Introduction
The Verlinde formula expresses the fusion rules of a rational vertex operator algebra through the modular S S S -matrix [38 , 28 ] . This paper concerns two extensions.
Twisted modules. For a finite automorphism group G G G of a rational, C 2 C_2 C 2 -cofinite vertex operator algebra, Dong and Lin [12 ] proved, for G G G of prime order, a twisted Verlinde formula for fusion rules among g g g -twisted modules, in terms of the orbifold S S S -matrix and a sum over G G G -stable untwisted modules.
Superalgebras. For V = V ˉ 0 ⊕ V ˉ 1 V = \bar V_0 \oplus\bar V_1 V = V ˉ 0 ⊕ V ˉ 1 with regular even part, Creutzig, Kanade and McRae [8 ] realized V V V as a superalgebra object (an extension by an order two simple current J = V ˉ 1 J = \bar V_1 J = V ˉ 1 ) in the modular tensor category of V ˉ 0 \bar V_0 V ˉ 0 -modules and obtained Verlinde formulas for dimensions N + N^+ N + and superdimensions N − N^- N − of spaces of intertwining operators among untwisted and parity-twisted V V V -modules.
In this paper, we combine the two: we extend the even-part formula to cyclic groups of arbitrary order and derive the super case formula, including J J J -fixed points, for the combined situation.
Previous works
The relation between the S S S -matrix of V G V^G V G and the S S S -matrix of the orbifold theory was determined in [21 ] for finite groups and used in [12 ] for prime order. Categorically, Verlinde formulas for G G G -crossed modular fusion categories are due to Turaev [37 ] (see also [1 ] ) and Deshpande-Mukhopadhyay [10 ] . It was shown in [35 , 12 ] that the category of twisted V ˉ 0 \bar V_0 V ˉ 0 -modules is G G G -crossed modular. Theorem 2.11 below is the vertex algebraic realization of the G G G -crossed Verlinde formula for cyclic G G G .
Our Lemma 2.8 and Theorem 2.11 are derived from modular invariance of twisted traces [16 ] and the classification of V G V^G V G -modules [23 , 36 , 20 ] . For ∣ G ∣ |G| ∣ G ∣ prime, Lemma 2.8 (b) (with identical conjugation conventions, see Remark 2.10, Theorem 2.11, and the case distinction in part 3 of its proof) corresponds to [12 ] .
For composite order new proven topics are: the treatment of arbitrary stabilizers G X G_X G X in the orbifold relation, the restriction to stable columns by character orthogonality in the stabilizers, and the scope of the individual coefficient formula (see Remark 2.13). Lemma 2.8 is connected to the relations of [21 ] for general finite G G G . We give a self-contained derivation which keeps track of the stabilizers. For permutation orbifolds see [22 ] .
Main results of this paper
Let G G G be cyclic of order n n n , acting faithfully on V ‾ 0 \overline{V}_0 V 0 (assumptions of Subsection 2.1). Let Q \mathcal{Q} Q be the set of G G G -stable irreducible untwisted V ‾ 0 \overline{V}_0 V 0 -modules, S X , Q S_{X,Q} S X , Q the orbifold S S S -matrix (X X X twisted, Q ∈ Q Q \in\mathcal{Q} Q ∈ Q ), and d = dim q J = ± 1 d = \dim_q J = \pm1 d = dim q J = ± 1 . We then have the following results.
Even part, arbitrary cyclic order:
Theorem 2.11. For g 1 g_1 g 1 , g 2 g_2 g 2 generating G G G , g 3 = g 1 g 2 g_3 = g_1g_2 g 3 = g 1 g 2 and M i ∈ M ( g i ) M_i \in\mathcal{M}(g_i) M i ∈ M ( g i ) , the average of N N 1 , N 2 M 3 N^{M_3}_{N_1,N_2} N N 1 , N 2 M 3 over N i N_i N i in the G G G -orbits of M i M_i M i equals ∑ Q ∈ Q S M 1 , Q S M 2 , Q S M 3 , Q ‾ / S V ‾ 0 , Q \sum_{Q \in\mathcal{Q}} S_{M_1,Q}S_{M_2,Q}\overline{S_{M_3,Q}}/S_{\overline{V}_0,Q} ∑ Q ∈ Q S M 1 , Q S M 2 , Q S M 3 , Q / S V 0 , Q . If two of the three modules are G G G -stable (automatic for n n n a prime power, or if all modules are stable) this is the individual fusion coefficient.
Proposition 2.14: for g g g generating G G G , the block ( S X , Q ) X ∈ M ( g ) , Q ∈ Q (S_{X,Q})_{X \in\mathcal{M}(g),Q \in\mathcal{Q}} ( S X , Q ) X ∈ M ( g ) , Q ∈ Q is square and unitary.
A consequence of Theorem 2.11 and the J J J -equivariance of Section 3:
Theorem 4.1 (a). For X i X_i X i as above, with N ± : = N X 1 , X 2 X 3 ± N X 1 , X 2 J ⊠ X 3 N^{\pm} := N^{X_3}_{X_1,X_2} \pm N^{J\boxtimes X_3}_{X_1,X_2} N ± := N X 1 , X 2 X 3 ± N X 1 , X 2 J ⊠ X 3 and π ( X ) ∈ Z / 2 Z \pi(X) \in\mathbb{Z}/2\mathbb{Z} π ( X ) ∈ Z /2 Z the parity type (see Definition 3.3),
N ± = δ π ( X 1 ) + π ( X 2 ) , π ( X 3 ) ( ∑ [ Q ] f r e e η Q = ± 1 f 0 ( Q ) + 2 ∑ Q f i x e d η Q = ± 1 f 0 ( Q ) ) , f 0 ( Q ) = S X 1 , Q S X 2 , Q S X 3 , Q ‾ S V ‾ 0 , Q , N^{\pm} = \delta_{\pi(X_1)+\pi(X_2),\pi(X_3)}\left( \sum_{\substack{[Q]\ \mathrm{free}\\ \eta_Q=\pm1}} f_0(Q) + 2\sum_{\substack{Q\ \mathrm{fixed}\\ \eta_Q=\pm1}} f_0(Q) \right), \qquad f_0(Q) = \frac{S_{X_1,Q}S_{X_2,Q}\overline{S_{X_3,Q}}}{S_{\overline{V}_0,Q}}, N ± = δ π ( X 1 ) + π ( X 2 ) , π ( X 3 ) [ Q ] free η Q = ± 1 ∑ f 0 ( Q ) + 2 Q fixed η Q = ± 1 ∑ f 0 ( Q ) , f 0 ( Q ) = S V 0 , Q S X 1 , Q S X 2 , Q S X 3 , Q ,
where the equality of parity types is in Z / 2 Z \mathbb{Z}/2\mathbb{Z} Z /2 Z (equivalently ϵ X 1 ϵ X 2 ϵ X 3 = 1 \epsilon_{X_1}\epsilon_{X_2}\epsilon_{X_3}=1 ϵ X 1 ϵ X 2 ϵ X 3 = 1 ). Here [ Q ] [Q] [ Q ] runs over J J J -orbits of size two in Q \mathcal{Q} Q , where “fixed” means J ⊠ Q ≃ Q J\boxtimes Q \simeq Q J ⊠ Q ≃ Q . Fixed points exist only for d = 1 d=1 d = 1 and have η Q = − 1 \eta_Q=-1 η Q = − 1 (see Lemma 3.9). For g X g_X g X generating G G G , X X X is J J J -fixed iff S X , Q = 0 S_{X,Q}=0 S X , Q = 0 whenever η Q = − 1 \eta_Q=-1 η Q = − 1 (Lemma 3.9 (iii)).
Dimension sum rule:
Proposition 5.3 and parity selection rule: Corollary 5.2 , and Lemma 2.1 : V G V^G V G is simple whenever V V V is.
Theorem 4.1 (b), Corollary 4.2: Under conditions 2.5.1, N + N^+ N + and N − N^- N − are the dimension and superdimension of the space of intertwining operators among the corresponding g g g -twisted and g P gP g P -twisted V V V -modules. The fusion-product statement is based in addition on conditions 2.5.2.
Theorems 2.11 and 4.1 (a), Proposition 2.14 and the sum rule use only the assumptions on V ‾ 0 \overline{V}_0 V 0 and U = V G U=V^G U = V G and on modular invariance of twisted traces. The statements that concern twisted V V V -modules (Theorem 4.1 (b), Corollary 4.2, Proposition 6.1 (iii)) use conditions 2.5.1-2.5.3. For inner automorphisms they reduce to the untwisted theory [8 , 30 ] provided Li Δ \Delta Δ -operator results [33 , 14 ] hold for superalgebras, including the fusion of two twisted modules (conditions 2.5.3).
The examples:
The numerical evidence is given in Section 7. The algebra of the proofs and the stated conventions are tested in all examples. The stable-column formula of Lemma 2.8 (b) is tested independently of S U S^U S U in one family (Example 7.1, from twisted traces). The stabilizer dependent parts are tested only for n = 2 n=2 n = 2 (Example 7.2). The composite order example with a proper stabilizer (Example 7.6) is a product and tests only the statements. Section 8 lists the remaining open problems.
Preliminaries
Vertex operator superalgebras
Let V = V ‾ 0 ⊕ J V=\overline{V}_0\oplus J V = V 0 ⊕ J be a vertex operator superalgebra with parity automorphism P P P . Automorphisms preserve the grading. Ideals are Z / 2 Z \mathbb{Z}/2\mathbb{Z} Z /2 Z -graded, and “simple” means that there are no non-trivial graded ideals. Twisted modules are as in [15 , 40 ] with the convention of [16 ] for g ↔ g − 1 g\leftrightarrow g^{-1} g ↔ g − 1 . g g g -twisted V V V -modules are called parity-untwisted , g P gP g P -twisted ones are called parity-twisted . In this paper we adopt the following two assumptions (in what follows we call them the first/second assumptions of Subsection 2.1):
First: V V V is a simple, 1 2 Z \frac{1}{2}\mathbb{Z} 2 1 Z -graded vertex operator superalgebra of CFT type. V ‾ 0 \overline{V}_0 V 0 is simple, rational, C 2 C_2 C 2 -cofinite, self-dual, of CFT type.
Second: G = ⟨ g ⟩ ⊂ Aut ( V ) G=\langle g\rangle\subset\operatorname{Aut}(V) G = ⟨ g ⟩ ⊂ Aut ( V ) is cyclic of order n ≥ 2 n\geq2 n ≥ 2 and acts faithfully on V ‾ 0 \overline{V}_0 V 0 .
