 1 4 ∑ i v i 2 , (276) \lambda_{\boldsymbol{v}}=-\frac{1}{2}\sum_{i=1}^{3}v_i\alpha_i,\qquad h^{\mathfrak{sl}_3}_{\boldsymbol{v}}=-\frac{1}{4}\sum_i v_i^2, \tag*{(276)} λ v = − 2 1 i = 1 ∑ 3 v i α i , h v sl 3 = − 4 1 i ∑ v i 2 , (276)
for any v ∈ K \boldsymbol{v}\in\mathcal{K} v ∈ K .
The affine characters of the four modules are given by
χ ^ v s l 3 ( q , x ) = q 1 3 − 1 4 ∑ i v i 2 ∏ i = 1 3 x i − v i / 2 PE [ P v s l 3 ( q , x ) 1 − q 2 ] . (277) \widehat{\chi}^{\mathfrak{sl}_3}_{\boldsymbol{v}}(q,x)=q^{\frac{1}{3}-\frac{1}{4}\sum_i v_i^2}\prod_{i=1}^{3}x_i^{-v_i/2}\operatorname{PE}\left[\frac{P^{\mathfrak{sl}_3}_{\boldsymbol{v}}(q,x)}{1-q^2}\right]. \tag*{(277)} χ v sl 3 ( q , x ) = q 3 1 − 4 1 ∑ i v i 2 i = 1 ∏ 3 x i − v i /2 PE [ 1 − q 2 P v sl 3 ( q , x ) ] . (277)
where x 1 x 2 x 3 = 1 x_1x_2x_3=1 x 1 x 2 x 3 = 1 and
P v s l 3 ( q , x ) = 2 q + ∑ i = 1 3 ( q 1 + v i x i + q 1 − v i x i − 1 ) . (278) P^{\mathfrak{sl}_3}_v(q,\mathbf{x})=2q+\sum_{i=1}^{3}\left(q^{1+v_i}x_i+q^{1-v_i}x_i^{-1}\right). \tag*{(278)} P v sl 3 ( q , x ) = 2 q + i = 1 ∑ 3 ( q 1 + v i x i + q 1 − v i x i − 1 ) . (278)
These characters can be derived from the general boundary admissible formulae [5 ] . The unrefined vacuum character also satisfies an MLDE, which is given by [10 ]
L 2 s l 3 χ ^ ( 0 , 0 , 0 ) ( q , 1 ) = 0 , L 2 s l 3 = D q ( 2 ) − 40 E 4 . (279) \mathcal{L}^{\mathfrak{sl}_3}_2\widehat{\chi}_{(0,0,0)}(q,1)=0,\qquad\mathcal{L}^{\mathfrak{sl}_3}_2=D_q^{(2)}-40E_4. \tag*{(279)} L 2 sl 3 χ ( 0 , 0 , 0 ) ( q , 1 ) = 0 , L 2 sl 3 = D q ( 2 ) − 40 E 4 . (279)
The relevant conventions are collected in Appendix § 12 .
We can also find other solutions to the MLDE. In particular, we can make the ansatz χ ^ = q α ( 1 + a 1 q + ⋯ ) \widehat{\chi}=q^\alpha(1+a_1q+\cdots) χ = q α ( 1 + a 1 q + ⋯ ) and substitute it into L 2 s l 3 χ ^ = 0 \mathcal{L}^{\mathfrak{sl}_3}_2\widehat{\chi}=0 L 2 sl 3 χ = 0 , whose leading-order term gives the indicial equation
( α − 1 3 ) ( α + 1 6 ) = 0. (280) \left(\alpha-\frac{1}{3}\right)\left(\alpha+\frac{1}{6}\right)=0. \tag*{(280)} ( α − 3 1 ) ( α + 6 1 ) = 0. (280)
The roots of the indicial equation above are { 1 3 , − 1 6 } \left\{\frac{1}{3},-\frac{1}{6}\right\} { 3 1 , − 6 1 } , which correspond to conformal weights
h s l 3 = { 0 , − 1 2 } , (281) h_{\mathfrak{sl}_3}=\left\{0,-\frac{1}{2}\right\}, \tag*{(281)} h sl 3 = { 0 , − 2 1 } , (281)
where we used the fact that h = α + c / 24 h=\alpha+c/24 h = α + c /24 .
It turns out the non-vacuum solution is given by
χ ^ n o n - v a c u u m = − 8 lim x → 1 ( χ ^ ( 1 , 0 , − 1 ) ( q , x ) + χ ^ ( 0 , 1 , − 1 ) ( q , x ) − χ ^ ( 1 , 1 , 0 ) ( q , x ) ) (282) \widehat{\chi}_{\mathrm{non\text{-}vacuum}}=-8\lim_{x\to1}\left(\widehat{\chi}_{(1,0,-1)}(q,\mathbf{x})+\widehat{\chi}_{(0,1,-1)}(q,\mathbf{x})-\widehat{\chi}_{(1,1,0)}(q,\mathbf{x})\right) \tag*{(282)} χ non - vacuum = − 8 x → 1 lim ( χ ( 1 , 0 , − 1 ) ( q , x ) + χ ( 0 , 1 , − 1 ) ( q , x ) − χ ( 1 , 1 , 0 ) ( q , x ) ) (282)
= q − 1 6 ( 1 + 28 q + 134 q 2 + 568 q 3 + 1809 q 4 + 5316 q 5 + 13990 q 6 + O ( q 7 ) ) . (283) =q^{-\frac{1}{6}}\left(1+28q+134q^2+568q^3+1809q^4+5316q^5+13990q^6+O(q^7)\right). \tag*{(283)} = q − 6 1 ( 1 + 28 q + 134 q 2 + 568 q 3 + 1809 q 4 + 5316 q 5 + 13990 q 6 + O ( q 7 ) ) . (283)
Note that in the unflavored limit, the vacuum character is finite, but the three non-vacuum characters are actually divergent. Nevertheless, taking the special linear combination above also gives a finite series in the unflavored limit.
BP modules
The module structures of BP algebra have been studied in detail in [33 ] . Here we restrict to the ordinary untwisted, or Neveu–Schwarz (NS), highest-weight modules.32 There are six simple ordinary untwisted modules which can be labelled by
( a , b ) , a , b ∈ Z ≥ 0 , a + b ≤ 2. (284) (a,b),\qquad a,b\in\mathbb{Z}_{\geq0},\qquad a+b\leq2. \tag*{(284)} ( a , b ) , a , b ∈ Z ≥ 0 , a + b ≤ 2. (284)
Their highest-weight eigenvalues are
j ( a , b ) B P = b − a 4 , h ( a , b ) B P = 2 ( a 2 + a b + b 2 ) − 5 ( a + b ) 8 . (285) j^{\mathrm{BP}}_{(a,b)}=\frac{b-a}{4},\qquad h^{\mathrm{BP}}_{(a,b)}=\frac{2(a^2+ab+b^2)-5(a+b)}{8}. \tag*{(285)} j ( a , b ) BP = 4 b − a , h ( a , b ) BP = 8 2 ( a 2 + ab + b 2 ) − 5 ( a + b ) . (285)
Thus their ( J 0 , L 0 ) (J_0,L_0) ( J 0 , L 0 ) highest weights are
( 0 , 0 ) , ( ± 1 2 , − 1 4 ) , ( 0 , − 1 2 ) , ( ± 1 4 , − 3 8 ) . (286) (0,0),\qquad\left(\pm\frac{1}{2},-\frac{1}{4}\right),\qquad\left(0,-\frac{1}{2}\right),\qquad\left(\pm\frac{1}{4},-\frac{3}{8}\right). \tag*{(286)} ( 0 , 0 ) , ( ± 2 1 , − 4 1 ) , ( 0 , − 2 1 ) , ( ± 4 1 , − 8 3 ) . (286)
The corresponding characters can again be derived from [5 ]
χ ^ ( a , b ) B P ( q , w ) = q h ( a , b ) B P + 23 / 48 w ( b − a ) / 4 PE [ P ( a , b ) B P ( q , w ) 1 − q 4 ] , (287) \widehat{\chi}^{\mathrm{BP}}_{(a,b)}(q,w)=q^{h^{\mathrm{BP}}_{(a,b)}+23/48}w^{(b-a)/4}\operatorname{PE}\left[\frac{P^{\mathrm{BP}}_{(a,b)}(q,w)}{1-q^4}\right], \tag*{(287)} χ ( a , b ) BP ( q , w ) = q h ( a , b ) BP + 23/48 w ( b − a ) /4 PE [ 1 − q 4 P ( a , b ) BP ( q , w ) ] , (287)
where
P a , b B P ( q , w ) = 2 ( q + q 2 + q 3 ) − q a + b + 1 − q 3 − a − b + ( w + w − 1 ) ( q 1 / 2 + q 3 / 2 + q 5 / 2 + q 7 / 2 ) − w − 1 q a + 1 / 2 − w q 7 / 2 − a − w q b + 1 / 2 − w − 1 q 7 / 2 − b . (288) P^{\mathrm{BP}}_{a,b}(q,w)=2(q+q^2+q^3)-q^{a+b+1}-q^{3-a-b}+(w+w^{-1})(q^{1/2}+q^{3/2}+q^{5/2}+q^{7/2})-w^{-1}q^{a+1/2}-wq^{7/2-a}-wq^{b+1/2}-w^{-1}q^{7/2-b}. \tag*{(288)} P a , b BP ( q , w ) = 2 ( q + q 2 + q 3 ) − q a + b + 1 − q 3 − a − b + ( w + w − 1 ) ( q 1/2 + q 3/2 + q 5/2 + q 7/2 ) − w − 1 q a + 1/2 − w q 7/2 − a − w q b + 1/2 − w − 1 q 7/2 − b . (288)
Independently, we can study their unrefined characters and modules using the MLDE. In this case, the character is a series in q 1 / 2 q^{1/2} q 1/2 , and it is described by a fourth-order Γ 0 ( 2 ) \Gamma^0(2) Γ 0 ( 2 ) differential operator given in [10 ]
L 4 B P = D q ( 4 ) + 5 24 X D q ( 3 ) − 25 E 4 D q ( 2 ) + ( − 115 18432 X 3 + 625 96 X E 4 ) D q ( 1 ) − 3065 5308416 X 4 + 1325 1536 X 2 E 4 − 7125 16 E 4 2 , (289) \mathcal{L}^{\mathrm{BP}}_4=D_q^{(4)}+\frac{5}{24}\mathsf{X}D_q^{(3)}-25E_4D_q^{(2)}+\left(-\frac{115}{18432}\mathsf{X}^3+\frac{625}{96}\mathsf{X}E_4\right)D_q^{(1)}-\frac{3065}{5308416}\mathsf{X}^4+\frac{1325}{1536}\mathsf{X}^2E_4-\frac{7125}{16}E_4^2, \tag*{(289)} L 4 BP = D q ( 4 ) + 24 5 X D q ( 3 ) − 25 E 4 D q ( 2 ) + ( − 18432 115 X 3 + 96 625 X E 4 ) D q ( 1 ) − 5308416 3065 X 4 + 1536 1325 X 2 E 4 − 16 7125 E 4 2 , (289)
which satisfies
L 4 B P χ ^ ( 0 , 0 ) B P ( q , 1 ) = 0. (290) \mathcal{L}^{\mathrm{BP}}_4\widehat{\chi}^{\mathrm{BP}}_{(0,0)}(q,1)=0. \tag*{(290)} L 4 BP χ ( 0 , 0 ) BP ( q , 1 ) = 0. (290)
The roots of the indicial equation are given by
{ 23 48 , 11 48 , 5 48 , − 1 48 } , (291) \left\{\frac{23}{48},\frac{11}{48},\frac{5}{48},-\frac{1}{48}\right\}, \tag*{(291)} { 48 23 , 48 11 , 48 5 , − 48 1 } , (291)
corresponding to conformal weights
h B P = { 0 , − 1 4 , − 3 8 , − 1 2 } . (292) h_{\mathrm{BP}}=\left\{0,-\frac{1}{4},-\frac{3}{8},-\frac{1}{2}\right\}. \tag*{(292)} h BP = { 0 , − 4 1 , − 8 3 , − 2 1 } . (292)
This is consistent with (286) . Note that in the unrefined limit w = 1 w=1 w = 1 , the character formulae in (287) are invariant under the exchange of a , b a,b a , b , which flips the sign of j j j but keeps h h h . Therefore the six ordinary modules only give four distinct unrefined characters above. Keeping the flavor fugacity distinguishes the charge-conjugate pairs, which could be detected by flavored MLDEs.
W3(3, 8) modules
The modules of the W 3 ( 3 , 8 ) W_3(3,8) W 3 ( 3 , 8 ) minimal model are labeled by
( r , s ) , ( r , s ) ∼ ( 5 − r − s , r ) ∼ ( s , 5 − r − s ) , r , s ∈ Z ≥ 0 , r + s ≤ 5 , (293) (r,s),\qquad(r,s)\sim(5-r-s,r)\sim(s,5-r-s),\qquad r,s\in\mathbb{Z}_{\geq0},\qquad r+s\leq5, \tag*{(293)} ( r , s ) , ( r , s ) ∼ ( 5 − r − s , r ) ∼ ( s , 5 − r − s ) , r , s ∈ Z ≥ 0 , r + s ≤ 5 , (293)
modulo cyclic permutation ( r , s , 5 − r − s ) (r,s,5-r-s) ( r , s , 5 − r − s ) . In total, there are seven inequivalent classes, which will be chosen as ( 0 , 0 ) , ( 0 , 1 ) , ( 1 , 0 ) , ( 0 , 2 ) , ( 2 , 0 ) , ( 1 , 1 ) , ( 1 , 2 ) (0,0),(0,1),(1,0),(0,2),(2,0),(1,1),(1,2) ( 0 , 0 ) , ( 0 , 1 ) , ( 1 , 0 ) , ( 0 , 2 ) , ( 2 , 0 ) , ( 1 , 1 ) , ( 1 , 2 ) . Their conformal weights are
h ( r , s ) W 3 = r 2 + r s + s 2 − 5 r − 5 s 8 . (294) h^{W_3}_{(r,s)}=\frac{r^2+rs+s^2-5r-5s}{8}. \tag*{(294)} h ( r , s ) W 3 = 8 r 2 + r s + s 2 − 5 r − 5 s . (294)
The characters can be obtained by principal Drinfel’d–Sokolov reduction of admissible affine characters [4 , 5 ] .
χ ^ ( r , s ) W 3 ( q ) = q h ( r , s ) W 3 + 23 / 24 PE [ P ( r , s ) W 3 ( q ) 1 − q 8 ] , (295) \widehat{\chi}^{W_3}_{(r,s)}(q)=q^{h^{W_3}_{(r,s)}+23/24}\operatorname{PE}\left[\frac{P^{W_3}_{(r,s)}(q)}{1-q^8}\right], \tag*{(295)} χ ( r , s ) W 3 ( q ) = q h ( r , s ) W 3 + 23/24 PE [ 1 − q 8 P ( r , s ) W 3 ( q ) ] , (295)
where
P ( r , s ) W 3 ( q ) = 2 ∑ n = 1 7 q n − ∑ d ∈ { r + 1 , s + 1 , r + s + 2 } ( q d + q 8 − d ) . (296) P^{W_3}_{(r,s)}(q)=2\sum_{n=1}^{7}q^n-\sum_{d\in\{r+1,s+1,r+s+2\}}\left(q^d+q^{8-d}\right). \tag*{(296)} P ( r , s ) W 3 ( q ) = 2 n = 1 ∑ 7 q n − d ∈ { r + 1 , s + 1 , r + s + 2 } ∑ ( q d + q 8 − d ) . (296)
Note that { r + 1 , s + 1 , r + s + 2 } \{r+1,s+1,r+s+2\} { r + 1 , s + 1 , r + s + 2 } is counted with multiplicity.
The modules can also be understood using the MLDE [8 ] :
L 5 W 3 = D q ( 5 ) − 125 E 4 D q ( 3 ) − 6825 8 E 6 D q ( 2 ) − 24125 16 E 4 2 D q ( 1 ) − 221375 32 E 4 E 6 , (297) \mathcal{L}^{W_3}_5=D_q^{(5)}-125E_4D_q^{(3)}-\frac{6825}{8}E_6D_q^{(2)}-\frac{24125}{16}E_4^2D_q^{(1)}-\frac{221375}{32}E_4E_6, \tag*{(297)} L 5 W 3 = D q ( 5 ) − 125 E 4 D q ( 3 ) − 8 6825 E 6 D q ( 2 ) − 16 24125 E 4 2 D q ( 1 ) − 32 221375 E 4 E 6 , (297)
which annihilates the vacuum character χ ^ ( 0 , 0 ) W 3 ( q ) \widehat{\chi}^{W_3}_{(0,0)}(q) χ ( 0 , 0 ) W 3 ( q ) . The roots of the indicial equation are
{ 23 24 , 11 24 , 5 24 , − 1 24 , 1 12 } , (298) \left\{\frac{23}{24},\frac{11}{24},\frac{5}{24},-\frac{1}{24},\frac{1}{12}\right\}, \tag*{(298)} { 24 23 , 24 11 , 24 5 , − 24 1 , 12 1 } , (298)
corresponding to conformal weights
h W 3 = { 0 , − 1 2 , − 3 4 , − 1 , − 7 8 } . (299) h_{W_3}=\left\{0,-\frac{1}{2},-\frac{3}{4},-1,-\frac{7}{8}\right\}. \tag*{(299)} h W 3 = { 0 , − 2 1 , − 4 3 , − 1 , − 8 7 } . (299)
The seven modules only give rise to five distinct characters, as ( 0 , s ) (0,s) ( 0 , s ) and ( s , 0 ) (s,0) ( s , 0 ) are different modules for s = 1 , 2 s=1,2 s = 1 , 2 , but share the same character.
We see that except for the last entry, the module conformal weights in (299) are twice those in (292) .
Module character identities
The first relation between module characters is:
h ( 2 a , 2 b ) W 3 = 2 h ( a , b ) B P , χ ^ ( 2 a , 2 b ) W 3 ( q ) = χ ^ a , b B P ( q 2 , 1 ) , a , b ∈ Z ≥ 0 , a + b ≤ 2 , (300) h^{W_3}_{(2a,2b)}=2h^{\mathrm{BP}}_{(a,b)},\qquad\widehat{\chi}^{W_3}_{(2a,2b)}(q)=\widehat{\chi}^{\mathrm{BP}}_{a,b}(q^2,1),\qquad a,b\in\mathbb{Z}_{\geq0},\qquad a+b\leq2, \tag*{(300)} h ( 2 a , 2 b ) W 3 = 2 h ( a , b ) BP , χ ( 2 a , 2 b ) W 3 ( q ) = χ a , b BP ( q 2 , 1 ) , a , b ∈ Z ≥ 0 , a + b ≤ 2 , (300)
which is the generalization of the vacuum character identity in (263) .
To generalize (264) , we can apply the same fugacity map q → q 2 q\to q^2 q → q 2 , x 1 = w q − 1 / 2 x_1=wq^{-1/2} x 1 = w q − 1/2 , x 2 = w − 1 q − 1 / 2 x_2=w^{-1}q^{-1/2} x 2 = w − 1 q − 1/2 , x 3 = q x_3=q x 3 = q . Then the affine character becomes
q − 3 / 16 χ ^ ν s l ^ 3 ( q 2 ; w q − 1 / 2 , w − 1 q − 1 / 2 , q ) = q 23 / 48 − ( v 1 + v 2 ) / 4 w ( v 2 − v 1 ) / 2 PE [ P ~ ν s l ^ 3 ( q , w ) 1 − q 4 ] , (301) q^{-3/16}\widehat{\chi}^{\widehat{\mathfrak{sl}}_3}_{\nu}\left(q^2;wq^{-1/2},w^{-1}q^{-1/2},q\right)=q^{23/48-(v_1+v_2)/4}w^{(v_2-v_1)/2}\operatorname{PE}\left[\frac{\widetilde{P}^{\widehat{\mathfrak{sl}}_3}_{\nu}(q,w)}{1-q^4}\right], \tag*{(301)} q − 3/16 χ ν sl 3 ( q 2 ; w q − 1/2 , w − 1 q − 1/2 , q ) = q 23/48 − ( v 1 + v 2 ) /4 w ( v 2 − v 1 ) /2 PE [ 1 − q 4 P ν sl 3 ( q , w ) ] , (301)
where
P ~ 0 s l 3 ( q , w ) = q + 2 q 2 + q 3 + w ( q 3 / 2 + 2 v 1 + q 5 / 2 − 2 v 2 ) + w − 1 ( q 5 / 2 − 2 v 1 + q 3 / 2 + 2 v 2 ) . (302) \widetilde{P}^{\mathfrak{sl}_3}_0(q,w)=q+2q^2+q^3+w\left(q^{3/2+2v_1}+q^{5/2-2v_2}\right)+w^{-1}\left(q^{5/2-2v_1}+q^{3/2+2v_2}\right). \tag*{(302)} P 0 sl 3 ( q , w ) = q + 2 q 2 + q 3 + w ( q 3/2 + 2 v 1 + q 5/2 − 2 v 2 ) + w − 1 ( q 5/2 − 2 v 1 + q 3/2 + 2 v 2 ) . (302)
One can explicitly check that
h a , b B P = 2 h λ s l 3 − 1 2 ( λ , ρ ) = 2 h λ s l 3 + v 1 + v 2 − 2 v 3 4 , P ~ ( v 1 , v 2 , v 3 ) s l 3 ( q , w ) = P ( a , b ) B P ( q , w ) , (303) h^{\mathrm{BP}}_{a,b}=2h^{\mathfrak{sl}_3}_{\lambda}-\frac{1}{2}(\lambda,\rho)=2h^{\mathfrak{sl}_3}_{\lambda}+\frac{v_1+v_2-2v_3}{4},\qquad\widetilde{P}^{\mathfrak{sl}_3}_{(v_1,v_2,v_3)}(q,w)=P^{\mathrm{BP}}_{(a,b)}(q,w), \tag*{(303)} h a , b BP = 2 h λ sl 3 − 2 1 ( λ , ρ ) = 2 h λ sl 3 + 4 v 1 + v 2 − 2 v 3 , P ( v 1 , v 2 , v 3 ) sl 3 ( q , w ) = P ( a , b ) BP ( q , w ) , (303)
if we identify ( a , b ) = ( v 1 ( 2 − v 2 ) , v 2 ( 2 − v 1 ) ) (a,b)=(v_1(2-v_2),v_2(2-v_1)) ( a , b ) = ( v 1 ( 2 − v 2 ) , v 2 ( 2 − v 1 )) for v ∈ K v\in\mathcal{K} v ∈ K . Therefore we have the character identity
q − 3 / 16 χ ^ V s l 3 ( q 2 ; w q − 1 / 2 , w − 1 q − 1 / 2 , q ) = χ ^ a , b B P ( q , w ) , (304) q^{-3/16}\widehat{\chi}^{\mathfrak{sl}_3}_{\mathcal{V}}\left(q^2;wq^{-1/2},w^{-1}q^{-1/2},q\right)=\widehat{\chi}^{\mathrm{BP}}_{a,b}(q,w), \tag*{(304)} q − 3/16 χ V sl 3 ( q 2 ; w q − 1/2 , w − 1 q − 1/2 , q ) = χ a , b BP ( q , w ) , (304)
where the non-trivial prefactor is q ( 2 c s l 3 − c B P ) / 24 q^{(2c_{\mathfrak{sl}_3}-c_{\mathrm{BP}})/24} q ( 2 c sl 3 − c BP ) /24 , with the central charges given in (269) .
We collect the relevant relations in Table 2 , which gives the character correspondences within the selected classes: four affine admissible highest-weight modules, six ordinary BP NS modules, and seven simple W 3 ( 3 , 8 ) W_3(3,8) W 3 ( 3 , 8 ) modules. The ( 1 , 1 ) (1,1) ( 1 , 1 ) module of W 3 W_3 W 3 algebra, with conformal weight h W = − 7 / 8 h_W=-7/8 h W = − 7/8 , has no counterpart among the modules paired by the two fixed regradings in Table 2 . Nevertheless, we find the following identity
Table 2. Modules and conformal weights for L − 3 / 2 ( s l 3 ) L_{-3/2}(\mathfrak{sl}_3) L − 3/2 ( sl 3 ) , B P − 9 / 4 \mathrm{BP}_{-9/4} BP − 9/4 , and W 3 ( 3 , 8 ) W_3(3,8) W 3 ( 3 , 8 ) . Non-empty entries in the same row satisfy the character identities, including the indicated regrading and affine prefactor.