Set U : = V 0 G U:=V^G_0 U := V 0 G , J : = V 1 J:=V_1 J := V 1 , J 0 : = V 1 G J_0:=V^G_1 J 0 := V 1 G . We have the following consequences of two assumptions above:
V ‾ 0 \overline{V}_0 V 0 is g g g -rational for every finite order automorphism g g g (with the assumption that V ‾ 0 \overline{V}_0 V 0 is rational and C 2 C_2 C 2 -cofinite, which is the first assumption above) [12 ] .
For a solvable (here cyclic) subgroup G ⊂ Aut ( V ‾ 0 ) G\subset\operatorname{Aut}(\overline{V}_0) G ⊂ Aut ( V 0 ) the fixed-point algebra U = V G U=V^G U = V G is again simple, rational, C 2 C_2 C 2 -cofinite, self-dual and of CFT type (with assumption that V ‾ 0 \overline{V}_0 V 0 is simple, rational, C 2 C_2 C 2 -cofinite, self-dual, of CFT type, i.e., the first assumption) applied to the vertex operator algebra V ‾ 0 \overline{V}_0 V 0 ), [12 , 7 , 34 ] .
For W ∈ { V ‾ 0 , U } W \in\{\overline{V}_0,U\} W ∈ { V 0 , U } the category C W \mathcal{C}_W C W is modular, the S S S -matrix is symmetric and unitary, and the Verlinde formula holds (with assumptions as above), [12 , 28 , 29 , 17 ] .
These results are statements about vertex operator algebras. The superalgebra V V V enters through the V ‾ 0 \overline{V}_0 V 0 -module J J J and the superalgebra object structure of [8 ] (see Proposition 2.3). The following lemma describes simplicity of V G V^G V G .
Lemma 2.1. Let V V V be a simple vertex operator superalgebra and G G G a finite group of automorphisms. Then V G = U ⊕ J 0 V^G=U\oplus J_0 V G = U ⊕ J 0 is a simple vertex operator superalgebra.
Proof. 1) Generation: Let M M M be a module for a vertex superalgebra V V V , c ∈ M c\in M c ∈ M homogeneous. We show that S : = span { u n c : u ∈ V , n ∈ Z } S:=\operatorname{span}\{u_n c:u\in V,\ n\in\mathbb{Z}\} S := span { u n c : u ∈ V , n ∈ Z } is a submodule. Let a , b ∈ V a,b\in V a , b ∈ V be homogeneous, ε = ( − 1 ) ∣ a ∣ ∣ b ∣ \varepsilon=(-1)^{\lvert a\rvert\lvert b\rvert} ε = ( − 1 ) ∣ a ∣ ∣ b ∣ . The Borcherds identity gives, for all m , n , k ∈ Z m,n,k\in\mathbb{Z} m , n , k ∈ Z ,
∑ j ≥ 0 ( m j ) ( a n + j b ) m + k − j c = ∑ j ≥ 0 ( − 1 ) j ( n j ) [ a m + n − j b k + j c − ( − 1 ) n ε b n + k − j a m + j c ] . (1) \sum_{j\geq0}\binom{m}{j}(a_{n+j}b)_{m+k-j}c=\sum_{j\geq0}(-1)^j\binom{n}{j}\left[a_{m+n-j}b_{k+j}c-(-1)^n\varepsilon b_{n+k-j}a_{m+j}c\right]. \tag*{(1)} j ≥ 0 ∑ ( j m ) ( a n + j b ) m + k − j c = j ≥ 0 ∑ ( − 1 ) j ( j n ) [ a m + n − j b k + j c − ( − 1 ) n ε b n + k − j a m + j c ] . (1)
where signs appear through ε \varepsilon ε . By the truncation axioms there are N N N with a i c = 0 a_i c=0 a i c = 0 for all i ≥ N i\geq N i ≥ N , and N ′ N' N ′ with b i c = 0 b_i c=0 b i c = 0 for all i ≥ N ′ i\geq N' i ≥ N ′ . Also, a n + j b = 0 a_{n+j}b=0 a n + j b = 0 for j ≫ 0 j\gg0 j ≫ 0 . Take m ≥ N m\geq N m ≥ N . Then the second sum on the right of (2.1) vanishes, and the left side belongs to S S S . Hence, for all m ≥ N m\geq N m ≥ N and all n , k n,k n , k :
∑ j = 0 J ( − 1 ) j ( n j ) a m + n − j b k + j c ∈ S , J = N ′ − k − 1 \sum_{j=0}^{J}(-1)^j\binom{n}{j}a_{m+n-j}b_{k+j}c\in S,
\qquad J=N'-k-1 j = 0 ∑ J ( − 1 ) j ( j n ) a m + n − j b k + j c ∈ S , J = N ′ − k − 1
(the terms with j > J j>J j > J vanish). Fix r ∈ Z r\in\mathbb{Z} r ∈ Z and k k k , and set n = r − m n=r-m n = r − m :
∑ j = 0 J p j ( m ) a r − j b k + j c ∈ S for all m ≥ N , p j ( m ) : = ( − 1 ) j ( r − m j ) . \sum_{j=0}^{J}p_j(m)a_{r-j}b_{k+j}c\in S\quad\text{for all }m\geq N,\qquad p_j(m):=(-1)^j\binom{r-m}{j}. j = 0 ∑ J p j ( m ) a r − j b k + j c ∈ S for all m ≥ N , p j ( m ) := ( − 1 ) j ( j r − m ) .
Each p j p_j p j is a polynomial in m m m of degree j j j , thus the ( J + 1 ) × ( J + 1 ) (J+1)\times(J+1) ( J + 1 ) × ( J + 1 ) matrix ( p j ( N + i ) ) 0 ≤ i , j ≤ J (p_j(N+i))_{0\leq i,j\leq J} ( p j ( N + i ) ) 0 ≤ i , j ≤ J is invertible. Hence, every vector a r − j b k + j c a_{r-j}b_{k+j}c a r − j b k + j c belongs to S S S . In particular, for j = 0 j=0 j = 0 , a r b k c ∈ S a_r b_k c\in S a r b k c ∈ S for all r , k r,k r , k . Thus, S S S is stable under all modes of all homogeneous elements. It contains c = 1 − 1 c c=1_{-1}c c = 1 − 1 c . This proves 1).
We show that I ′ I' I ′ is a nonzero graded ideal of V V V : let 0 ≠ I ⊂ V G 0\ne I\subset V^G 0 = I ⊂ V G be a graded ideal and I ′ : = span { u n i : u ∈ V , n ∈ Z , i ∈ I } I':=\operatorname{span}\{u_n i:u\in V,\ n\in\mathbb{Z},\ i\in I\} I ′ := span { u n i : u ∈ V , n ∈ Z , i ∈ I } . Applying 1) to M = V M=V M = V as adjoint module, I ′ = ∑ i ∈ I span { u n i } I'=\sum_{i\in I}\operatorname{span}\{u_n i\} I ′ = ∑ i ∈ I span { u n i } is a sum of left ideals. Hence, it is stable under action all v m v_m v m , v ∈ V v\in V v ∈ V . It is spanned by homogeneous elements, hence it is graded. It contains I I I (u = 1 u=1 u = 1 ), hence is nonzero. It is stable under action of D = L ( − 1 ) = ω 0 D=L(-1)=\omega_0 D = L ( − 1 ) = ω 0 which is a mode of ω ∈ V \omega\in V ω ∈ V . By skew-symmetry,
i n ′ v = ± ∑ k ≥ 0 ( − 1 ) n + 1 + k k ! D k ( v n + k i ′ ) i'_n v=\pm\sum_{k\geq0}\frac{(-1)^{n+1+k}}{k!}D^k(v_{n+k}i') i n ′ v = ± k ≥ 0 ∑ k ! ( − 1 ) n + 1 + k D k ( v n + k i ′ )
for i ′ ∈ I ′ i'\in I' i ′ ∈ I ′ , v ∈ V v\in V v ∈ V , which belongs to I ′ I' I ′ since I ′ I' I ′ is stable under v m v_m v m and D D D . Thus, I ′ I' I ′ is a two-sided graded ideal. By simplicity of V V V , I ′ = V I'=V I ′ = V .
Averaging and both inclusions: let av = ∣ G ∣ − 1 ∑ g ∈ G g \operatorname{av}=\lvert G\rvert^{-1}\sum_{g\in G}g av = ∣ G ∣ − 1 ∑ g ∈ G g . For u ∈ V u\in V u ∈ V , i ∈ I ⊂ V G i\in I\subset V^G i ∈ I ⊂ V G , g ∈ G g\in G g ∈ G : g ( u n i ) = ( g u ) n ( g i ) = ( g u ) n i g(u_n i)=(gu)_n(gi)=(gu)_n i g ( u n i ) = ( g u ) n ( g i ) = ( g u ) n i . Hence, av ( u n i ) = av ( u ) n i ∈ I \operatorname{av}(u_n i)=\operatorname{av}(u)_n i\in I av ( u n i ) = av ( u ) n i ∈ I , since av ( u ) ∈ V G \operatorname{av}(u)\in V^G av ( u ) ∈ V G and I I I is an ideal of V G V^G V G . Thus, av ( I ′ ) ⊂ I \operatorname{av}(I')\subset I av ( I ′ ) ⊂ I . Since I ′ = V I'=V I ′ = V , V G = av ( V ) = av ( I ′ ) ⊂ I V^G=\operatorname{av}(V)=\operatorname{av}(I')\subset I V G = av ( V ) = av ( I ′ ) ⊂ I . Together with I ⊂ V G I\subset V^G I ⊂ V G this gives I = V G I=V^G I = V G .
Lemma 2.2. Under the second assumption of Subsection 2.1, G ∩ ⟨ P ⟩ = { 1 } G\cap\langle P\rangle=\{1\} G ∩ ⟨ P ⟩ = { 1 } . Hence, G × ⟨ P ⟩ ≅ Z / n × Z / 2 G\times\langle P\rangle\cong\mathbb{Z}/n\times\mathbb{Z}/2 G × ⟨ P ⟩ ≅ Z / n × Z /2 acts faithfully on V V V and the sectors g p g_p g p (with g ∈ G g\in G g ∈ G ) are all distinct.
Proof. P P P acts trivially on V ‾ 0 \overline{V}_0 V 0 and G → Aut ( V ‾ 0 ) G\to\operatorname{Aut}(\overline{V}_0) G → Aut ( V 0 ) is injective, thus P ∈ G P\in G P ∈ G would lead to P = 1 P=1 P = 1 .