χ ^ ( 1 , 1 ) W 3 ( q ) = χ ^ v a c s l 3 ( q 1 / 4 ; i , i , − 1 ) , (305) \widehat{\chi}^{W_3}_{(1,1)}(q)=\widehat{\chi}^{\mathfrak{sl}_3}_{\mathrm{vac}}(q^{1/4};i,i,-1), \tag*{(305)} χ ( 1 , 1 ) W 3 ( q ) = χ vac sl 3 ( q 1/4 ; i , i , − 1 ) , (305)
where the vacuum character of the affine algebra χ ^ v a c s l 3 ( q ; x 1 , x 2 , x 3 ) \widehat{\chi}^{\mathfrak{sl}_3}_{\mathrm{vac}}(q;x_1,x_2,x_3) χ vac sl 3 ( q ; x 1 , x 2 , x 3 ) is given in (277) subject to x 1 x 2 x 3 = 1 x_1x_2x_3=1 x 1 x 2 x 3 = 1 . The identity (305) is very different from other character identities in the rest of the paper. It would be interesting to understand whether it is a coincidence or points towards another relation between VOAs. Nevertheless, we find that it can be generalized to other W 3 W_3 W 3 modules:
χ ^ ( 0 , 0 ) W 3 ( q ) = q − 3 / 8 χ ^ v a c s l 3 ( q 4 ; q − 1 , q − 1 , q 2 ) , (306) \widehat{\chi}^{W_3}_{(0,0)}(q)=q^{-3/8}\widehat{\chi}^{\mathfrak{sl}_3}_{\mathrm{vac}}(q^4;q^{-1},q^{-1},q^2), \tag*{(306)} χ ( 0 , 0 ) W 3 ( q ) = q − 3/8 χ vac sl 3 ( q 4 ; q − 1 , q − 1 , q 2 ) , (306)
χ ^ ( 0 , 1 ) W 3 ( q ) = χ ^ ( 1 , 0 ) W 3 ( q ) = q − 7 / 8 χ ^ v a c s l 3 ( q 4 ; q , q 2 , q − 3 ) , (307) \widehat{\chi}^{W_3}_{(0,1)}(q)=\widehat{\chi}^{W_3}_{(1,0)}(q)=q^{-7/8}\widehat{\chi}^{\mathfrak{sl}_3}_{\mathrm{vac}}(q^4;q,q^2,q^{-3}), \tag*{(307)} χ ( 0 , 1 ) W 3 ( q ) = χ ( 1 , 0 ) W 3 ( q ) = q − 7/8 χ vac sl 3 ( q 4 ; q , q 2 , q − 3 ) , (307)
χ ^ ( 2 , 0 ) W 3 ( q ) = χ ^ ( 0 , 2 ) W 3 ( q ) = q − 1 / 8 χ ^ v a c s l 3 ( q ; − 1 , − q − 1 / 2 , q 1 / 2 ) , (308) \widehat{\chi}^{W_3}_{(2,0)}(q)=\widehat{\chi}^{W_3}_{(0,2)}(q)=q^{-1/8}\widehat{\chi}^{\mathfrak{sl}_3}_{\mathrm{vac}}(q;-1,-q^{-1/2},q^{1/2}), \tag*{(308)} χ ( 2 , 0 ) W 3 ( q ) = χ ( 0 , 2 ) W 3 ( q ) = q − 1/8 χ vac sl 3 ( q ; − 1 , − q − 1/2 , q 1/2 ) , (308)
χ ^ ( 2 , 1 ) W 3 ( q ) = χ ^ ( 2 , 2 ) W 3 ( q ) = − q − 19 / 8 χ ^ v a c s l 3 ( q 4 ; q − 5 , q 3 , q 2 ) . (309) \widehat{\chi}^{W_3}_{(2,1)}(q)=\widehat{\chi}^{W_3}_{(2,2)}(q)=-q^{-19/8}\widehat{\chi}^{\mathfrak{sl}_3}_{\mathrm{vac}}(q^4;q^{-5},q^3,q^2). \tag*{(309)} χ ( 2 , 1 ) W 3 ( q ) = χ ( 2 , 2 ) W 3 ( q ) = − q − 19/8 χ vac sl 3 ( q 4 ; q − 5 , q 3 , q 2 ) . (309)
These identities relate the modules of W 3 W_3 W 3 algebra to the vacuum module of the affine vertex algebra. Note that the last one has an overall sign and its fugacity specialization requires meromorphic continuation.
The simple relation in (300) also allows us to establish a connection between the two modular linear differential operators (7.35 and (289) .
For this purpose, we introduce the scaling operator R 1 / 2 R_{1/2} R 1/2 which acts as
( R 1 / 2 f ) ( q ) = f ( q 1 / 2 ) . (310) (R_{1/2}f)(q)=f(q^{1/2}). \tag*{(310)} ( R 1/2 f ) ( q ) = f ( q 1/2 ) . (310)
The covariant derivative transforms as
1 2 R 1 / 2 ∂ ( k ) R 1 / 2 − 1 = ∂ ( k ) + k X 24 . (311) \frac{1}{2}R_{1/2}\partial_{(k)}R_{1/2}^{-1}=\partial_{(k)}+\frac{kX}{24}. \tag*{(311)} 2 1 R 1/2 ∂ ( k ) R 1/2 − 1 = ∂ ( k ) + 24 k X . (311)
This can be checked using (B.4) - (B.7):
1 2 ( ∂ ( k ) f ) ( q 1 / 2 ) = q ∂ q [ f ( q 1 / 2 ) ] + k 2 E 2 ( q 1 / 2 ) f ( q 1 / 2 ) (312) \frac{1}{2}(\partial_{(k)}f)(q^{1/2})=q\partial_q[f(q^{1/2})]+\frac{k}{2}E_2(q^{1/2})f(q^{1/2}) \tag*{(312)} 2 1 ( ∂ ( k ) f ) ( q 1/2 ) = q ∂ q [ f ( q 1/2 )] + 2 k E 2 ( q 1/2 ) f ( q 1/2 ) (312)
= q ∂ q [ f ( q 1 / 2 ) ] + k E 2 ( q ) f ( q 1 / 2 ) + k 24 X ( q ) f ( q 1 / 2 ) . (313) =q\partial_q[f(q^{1/2})]+kE_2(q)f(q^{1/2})+\frac{k}{24}X(q)f(q^{1/2}). \tag*{(313)} = q ∂ q [ f ( q 1/2 )] + k E 2 ( q ) f ( q 1/2 ) + 24 k X ( q ) f ( q 1/2 ) . (313)
We can then define
L ^ 5 W 3 = 2 − 5 R 1 / 2 L 5 W 3 R 1 / 2 − 1 . (314) \widehat{\mathcal{L}}^{W_3}_5=2^{-5}R_{1/2}\mathcal{L}^{W_3}_5R_{1/2}^{-1}. \tag*{(314)} L 5 W 3 = 2 − 5 R 1/2 L 5 W 3 R 1/2 − 1 . (314)
This operator can be rewritten as
L ^ 5 W 3 = ( ∂ ( 8 ) + 5 8 X ) ∘ L 4 B P , (315) \widehat{\mathcal{L}}^{W_3}_5=\left(\partial_{(8)}+\frac{5}{8}X\right)\circ\mathcal{L}^{\mathrm{BP}}_4, \tag*{(315)} L 5 W 3 = ( ∂ ( 8 ) + 8 5 X ) ∘ L 4 BP , (315)
by using (B.4) - (B.7) and replacing ∂ ( k ) \partial_{(k)} ∂ ( k ) with ∂ ‾ ( k ) = ∂ ( k ) + k X / 24 \overline{\partial}_{(k)}=\partial_{(k)}+kX/24 ∂ ( k ) = ∂ ( k ) + k X /24 . The factorization in (315) gives an inclusion of solution spaces after rescaling, and thus provides a natural explanation for (300) .
In section 6.2, we established the GVISO between s l 3 \mathfrak{sl}_3 sl 3 and BP in the vacuum module. It is then natural to ask whether this construction can be generalized to non-vacuum modules. This requires a deeper understanding of the modules and is beyond the scope of this paper. Nevertheless, we believe the GVISO construction in the vacuum case should be generalizable to a special class of modules, which are obtained from the vacuum module by spectral flow.
Note that the two VOAs have U(1) symmetries which can be used to generate spectral flows. We leave a detailed study to the future.
General conjecture
The concrete example above suggests that the non-vacuum module character identity is a general feature. We therefore conjecture the following character identity between an affine vertex algebra and a principal W W W -algebra:
Tr M ( L k ( g ) , λ ) q ( h ∨ + 1 ) L 0 − J 0 ρ = Tr M ( W k ′ ( g , f p r i n ) , μ ) q L 0 (316) \operatorname{Tr}_{\mathcal{M}(L_k(\mathfrak{g}),\lambda)}q^{(h^\vee+1)L_0-J_0^\rho}=\operatorname{Tr}_{\mathcal{M}(W_{k'}(\mathfrak{g},f_{\mathrm{prin}}),\mu)}q^{L_0} \tag*{(316)} Tr M ( L k ( g ) , λ ) q ( h ∨ + 1 ) L 0 − J 0 ρ = Tr M ( W k ′ ( g , f prin ) , μ ) q L 0 (316)
for general g \mathfrak{g} g at general admissible levels subject to ( k + h ∨ ) = ( k ′ + h ∨ ) ( h + 1 ) (k+h^\vee)=(k'+h^\vee)(h+1) ( k + h ∨ ) = ( k ′ + h ∨ ) ( h + 1 ) . We expect the conjecture holds for a selected class of modules with a suitable map λ ↦ μ \lambda\mapsto\mu λ ↦ μ . This generalizes the vacuum-module identity in the boundary admissible case that we discussed extensively in the previous sections. We conjecture that similar module character identities also hold beyond the affine/principal setting.
We defer the detailed discussions to [34 ] , but present the key module character formulae in the boundary admissible case with g = s l N \mathfrak{g}=\mathfrak{sl}_N g = sl N , which are particularly relevant for 4d SCFTs and provide a starting point for proving the above conjecture in this case.
The two relevant VOAs are given by
W N ( N , U ) : = W − N + N U ( s l N , f p r i n ) , U > N + 1 , gcd ( N , U ) = 1 , (317) W_N(N,U):=W_{-N+\frac{N}{U}}(\mathfrak{sl}_N,f_{\mathrm{prin}}),\qquad U>N+1,\qquad\gcd(N,U)=1, \tag*{(317)} W N ( N , U ) := W − N + U N ( sl N , f prin ) , U > N + 1 , g cd( N , U ) = 1 , (317)
and
L − N + N u ( s l N ) , u > 1 , gcd ( N , u ) = 1. (318) L_{-N+\frac{N}{u}}(\mathfrak{sl}_N),\qquad u>1,\qquad\gcd(N,u)=1. \tag*{(318)} L − N + u N ( sl N ) , u > 1 , g cd( N , u ) = 1. (318)
For N ≥ 2 N\ge2 N ≥ 2 , the former is the W N W_N W N minimal model associated with ( A N − 1 , A U − N − 1 ) (A_{N-1},A_{U-N-1}) ( A N − 1 , A U − N − 1 ) , and the latter is the affine algebra associated with D u ( S U ( N ) ) D_u(SU(N)) D u ( S U ( N )) when u > 1 u>1 u > 1 . Note that the endpoints u = 1 u=1 u = 1 and U = N + 1 U=N+1 U = N + 1 give trivial VOAs at zero central charges.
Their central charges are
c a f f = ( 1 − u ) ( N 2 − 1 ) , c W = ( N − 1 ) − ( N 2 − 1 ) ( U − N ) 2 U . (319) c_{\mathrm{aff}}=(1-u)(N^2-1),\qquad c_W=(N-1)-\frac{(N^2-1)(U-N)^2}{U}. \tag*{(319)} c aff = ( 1 − u ) ( N 2 − 1 ) , c W = ( N − 1 ) − U ( N 2 − 1 ) ( U − N ) 2 . (319)
Their modules are specified by weights 33
λ = ∑ i = 1 N − 1 λ i ω i , μ = ∑ i = 1 N − 1 μ i ω i . (320) \lambda=\sum_{i=1}^{N-1}\lambda_i\omega_i,\qquad\mu=\sum_{i=1}^{N-1}\mu_i\omega_i. \tag*{(320)} λ = i = 1 ∑ N − 1 λ i ω i , μ = i = 1 ∑ N − 1 μ i ω i . (320)
On the W W W -algebra side, all simple ordinary modules are labelled by nonnegative tuples
( μ 0 , μ 1 , … , μ N − 1 ) , ∑ i = 0 N − 1 μ i = U − N , μ i ∈ Z ≥ 0 , (321) (\mu_0,\mu_1,\ldots,\mu_{N-1}),\qquad\sum_{i=0}^{N-1}\mu_i=U-N,\qquad\mu_i\in\mathbb{Z}_{\geq0}, \tag*{(321)} ( μ 0 , μ 1 , … , μ N − 1 ) , i = 0 ∑ N − 1 μ i = U − N , μ i ∈ Z ≥ 0 , (321)
modulo cyclic rotation. The conformal weight is
h μ W = N 2 U ( μ , μ + 2 ρ ) − ( ρ , μ ) . (322) h_{\mu}^{W}=\frac{N}{2U}(\mu,\mu+2\rho)-(\rho,\mu). \tag*{(322)} h μ W = 2 U N ( μ , μ + 2 ρ ) − ( ρ , μ ) . (322)
The character is given by
χ ^ [ μ ] W ( q ) = q h μ W − c W / 24 PE [ A μ ( q ) 1 − q U ] , (323) \widehat{\chi}_{[\mu]}^{W}(q)=q^{h_{\mu}^{W}-c_{W}/24}\operatorname{PE}\left[\frac{A_{\mu}(q)}{1-q^{U}}\right], \tag*{(323)} χ [ μ ] W ( q ) = q h μ W − c W /24 PE [ 1 − q U A μ ( q ) ] , (323)
where
A μ ( q ) = ( N − 1 ) ∑ d = 1 U − 1 q d − ∑ 1 ≤ i < j ≤ N ( q d i j + q U − d i j ) , d i j = ∑ a = i j − 1 ( μ a + 1 ) . (324) A_{\mu}(q)=(N-1)\sum_{d=1}^{U-1}q^{d}-\sum_{1\leq i<j\leq N}\left(q^{d_{ij}}+q^{U-d_{ij}}\right),\qquad d_{ij}=\sum_{a=i}^{j-1}(\mu_a+1). \tag*{(324)} A μ ( q ) = ( N − 1 ) d = 1 ∑ U − 1 q d − 1 ≤ i < j ≤ N ∑ ( q d ij + q U − d ij ) , d ij = a = i ∑ j − 1 ( μ a + 1 ) . (324)
On the affine side, the irreducible admissible highest-weight modules can be labelled by λ i \lambda_i λ i subject to
S i j : = ∑ s = i j − 1 ( λ s + 1 ) ∈ P N , u , ∀ 1 ≤ i < j ≤ N , S l l = 0 , ∀ 1 ≤ l ≤ N (325) S_{ij}:=\sum_{s=i}^{j-1}(\lambda_s+1)\in\mathcal{P}_{N,u},\qquad\forall\,1\leq i<j\leq N,\qquad S_{ll}=0,\qquad\forall\,1\leq l\leq N \tag*{(325)} S ij := s = i ∑ j − 1 ( λ s + 1 ) ∈ P N , u , ∀ 1 ≤ i < j ≤ N , S l l = 0 , ∀ 1 ≤ l ≤ N (325)
where
P N , u = { m − N n u | m , n ∈ Z , 1 ≤ m ≤ N − 1 , 0 ≤ n ≤ u − 1 } . (326) \mathcal{P}_{N,u}=\left\{m-\frac{Nn}{u}\,\middle|\,m,n\in\mathbb{Z},\quad1\leq m\leq N-1,\quad0\leq n\leq u-1\right\}. \tag*{(326)} P N , u = { m − u N n m , n ∈ Z , 1 ≤ m ≤ N − 1 , 0 ≤ n ≤ u − 1 } . (326)
Note that the affine Dynkin label is given by λ 0 = k − ∑ i = 1 N − 1 λ i \lambda_0=k-\sum_{i=1}^{N-1}\lambda_i λ 0 = k − ∑ i = 1 N − 1 λ i for k = − N + N u k=-N+\frac{N}{u} k = − N + u N . The conformal weight of the module is given by
h λ a f f = u 2 N ( λ , λ + 2 ρ ) . (327) h_{\lambda}^{\mathrm{aff}}=\frac{u}{2N}(\lambda,\lambda+2\rho). \tag*{(327)} h λ aff = 2 N u ( λ , λ + 2 ρ ) . (327)
The module character is
χ ^ λ a f f ( Q , x ) = Q h λ a f f − c a f f / 24 x λ PE [ N λ ( Q , x ) 1 − Q u ] , (328) \widehat{\chi}_{\lambda}^{\mathrm{aff}}(Q,x)=Q^{h_{\lambda}^{\mathrm{aff}}-c_{\mathrm{aff}}/24}x^{\lambda}\operatorname{PE}\left[\frac{N_{\lambda}(Q,x)}{1-Q^{u}}\right], \tag*{(328)} χ λ aff ( Q , x ) = Q h λ aff − c aff /24 x λ PE [ 1 − Q u N λ ( Q , x ) ] , (328)
where
N λ ( Q , x ) = ( N − 1 ) ( B u ( Q ) − 1 ) + ∑ 1 ≤ i < j ≤ N [ ( Q B u ( Q ) − Q u − ν i j ) x i x j + ( B u ( Q ) − Q ν i j ) x j x i ] . (329) N_{\lambda}(Q,x)=(N-1)(B_u(Q)-1)+\sum_{1\leq i<j\leq N}\left[(QB_u(Q)-Q^{u-\nu_{ij}})\frac{x_i}{x_j}+(B_u(Q)-Q^{\nu_{ij}})\frac{x_j}{x_i}\right]. \tag*{(329)} N λ ( Q , x ) = ( N − 1 ) ( B u ( Q ) − 1 ) + 1 ≤ i < j ≤ N ∑ [ ( Q B u ( Q ) − Q u − ν ij ) x j x i + ( B u ( Q ) − Q ν ij ) x i x j ] . (329)
and
ν i j = [ − a u S i j ] u , x λ = ∏ a = 1 N − 1 ( x 1 ⋯ x a ) λ a , B u ( Q ) = 1 + Q + ⋯ + Q u − 1 . (330) \nu_{ij}=[-a u S_{ij}]_{u},\qquad\mathbf{x}^{\lambda}=\prod_{a=1}^{N-1}(x_1\cdots x_a)^{\lambda_a},\qquad B_u(Q)=1+Q+\cdots+Q^{u-1}. \tag*{(330)} ν ij = [ − a u S ij ] u , x λ = a = 1 ∏ N − 1 ( x 1 ⋯ x a ) λ a , B u ( Q ) = 1 + Q + ⋯ + Q u − 1 . (330)
Here a ≡ N − 1 ( m o d u ) a\equiv N^{-1}\pmod{u} a ≡ N − 1 ( mod u ) , and [ m ] u ∈ { 0 , ⋯ , u − 1 } [m]_u\in\{0,\cdots,u-1\} [ m ] u ∈ { 0 , ⋯ , u − 1 } denotes the mod u u u reduction of m m m . Note that x 1 ⋯ x N = 1 x_1\cdots x_N=1 x 1 ⋯ x N = 1 .
The character formulae (323) and (328) can be derived from [5 ] .35 Their simple PE form is related to the boundary admissibility of the levels. One can verify that they reduce to the formulae in (78) and (77) for the vacuum module.
Using these character formulae, one can show that the following character identity holds
χ ^ [ μ ] W ( q ) = q k N ( N − 1 ) / 24 χ λ a f f ( q N + 1 ; x i = q i − ( N + 1 ) / 2 ) , (331) \widehat{\chi}^{W}_{[\mu]}(q)=q^{kN(N-1)/24}\chi^{\mathrm{aff}}_{\lambda}\left(q^{N+1};x_i=q^{i-(N+1)/2}\right), \tag*{(331)} χ [ μ ] W ( q ) = q k N ( N − 1 ) /24 χ λ aff ( q N + 1 ; x i = q i − ( N + 1 ) /2 ) , (331)
under identification U = ( N + 1 ) u U=(N+1)u U = ( N + 1 ) u and the module map: μ i = p i + 1 − p i − 1 \mu_i=p_{i+1}-p_i-1 μ i = p i + 1 − p i − 1 for i = 1 , … , N − 1 i=1,\ldots,N-1 i = 1 , … , N − 1 , where the ordered set 0 = p 1 < p 2 < ⋯ < p N < U 0=p_1<p_2<\cdots<p_N<U 0 = p 1 < p 2 < ⋯ < p N < U is just a permutation of b i : = i − 1 + ( N + 1 ) ν 1 i b_i:=i-1+(N+1)\nu_{1i} b i := i − 1 + ( N + 1 ) ν 1 i , namely p i = b σ ( i ) p_i=b_{\sigma(i)} p i = b σ ( i ) for a permutation σ ∈ S N \sigma\in S_N σ ∈ S N .
We will derive the character formulae (323) and (328) , discuss the proof of (331) , and explore further generalization to modules for other Lie algebras at general admissible levels in detail in [34 ] .
Character identities beyond SCFT/VOA
Our discussions so far have been motivated by the SCFT/VOA correspondence, where the VOAs feature quasi-lisse properties and non-unitarity. In this section, we generalize the character identity and GVISO beyond the realm of the SCFT/VOA correspondence.
Affine and Ising from Majorana
We begin by discussing this relation in a simple unitary setting involving the Ising VOA and the level-one affine VOA.
For n ≥ 2 n\geq2 n ≥ 2 ,36 we now consider the following pair of VOAs
V = L 1 ( s o 2 n + 1 ) , W = Vir ( 3 , 4 ) , (332) V=L_1(\mathfrak{so}_{2n+1}),\qquad W=\operatorname{Vir}(3,4), \tag*{(332)} V = L 1 ( so 2 n + 1 ) , W = Vir ( 3 , 4 ) , (332)
which correspond to the level one affine vertex algebra and the Ising VOA, respectively. Their central charges are
c V = 2 n + 1 2 , c W = 1 2 . (333) c_V=\frac{2n+1}{2},\qquad c_W=\frac{1}{2}. \tag*{(333)} c V = 2 2 n + 1 , c W = 2 1 . (333)
We will show that there is a GVISO between them
( L 1 ( s o 2 n + 1 ) , H ) ≅ g r ( Vir ( 3 , 4 ) , S 0 ) , (334) (L_1(\mathfrak{so}_{2n+1}),H)\cong_{\mathrm{gr}}(\operatorname{Vir}(3,4),S_0), \tag*{(334)} ( L 1 ( so 2 n + 1 ) , H ) ≅ gr ( Vir ( 3 , 4 ) , S 0 ) , (334)
where
H = ℓ L 0 V − J 0 V , ℓ = 2 n + 1. (335) H = \ell L_0^V - J_0^V,\qquad\ell= 2n + 1. \tag*{(335)} H = ℓ L 0 V − J 0 V , ℓ = 2 n + 1. (335)
The two VOAs admit free-fermion realizations
V ≅ ( F ⊗ ( 2 n + 1 ) ) 0 ‾ , W ≅ F 0 ‾ , (336) V \cong\left(F^{\otimes(2n+1)}\right)_{\overline{0}},\qquad W \cong F_{\overline{0}}, \tag*{(336)} V ≅ ( F ⊗ ( 2 n + 1 ) ) 0 , W ≅ F 0 , (336)
where F F F is the free Majorana fermion and the subscript 0 ‾ \overline{0} 0 implies that we only consider the even fermion parity sector.
On the W W W side, we choose one Majorana χ \chi χ , with modes χ b \chi_b χ b , b ∈ Z + 1 2 b \in\mathbb{Z} + \frac{1}{2} b ∈ Z + 2 1 with
{ χ b 1 , χ b 2 } = δ b 1 + b 2 , 0 . (337) \{\chi_{b_1},\chi_{b_2}\} = \delta_{b_1+b_2,0}. \tag*{(337)} { χ b 1 , χ b 2 } = δ b 1 + b 2 , 0 . (337)
On the V V V side, we choose 2 n + 1 2n+1 2 n + 1 Majorana fermions and linearly combine them such that
ψ a r , r = − n , … , n , a ∈ Z + 1 2 , (338) \psi_a^r,\qquad r=-n,\ldots,n,\qquad a\in\mathbb{Z}+\frac{1}{2}, \tag*{(338)} ψ a r , r = − n , … , n , a ∈ Z + 2 1 , (338)
with
{ ψ a 1 r , ψ a 2 s } = δ r + s , 0 δ a 1 + a 2 , 0 , [ J 0 V , ψ a r ] = r ψ a r , (339) \{\psi_{a_1}^r,\psi_{a_2}^s\}=\delta_{r+s,0}\delta_{a_1+a_2,0},\qquad[J_0^V,\psi_a^r]=r\psi_a^r, \tag*{(339)} { ψ a 1 r , ψ a 2 s } = δ r + s , 0 δ a 1 + a 2 , 0 , [ J 0 V , ψ a r ] = r ψ a r , (339)
where the explicit form of the current J 0 V J_0^V J 0 V is given below in (355) .
The two Fock vacua are defined as
χ b > 0 ∣ 1 W ⟩ = 0 , ψ a > 0 r ∣ 1 V ⟩ = 0 , ∀ r . (340) \chi_{b>0}\lvert1_W\rangle=0,\qquad\psi_{a>0}^r\lvert1_V\rangle=0,\qquad\forall r. \tag*{(340)} χ b > 0 ∣ 1 W ⟩ = 0 , ψ a > 0 r ∣ 1 V ⟩ = 0 , ∀ r . (340)
We introduce the following intertwiner map:
Φ F ∣ 1 V ⟩ = ∣ 1 W ⟩ , Φ F ψ a r = χ ℓ a + r Φ F . (341) \Phi_F\lvert1_V\rangle=\lvert1_W\rangle,\qquad\Phi_F\psi_a^r=\chi_{\ell a+r}\Phi_F. \tag*{(341)} Φ F ∣ 1 V ⟩ = ∣ 1 W ⟩ , Φ F ψ a r = χ ℓ a + r Φ F . (341)
It has the following four properties:
The indices are in bijection. Indeed, one can introduce a ′ = a − 1 2 a'=a-\frac{1}{2} a ′ = a − 2 1 , b ′ = b − 1 2 b'=b-\frac{1}{2} b ′ = b − 2 1 and r ′ = r + ℓ 2 − 1 2 = r + n r'=r+\frac{\ell}{2}-\frac{1}{2}=r+n r ′ = r + 2 ℓ − 2 1 = r + n , so that the map between ( a , r ) (a,r) ( a , r ) and b b b reduces to b ′ = ℓ a ′ + r ′ b'=\ell a'+r' b ′ = ℓ a ′ + r ′ with 0 ≤ r ′ < ℓ 0\le r'<\ell 0 ≤ r ′ < ℓ . This gives an explicit bijection of indices between Z \mathbb{Z} Z and Z × Z ℓ \mathbb{Z}\times\mathbb{Z}_{\ell} Z × Z ℓ .
Creation and annihilation modes remain creation and annihilation modes. Indeed, if a > 0 a>0 a > 0 , then a ′ = a − 1 / 2 ≥ 0 a'=a-1/2\ge0 a ′ = a − 1/2 ≥ 0 , and b ′ = ℓ a ′ + r ′ ≥ 0 b'=\ell a'+r'\ge0 b ′ = ℓ a ′ + r ′ ≥ 0 . If a < 0 a<0 a < 0 , then ℓ a + r < 0 \ell a+r<0 ℓ a + r < 0 . Similarly, if a < 0 a<0 a < 0 , then a ′ = a − 1 / 2 ≤ − 1 a'=a-1/2\le-1 a ′ = a − 1/2 ≤ − 1 and b ′ = ℓ a ′ + r ′ ≤ − ℓ + r ′ < 0 b'=\ell a'+r'\le-\ell+r'<0 b ′ = ℓ a ′ + r ′ ≤ − ℓ + r ′ < 0 as r ′ ∈ { 0 , ⋯ , ℓ − 1 } r'\in\{0,\cdots,\ell-1\} r ′ ∈ { 0 , ⋯ , ℓ − 1 } . Therefore, the creation and annihilation modes on the V V V side are mapped to creation and annihilation modes on the W W W side.
The anticommutators agree. This follows from
ℓ ( a 1 + a 2 ) + r + s = 0 ⟺ a 1 + a 2 = 0 , r + s = 0 , (342) \ell(a_1+a_2)+r+s=0\Longleftrightarrow a_1+a_2=0,\qquad r+s=0, \tag*{(342)} ℓ ( a 1 + a 2 ) + r + s = 0 ⟺ a 1 + a 2 = 0 , r + s = 0 , (342)
due to ∣ r + s ∣ ≤ ℓ − 1 \lvert r+s\rvert\le\ell-1 ∣ r + s ∣ ≤ ℓ − 1 .