Note that J J J is a G G G -stable V ‾ 0 \overline{V}_0 V 0 -module (( σ ∣ J : J → J ) ∘ σ (\sigma|_J:J\to J)\circ\sigma ( σ ∣ J : J → J ) ∘ σ is an isomorphism).
Let us adopt the following conventions. In the braided category C V ‾ 0 \mathcal{C}_{\overline{V}_0} C V 0 of [29 ] the ribbon twist of a simple module A A A is θ A = e 2 π i h A \theta_A=e^{2\pi i h_A} θ A = e 2 π i h A (where h A h_A h A the conformal weight), dim q \dim_q dim q is the ribbon-categorical dimension, and c c c is the braiding. For an invertible object J J J with J ⊗ J ≅ 1 J\otimes J\cong1 J ⊗ J ≅ 1 , c J , J ∈ C c_{J,J}\in\mathbb{C} c J , J ∈ C is a scalar, and the ribbon relation (curl) gives θ J = c J , J dim q J \theta_J=c_{J,J}\dim_q J θ J = c J , J dim q J [2 ] , since θ J ⊗ J = θ 1 = 1 = θ J 2 c J , J 2 \theta_{J\otimes J}=\theta_1=1=\theta_J^2c_{J,J}^2 θ J ⊗ J = θ 1 = 1 = θ J 2 c J , J 2 and dim q J = ± 1 \dim_q J=\pm1 dim q J = ± 1 , c J , J = ± 1 c_{J,J}=\pm1 c J , J = ± 1 . Then we recall the following proposition.
Proposition 2.3.
(i) J J J is a simple current of order two for V ‾ 0 \overline{V}_0 V 0 . Since V G = U ⊕ J 0 V^G=U\oplus J_0 V G = U ⊕ J 0 is a simple vertex operator superalgebra (Lemma 2.1) with regular even part U U U , J 0 J_0 J 0 is a simple current of order two for U U U [8 ] .
(ii) θ J = dim q J ∈ { ± 1 } \theta_J=\dim_q J\in\{\pm1\} θ J = dim q J ∈ { ± 1 } .
(iii) for a simple object A A A of C V ‾ 0 \mathcal{C}_{\overline{V}_0} C V 0 or C U \mathcal{C}_U C U , dim q A = S A , 0 / S 0 , 0 \dim_q A=S_{A,0}/S_{0,0} dim q A = S A , 0 / S 0 , 0 [29 , 11 ] .
For (ii), indeed, V = V ˉ 0 ⊕ J V=\bar{V}_{0}\oplus J V = V ˉ 0 ⊕ J is a superalgebra object of C V ˉ 0 \mathcal{C}_{\bar{V}_{0}} C V ˉ 0 , [8 ] . The multiplication μ : J ⊗ J → V ˉ 0 \mu:J\otimes J\to\bar{V}_{0} μ : J ⊗ J → V ˉ 0 (nonzero, since J ⊗ J ≃ V ˉ 0 J\otimes J\simeq\bar{V}_{0} J ⊗ J ≃ V ˉ 0 is simple and V V V is simple) satisfies μ ∘ c J , J = − μ \mu\circ c_{J,J}=-\mu μ ∘ c J , J = − μ (supercommutativity on the odd part corresponding to the sign ( − 1 ) ∣ u ∣ ∣ v ∣ (-1)^{\lvert u\rvert\lvert v\rvert} ( − 1 ) ∣ u ∣ ∣ v ∣ in skew-symmetry). Hence, c J , J = − 1 c_{J,J}=-1 c J , J = − 1 and θ J = − dim q J \theta_{J}=-\dim_{q}J θ J = − dim q J . In particular, M J , J = id M_{J,J}=\operatorname{id} M J , J = id . Note that (iii) is the identification between categorical dimensions and modular-transformation data. Below d : = S J , V ˉ 0 / S V ˉ 0 , V ˉ 0 d:=S_{J,\bar{V}_{0}}/S_{\bar{V}_{0},\bar{V}_{0}} d := S J , V ˉ 0 / S V ˉ 0 , V ˉ 0 is defined through S S S , and (iii) identifies it with dim q J \dim_{q}J dim q J . It is also used in (ii) to relate d d d to θ J \theta_{J} θ J .
The categories involved
C V ˉ 0 \mathcal{C}_{\bar{V}_{0}} C V ˉ 0 , C U \mathcal{C}_{U} C U are modular tensor categories of ordinary modules [29 ] . S U S^{U} S U is the (symmetric, unitary) S S S -matrix of C U \mathcal{C}_{U} C U , defined through modular transformation of the trace functions Z Y ( v , τ ) Z_{Y}(v,\tau) Z Y ( v , τ ) , v ∈ U v\in U v ∈ U . Rep ( V ˉ 0 ) \operatorname{Rep}(\bar{V}_{0}) Rep ( V ˉ 0 ) are modules of the commutative algebra object V ˉ 0 ∈ C U \bar{V}_{0}\in\mathcal{C}_{U} V ˉ 0 ∈ C U [12 ] . Its simple objects are the irreducible g g g -twisted V ˉ 0 \bar{V}_{0} V ˉ 0 -modules, g ∈ G g\in G g ∈ G . F 0 F_{0} F 0 is induction F 0 = V ˉ 0 ⊠ U ( ⋅ ) : C U → Rep ( V ˉ 0 ) F_{0}=\bar{V}_{0}\boxtimes_{U}(\cdot):\mathcal{C}_{U}\to\operatorname{Rep}(\bar{V}_{0}) F 0 = V ˉ 0 ⊠ U ( ⋅ ) : C U → Rep ( V ˉ 0 ) , left adjoint to the forgetful functor. Rep ( V ) \operatorname{Rep}(V) Rep ( V ) , F V F_{V} F V are Z / 2 Z \mathbb{Z}/2\mathbb{Z} Z /2 Z -graded modules of the superalgebra A = V ˉ 0 ⊕ J A=\bar{V}_{0}\oplus J A = V ˉ 0 ⊕ J in Rep ( V ˉ 0 ) \operatorname{Rep}(\bar{V}_{0}) Rep ( V ˉ 0 ) and induction F V ( M ) = M ⊕ J ⊠ M F_{V}(M)=M\oplus J\boxtimes M F V ( M ) = M ⊕ J ⊠ M . Note that it is used under conditions 2.5.1.
Next, we recall from [12 ] (see also [31 , 18 ] )
Proposition 2.4. Assume that V ˉ 0 \bar{V}_{0} V ˉ 0 simple, rational, C 2 C_{2} C 2 -cofinite, self-dual, of CFT type, and G G G finite abelian group of automorphisms. Then:
(a) V ˉ 0 \bar{V}_{0} V ˉ 0 is a commutative algebra in C U \mathcal{C}_{U} C U . Rep ( V ˉ 0 ) \operatorname{Rep}(\bar{V}_{0}) Rep ( V ˉ 0 ) is a tensor category (a fusion category, in particular semisimple with finitely many simple objects). Every irreducible g g g -twisted V ˉ 0 \bar{V}_{0} V ˉ 0 -module (g ∈ G g\in G g ∈ G ) is an object of Rep ( V ˉ 0 ) \operatorname{Rep}(\bar{V}_{0}) Rep ( V ˉ 0 ) , two such are equivalent objects iff they have the same g g g and are isomorphic as twisted modules, and every simple object of Rep ( V ˉ 0 ) \operatorname{Rep}(\bar{V}_{0}) Rep ( V ˉ 0 ) is an irreducible g g g -twisted V ˉ 0 \bar{V}_{0} V ˉ 0 -module for some g g g .
(b) the multiplicity N M 1 , M 2 M 3 N_{M_{1},M_{2}}^{M_{3}} N M 1 , M 2 M 3 of M 3 M_{3} M 3 in M 1 ⊠ M 2 M_{1}\boxtimes M_{2} M 1 ⊠ M 2 equals dim I ( M 3 ; M 1 , M 2 ) \dim I(M_{3};M_{1},M_{2}) dim I ( M 3 ; M 1 , M 2 ) , the dimension of the space of intertwining operators of type ( M 3 M 1 M 2 ) \binom{M_{3}}{M_{1}\ M_{2}} ( M 1 M 2 M 3 ) among twisted modules.
(c) N M 1 , M 2 M 3 = 0 N_{M_{1},M_{2}}^{M_{3}}=0 N M 1 , M 2 M 3 = 0 unless g 3 = g 1 g 2 g_{3}=g_{1}g_{2} g 3 = g 1 g 2 . Hence, Rep ( V ˉ 0 ) = ⨁ g Rep ( V ˉ 0 ) g \operatorname{Rep}(\bar{V}_{0})=\bigoplus_{g}\operatorname{Rep}(\bar{V}_{0})_{g} Rep ( V ˉ 0 ) = ⨁ g Rep ( V ˉ 0 ) g is G G G -graded.
(d) N M 1 ∘ h , M 2 ∘ h M 3 ∘ h = N M 1 , M 2 M 3 N_{M_{1}\circ h,M_{2}\circ h}^{M_{3}\circ h}=N_{M_{1},M_{2}}^{M_{3}} N M 1 ∘ h , M 2 ∘ h M 3 ∘ h = N M 1 , M 2 M 3 for h ∈ G h\in G h ∈ G . Consequently, comparing multiplicities in the semisimple category Rep ( V ˉ 0 ) \operatorname{Rep}(\bar{V}_{0}) Rep ( V ˉ 0 ) , one obtains ( M 1 ⊠ M 2 ) ∘ h ≃ ( M 1 ∘ h ) ⊠ ( M 2 ∘ h ) (M_{1}\boxtimes M_{2})\circ h\simeq(M_{1}\circ h)\boxtimes(M_{2}\circ h) ( M 1 ⊠ M 2 ) ∘ h ≃ ( M 1 ∘ h ) ⊠ ( M 2 ∘ h ) as objects.
(e) the induction functor F 0 : C U → Rep ( V ˉ 0 ) F_{0}:\mathcal{C}_{U}\to\operatorname{Rep}(\bar{V}_{0}) F 0 : C U → Rep ( V ˉ 0 ) , F 0 ( Y ) = V ˉ 0 ⊠ U Y F_{0}(Y)=\bar{V}_{0}\boxtimes_{U}Y F 0 ( Y ) = V ˉ 0 ⊠ U Y , is exact and a tensor functor, and Hom Rep ( V ˉ 0 ) ( F 0 ( Y ) , N ) = Hom C U ( Y , N ) \operatorname{Hom}_{\operatorname{Rep}(\bar{V}_{0})}(F_{0}(Y),N)=\operatorname{Hom}_{\mathcal{C}_{U}}(Y,N) Hom Rep ( V ˉ 0 ) ( F 0 ( Y ) , N ) = Hom C U ( Y , N ) for N ∈ Rep ( V ˉ 0 ) N\in\operatorname{Rep}(\bar{V}_{0}) N ∈ Rep ( V ˉ 0 ) regarded as a U U U -module.