The fermion parity is preserved.
These properties guarantee that (341) is indeed a bijection between the two Fock-spaces. It provides a complete specification of the linear map on the two fermion Fock spaces
Φ F ( ψ − a 1 r 1 ⋯ ψ − a s r s 1 ) = χ − ℓ a 1 + r 1 ⋯ χ − ℓ a s + r s 1 , a j > 0 , (343) \Phi_F\left(\psi_{-a_1}^{r_1}\cdots\psi_{-a_s}^{r_s}1\right)=\chi_{-\ell a_1+r_1}\cdots\chi_{-\ell a_s+r_s}1,\qquad a_j>0, \tag*{(343)} Φ F ( ψ − a 1 r 1 ⋯ ψ − a s r s 1 ) = χ − ℓ a 1 + r 1 ⋯ χ − ℓ a s + r s 1 , a j > 0 , (343)
Moreover, using (335) and (355) , one can see that
[ H , ψ a r ] = − ( ℓ a + r ) ψ a r , [ S 0 , χ b ] = − b χ b . (344) [H,\psi_a^r]=-(\ell a+r)\psi_a^r,\qquad[S_0,\chi_b]=-b\chi_b. \tag*{(344)} [ H , ψ a r ] = − ( ℓ a + r ) ψ a r , [ S 0 , χ b ] = − b χ b . (344)
Combining with (341) , we thus have
Φ F H = S 0 Φ F . (345) \Phi_FH=S_0\Phi_F. \tag*{(345)} Φ F H = S 0 Φ F . (345)
Since our two VOAs are the even fermion-parity sectors of the corresponding free Majorana algebras (336) , we thus have the following graded vector space isomorphism
( L 1 ( s o 2 n + 1 ) , ( 2 n + 1 ) L 0 V − J 0 V ) ≅ g r ( Vir ( 3 , 4 ) , S 0 ) . (346) \left(L_1(\mathfrak{so}_{2n+1}),(2n+1)L_0^V-J_0^V\right)\cong_{\mathrm{gr}}(\operatorname{Vir}(3,4),S_0). \tag*{(346)} ( L 1 ( so 2 n + 1 ) , ( 2 n + 1 ) L 0 V − J 0 V ) ≅ gr ( Vir ( 3 , 4 ) , S 0 ) . (346)
This realization also gives further correspondences. The affine currents in L 1 ( s o 2 n + 1 ) L_1(\mathfrak{so}_{2n+1}) L 1 ( so 2 n + 1 ) admit the following description
J m r s = ∑ a ∈ Z + 1 / 2 : ψ a r ψ m − a s : . (347) J_m^{rs}=\sum_{a\in\mathbb{Z}+1/2}:\psi_a^r\psi_{m-a}^s:. \tag*{(347)} J m r s = a ∈ Z + 1/2 ∑ : ψ a r ψ m − a s : . (347)
Under (341) , these currents are mapped to
J ~ m r s = ∑ a ∈ Z + 1 / 2 : χ ℓ a + r χ ℓ ( m − a ) + s : , Φ F J m r s = J ~ m r s Φ F . (348) \widetilde{J}_m^{rs}=\sum_{a\in\mathbb{Z}+1/2}:\chi_{\ell a+r}\chi_{\ell(m-a)+s}:,\qquad\Phi_FJ_m^{rs}=\widetilde{J}_m^{rs}\Phi_F. \tag*{(348)} J m r s = a ∈ Z + 1/2 ∑ : χ ℓ a + r χ ℓ ( m − a ) + s : , Φ F J m r s = J m r s Φ F . (348)
Therefore, for every current relation, we have
P ( J ) ∣ 0 ⟩ = 0 ⟺ P ( J ~ ) ∣ 0 ⟩ = 0 , (349) P(J)|0\rangle=0\quad\Longleftrightarrow\quad P(\widetilde{J})|0\rangle=0, \tag*{(349)} P ( J ) ∣0 ⟩ = 0 ⟺ P ( J ) ∣0 ⟩ = 0 , (349)
where P ( J ) P(J) P ( J ) , P ( J ~ ) P(\widetilde{J}) P ( J ) are the same finite noncommutative polynomial in the respective current modes. This resolves the question of all null relations in this case: the invertible oscillator map transports every relation.
The Virasoro modes on the Ising VOA are
S m = 1 4 ∑ t + v = m ( v − t ) : χ t χ v : . (350) S_m=\frac{1}{4}\sum_{t+v=m}(v-t):\chi_t\chi_v:. \tag*{(350)} S m = 4 1 t + v = m ∑ ( v − t ) : χ t χ v : . (350)
Under the inverse of (341) , these modes are mapped to
S m V = 1 4 ∑ a , b ∈ Z + 1 2 , − n ≤ r , s ≤ n ℓ ( a + b ) + r + s = m [ ℓ ( b − a ) + s − r ] : ψ a r ψ b s : . (351) S_m^V=\frac{1}{4}\sum_{\substack{a,b\in\mathbb{Z}+\frac{1}{2},\ -n\le r,s\le n\\ \ell(a+b)+r+s=m}}\left[\ell(b-a)+s-r\right]:\psi_a^r\psi_b^s:. \tag*{(351)} S m V = 4 1 a , b ∈ Z + 2 1 , − n ≤ r , s ≤ n ℓ ( a + b ) + r + s = m ∑ [ ℓ ( b − a ) + s − r ] : ψ a r ψ b s : . (351)
They satisfy the Virasoro algebra
[ S m V , S j V ] = ( m − j ) S m + j V + m ( m 2 − 1 ) 24 δ m + j , 0 , (352) [S^{V}_{m},S^{V}_{j}]=(m-j)S^{V}_{m+j}+\frac{m(m^{2}-1)}{24}\delta_{m+j,0}, \tag*{(352)} [ S m V , S j V ] = ( m − j ) S m + j V + 24 m ( m 2 − 1 ) δ m + j , 0 , (352)
with central charge 1 / 2 1/2 1/2 . This should be distinguished from the original Sugawara stress tensor in the affine algebra. In addition, we have
S 0 V = − 1 2 ∑ a ∈ Z + 1 2 ∑ r = − n n ( ℓ a + r ) : ψ a r ψ − a − r : (353) S^{V}_{0}=-\frac{1}{2}\sum_{a\in\mathbb{Z}+\frac{1}{2}}\sum_{r=-n}^{n}(\ell a+r):\psi^{r}_{a}\psi^{-r}_{-a}: \tag*{(353)} S 0 V = − 2 1 a ∈ Z + 2 1 ∑ r = − n ∑ n ( ℓ a + r ) : ψ a r ψ − a − r : (353)
= − ∑ a > 0 a ∈ Z + 1 2 ∑ r = − n n ( ℓ a + r ) : ψ a r ψ − a − r : = ℓ L 0 V − J 0 V = H . (354) =-\sum_{\substack{a>0\\a\in\mathbb{Z}+\frac{1}{2}}}\sum_{r=-n}^{n}(\ell a+r):\psi^{r}_{a}\psi^{-r}_{-a}:=\ell L^{V}_{0}-J^{\mathrm{V}}_{0}=H. \tag*{(354)} = − a > 0 a ∈ Z + 2 1 ∑ r = − n ∑ n ( ℓ a + r ) : ψ a r ψ − a − r := ℓ L 0 V − J 0 V = H . (354)
Here we used
J 0 V = ∑ a > 0 a ∈ Z + 1 2 ∑ r = − n n r : ψ a r ψ − a − r : = 1 2 ∑ r = − n n r J 0 r , − r . (355) J^{\mathrm{V}}_{0}=\sum_{\substack{a>0\\a\in\mathbb{Z}+\frac{1}{2}}}\sum_{r=-n}^{n}r:\psi^{r}_{a}\psi^{-r}_{-a}:=\frac{1}{2}\sum_{r=-n}^{n}rJ^{r,-r}_{0}. \tag*{(355)} J 0 V = a > 0 a ∈ Z + 2 1 ∑ r = − n ∑ n r : ψ a r ψ − a − r := 2 1 r = − n ∑ n r J 0 r , − r . (355)
The universal Virasoro vacuum algebra at c = 1 / 2 c=1/2 c = 1/2 contains the nonzero weight-six singular vector
N I s i n g = ( S − 2 3 − 33 8 S − 4 S − 2 + 93 64 S − 3 2 − 27 16 S − 6 ) ∣ 0 ⟩ . (356) \mathcal{N}_{\mathrm{Ising}}=\left(S_{-2}^{3}-\frac{33}{8}S_{-4}S_{-2}+\frac{93}{64}S_{-3}^{2}-\frac{27}{16}S_{-6}\right)|0\rangle. \tag*{(356)} N Ising = ( S − 2 3 − 8 33 S − 4 S − 2 + 64 93 S − 3 2 − 16 27 S − 6 ) ∣0 ⟩ . (356)
Its image is zero in the simple Ising VOA and in its fermion realization. Therefore replacing every S m S_m S m by S m V S^{V}_m S m V gives zero in the source fermion realization as well, by conjugation with Φ F \Phi_F Φ F .
We now extend this discussion to modules. First, the affine algebra L 1 ( s o 2 n + 1 ) L_1(\mathfrak{so}_{2n+1}) L 1 ( so 2 n + 1 ) has three affine primaries, denoted as 0 , ω 1 , ω n 0,\omega_1,\omega_n 0 , ω 1 , ω n , corresponding to the trivial, vector and spinor representations. Similarly, the Ising VOA has three Virasoro primaries, 1 , ϵ , σ 1,\epsilon,\sigma 1 , ϵ , σ . They are related as follows:
Table 8.26.
The modules of the Ising VOA also admit free field realizations
I 0 = ( F χ N S ) e v e n = span { χ − a 1 ⋯ χ − a s ∣ 0 ⟩ : a 1 > ⋯ > a s > 0 , a i ∈ Z + 1 2 , s ∈ 2 N } , (357) I_{0}=(\mathcal{F}^{\mathrm{NS}}_{\chi})_{\mathrm{even}}=\operatorname{span}\{\chi_{-a_{1}}\cdots\chi_{-a_{s}}|0\rangle:a_{1}>\cdots>a_{s}>0,\ a_{i}\in\mathbb{Z}+\frac{1}{2},\ s\in2\mathbb{N}\}, \tag*{(357)} I 0 = ( F χ NS ) even = span { χ − a 1 ⋯ χ − a s ∣0 ⟩ : a 1 > ⋯ > a s > 0 , a i ∈ Z + 2 1 , s ∈ 2 N } , (357)
I ϵ = ( F χ N S ) o d d = span { χ − a 1 ⋯ χ − a s ∣ 0 ⟩ : a 1 > ⋯ > a s > 0 , a i ∈ Z + 1 2 , s ∈ 2 N + 1 } , (358) I_{\epsilon}=(\mathcal{F}^{\mathrm{NS}}_{\chi})_{\mathrm{odd}}=\operatorname{span}\{\chi_{-a_{1}}\cdots\chi_{-a_{s}}|0\rangle:a_{1}>\cdots>a_{s}>0,\ a_{i}\in\mathbb{Z}+\frac{1}{2},\ s\in2\mathbb{N}+1\}, \tag*{(358)} I ϵ = ( F χ NS ) odd = span { χ − a 1 ⋯ χ − a s ∣0 ⟩ : a 1 > ⋯ > a s > 0 , a i ∈ Z + 2 1 , s ∈ 2 N + 1 } , (358)
I σ = ( F χ R ) = span { χ − a 1 ⋯ χ − a s ∣ σ ⟩ : a 1 > ⋯ > a s > 0 , a i ∈ Z , s ∈ N } , (359) I_{\sigma}=(\mathcal{F}^{\mathrm{R}}_{\chi})=\operatorname{span}\{\chi_{-a_{1}}\cdots\chi_{-a_{s}}|\sigma\rangle:a_{1}>\cdots>a_{s}>0,\ a_{i}\in\mathbb{Z},\ s\in\mathbb{N}\}, \tag*{(359)} I σ = ( F χ R ) = span { χ − a 1 ⋯ χ − a s ∣ σ ⟩ : a 1 > ⋯ > a s > 0 , a i ∈ Z , s ∈ N } , (359)
where the ground state ∣ σ ⟩ |\sigma\rangle ∣ σ ⟩ satisfies
χ a ∣ σ ⟩ = 0 ( a > 0 ) , χ 0 ∣ σ ⟩ = 1 2 ∣ σ ⟩ , S 0 ∣ σ ⟩ = 1 16 ∣ σ ⟩ . (360) \chi_a|\sigma\rangle= 0 \quad(a>0), \qquad\chi_0|\sigma\rangle= \frac{1}{\sqrt{2}}|\sigma\rangle, \qquad S_0|\sigma\rangle= \frac{1}{16}|\sigma\rangle. \tag*{(360)} χ a ∣ σ ⟩ = 0 ( a > 0 ) , χ 0 ∣ σ ⟩ = 2 1 ∣ σ ⟩ , S 0 ∣ σ ⟩ = 16 1 ∣ σ ⟩ . (360)
The ground state in I ϵ I_\epsilon I ϵ is ∣ ϵ ⟩ = χ − 1 2 ∣ 0 ⟩ |\epsilon\rangle= \chi_{-\frac{1}{2}}|0\rangle ∣ ϵ ⟩ = χ − 2 1 ∣0 ⟩ which has conformal weight 1 / 2 1/2 1/2 . A similar construction also applies for level 1 affine algebra L 1 ( s o 2 n + 1 ) L_1(\mathfrak{so}_{2n+1}) L 1 ( so 2 n + 1 ) [35 ] . See e.g. [36 ] .
The Ising characters are given in [35 ]
χ ^ I 0 ( q ) = q − 1 / 48 2 [ P + + P − ] , χ ^ I ϵ ( q ) = q − 1 / 48 2 [ P + − P − ] , χ ^ I σ ( q ) = q 1 / 24 P R , (361) \widehat{\chi}_{I_0}(q) = \frac{q^{-1/48}}{2}\left[P_+ + P_-\right], \qquad\widehat{\chi}_{I_\epsilon}(q) = \frac{q^{-1/48}}{2}\left[P_+ - P_-\right], \qquad\widehat{\chi}_{I_\sigma}(q) = q^{1/24}P_R, \tag*{(361)} χ I 0 ( q ) = 2 q − 1/48 [ P + + P − ] , χ I ϵ ( q ) = 2 q − 1/48 [ P + − P − ] , χ I σ ( q ) = q 1/24 P R , (361)
where
P ± ( q ) = ( ∓ q 1 / 2 ; q ) ∞ , P R ( q ) = ( − q ; q ) ∞ , ( a ; q ) ∞ = ∏ i = 0 ∞ ( 1 − a q i ) . (362) P_\pm(q) = (\mp q^{1/2};q)_\infty, \qquad P_R(q) = (-q;q)_\infty, \qquad(a;q)_\infty= \prod_{i=0}^{\infty}(1-aq^i). \tag*{(362)} P ± ( q ) = ( ∓ q 1/2 ; q ) ∞ , P R ( q ) = ( − q ; q ) ∞ , ( a ; q ) ∞ = i = 0 ∏ ∞ ( 1 − a q i ) . (362)
The affine characters of the three modules are given by [37 ]
χ ^ M 0 ( Q , x ) = Q − ℓ / 48 2 [ A + ( Q , x ) + A − ( Q , x ) ] , χ ^ M v ( Q , x ) = Q − ℓ / 48 2 [ A + ( Q , x ) − A − ( Q , x ) ] , χ ^ M s ( Q , x ) = Q ℓ / 24 B ( Q , x ) , (363) \begin{aligned}
\widehat{\chi}_{M_0}(Q,\boldsymbol{x}) &= \frac{Q^{-\ell/48}}{2}\left[A_+(Q,\boldsymbol{x})+A_-(Q,\boldsymbol{x})\right],\\
\widehat{\chi}_{M_v}(Q,\boldsymbol{x}) &= \frac{Q^{-\ell/48}}{2}\left[A_+(Q,\boldsymbol{x})-A_-(Q,\boldsymbol{x})\right],\\
\widehat{\chi}_{M_s}(Q,\boldsymbol{x}) &= Q^{\ell/24}B(Q,\boldsymbol{x}), \tag*{(363)}
\end{aligned} χ M 0 ( Q , x ) χ M v ( Q , x ) χ M s ( Q , x ) = 2 Q − ℓ /48 [ A + ( Q , x ) + A − ( Q , x ) ] , = 2 Q − ℓ /48 [ A + ( Q , x ) − A − ( Q , x ) ] , = Q ℓ /24 B ( Q , x ) , (363)
where
A ± ( Q , x ) = ( ∓ Q 1 / 2 ; Q ) ∞ ∏ r = 1 n ( ∓ x r Q 1 / 2 ; Q ) ∞ ( ∓ x r − 1 Q 1 / 2 ; Q ) ∞ , (364) A_\pm(Q,\boldsymbol{x}) = (\mp Q^{1/2};Q)_\infty\prod_{r=1}^{n}(\mp x_rQ^{1/2};Q)_\infty(\mp x_r^{-1}Q^{1/2};Q)_\infty, \tag*{(364)} A ± ( Q , x ) = ( ∓ Q 1/2 ; Q ) ∞ r = 1 ∏ n ( ∓ x r Q 1/2 ; Q ) ∞ ( ∓ x r − 1 Q 1/2 ; Q ) ∞ , (364)
B ( Q , x ) = ( − Q ; Q ) ∞ ∏ r = 1 n ( x r 1 / 2 + x r − 1 / 2 ) ( − x r Q ; Q ) ∞ ( − x r − 1 Q ; Q ) ∞ . B(Q,\boldsymbol{x}) = (-Q;Q)_\infty\prod_{r=1}^{n}(x_r^{1/2}+x_r^{-1/2})(-x_rQ;Q)_\infty(-x_r^{-1}Q;Q)_\infty. B ( Q , x ) = ( − Q ; Q ) ∞ r = 1 ∏ n ( x r 1/2 + x r − 1/2 ) ( − x r Q ; Q ) ∞ ( − x r − 1 Q ; Q ) ∞ .
Setting Q = q ℓ Q=q^\ell Q = q ℓ and x r = q − r x_r=q^{-r} x r = q − r , we have
A ± ( q ℓ , x r = q − r ) = P ± ( q ) , B ( q ℓ , x r = q − r ) = q − n ( n + 1 ) / 4 P R ( q ) . (365) A_\pm(q^\ell,x_r=q^{-r}) = P_\pm(q), \qquad B(q^\ell,x_r=q^{-r}) = q^{-n(n+1)/4}P_R(q). \tag*{(365)} A ± ( q ℓ , x r = q − r ) = P ± ( q ) , B ( q ℓ , x r = q − r ) = q − n ( n + 1 ) /4 P R ( q ) . (365)
As a result, we obtain the character identities
χ ^ I α ( q ) = q ( ℓ c V − c W ) / 24 χ ^ M α ( q ℓ , x r = q − r ) , ℓ c V − c W 24 = ℓ 2 − 1 48 , (366) \widehat{\chi}_{I_\alpha}(q) = q^{(\ell c_V-c_W)/24}\widehat{\chi}_{M_\alpha}(q^\ell,x_r=q^{-r}), \qquad\frac{\ell c_V-c_W}{24} = \frac{\ell^2-1}{48}, \tag*{(366)} χ I α ( q ) = q ( ℓ c V − c W ) /24 χ M α ( q ℓ , x r = q − r ) , 24 ℓ c V − c W = 48 ℓ 2 − 1 , (366)
for the three pairs of modules ( M 0 , I 0 ) (M_0,I_0) ( M 0 , I 0 ) , ( M v , I ϵ ) (M_v,I_\epsilon) ( M v , I ϵ ) , ( M s , I σ ) (M_s,I_\sigma) ( M s , I σ ) . Equivalently, we have the equality Tr M α q H = Tr I α q S 0 \operatorname{Tr}_{M_\alpha}q^H = \operatorname{Tr}_{I_\alpha}q^{S_0} Tr M α q H = Tr I α q S 0 . The product formulae also show that all eigenspaces of these regraded operators are finite-dimensional and that their spectra are bounded below.
These character identities imply the following graded vector-space isomorphisms for the modules
( L 1 ( s o 2 n + 1 , 0 ) , H ) ≅ g r ( I 0 , S 0 ) , ( L 1 ( s o 2 n + 1 , ω 1 ) , H ) ≅ g r ( I ε , S 0 ) , ( L 1 ( s o 2 n + 1 , ω n ) , H ) ≅ g r ( I σ , S 0 ) . (367) \begin{aligned}
(L_1(\mathfrak{so}_{2n+1},0),H)&\cong_{\mathrm{gr}}(I_0,S_0),\\
(L_1(\mathfrak{so}_{2n+1},\omega_1),H)&\cong_{\mathrm{gr}}(I_{\varepsilon},S_0),\\
(L_1(\mathfrak{so}_{2n+1},\omega_n),H)&\cong_{\mathrm{gr}}(I_{\sigma},S_0). \tag*{(367)}
\end{aligned} ( L 1 ( so 2 n + 1 , 0 ) , H ) ( L 1 ( so 2 n + 1 , ω 1 ) , H ) ( L 1 ( so 2 n + 1 , ω n ) , H ) ≅ gr ( I 0 , S 0 ) , ≅ gr ( I ε , S 0 ) , ≅ gr ( I σ , S 0 ) . (367)
Further generalization
It turns out the preceding example admits broad generalizations, which will be explored in more detail in [38 ] . Below we present several interesting results.
First, for m ≥ 2 m \ge2 m ≥ 2 , gcd ( m + 2 , 3 ) = 1 \gcd(m+2,3)=1 g cd( m + 2 , 3 ) = 1 , we claim that
( L 1 ( s p 2 m ) , ( 2 m + 1 ) L 0 V − J 0 ρ V ) ≅ g r ( Vir ( 3 , m + 2 ) , S 0 ) . (368) \left(L_1(\mathfrak{sp}_{2m}),(2m+1)L_0^V-J_0^{\rho^V}\right)\cong_{\mathrm{gr}}\left(\operatorname{Vir}(3,m+2),S_0\right). \tag*{(368)} ( L 1 ( sp 2 m ) , ( 2 m + 1 ) L 0 V − J 0 ρ V ) ≅ gr ( Vir ( 3 , m + 2 ) , S 0 ) . (368)
This follows from the specialized character formula in [39 Theorem 7.14] , which gives
Tr L 1 ( s p 2 m ) q ( 2 m + 1 ) L 0 V − J 0 ρ V = ( q , q 2 m + 3 , q 2 m + 4 ; q 2 m + 4 ) ∞ ( q 2 m + 2 , q 2 m + 6 ; q 4 m + 8 ) ∞ ( q ; q ) ∞ (369) \operatorname{Tr}_{L_1(\mathfrak{sp}_{2m})}q^{(2m+1)L_0^V-J_0^{\rho^V}}=
\frac{(q,q^{2m+3},q^{2m+4};q^{2m+4})_\infty(q^{2m+2},q^{2m+6};q^{4m+8})_\infty}{(q;q)_\infty} \tag*{(369)} Tr L 1 ( sp 2 m ) q ( 2 m + 1 ) L 0 V − J 0 ρ V = ( q ; q ) ∞ ( q , q 2 m + 3 , q 2 m + 4 ; q 2 m + 4 ) ∞ ( q 2 m + 2 , q 2 m + 6 ; q 4 m + 8 ) ∞ (369)
= 1 ( q ; q ) ∞ ∑ n = − ∞ ∞ ( q 3 ( m + 2 ) n 2 + ( m − 1 ) n − q 3 ( m + 2 ) n 2 + ( m + 5 ) n + 1 ) (370) =\frac{1}{(q;q)_\infty}\sum_{n=-\infty}^{\infty}\left(q^{3(m+2)n^2+(m-1)n}-q^{3(m+2)n^2+(m+5)n+1}\right) \tag*{(370)} = ( q ; q ) ∞ 1 n = − ∞ ∑ ∞ ( q 3 ( m + 2 ) n 2 + ( m − 1 ) n − q 3 ( m + 2 ) n 2 + ( m + 5 ) n + 1 ) (370)
= Tr Vir ( 3 , m + 2 ) q L 0 , (371) =\operatorname{Tr}_{\operatorname{Vir}(3,m+2)}q^{L_0}, \tag*{(371)} = Tr Vir ( 3 , m + 2 ) q L 0 , (371)
where ( a 1 , ⋯ , a s ; q ) ∞ = ( a 1 ; q ) ∞ ⋯ ( a s ; q ) ∞ (a_1,\cdots,a_s;q)_\infty=(a_1;q)_\infty\cdots(a_s;q)_\infty ( a 1 , ⋯ , a s ; q ) ∞ = ( a 1 ; q ) ∞ ⋯ ( a s ; q ) ∞ and we used the quintuple product identity in the second equality. 37 The resulting infinite sum in (370) is exactly the vacuum character of the Virasoro minimal model. This thus establishes (368) . Setting m = 2 m=2 m = 2 in (368) reduces to (334) with n = 2 n=2 n = 2 , due to the isomorphism s p 4 ≅ s o 5 \mathfrak{sp}_4\cong\mathfrak{so}_5 sp 4 ≅ so 5 .