Note that (d) is stated and proved in [12 ] for arbitrary g 1 g_{1} g 1 , g 2 g_{2} g 2 , g 3 g_{3} g 3 , h ∈ G h\in G h ∈ G . The proof is the application of the Jacobi identity and does not use that ∣ G ∣ \lvert G\rvert ∣ G ∣ is prime.
Definition 2.5. Fusion coefficients N M 1 , M 2 M 3 N_{M_{1},M_{2}}^{M_{3}} N M 1 , M 2 M 3 are defined as dim I ( M 3 ; M 1 , M 2 ) \dim I(M_{3};M_{1},M_{2}) dim I ( M 3 ; M 1 , M 2 ) , equivalently, by (b), multiplicities in Rep ( V ˉ 0 ) \operatorname{Rep}(\bar{V}_{0}) Rep ( V ˉ 0 ) . The tensor product ⊠ \boxtimes ⊠ of Rep ( V ˉ 0 ) \operatorname{Rep}(\bar{V}_{0}) Rep ( V ˉ 0 ) is used through (a), (e) and for the invertible object J J J . A tensor product ⊠ V \boxtimes_{V} ⊠ V of twisted V V V -modules assumes 2.5.1.
Twisted modules, stabilizers and U-modules
Let M ( g ) \mathcal{M}(g) M ( g ) be the irreducible g g g -twisted V ˉ 0 \bar{V}_{0} V ˉ 0 -modules and R = ⨆ g ∈ G M ( g ) \mathcal{R}=\bigsqcup_{g\in G}\mathcal{M}(g) R = ⨆ g ∈ G M ( g ) . For X ∈ M ( g ) X\in\mathcal{M}(g) X ∈ M ( g ) let G X = { h ∈ G : X ∘ h ≃ X } G_{X}=\{h\in G:X\circ h\simeq X\} G X = { h ∈ G : X ∘ h ≃ X } (cyclic, and g ∈ G X g\in G_{X} g ∈ G X by [12 ] ) and O X = { X ∘ a } O_{X}=\{X\circ a\} O X = { X ∘ a } its G G G -orbit, ∣ O X ∣ = n / ∣ G X ∣ \lvert O_{X}\rvert=n/\lvert G_{X}\rvert ∣ O X ∣ = n / ∣ G X ∣ . X X X is stable if G X = G G_{X}=G G X = G . Set Q : = { Q ∈ M ( 1 ) : G Q = G } \mathcal{Q}:=\{Q\in\mathcal{M}(1):G_{Q}=G\} Q := { Q ∈ M ( 1 ) : G Q = G } .
Choose a representation φ X \varphi_{X} φ X of G X G_{X} G X on X X X with φ X ( h ) Y X ( u , z ) φ X ( h ) − 1 = Y X ( h u , z ) \varphi_{X}(h)Y_{X}(u,z)\varphi_{X}(h)^{-1}=Y_{X}(hu,z) φ X ( h ) Y X ( u , z ) φ X ( h ) − 1 = Y X ( h u , z ) . It is a projective representation of cyclic groups lift, φ X \varphi_{X} φ X is unique up to G X G_{X} G X . Let ψ s ( σ ) = e − 2 π i s / n ψ s \psi_{s}(\sigma)=e^{-2\pi is/n}\psi_{s} ψ s ( σ ) = e − 2 π i s / n ψ s , thus G ^ = { ψ s } \widehat{G}=\{\psi_{s}\} G = { ψ s } , and set X = ⨁ χ ∈ G ^ X X χ X=\bigoplus_{\chi\in\widehat{G}_{X}}X_{\chi} X = ⨁ χ ∈ G X X χ . For stable X X X set X s = X ψ s X_{s}=X_{\psi_{s}} X s = X ψ s . For V ˉ 0 \bar{V}_{0} V ˉ 0 we normalize φ \varphi φ as the action of σ \sigma σ . Then ( V ˉ 0 ) 0 = U (\bar{V}_{0})_{0}=U ( V ˉ 0 ) 0 = U and J 0 = V 1 G J_{0}=V_{1}^{G} J 0 = V 1 G . Recall the following proposition [20 , 23 , 36 ] (and [12 ] for n n n prime).
Proposition 2.6. The statements below hold for every g ∈ G g\in G g ∈ G and every X ∈ M ( g ) X\in\mathcal{M}(g) X ∈ M ( g ) with the stabilizer G X G_{X} G X :
(i) each X χ X_{\chi} X χ , χ ∈ G ^ X \chi\in\widehat{G}_{X} χ ∈ G X , is a nonzero irreducible U U U -module, and X χ ≄ X χ ′ X_{\chi}\not\simeq X_{\chi'} X χ ≃ X χ ′ for χ ≠ χ ′ \chi\ne\chi' χ = χ ′ .
(ii) for X ∘ a ∈ O X X\circ a\in O_{X} X ∘ a ∈ O X one has ( X ∘ a ) χ = X χ (X\circ a)_{\chi}=X_{\chi} ( X ∘ a ) χ = X χ , with the convention φ X ∘ a = φ X \varphi_{X\circ a}=\varphi_{X} φ X ∘ a = φ X .
(iii) X χ ≃ Y ψ X_{\chi}\simeq Y_{\psi} X χ ≃ Y ψ iff Y ∈ O X Y\in O_{X} Y ∈ O X and ψ = χ \psi=\chi ψ = χ . For Y ∈ O X Y\in O_{X} Y ∈ O X the groups G Y = G X G_{Y}=G_{X} G Y = G X coincide and χ , ψ \chi,\psi χ , ψ are compared as characters of this group.
(iv) every irreducible U U U -module is of the form X χ X_{\chi} X χ .
In [12 ] these statements are quoted for ∣ G ∣ \lvert G\rvert ∣ G ∣ prime (Theorems 5.2-5.3, where G X ∈ { 1 , G } G_{X}\in\{1,G\} G X ∈ { 1 , G } ). In [23 , 36 , 20 ] treat finite groups with a projective representation of the stabilizer on X X X in general. For a cyclic stabilizer G X G_X G X every projective representation lifts to a linear one. H 2 H^2 H 2 of a cyclic group with values in C × \mathbb{C}^{\times} C × vanishes. Thus φ X \varphi_X φ X may be taken to be a representation, unique up to G ^ X \widehat{G}_X G X . We use (i)–(iv) for cyclic G G G of arbitrary order in this form (see also [20 , 36 ] ).
Concerning the convention in (ii): X ∘ a X \circ a X ∘ a has the same underlying space and Y X ∘ a ( v , z ) = Y X ( a v , z ) Y_{X\circ a}(v,z)=Y_X(av,z) Y X ∘ a ( v , z ) = Y X ( a v , z ) . For v ∈ U v\in U v ∈ U , Y X ∘ a ( v , z ) = Y X ( v , z ) Y_{X\circ a}(v,z)=Y_X(v,z) Y X ∘ a ( v , z ) = Y X ( v , z ) , thus the U U U -actions agree. For k ∈ G X k\in G_X k ∈ G X , φ X ( k ) Y X ∘ a ( v , z ) φ X ( k ) − 1 = Y X ( k a v , z ) = Y X ( a k v , z ) = Y X ∘ a ( k v , z ) \varphi_X(k)Y_{X\circ a}(v,z)\varphi_X(k)^{-1}=Y_X(kav,z)=Y_X(akv,z)=Y_{X\circ a}(kv,z) φ X ( k ) Y X ∘ a ( v , z ) φ X ( k ) − 1 = Y X ( k a v , z ) = Y X ( ak v , z ) = Y X ∘ a ( k v , z ) for abelian G G G . Thus, φ X ( k ) \varphi_X(k) φ X ( k ) satisfies the defining relation of a representation on X ∘ a X\circ a X ∘ a with the same stabilizer G X ∘ a = G X G_{X\circ a}=G_X G X ∘ a = G X . Hence, the eigenspaces X χ X_\chi X χ are the same subspaces with the same U U U -action, and the labels χ \chi χ are preserved. By (i) and (iv), a U U U -module Y Y Y determines ( O X , χ ) (\mathcal{O}_X,\chi) ( O X , χ ) uniquely by (iii). In each orbit there is one irreducible U U U -module for each χ ∈ G ^ X \chi\in\widehat{G}_X χ ∈ G X .
Lemma 2.7. F 0 ( X χ ) = ⨁ N ∈ O X N F_0(X_\chi)=\bigoplus_{N\in\mathcal{O}_X}N F 0 ( X χ ) = ⨁ N ∈ O X N , each with multiplicity one.
Proof. For a simple N ∈ Rep ( V ‾ 0 ) N\in\operatorname{Rep}(\overline{V}_0) N ∈ Rep ( V 0 ) (an irreducible twisted module, in any sector h ∈ G h\in G h ∈ G ) Proposition 2.4 (e) gives Hom ( F 0 ( X χ ) , N ) = Hom U ( X χ , ⨁ ψ N ψ ) \operatorname{Hom}(F_0(X_\chi),N)=\operatorname{Hom}_U(X_\chi,\bigoplus_\psi N_\psi) Hom ( F 0 ( X χ ) , N ) = Hom U ( X χ , ⨁ ψ N ψ ) . By Proposition 2.6 (i)–(iii) it has dimension one if N ∈ O X N\in\mathcal{O}_X N ∈ O X and 0 otherwise. Then we use semisimplicity.