Under the same conditions m ≥ 2 m \ge2 m ≥ 2 and gcd ( m + 2 , 3 ) = 1 \gcd(m+2,3)=1 g cd( m + 2 , 3 ) = 1 , there is also a GVISO between s l 2 \mathfrak{sl}_2 sl 2 and the Virasoro algebra [6 ] :
( L m ( s l 2 ) , 3 L 0 − 1 2 J 0 h ) ≅ g r ( Vir ( 3 , m + 2 ) , S 0 ) . (372) \left(L_m(\mathfrak{sl}_2),3L_0-\frac{1}{2}J_0^h\right)\cong_{\mathrm{gr}}\left(\operatorname{Vir}(3,m+2),S_0\right). \tag*{(372)} ( L m ( sl 2 ) , 3 L 0 − 2 1 J 0 h ) ≅ gr ( Vir ( 3 , m + 2 ) , S 0 ) . (372)
Combining (368) and (372) gives
( L 1 ( s p 2 m ) , ( 2 m + 1 ) L 0 − J 0 ρ V ) ≅ g r ( L m ( s l 2 ) , 3 L 0 − 1 2 J 0 h ) = ( L m ( s p 2 ) , 3 L 0 − J 0 ρ V ) . (373) \left(L_1(\mathfrak{sp}_{2m}),(2m+1)L_0-J_0^{\rho^V}\right)\cong_{\mathrm{gr}}\left(L_m(\mathfrak{sl}_2),3L_0-\frac{1}{2}J_0^h\right)=\left(L_m(\mathfrak{sp}_2),3L_0-J_0^{\rho^V}\right). \tag*{(373)} ( L 1 ( sp 2 m ) , ( 2 m + 1 ) L 0 − J 0 ρ V ) ≅ gr ( L m ( sl 2 ) , 3 L 0 − 2 1 J 0 h ) = ( L m ( sp 2 ) , 3 L 0 − J 0 ρ V ) . (373)
This admits a further generalization to modules. It turns out that this is just the special case of the following general family of relations:
( L m ( s p 2 n ) , ( 2 n + 1 ) L 0 − J 0 ρ n V ) ≅ g r ( L n ( s p 2 m ) , ( 2 m + 1 ) L 0 − J 0 ρ m V ) . (374) \left(L_m(\mathfrak{sp}_{2n}),(2n+1)L_0-J_0^{\rho_n^V}\right)\cong_{\mathrm{gr}}\left(L_n(\mathfrak{sp}_{2m}),(2m+1)L_0-J_0^{\rho_m^V}\right). \tag*{(374)} ( L m ( sp 2 n ) , ( 2 n + 1 ) L 0 − J 0 ρ n V ) ≅ gr ( L n ( sp 2 m ) , ( 2 m + 1 ) L 0 − J 0 ρ m V ) . (374)
Actually, the formula in [39 Theorem 7.14] allows us to compute the graded character explicitly
Tr L m ( s p 2 n ) q ( 2 n + 1 ) L 0 − J 0 ν = PE [ q 2 ( 1 − q 2 m ) ( 1 − q 2 n ) ( 1 + q 2 m + 2 n + 3 ) ( 1 − q ) ( 1 − q 2 ) ( 1 − q 4 m + 4 n + 4 ) ] (375) \operatorname{Tr}_{L_m(\mathfrak{sp}_{2n})} q^{(2n+1)L_0-J_0^{\nu}}=\operatorname{PE}\left[\frac{q^2(1-q^{2m})(1-q^{2n})(1+q^{2m+2n+3})}{(1-q)(1-q^2)(1-q^{4m+4n+4})}\right] \tag*{(375)} Tr L m ( sp 2 n ) q ( 2 n + 1 ) L 0 − J 0 ν = PE [ ( 1 − q ) ( 1 − q 2 ) ( 1 − q 4 m + 4 n + 4 ) q 2 ( 1 − q 2 m ) ( 1 − q 2 n ) ( 1 + q 2 m + 2 n + 3 ) ] (375)
which is obviously symmetric in m m m and n n n , thus proving the GVISO (374) . Alternatively, (374) can be understood by combining the VOA isomorphism [40, Theorem 9.3[2 ] }
W − ( n + 1 ) + ( m + n + 1 ) / ( 2 n + 1 ) ( s p 2 n ) ≅ W − ( m + 1 ) + ( m + n + 1 ) / ( 2 m + 1 ) ( s p 2 m ) , (376) \mathcal{W}_{-(n+1)+(m+n+1)/(2n+1)}(\mathfrak{sp}_{2n}) \cong\mathcal{W}_{-(m+1)+(m+n+1)/(2m+1)}(\mathfrak{sp}_{2m}), \tag*{(376)} W − ( n + 1 ) + ( m + n + 1 ) / ( 2 n + 1 ) ( sp 2 n ) ≅ W − ( m + 1 ) + ( m + n + 1 ) / ( 2 m + 1 ) ( sp 2 m ) , (376)
together with a generalization of affine/principal-W W W character identity in (93) beyond the boundary admissible regime.
Eq. (374) gives rise to the character identity between two affine algebras whose rank and level are exchanged. Because of this property, this kind of relation will be referred to as spectral level-rank duality . This is reminiscent of the standard level-rank duality in WZW and CS theory, which shall be referred to as topological level-rank duality . Detailed derivations of (374) and (375) , and further generalizations to other pairs of VOAs and modules will be presented in [38 ] .
Conclusion
In this paper, we establish infinite families of index identities among AD theories and, equivalently, character identities for the associated affine and W W W -algebras. At the VOA level, we construct an explicit graded vector space isomorphism (GVISO) for a concrete pair of VOAs. We also establish the correspondence for universal affine and principal W W W -algebras and test its compatibility with their simple quotients by matching the lowest-degree null states. We extend our discussion to non-vacuum modules and to examples beyond the SCFT/VOA correspondence. In the latter setting, free Majorana fermion realizations provide a direct explanation of identities relating level-one affine and Virasoro algebras. These examples finally lead us to a precise and new proposal of spectral level-rank duality .
Taken together, these results suggest a broader conjectural framework in which distinct VOAs and selected representations are related by graded vector space isomorphisms after suitable regradings. Such correspondences identify graded state spaces without requiring preservation of the OPEs and are therefore weaker than VOA isomorphisms. They would account for identities between appropriately specialized characters of selected modules, while allowing the underlying operator algebras to remain distinct.
Our main constructions concern boundary admissible affine and W W W -algebras associated with four-dimensional AD theories. A central family relates affine algebras to principal W W W -algebras. Non-principal W W W -algebras also appear, but our examples are far from exhaustive. It would be valuable to develop systematic tools for constructing further graded isomorphisms between affine and more general W W W -algebras. From a complementary perspective, one may seek to classify the specializations of flavored affine characters that yield series with nice modular transformation properties. Character identities with W W W -algebras provide one mechanism for obtaining such specializations, thus naturally ensuring the requisite modular behavior. The new perspective allows one to investigate the modular structure of affine characters directly, without first identifying a corresponding W W W -algebra. Such an investigation is of independent interest and may also guide the discovery of new correspondences.38
Another important direction is to extend these graded isomorphisms beyond boundary admissible levels, encompassing non-boundary admissible and even non-admissible algebras and their modules. We will address some of these questions in [34 ] and [38 ] . Such extensions may reveal deeper structural principles underlying the identities studied here. In particular, spectral level-rank duality may provide a framework for exploring further relations with WZW models and Chern–Simons theories, including possible connections to topological quantum field theory and its categorical structures. A central question is whether, and under what additional conditions, these spectral correspondences admit a categorical interpretation.
The high-temperature limits of the index identities constrain combinations of central charges and flavor levels, together with the modular data governing these asymptotics. These relations provide useful criteria for identifying candidate GVISOs and guiding the search for new examples. It would be interesting to determine whether these asymptotic constraints, supplemented by non-vacuum character identities, yield direct relations between the corresponding modular S S S and T T T matrices, as well as between the (flavored) MLDEs satisfied by the characters. More broadly, one may ask whether the SCFTs connected by these index identities also exhibit relations in other sectors, in particular their Coulomb branches. Such connections could clarify the physical origin of the index identities and may help build more fundamental connections between the underlying SCFTs.
Finally, the Macdonald index provides another useful and natural direction to explore, as it incorporates more refined information beyond the Schur index. Although a universal intrinsic prescription for recovering the Macdonald index from the VOA alone is not available, a distinguished non-Schur limit does admit an intrinsic VOA interpretation [41 ] . It would be interesting to establish analogous identities for these indices and to determine whether the corresponding GVISOs can be refined to respect the additional gradings or filtrations involved. Such refinements could reveal further structure in the relations between the underlying state spaces of VOAs.
We hope to report further progress on these questions in future work.
Acknowledgments
This work was supported by the startup grant at SIMIS and the Shanghai Pujiang Program (No. 25PJА128). We acknowledge the assistance of AI tools with computations and language improvement.
Bershadsky–Polyakov algebra
The Bershadsky–Polyakov algebra has strong generators given by T , G + , G − , J T,G^{+},G^{-},J T , G + , G − , J , with conformal weights 2 , 3 2 , 3 2 , 1 2,\frac{3}{2},\frac{3}{2},1 2 , 2 3 , 2 3 , 1 . They are all bosonic and obey the OPEs [10 ]
T ( z ) T ( w ) ∼ k ( 14 − 9 k ) 4 ( 2 + k ) ( z − w ) 4 + 2 T ( w ) ( z − w ) 2 + T ′ ( w ) z − w , T ( z ) G + ( w ) ∼ 3 2 G + ( w ) ( z − w ) 2 + G + ′ ( w ) z − w , T ( z ) G − ( w ) ∼ 3 2 G − ( w ) ( z − w ) 2 + G − ′ ( w ) z − w , T ( z ) J ( w ) ∼ J ( w ) ( z − w ) 2 + J ′ ( w ) z − w , J ( z ) J ( w ) ∼ k / 2 ( z − w ) 2 , J ( z ) G + ( w ) ∼ G + ( w ) z − w , J ( z ) G − ( w ) ∼ − G − ( w ) z − w , G + ( z ) G − ( w ) ∼ 3 8 ( 3 k 2 − 2 k ) ( z − w ) 3 + 3 4 ( 3 k − 2 ) J ( w ) ( z − w ) 2 + 3 ( J J ) ( w ) + 3 8 ( 3 k − 2 ) J ′ ( w ) − 3 4 ( k + 2 ) T ( w ) z − w , G ± ( z ) G ± ( w ) ∼ 0. \begin{aligned}
T(z)T(w) &\sim\frac{k(14-9k)}{4(2+k)(z-w)^{4}}+\frac{2T(w)}{(z-w)^{2}}+\frac{T'(w)}{z-w},\\
T(z)G^{+}(w) &\sim\frac{\frac{3}{2}G^{+}(w)}{(z-w)^{2}}+\frac{G^{+\prime}(w)}{z-w},\\
T(z)G^{-}(w) &\sim\frac{\frac{3}{2}G^{-}(w)}{(z-w)^{2}}+\frac{G^{-\prime}(w)}{z-w},\\
T(z)J(w) &\sim\frac{J(w)}{(z-w)^{2}}+\frac{J'(w)}{z-w},\\
J(z)J(w) &\sim\frac{k/2}{(z-w)^{2}},\\
J(z)G^{+}(w) &\sim\frac{G^{+}(w)}{z-w},\\
J(z)G^{-}(w) &\sim\frac{-G^{-}(w)}{z-w},\\
G^{+}(z)G^{-}(w) &\sim\frac{\frac{3}{8}(3k^{2}-2k)}{(z-w)^{3}}+\frac{\frac{3}{4}(3k-2)J(w)}{(z-w)^{2}}+\frac{3(JJ)(w)+\frac{3}{8}(3k-2)J'(w)-\frac{3}{4}(k+2)T(w)}{z-w},\\
G^{\pm}(z)G^{\pm}(w) &\sim0.
\end{aligned} T ( z ) T ( w ) T ( z ) G + ( w ) T ( z ) G − ( w ) T ( z ) J ( w ) J ( z ) J ( w ) J ( z ) G + ( w ) J ( z ) G − ( w ) G + ( z ) G − ( w ) G ± ( z ) G ± ( w ) ∼ 4 ( 2 + k ) ( z − w ) 4 k ( 14 − 9 k ) + ( z − w ) 2 2 T ( w ) + z − w T ′ ( w ) , ∼ ( z − w ) 2 2 3 G + ( w ) + z − w G + ′ ( w ) , ∼ ( z − w ) 2 2 3 G − ( w ) + z − w G − ′ ( w ) , ∼ ( z − w ) 2 J ( w ) + z − w J ′ ( w ) , ∼ ( z − w ) 2 k /2 , ∼ z − w G + ( w ) , ∼ z − w − G − ( w ) , ∼ ( z − w ) 3 8 3 ( 3 k 2 − 2 k ) + ( z − w ) 2 4 3 ( 3 k − 2 ) J ( w ) + z − w 3 ( J J ) ( w ) + 8 3 ( 3 k − 2 ) J ′ ( w ) − 4 3 ( k + 2 ) T ( w ) , ∼ 0.
The Bershadsky–Polyakov algebra can be obtained as the subregular Drinfel’d-Sokolov reduction of s l 3 \mathfrak{sl}_{3} sl 3 , namely
B P ~ k = W k ~ ( s l 3 , f s u b r e g ) . (377) \widetilde{\mathrm{BP}}_{k}=W_{\widetilde{k}}(\mathfrak{sl}_{3},f_{\mathrm{subreg}}). \tag*{(377)} BP k = W k ( sl 3 , f subreg ) . (377)
Here k k k is twice the u ( 1 ) \mathfrak{u}(1) u ( 1 ) level. We assume k ≠ − 2 k\ne-2 k = − 2 , equivalently k ~ ≠ − 3 \widetilde{k}\ne-3 k = − 3 , so that the conformal vector is defined. The parameter k k k and central charge c c c are related to the parent affine s l 3 \mathfrak{sl}_{3} sl 3 level k ~ \widetilde{k} k by
k = 2 3 ( 2 k ~ + 3 ) , c = k ( 14 − 9 k ) 2 ( 2 + k ) = 7 − 6 k ~ − 24 k ~ + 3 . (378) k=\frac{2}{3}(2\widetilde{k}+3),\qquad c=\frac{k(14-9k)}{2(2+k)}=7-6\widetilde{k}-\frac{24}{\widetilde{k}+3}. \tag*{(378)} k = 3 2 ( 2 k + 3 ) , c = 2 ( 2 + k ) k ( 14 − 9 k ) = 7 − 6 k − k + 3 24 . (378)
Modular forms and derivatives
The Eisenstein series are given by
E 2 k ( q ) = − B 2 k ( 2 k ) ! + 2 ( 2 k − 1 ) ! ∑ n ≥ 1 n 2 k − 1 q n 1 − q n , q = e 2 π i τ , (379) E_{2k}(q)=-\frac{B_{2k}}{(2k)!}+\frac{2}{(2k-1)!}\sum_{n\ge1}\frac{n^{2k-1}q^{n}}{1-q^{n}},\qquad q=e^{2\pi i\tau}, \tag*{(379)} E 2 k ( q ) = − ( 2 k )! B 2 k + ( 2 k − 1 )! 2 n ≥ 1 ∑ 1 − q n n 2 k − 1 q n , q = e 2 π i τ , (379)
where B 2 k B_{2k} B 2 k are Bernoulli numbers. We also introduce
X ( q ) = 12 [ E 2 ( q 1 / 2 ) − 2 E 2 ( q ) ] = θ 2 ( q ) 4 + θ 3 ( q ) 4 , (380) X(q)=12\left[\mathbb{E}_{2}(q^{1/2})-2\mathbb{E}_{2}(q)\right]=\theta_{2}(q)^{4}+\theta_{3}(q)^{4}, \tag*{(380)} X ( q ) = 12 [ E 2 ( q 1/2 ) − 2 E 2 ( q ) ] = θ 2 ( q ) 4 + θ 3 ( q ) 4 , (380)
where the theta functions are defined by
θ 2 ( q ) = ∑ n ∈ Z q ( n + 1 / 2 ) 2 / 2 , θ 3 ( q ) = ∑ n ∈ Z q n 2 / 2 . (381) \theta_2(q)=\sum_{n\in\mathbb{Z}}q^{(n+1/2)^2/2},\qquad\theta_3(q)=\sum_{n\in\mathbb{Z}}q^{n^2/2}. \tag*{(381)} θ 2 ( q ) = n ∈ Z ∑ q ( n + 1/2 ) 2 /2 , θ 3 ( q ) = n ∈ Z ∑ q n 2 /2 . (381)
We also have the following identities:
E 2 ( q 1 / 2 ) = 2 E 2 ( q ) + X ( q ) 12 , (382) \mathbb{E}_2(q^{1/2})=2\mathbb{E}_2(q)+\frac{\mathsf{X}(q)}{12}, \tag*{(382)} E 2 ( q 1/2 ) = 2 E 2 ( q ) + 12 X ( q ) , (382)
E 4 ( q 1 / 2 ) = X ( q ) 2 144 − 4 E 4 ( q ) , (383) \mathbb{E}_4(q^{1/2})=\frac{\mathsf{X}(q)^2}{144}-4\mathbb{E}_4(q), \tag*{(383)} E 4 ( q 1/2 ) = 144 X ( q ) 2 − 4 E 4 ( q ) , (383)
E 6 ( q 1 / 2 ) = 11 X ( q ) 3 30240 − 2 7 X ( q ) E 4 ( q ) . (384) \mathbb{E}_6(q^{1/2})=\frac{11\mathsf{X}(q)^3}{30240}-\frac{2}{7}\mathsf{X}(q)\mathbb{E}_4(q). \tag*{(384)} E 6 ( q 1/2 ) = 30240 11 X ( q ) 3 − 7 2 X ( q ) E 4 ( q ) . (384)
The Serre derivatives are
∂ ( w ) = q d d q + w E 2 , D q ( r ) = ∂ ( 2 r − 2 ) ∘ ⋯ ∘ ∂ ( 2 ) ∘ ∂ ( 0 ) , D q ( 0 ) = 1. (385) \partial_{(w)}=q\frac{d}{dq}+w\mathbb{E}_2,\qquad D_q^{(r)}=\partial_{(2r-2)}\circ\cdots\circ\partial_{(2)}\circ\partial_{(0)},\qquad D_q^{(0)}=1. \tag*{(385)} ∂ ( w ) = q d q d + w E 2 , D q ( r ) = ∂ ( 2 r − 2 ) ∘ ⋯ ∘ ∂ ( 2 ) ∘ ∂ ( 0 ) , D q ( 0 ) = 1. (385)
A character summation formula
Our goal in this appendix is to evaluate the following sum:
χ ( q , w ) = ∑ m ≥ 0 ∑ ∣ ν ∣ ≤ m ν − m ∈ 2 Z w ν Δ m , ν 1 − q m + 1 ( q ; q ) ∞ , (386) \chi(q,w)=\sum_{m\geq0}\sum_{\substack{|\nu|\leq m\\ \nu-m\in2\mathbb{Z}}}w^\nu\Delta_{m,\nu}\frac{1-q^{m+1}}{(q;q)_\infty}, \tag*{(386)} χ ( q , w ) = m ≥ 0 ∑ ∣ ν ∣ ≤ m ν − m ∈ 2 Z ∑ w ν Δ m , ν ( q ; q ) ∞ 1 − q m + 1 , (386)
where Δ m , ν \Delta_{m,\nu} Δ m , ν is given in (209) .
Set
a = ( m + ν ) / 2 , b = ( m − ν ) / 2 , Z = q 2 p , u = w q p − 1 2 , v = w − 1 q p − 1 2 , u v = Z q . (387) a=(m+\nu)/2,\qquad b=(m-\nu)/2,\qquad Z=q^{2p},\qquad u=wq^{p-\frac{1}{2}},\qquad v=w^{-1}q^{p-\frac{1}{2}},\qquad uv=\frac{Z}{q}. \tag*{(387)} a = ( m + ν ) /2 , b = ( m − ν ) /2 , Z = q 2 p , u = w q p − 2 1 , v = w − 1 q p − 2 1 , uv = q Z . (387)
The numerator in (C.1) becomes
( q ; q ) ∞ 2 χ ( q , w ) = ∑ a , b ≥ 0 u a v b Z a b ( 1 − q a + b + 1 ) = ∑ a ≥ 0 u a 1 − v Z a − q ∑ b ≥ 0 ( q v ) b 1 − q u Z b = ∑ n ∈ Z u n 1 − v Z n . (388) \begin{aligned}
(q;q)_\infty^2\chi(q,w)=\sum_{a,b\geq0}u^av^bZ^{ab}(1-q^{a+b+1})=\sum_{a\geq0}\frac{u^a}{1-vZ^a}-q\sum_{b\geq0}\frac{(qv)^b}{1-quZ^b} \\
&=\sum_{n\in\mathbb{Z}}\frac{u^n}{1-vZ^n}. \tag*{(388)}
\end{aligned} ( q ; q ) ∞ 2 χ ( q , w ) = a , b ≥ 0 ∑ u a v b Z ab ( 1 − q a + b + 1 ) = a ≥ 0 ∑ 1 − v Z a u a − q b ≥ 0 ∑ 1 − q u Z b ( q v ) b = n ∈ Z ∑ 1 − v Z n u n . (388)
For the last equality, write n = − b − 1 n=-b-1 n = − b − 1 in the negative part of the bilateral sum and use Z / u = q v Z/u=qv Z / u = q v and Z / v = q u Z/v=qu Z / v = q u :
u − b − 1 1 − v Z − b − 1 = − Z u v ( Z / u ) b 1 − ( Z / v ) Z b = − q ( q v ) b 1 − q u Z b . (389) \frac{u^{-b-1}}{1-vZ^{-b-1}}=-\frac{Z}{uv}\frac{(Z/u)^b}{1-(Z/v)Z^b}=-q\frac{(qv)^b}{1-quZ^b}. \tag*{(389)} 1 − v Z − b − 1 u − b − 1 = − uv Z 1 − ( Z / v ) Z b ( Z / u ) b = − q 1 − q u Z b ( q v ) b . (389)
Next we use Kronecker’s identity:
∑ n ∈ Z u n 1 − v Z n = ( Z ; Z ) ∞ 2 ( u v ; Z ) ∞ ( Z / ( u v ) ; Z ) ∞ ( u ; Z ) ∞ ( Z / u ; Z ) ∞ ( v ; Z ) ∞ ( Z / v ; Z ) ∞ , (390) \sum_{n\in\mathbb{Z}} \frac{u^n}{1-vZ^n} = \frac{(Z;Z)_\infty^2(uv;Z)_\infty(Z/(uv);Z)_\infty}{(u;Z)_\infty(Z/u;Z)_\infty(v;Z)_\infty(Z/v;Z)_\infty}, \tag*{(390)} n ∈ Z ∑ 1 − v Z n u n = ( u ; Z ) ∞ ( Z / u ; Z ) ∞ ( v ; Z ) ∞ ( Z / v ; Z ) ∞ ( Z ; Z ) ∞ 2 ( uv ; Z ) ∞ ( Z / ( uv ) ; Z ) ∞ , (390)
which follows by setting ( A , B , ξ ) = ( v , Z v , u ) (A, B, \xi)=(v, Zv, u) ( A , B , ξ ) = ( v , Z v , u ) in Ramanujan’s 1 ψ 1 1\psi_1 1 ψ 1 summation [42 ] :
∑ n ∈ Z ( A ; Z ) n ( B ; Z ) n ξ n = ( Z ; Z ) ∞ ( B / A ; Z ) ∞ ( A ξ ; Z ) ∞ ( Z / ( A ξ ) ; Z ) ∞ ( B ; Z ) ∞ ( Z / A ; Z ) ∞ ( ξ ; Z ) ∞ ( B / ( A ξ ) ; Z ) ∞ . (391) \sum_{n\in\mathbb{Z}} \frac{(A;Z)_n}{(B;Z)_n}\xi^n = \frac{(Z;Z)_\infty(B/A;Z)_\infty(A\xi;Z)_\infty(Z/(A\xi);Z)_\infty}{(B;Z)_\infty(Z/A;Z)_\infty(\xi;Z)_\infty(B/(A\xi);Z)_\infty}. \tag*{(391)} n ∈ Z ∑ ( B ; Z ) n ( A ; Z ) n ξ n = ( B ; Z ) ∞ ( Z / A ; Z ) ∞ ( ξ ; Z ) ∞ ( B / ( A ξ ) ; Z ) ∞ ( Z ; Z ) ∞ ( B / A ; Z ) ∞ ( A ξ ; Z ) ∞ ( Z / ( A ξ ) ; Z ) ∞ . (391)
Applying the identity (C.5) to (C.3) gives
( q ; q ) ∞ 2 χ ( q , w ) = ( Z ; Z ) ∞ 2 ( u v ; Z ) ∞ ( Z / ( u v ) ; Z ) ∞ ( u ; Z ) ∞ ( Z / u ; Z ) ∞ ( v ; Z ) ∞ ( Z / v ; Z ) ∞ . (392) (q;q)_\infty^2\chi(q,w)=\frac{(Z;Z)_\infty^2(uv;Z)_\infty(Z/(uv);Z)_\infty}{(u;Z)_\infty(Z/u;Z)_\infty(v;Z)_\infty(Z/v;Z)_\infty}. \tag*{(392)} ( q ; q ) ∞ 2 χ ( q , w ) = ( u ; Z ) ∞ ( Z / u ; Z ) ∞ ( v ; Z ) ∞ ( Z / v ; Z ) ∞ ( Z ; Z ) ∞ 2 ( uv ; Z ) ∞ ( Z / ( uv ) ; Z ) ∞ . (392)
As a result, we obtain
χ ( q , w ) = ( q 2 p ; q 2 p ) ∞ 2 ( q ; q 2 p ) ∞ ( q 2 p − 1 ; q 2 p ) ∞ ( q ; q ) ∞ 2 ∏ ϵ = ± 1 ( w ϵ q p − 1 2 ; q 2 p ) ∞ ( w ϵ q p + 1 2 ; q 2 p ) ∞ . (393) \chi(q,w)=\frac{(q^{2p};q^{2p})_\infty^2(q;q^{2p})_\infty(q^{2p-1};q^{2p})_\infty}{(q;q)_\infty^2\prod_{\epsilon=\pm1}(w^\epsilon q^{p-\frac{1}{2}};q^{2p})_\infty(w^\epsilon q^{p+\frac{1}{2}};q^{2p})_\infty}. \tag*{(393)} χ ( q , w ) = ( q ; q ) ∞ 2 ∏ ϵ = ± 1 ( w ϵ q p − 2 1 ; q 2 p ) ∞ ( w ϵ q p + 2 1 ; q 2 p ) ∞ ( q 2 p ; q 2 p ) ∞ 2 ( q ; q 2 p ) ∞ ( q 2 p − 1 ; q 2 p ) ∞ . (393)
Using
log ( t ; Z ) ∞ = − ∑ r ≥ 1 t r r ( 1 − Z r ) , (394) \log(t;Z)_\infty=-\sum_{r\geq1}\frac{t^r}{r(1-Z^r)}, \tag*{(394)} log ( t ; Z ) ∞ = − r ≥ 1 ∑ r ( 1 − Z r ) t r , (394)
we can write (C.8) as
χ ( q , w ) = PE [ q + q 2 − q 2 p − 1 − q 2 p + ( w + w − 1 ) ( q p − 1 2 − q p + 3 2 ) ( 1 − q ) ( 1 − q 2 p ) ] . (395) \chi(q,w)=\operatorname{PE}\left[\frac{q+q^2-q^{2p-1}-q^{2p}+(w+w^{-1})(q^{p-\frac{1}{2}}-q^{p+\frac{3}{2}})}{(1-q)(1-q^{2p})}\right]. \tag*{(395)} χ ( q , w ) = PE [ ( 1 − q ) ( 1 − q 2 p ) q + q 2 − q 2 p − 1 − q 2 p + ( w + w − 1 ) ( q p − 2 1 − q p + 2 3 ) ] . (395)