The orbifold S S S -matrix
We then use the following notations [12 ] . For commuting g , h ∈ G g,h\in G g , h ∈ G , let M ( g , h ) = { M ∈ M ( g ) : M ∘ h ≃ M } \mathcal{M}(g,h)=\{M\in\mathcal{M}(g):M\circ h\simeq M\} M ( g , h ) = { M ∈ M ( g ) : M ∘ h ≃ M } , a set of isomorphism classes of individual modules. For M ∈ M ( g , h ) M\in\mathcal{M}(g,h) M ∈ M ( g , h ) and v ∈ V ‾ 0 v\in\overline{V}_0 v ∈ V 0 set Z M ( v , ( g , h ) , τ ) = tr M o ( v ) φ M ( h ) q L ( 0 ) − c / 24 Z_M(v,(g,h),\tau)=\operatorname{tr}_M o(v)\varphi_M(h)q^{L(0)-c/24} Z M ( v , ( g , h ) , τ ) = tr M o ( v ) φ M ( h ) q L ( 0 ) − c /24 . By [16 ] , the functions Z M ( v , ( g , h ) , τ ) Z_M(v,(g,h),\tau) Z M ( v , ( g , h ) , τ ) (v ∈ V ‾ 0 v\in\overline{V}_0 v ∈ V 0 ), for all commuting pairs ( g , h ) (g,h) ( g , h ) and all M ∈ M ( g , h ) M\in\mathcal{M}(g,h) M ∈ M ( g , h ) , are linearly independent, and the transformation τ ↦ − 1 / τ \tau\mapsto-1/\tau τ ↦ − 1/ τ acts by
Z X ( v , ( g , k ) , − 1 τ ) = τ wt [ v ] ∑ L ∈ M ( k , g − 1 ) S [ X ; ( g , k ) ] , [ L ; ( k , g − 1 ) ] Z L ( v , ( k , g − 1 ) , τ ) , X ∈ M ( g , k ) , (2) Z_X\left(v,(g,k),-\frac{1}{\tau}\right)=\tau^{\operatorname{wt}[v]}\sum_{L\in\mathcal{M}(k,g^{-1})}S_{[X;(g,k)],[L;(k,g^{-1})]}Z_L(v,(k,g^{-1}),\tau),\qquad X\in\mathcal{M}(g,k), \tag*{(2)} Z X ( v , ( g , k ) , − τ 1 ) = τ wt [ v ] L ∈ M ( k , g − 1 ) ∑ S [ X ; ( g , k )] , [ L ; ( k , g − 1 )] Z L ( v , ( k , g − 1 ) , τ ) , X ∈ M ( g , k ) , (2)
where M ( k , g − 1 ) = { L ∈ M ( k ) : g ∈ G L } \mathcal{M}(k,g^{-1})=\{L\in\mathcal{M}(k):g\in G_L\} M ( k , g − 1 ) = { L ∈ M ( k ) : g ∈ G L } and the bracket [ X ; ( g , k ) ] [X;(g,k)] [ X ; ( g , k )] denotes the module X X X together with the sector ( g , k ) (g,k) ( g , k ) and the chosen φ X ( k ) \varphi_X(k) φ X ( k ) . The coefficients are unique since the functions on the right hand side are linearly independent as functions of ( v , τ ) (v,\tau) ( v , τ ) with v ∈ V ‾ 0 v\in\overline{V}_0 v ∈ V 0 . For Q ∈ Q Q\in\mathcal{Q} Q ∈ Q set S X , Q = S [ X ; ( g X , 1 ) ] , [ Q ; ( 1 , g X ) ] S_{X,Q}=S_{[X;(g_X,1)],[Q;(1,g_X)]} S X , Q = S [ X ; ( g X , 1 )] , [ Q ; ( 1 , g X )] (row X X X , column Q Q Q ). For g X = 1 g_X=1 g X = 1 it is the ordinary S S S -matrix of V ‾ 0 \overline{V}_0 V 0 . For a stable W ∈ M ( h ) W\in\mathcal{M}(h) W ∈ M ( h ) and h ∈ G X h\in G_X h ∈ G X set S X , W [ h ] = S [ X ; ( g X , h ) ] , [ W ; ( h , g X − 1 ) ] S^{[h]}_{X,W}=S_{[X;(g_X,h)],[W;(h,g_X^{-1})]} S X , W [ h ] = S [ X ; ( g X , h )] , [ W ; ( h , g X − 1 )] .
Let us now describe the restriction to U U U . For v ∈ U ⊂ V ‾ 0 v\in U\subset\overline{V}_0 v ∈ U ⊂ V 0 and a ∈ G a\in G a ∈ G one has Z L ∘ a ( v , ⋅ ) = Z L ( v , ⋅ ) Z_{L\circ a}(v,\cdot)=Z_L(v,\cdot) Z L ∘ a ( v , ⋅ ) = Z L ( v , ⋅ ) (see Proposition 2.6 (ii)). Thus, restricted to v ∈ U v\in U v ∈ U , the functions of the modules in one G G G -orbit coincide, and the sum of the coefficients over an orbit can be read off. Lemma 2.8 is a statement about v ∈ U v\in U v ∈ U and uses these orbit sums.
We will use the following conventions. (i) For twisted modules we follow [16 ] . (ii) ψ σ ( g ) = e − 2 π i s / n \psi_\sigma(g)=e^{-2\pi i s/n} ψ σ ( g ) = e − 2 π i s / n , X χ = { w : φ X ( h ) w = χ ( h ) w ∀ h ∈ G X } X_\chi=\{w:\varphi_X(h)w=\chi(h)w\ \forall h\in G_X\} X χ = { w : φ X ( h ) w = χ ( h ) w ∀ h ∈ G X } . (iii) The transformation S S S sends the sector ( g , k ) (g,k) ( g , k ) to ( k , g − 1 ) (k,g^{-1}) ( k , g − 1 ) . All factors ψ ( g − 1 ) \psi(g^{-1}) ψ ( g − 1 ) below arise from Z L ( v , ( k , g − 1 ) , τ ) = ∑ ψ ψ ( g − 1 ) Z L ψ ( v , τ ) Z_L(v,(k,g^{-1}),\tau)=\sum_\psi\psi(g^{-1})Z_{L_\psi}(v,\tau) Z L ( v , ( k , g − 1 ) , τ ) = ∑ ψ ψ ( g − 1 ) Z L ψ ( v , τ ) , and all factors χ ( k ) ‾ \overline{\chi(k)} χ ( k ) from the projection Z X χ = ∣ G X ∣ − 1 ∑ k χ ( k ) ‾ Z X ( ⋅ , ( g , k ) , ⋅ ) Z_{X_\chi}=|G_X|^{-1}\sum_k\overline{\chi(k)}Z_X(\cdot,(g,k),\cdot) Z X χ = ∣ G X ∣ − 1 ∑ k χ ( k ) Z X ( ⋅ , ( g , k ) , ⋅ ) . (iv) S U S^U S U is defined by Z Y ( v , − 1 / τ ) = τ wt [ v ] ∑ Y ′ S Y , Y ′ U Z Y ′ ( v , τ ) Z_Y(v,-1/\tau)=\tau^{\operatorname{wt}[v]}\sum_{Y'}S^U_{Y,Y'}Z_{Y'}(v,\tau) Z Y ( v , − 1/ τ ) = τ wt [ v ] ∑ Y ′ S Y , Y ′ U Z Y ′ ( v , τ ) , v ∈ U v\in U v ∈ U . (v) T T T acts by e 2 π i ( h − c / 24 ) e^{2\pi i(h-c/24)} e 2 π i ( h − c /24 ) . (vi) Three kinds of S S S occur: the orbifold entries S X , Q S_{X,Q} S X , Q (row twisted or untwisted, column Q ∈ Q Q\in\mathcal{Q} Q ∈ Q ), the matrix S U S^U S U of C U C_U C U , and the untwisted S S S -matrix of V ‾ 0 \overline{V}_0 V 0 (the case g X = 1 g_X=1 g X = 1 of the first). These conventions are confirmed independently in Example 7.1 from twisted traces (see Remark 2.10). Then the following lemma concerning the orbifold relation holds.
Lemma 2.8. Let X ∈ M ( g ) X\in\mathcal{M}(g) X ∈ M ( g ) , χ ∈ G ^ X \chi\in\widehat{G}_X χ ∈ G X .
(a) If N ∈ M ( h ) N\in\mathcal{M}(h) N ∈ M ( h ) and ψ ∈ G ^ N \psi\in\widehat{G}_N ψ ∈ G N , then S X χ , N ψ U = 0 S^U_{X_\chi,N_\psi}=0 S X χ , N ψ U = 0 unless h ∈ G X h\in G_X h ∈ G X and g ∈ G N g\in G_N g ∈ G N .
(b) If W ∈ M ( h ) W\in\mathcal{M}(h) W ∈ M ( h ) is stable, h ∈ G X h\in G_X h ∈ G X and t ∈ Z / n t\in\mathbb{Z}/n t ∈ Z / n , then
S X χ , W t U = ∣ G X ∣ − 1 χ ( h ) ‾ ψ t ( g − 1 ) S X , W [ h ] . S^U_{X_\chi,W_t}=|G_X|^{-1}\overline{\chi(h)}\psi_t(g^{-1})S^{[h]}_{X,W}. S X χ , W t U = ∣ G X ∣ − 1 χ ( h ) ψ t ( g − 1 ) S X , W [ h ] . In particular, for Q ∈ Q Q\in\mathcal{Q} Q ∈ Q :
S X χ , Q t U = ∣ G X ∣ − 1 ψ t ( g − 1 ) S X , Q . S^U_{X_\chi,Q_t}=|G_X|^{-1}\psi_t(g^{-1})S_{X,Q}. S X χ , Q t U = ∣ G X ∣ − 1 ψ t ( g − 1 ) S X , Q . (c) If N ∈ M ( h ) N\in\mathcal{M}(h) N ∈ M ( h ) has stabilizer G N ∋ g G_N\ni g G N ∋ g , h ∈ G X h\in G_X h ∈ G X , ψ ∈ G ^ N \psi\in\widehat{G}_N ψ ∈ G N , then, with the convention φ L = φ N \varphi_L=\varphi_N φ L = φ N for L ∈ O N L\in\mathcal{O}_N L ∈ O N ,
S X χ , N ψ U = ∣ G X ∣ − 1 χ ( h ) ‾ ψ ( g − 1 ) ∑ L ∈ O N S [ X ; ( g , h ) ] , [ L ; ( h , g − 1 ) ] . S^U_{X_\chi,N_\psi}=|G_X|^{-1}\overline{\chi(h)}\psi(g^{-1})\sum_{L\in\mathcal{O}_N}S_{[X;(g,h)],[L;(h,g^{-1})]}. S X χ , N ψ U = ∣ G X ∣ − 1 χ ( h ) ψ ( g − 1 ) L ∈ O N ∑ S [ X ; ( g , h )] , [ L ; ( h , g − 1 )] . (b) is the case O N = { W } \mathcal{O}_N=\{W\} O N = { W } .
Proof. In what follows let v ∈ U v \in U v ∈ U .
Projection onto X χ X_\chi X χ : for v ∈ U v \in U v ∈ U the operator o ( v ) o(v) o ( v ) commutes with φ X ( k ) \varphi_X(k) φ X ( k ) , k ∈ G X k \in G_X k ∈ G X . The projection onto X χ X_\chi X χ is P X = ∣ G X ∣ − 1 ∑ k ∈ G X χ ( k ) ‾ φ X ( k ) P_X=|G_X|^{-1}\sum_{k\in G_X}\overline{\chi(k)}\varphi_X(k) P X = ∣ G X ∣ − 1 ∑ k ∈ G X χ ( k ) φ X ( k ) . Note that G X G_X G X is cyclic, thus this is the orthogonal projection for the characters of G X G_X G X . Hence, Z X χ ( v , τ ) = ∣ G X ∣ − 1 ∑ k ∈ G X χ ( k ) ‾ Z X ( v , ( g , k ) , τ ) Z_{X_\chi}(v,\tau)=|G_X|^{-1}\sum_{k\in G_X}\overline{\chi(k)}Z_X(v,(g,k),\tau) Z X χ ( v , τ ) = ∣ G X ∣ − 1 ∑ k ∈ G X χ ( k ) Z X ( v , ( g , k ) , τ ) . This is the origin of the factor ∣ G X ∣ − 1 |G_X|^{-1} ∣ G X ∣ − 1 and of χ ( k ) ‾ \overline{\chi(k)} χ ( k ) . The sum is over k ∈ G X k \in G_X k ∈ G X , thus X ∈ M ( g , k ) X \in\mathcal{M}(g,k) X ∈ M ( g , k ) for every term.
Transformation: apply (2.2) to each term. The sector ( g , k ) (g,k) ( g , k ) becomes ( k , g − 1 ) (k,g^{-1}) ( k , g − 1 ) , and L L L ranges over M ( k , g − 1 ) \mathcal{M}(k,g^{-1}) M ( k , g − 1 ) , i.e., over k k k -twisted modules with g ∈ G L g \in G_L g ∈ G L .
Decomposition of the transformed traces: for such L L L , φ L ( g − 1 ) \varphi_L(g^{-1}) φ L ( g − 1 ) acts on L ψ L_\psi L ψ by ψ ( g − 1 ) \psi(g^{-1}) ψ ( g − 1 ) , thus Z L ( v , ( k , g − 1 ) , τ ) = ∑ ψ ∈ G L ^ ψ ( g − 1 ) Z L ψ ( v , τ ) Z_L(v,(k,g^{-1}),\tau)=\sum_{\psi\in\widehat{G_L}}\psi(g^{-1})Z_{L_\psi}(v,\tau) Z L ( v , ( k , g − 1 ) , τ ) = ∑ ψ ∈ G L ψ ( g − 1 ) Z L ψ ( v , τ ) . This is the origin of ψ ( g − 1 ) \psi(g^{-1}) ψ ( g − 1 ) . Altogether,
Z X χ ( v , − 1 τ ) = τ wt [ v ] ∣ G X ∣ − 1 ∑ k ∈ G X χ ( k ) ‾ ∑ L ∈ M ( k , g − 1 ) S [ X ; ( g , k ) ] , [ L ; ( k , g − 1 ) ] ∑ ψ ∈ G L ^ ψ ( g − 1 ) Z L ψ ( v , τ ) . Z_{X_\chi}\left(v,-\frac{1}{\tau}\right)=\tau^{\operatorname{wt}[v]}|G_X|^{-1}\sum_{k\in G_X}\overline{\chi(k)}\sum_{L\in\mathcal{M}(k,g^{-1})}S_{[X;(g,k)],[L;(k,g^{-1})]}\sum_{\psi\in\widehat{G_L}}\psi(g^{-1})Z_{L_\psi}(v,\tau). Z X χ ( v , − τ 1 ) = τ wt [ v ] ∣ G X ∣ − 1 k ∈ G X ∑ χ ( k ) L ∈ M ( k , g − 1 ) ∑ S [ X ; ( g , k )] , [ L ; ( k , g − 1 )] ψ ∈ G L ∑ ψ ( g − 1 ) Z L ψ ( v , τ ) .
Comparison of coefficients: by definition, Z X χ ( v , − 1 τ ) = τ wt [ v ] ∑ Y S X χ , Y U Z Y ( v , τ ) Z_{X_\chi}\left(v,-\frac{1}{\tau}\right)=\tau^{\operatorname{wt}[v]}\sum_Y S^U_{X_\chi,Y}Z_Y(v,\tau) Z X χ ( v , − τ 1 ) = τ wt [ v ] ∑ Y S X χ , Y U Z Y ( v , τ ) over the irreducible U U U -modules Y Y Y , and the functions Z Y ( v , τ ) Z_Y(v,\tau) Z Y ( v , τ ) , v ∈ U v \in U v ∈ U , for pairwise inequivalent Y Y Y are linearly independent [41 , 28 ] . By Proposition 2.6 (iii), (iv), the U U U -module L ψ L_\psi L ψ is equivalent to N ψ ′ N_{\psi'} N ψ ′ iff L ∈ O N L \in\mathcal{O}_N L ∈ O N (hence, k = h k=h k = h , since the orbit belongs to one sector) and ψ = ψ ′ \psi=\psi' ψ = ψ ′ . Distinct pairs (sector k k k , orbit, and character) give inequivalent U U U -modules. Thus, no cancellation between different terms occurs. Consequently, the coefficient of Z N ψ Z_{N_{\psi}} Z N ψ is obtained from the terms with k = h k=h k = h and L ∈ O N L \in\mathcal{O}_N L ∈ O N .
If h ∉ G X h \notin G_X h ∈ / G X , no term has k = h k=h k = h . Thus, S X χ , N ψ U = 0 S^U_{X_\chi,N_\psi}=0 S X χ , N ψ U = 0 . If g ∉ G N g \notin G_N g ∈ / G N then no L ∈ O N L \in\mathcal{O}_N L ∈ O N belongs to M ( h , g − 1 ) \mathcal{M}(h,g^{-1}) M ( h , g − 1 ) (as G L = G N G_L=G_N G L = G N ), thus again 0 0 0 . This gives (a). Otherwise, the terms L ∈ O N L \in\mathcal{O}_N L ∈ O N all equal Z N ψ Z_{N_\psi} Z N ψ on U U U , their coefficients add, and we obtain (c). For O N = { W } \mathcal{O}_N=\{W\} O N = { W } this gives (b). □
Note that whole orbifold ρ ( S ) \rho(S) ρ ( S ) depends on the choices of the representations φ X \varphi_X φ X in every sector ( g , k ) (g,k) ( g , k ) which is proven using the existence of the matrix (2.2) in the chosen bases. The following lemma describes orbit invariance and dependence on G G G .
Lemma 2.9. (a) For X ∈ M ( g ) X \in\mathcal{M}(g) X ∈ M ( g ) , a ∈ G a \in G a ∈ G and Q ∈ Q Q \in\mathcal{Q} Q ∈ Q : S X ∘ a , Q = S X , Q S_{X\circ a,Q}=S_{X,Q} S X ∘ a , Q = S X , Q .
(b) The entry S X , Q S_{X,Q} S X , Q depends only on X X X , on g = g X g=g_X g = g X and on the operator φ Q ( g − 1 ) \varphi_Q(g^{-1}) φ Q ( g − 1 ) . In particular, if H ≤ G H \leq G H ≤ G contains g X g_X g X and Q Q Q is G G G -stable, the entry calculated with respect to H H H (with φ Q ∣ H \varphi_Q|_H φ Q ∣ H ) equals the entry computed with respect to G G G . The set Q = M ( 1 , G ) \mathcal{Q}=\mathcal{M}(1,G) Q = M ( 1 , G ) , however, depends on the group.
Proof. (a) By Proposition 2.6 (ii), ( X ∘ a ) χ = X χ (X\circ a)_\chi=X_\chi ( X ∘ a ) χ = X χ and G X ∘ a = G X G_{X\circ a}=G_X G X ∘ a = G X . Lemma 2.8 (b) gives ∣ G X ∣ − 1 ψ t ( g − 1 ) S X ∘ a , Q = S X χ , Q t U = ∣ G X ∣ − 1 ψ t ( g − 1 ) S X , Q |G_X|^{-1}\psi_t(g^{-1})S_{X\circ a,Q}=S^U_{X_\chi,Q_t}=|G_X|^{-1}\psi_t(g^{-1})S_{X,Q} ∣ G X ∣ − 1 ψ t ( g − 1 ) S X ∘ a , Q = S X χ , Q t U = ∣ G X ∣ − 1 ψ t ( g − 1 ) S X , Q . (b) S X , Q S_{X,Q} S X , Q is the coefficient of Z Q ( v , ( 1 , g − 1 ) , τ ) Z_Q(v,(1,g^{-1}),\tau) Z Q ( v , ( 1 , g − 1 ) , τ ) in the expansion (2.2) of Z X ( v , ( g , 1 / τ ) ) Z_X(v,(g,1/\tau)) Z X ( v , ( g , 1/ τ )) over the linearly independent functions Z L ( v , ( 1 , g − 1 ) , τ ) Z_L(v,(1,g^{-1}),\tau) Z L ( v , ( 1 , g − 1 ) , τ ) , L ∈ M ( 1 , g − 1 ) L \in\mathcal{M}(1,g^{-1}) L ∈ M ( 1 , g − 1 ) . This expansion involves only g g g .
Let us make a remark concerning two different normalization issues. (1) Orbit convention for φ X ∘ a = φ X \varphi_{X\circ a}=\varphi_X φ X ∘ a = φ X , Proposition 2.6 (ii): it affects only the identification of U U U -components inside an orbit and leaves all entries S X , Q S_{X,Q} S X , Q unchanged (see Lemma 2.9 (a)). (2) Column normalization: the entries S X , Q S_{X,Q} S X , Q with g X ≠ 1 g_X\ne1 g X = 1 depend on φ Q \varphi_Q φ Q . If φ Q \varphi_Q φ Q is replaced by λ ( σ ) φ Q \lambda(\sigma)\varphi_Q λ ( σ ) φ Q , λ = ψ a ∈ G ^ \lambda=\psi_a\in\widehat{G} λ = ψ a ∈ G , then Q t = Q t + a ′ Q_t=Q'_{t+a} Q t = Q t + a ′ . The eigenvalue ψ t ( σ ) \psi_t(\sigma) ψ t ( σ ) of φ Q ( σ ) \varphi_Q(\sigma) φ Q ( σ ) becomes ψ t + a ( σ ) \psi_{t+a}(\sigma) ψ t + a ( σ ) for φ Q ′ ( σ ) \varphi_{Q'}(\sigma) φ Q ′ ( σ ) . Comparing Lemma 2.8 (b) for the two labelings gives
S X , Q ′ = λ ( g X ) S X , Q . (3) S'_{X,Q}=\lambda(g_X)S_{X,Q}. \tag*{(3)} S X , Q ′ = λ ( g X ) S X , Q . (3)
Entries with g X = 1 g_X=1 g X = 1 and S V 0 ˉ , Q S_{V_{\bar{0}},Q} S V 0 ˉ , Q do not change (λ ( 1 ) = 1 \lambda(1)=1 λ ( 1 ) = 1 ). The row module X X X does not require normalization.
Remark 2.10. Let us make remarks concerning conventions and comparison with [12 ] . We distinguish S X , Q S_{X,Q} S X , Q (row X X X ), its complex conjugate S X , Q ‾ \overline{S_{X,Q}} S X , Q , and the entry ( S − 1 ) Q , X (S^{-1})_{Q,X} ( S − 1 ) Q , X of the inverse of ρ ( S ) \rho(S) ρ ( S ) . For the blocks used here ( S − 1 ) Q , X = S X , Q ‾ (S^{-1})_{Q,X}=\overline{S_{X,Q}} ( S − 1 ) Q , X = S X , Q by Proposition 2.14. Symmetry is used for S U S^U S U and the untwisted S S S of V 0 ˉ V_{\bar{0}} V 0 ˉ and never for entries with a twisted index. For G G G of prime order p p p , Lemma 2.8 (b) reads S M λ s 1 , M λ t 2 U = 1 p S M 1 , M 2 V λ s ( g 2 ) λ t ( g 1 − 1 ) S^U_{M^1_{\lambda_s},M^2_{\lambda_t}}=\frac{1}{p}S^V_{M^1,M^2}\lambda_s(g_2)\lambda_t(g_1^{-1}) S M λ s 1 , M λ t 2 U = p 1 S M 1 , M 2 V λ s ( g 2 ) λ t ( g 1 − 1 ) . Our conjugation, the argument g 1 − 1 g_1^{-1} g 1 − 1 , and the factor 1 / p = ∣ G X ∣ − 1 1/p=|G_X|^{-1} 1/ p = ∣ G X ∣ − 1 coincide with those of [12 ] with λ s = ψ s \lambda_s=\psi_s λ s = ψ s . The Verlinde formula [12 ] also carries the conjugate on the third factor as ours.
Let us explain why the conjugate matters. Without it the summand of Theorem 2.11 would be multiplied by λ ( g 1 g 2 g 3 ) = λ ( g 3 2 ) \lambda(g_{1}g_{2}g_{3})=\lambda(g_{3}^{2}) λ ( g 1 g 2 g 3 ) = λ ( g 3 2 ) under a renormalization of the columns by (2.3). In Example 7.1, n = 3 n=3 n = 3 , for ( T 0 ( 1 ) , T 0 ( 1 ) , T 0 ( 2 ) ) (T_{0}^{(1)},T_{0}^{(1)},T_{0}^{(2)}) ( T 0 ( 1 ) , T 0 ( 1 ) , T 0 ( 2 ) ) , whose true fusion coefficient is one, the sum with the conjugate equals one, and remains one after multiplying all column normalizations by λ ( σ ) = e 2 π i / 3 \lambda(\sigma)=e^{2\pi i/3} λ ( σ ) = e 2 π i /3 , whereas the sum without it equals 1 / 4 1/4 1/4 and becomes − 0.125 + 0.217 i -0.125+0.217\,i − 0.125 + 0.217 i (scripts verify_all.py, verify_conv.py).
In Example 7.1 the twisted trace functions are theta functions. For n = 3 , 9 n=3,9 n = 3 , 9 and all sectors a ≠ 0 a\ne0 a = 0 , the transformation law (2.2) with the entries S T m ( a ) , Q k = 1 2 e − 2 π i m k / 4 e − 2 π i a k / 2 n S_{T_{m}^{(a)},Q_{k}}=\frac{1}{2}e^{-2\pi imk/4}e^{-2\pi iak/2n} S T m ( a ) , Q k = 2 1 e − 2 π imk /4 e − 2 π iak /2 n of Lemma 2.8 (computed from S U S^{U} S U ) holds to 10 − 15 10^{-15} 1 0 − 15 at three values of τ \tau τ (script verify_twisted_traces.py). Reversing the sign of a a a or taking complex conjugates produces errors of order 0.7 0.7 0.7 -1.2 1.2 1.2 . This tests the identification of the entries with the transformation law of twisted traces, including the placement of g g g versus g − 1 g^{-1} g − 1 and the normalization φ Q k = e − 2 π i k / 2 n σ \varphi_{Q_k}=e^{-2\pi ik/2n}\sigma φ Q k = e − 2 π ik /2 n σ .
The twisted Verlinde formula for V for cyclic groups
For M i ∈ M ( g i ) M_i\in\mathcal{M}(g_i) M i ∈ M ( g i ) , g 3 = g 1 g 2 g_3=g_1g_2 g 3 = g 1 g 2 , set
T ( M 1 , M 2 , M 3 ) : = ∑ Q ∈ Q S M 1 , Q S M 2 , Q S M 3 , Q ‾ S V ‾ 0 , Q . \mathrm{T}(M_1,M_2,M_3):=\sum_{Q\in\mathcal{Q}}\frac{S_{M_1,Q}S_{M_2,Q}\overline{S_{M_3,Q}}}{S_{\overline{V}_0,Q}}. T ( M 1 , M 2 , M 3 ) := Q ∈ Q ∑ S V 0 , Q S M 1 , Q S M 2 , Q S M 3 , Q .
By (2.3), each summand is multiplied by λ ( g 1 ) λ ( g 2 ) λ ( g 3 ) ‾ = λ ( g 1 g 2 g 3 − 1 ) = 1 \lambda(g_1)\lambda(g_2)\overline{\lambda(g_3)}=\lambda(g_1g_2g_3^{-1})=1 λ ( g 1 ) λ ( g 2 ) λ ( g 3 ) = λ ( g 1 g 2 g 3 − 1 ) = 1 under a change of the normalizations of Q Q Q . This also holds if some g i = 1 g_i=1 g i = 1 , where λ ( g i ) = 1 \lambda(g_i)=1 λ ( g i ) = 1 . Thus, T is well defined, and it is constant on orbits in each argument by Lemma 2.9 (a). The theorem below provides the first main result of this paper, the twisted Verlinde formula for V ‾ 0 \overline{V}_0 V 0 .
Theorem 2.11. Adopt both assumptions of Subsection 2.1. Let g 1 , g 2 ∈ G g_1,g_2\in G g 1 , g 2 ∈ G with ⟨ g 1 , g 2 ⟩ = G \langle g_1,g_2\rangle=G ⟨ g 1 , g 2 ⟩ = G , g 3 = g 1 g 2 g_3=g_1g_2 g 3 = g 1 g 2 , M i ∈ M ( g i ) M_i\in\mathcal{M}(g_i) M i ∈ M ( g i ) with orbits O i O_i O i . Then:
1 ∣ O 1 ∣ ∣ O 2 ∣ ∑ N 1 ∈ O 1 , N 2 ∈ O 2 N N 1 , N 2 M 3 = T ( M 1 , M 2 , M 3 ) \frac{1}{|O_1||O_2|}\sum_{N_1\in O_1,N_2\in O_2}N_{N_1,N_2}^{M_3}=\mathrm{T}(M_1,M_2,M_3) ∣ O 1 ∣∣ O 2 ∣ 1 ∑ N 1 ∈ O 1 , N 2 ∈ O 2 N N 1 , N 2 M 3 = T ( M 1 , M 2 , M 3 ) .
If at least two of M 1 M_1 M 1 , M 2 M_2 M 2 , M 3 M_3 M 3 are G G G -stable, then N M 1 , M 2 M 3 = T ( M 1 , M 2 , M 3 ) N_{M_1,M_2}^{M_3}=\mathrm{T}(M_1,M_2,M_3) N M 1 , M 2 M 3 = T ( M 1 , M 2 , M 3 ) .
Part (ii) always applies if n n n is a prime power, and if every irreducible twisted V ‾ 0 \overline{V}_0 V 0 -module is G G G -stable (e.g., for inner automorphisms, Section 6). For n n n prime it is given in [12 ] .
Proof. In this proof ⊠ \boxtimes ⊠ denotes the tensor product of Rep ( V ‾ 0 ) \operatorname{Rep}(\overline{V}_0) Rep ( V 0 ) and ⊠ U \boxtimes_U ⊠ U that of C U C_U C U . No tensor product of twisted V V V -modules occurs (see Definition 2.5).
0 ) 0) 0 ) Reduction to U U U and multiplicities: by Lemma 2.7 and Proposition 2.4 (e), for arbitrary χ 1 ∈ G ^ M 1 \chi_1\in\widehat{G}_{M_1} χ 1 ∈ G M 1 , χ 2 ∈ G ^ M 2 \chi_2\in\widehat{G}_{M_2} χ 2 ∈ G M 2 ,
⨁ N 1 ∈ O 1 , N 2 ∈ O 2 N 1 ⊠ N 2 = F 0 ( M 1 , χ 1 ) ⊠ F 0 ( M 2 , χ 2 ) = F 0 ( M 1 , χ 1 ⊠ U M 2 , χ 2 ) = ⨁ Y N M 1 , χ 1 , M 2 , χ 2 Y F 0 ( Y ) . \bigoplus_{N_1\in O_1,N_2\in O_2}N_1\boxtimes N_2=F_0(M_1,\chi_1)\boxtimes F_0(M_2,\chi_2)=F_0(M_1,\chi_1\boxtimes_U M_2,\chi_2)=\bigoplus_Y N_{M_1,\chi_1,M_2,\chi_2}^{Y}F_0(Y). N 1 ∈ O 1 , N 2 ∈ O 2 ⨁ N 1 ⊠ N 2 = F 0 ( M 1 , χ 1 ) ⊠ F 0 ( M 2 , χ 2 ) = F 0 ( M 1 , χ 1 ⊠ U M 2 , χ 2 ) = Y ⨁ N M 1 , χ 1 , M 2 , χ 2 Y F 0 ( Y ) .
where Y Y Y runs over the irreducible U U U -modules and each F 0 ( Y ) = ⨁ N ∈ O Z N F_0(Y)=\bigoplus_{N\in O_Z}N F 0 ( Y ) = ⨁ N ∈ O Z N , Y = Z ψ Y=Z_\psi Y = Z ψ , is multiplicity free (see Lemma 2.7). Fix the simple object M 3 M_3 M 3 . It occurs in F 0 ( Y ) F_0(Y) F 0 ( Y ) iff M 3 ∈ O Z M_3\in O_Z M 3 ∈ O Z , i.e., iff Y ≅ ( M 3 ) χ 3 Y\cong(M_3)_{\chi_3} Y ≅ ( M 3 ) χ 3 for some χ 3 ∈ G ^ M 3 \chi_3\in\widehat{G}_{M_3} χ 3 ∈ G M 3 (Proposition 2.6 (iii)), and then exactly once. The modules ( M 3 ) χ 3 (M_3)_{\chi_3} ( M 3 ) χ 3 , χ 3 ∈ G ^ M 3 \chi_3\in\widehat{G}_{M_3} χ 3 ∈ G M 3 , are pairwise inequivalent (Proposition 2.6 (ii)). Thus, the U U U -modules Y Y Y whose induction contains M 3 M_3 M 3 are these ∣ G M 3 ∣ |G_{M_3}| ∣ G M 3 ∣ modules, each contributing once. Comparing the multiplicities of M 3 M_3 M 3 we obtain
∑ N 1 ∈ O 1 , N 2 ∈ O 2 N N 1 , N 2 M 3 = ∑ χ 3 ∈ G ^ M 3 N M 1 , χ 1 , M 2 , χ 2 ( M 3 ) χ 3 . (4) \sum_{N_1\in O_1,N_2\in O_2}N_{N_1,N_2}^{M_3}=\sum_{\chi_3\in\widehat{G}_{M_3}}N_{M_1,\chi_1,M_2,\chi_2}^{(M_3)_{\chi_3}}. \tag*{(4)} N 1 ∈ O 1 , N 2 ∈ O 2 ∑ N N 1 , N 2 M 3 = χ 3 ∈ G M 3 ∑ N M 1 , χ 1 , M 2 , χ 2 ( M 3 ) χ 3 . (4)
1 ) 1) 1 ) Verlinde formula for U U U : by Huang version of Verlinde formula [28 ] for C U C_U C U , N Y 1 , Y 2 Y 3 = ∑ Y S Y 1 , Y U S Y 2 , Y U S Y 3 , Y U ‾ S Y , Y U N_{Y_1,Y_2}^{Y_3}=\sum_Y\frac{S^U_{Y_1,Y}S^U_{Y_2,Y}\overline{S^U_{Y_3,Y}}}{S^U_{Y,Y}} N Y 1 , Y 2 Y 3 = ∑ Y S Y , Y U S Y 1 , Y U S Y 2 , Y U S Y 3 , Y U . Apply it with Y i = ( M i ) χ i Y_i=(M_i)_{\chi_i} Y i = ( M i ) χ i . By Lemma 2.8 (a), a summand with Y = N ψ Y=N_\psi Y = N ψ vanishes unless g 1 , g 2 , g 3 ∈ G N g_1,g_2,g_3\in G_N g 1 , g 2 , g 3 ∈ G N . Since G N G_N G N is a subgroup and ⟨ g 1 , g 2 ⟩ = G \langle g_1,g_2\rangle=G ⟨ g 1 , g 2 ⟩ = G this leads to G N = G G_N=G G N = G , i.e., Y = W t Y=W_t Y = W t with W ∈ M ( h ) W\in\mathcal{M}(h) W ∈ M ( h ) stable, h ∈ G h\in G h ∈ G , t ∈ Z / n t\in\mathbb{Z}/n t ∈ Z / n .
2 ) 2) 2 ) Evaluation, characters and domains: fix χ 1 , χ 2 \chi_1,\chi_2 χ 1 , χ 2 . For stable W ∈ M ( h ) W\in\mathcal{M}(h) W ∈ M ( h ) and t ∈ Z / n t\in\mathbb{Z}/n t ∈ Z / n , Lemma 2.8 (b) applies to the rows ( M i ) χ i (M_i)_{\chi_i} ( M i ) χ i , i = 1 , 2 i=1,2 i = 1 , 2 , and to the complex conjugate of the row ( M 3 ) χ 3 (M_3)_{\chi_3} ( M 3 ) χ 3 , i.e., S ‾ ( M 3 ) χ 3 , W t U = ∣ G M 3 ∣ − 1 χ 3 ( h ) ‾ ψ t ( g 3 − 1 ) ‾ S ‾ M 3 , W [ h ] \overline{S}^{U}_{(M_{3})_{\chi_{3},W_{t}}}=|G_{M_{3}}|^{-1}\overline{\chi_{3}(h)}\,\overline{\psi_{t}(g_{3}^{-1})}\,\overline{S}^{[h]}_{M_{3},W} S ( M 3 ) χ 3 , W t U = ∣ G M 3 ∣ − 1 χ 3 ( h ) ψ t ( g 3 − 1 ) S M 3 , W [ h ] . Moreover, S U ‾ , W t U = n − 1 S V ‾ 0 , W [ h ] S^{U}_{\overline{U},W_{t}}=n^{-1}S^{[h]}_{\overline{V}_{0},W} S U , W t U = n − 1 S V 0 , W [ h ] for X = V ‾ 0 X=\overline{V}_{0} X = V 0 , χ = ψ 0 \chi=\psi_{0} χ = ψ 0 , g = 1 g=1 g = 1 . The ( W , t ) (W,t) ( W , t ) -summand therefore vanishes unless h ∈ G M 1 ∩ G M 2 ∩ G M 3 h\in G_{M_{1}}\cap G_{M_{2}}\cap G_{M_{3}} h ∈ G M 1 ∩ G M 2 ∩ G M 3 , and in that case equals
∏ i = 1 3 ∣ G M i ∣ − 1 χ 1 ( h ) ‾ χ 2 ( h ) ‾ χ 3 ( h ) ‾ ψ t ( g 1 − 1 ) ‾ ψ t ( g 2 − 1 ) ‾ ψ t ( g 3 − 1 ) ‾ S ‾ M 1 , W [ h ] S ‾ M 2 , W [ h ] S ‾ M 3 , W [ h ] n − 1 S V ‾ 0 , W [ h ] . (5) \frac{\prod_{i=1}^{3}|G_{M_{i}}|^{-1}\overline{\chi_{1}(h)}\,\overline{\chi_{2}(h)}\,\overline{\chi_{3}(h)}\,\overline{\psi_{t}(g_{1}^{-1})}\,\overline{\psi_{t}(g_{2}^{-1})}\,\overline{\psi_{t}(g_{3}^{-1})}\,\overline{S}^{[h]}_{M_{1},W}\,\overline{S}^{[h]}_{M_{2},W}\,\overline{S}^{[h]}_{M_{3},W}}{n^{-1}S^{[h]}_{\overline{V}_{0},W}}. \tag*{(5)} n − 1 S V 0 , W [ h ] ∏ i = 1 3 ∣ G M i ∣ − 1 χ 1 ( h ) χ 2 ( h ) χ 3 ( h ) ψ t ( g 1 − 1 ) ψ t ( g 2 − 1 ) ψ t ( g 3 − 1 ) S M 1 , W [ h ] S M 2 , W [ h ] S M 3 , W [ h ] . (5)
(i) For t t t -sum: the t t t -dependence is ψ t ( g 1 − 1 ) ‾ ψ t ( g 2 − 1 ) ‾ ψ t ( g 3 − 1 ) ‾ = ψ t ( g 1 − 1 g 2 − 1 g 3 ) ‾ = ψ t ( 1 ) ‾ = 1 \overline{\psi_{t}(g_{1}^{-1})}\overline{\psi_{t}(g_{2}^{-1})}\overline{\psi_{t}(g_{3}^{-1})}=\overline{\psi_{t}(g_{1}^{-1}g_{2}^{-1}g_{3})}=\overline{\psi_{t}(1)}=1 ψ t ( g 1 − 1 ) ψ t ( g 2 − 1 ) ψ t ( g 3 − 1 ) = ψ t ( g 1 − 1 g 2 − 1 g 3 ) = ψ t ( 1 ) = 1 since g 3 = g 1 g 2 g_{3}=g_{1}g_{2} g 3 = g 1 g 2 . Hence, ∑ t ∈ Z / n \sum_{t\in\mathbb{Z}/n} ∑ t ∈ Z / n gives n n n .
(ii) For χ 3 \chi_{3} χ 3 -sum: the dependence on χ 3 ∈ G ^ M 3 \chi_{3}\in\widehat{G}_{M_{3}} χ 3 ∈ G M 3 is in the factor χ 3 ( h ) \chi_{3}(h) χ 3 ( h ) with h ∈ G M 3 h\in G_{M_{3}} h ∈ G M 3 . We use the orthogonality of the cyclic group G M 3 G_{M_{3}} G M 3 characters of order ∣ G M 3 ∣ |G_{M_{3}}| ∣ G M 3 ∣ , i.e., ∑ χ 3 ∈ G ^ M 3 χ 3 ( h ) = ∣ G M 3 ∣ \sum_{\chi_{3}\in\widehat{G}_{M_{3}}}\chi_{3}(h)=|G_{M_{3}}| ∑ χ 3 ∈ G M 3 χ 3 ( h ) = ∣ G M 3 ∣ if h = 1 h=1 h = 1 and 0 0 0 if h ∈ G M 3 ∖ { 1 } h\in G_{M_{3}}\setminus\{1\} h ∈ G M 3 ∖ { 1 } . Terms with h ∉ G M 3 h\notin G_{M_{3}} h ∈ / G M 3 vanish already. Thus, only h = 1 h=1 h = 1 , i.e., W ∈ Q W\in\mathcal{Q} W ∈ Q , contributes. Then S [ 1 ] = S S^{[1]}=S S [ 1 ] = S and all character factors in (2.5) equal one.
(iii) Summing (2.5) over t t t , χ 3 \chi_{3} χ 3 and W ∈ Q W\in\mathcal{Q} W ∈ Q , we obtain
∑ χ 3 N M 1 , χ 1 , M 2 , χ 2 = n ⋅ n ⋅ ∣ G M 3 ∣ ∣ G M 1 ∣ ∣ G M 2 ∣ ∣ G M 3 ∣ T ( M 1 , M 2 , M 3 ) = ∣ O 1 ∣ ∣ O 2 ∣ T ( M 1 , M 2 , M 3 ) , ( M 3 ) χ 3 \sum_{\chi_{3}}N^{(M_{3})_{\chi_{3}}}_{M_{1,\chi_{1},M_{2,\chi_{2}}}
=\frac{n\cdot n\cdot|G_{M_{3}}|}{|G_{M_{1}}||G_{M_{2}}||G_{M_{3}}|}T(M_{1},M_{2},M_{3})=|O_{1}||O_{2}|T(M_{1},M_{2},M_{3}),} χ 3 ∑ N M 1 , χ 1 , M 2 , χ 2 = ∣ G M 1 ∣∣ G M 2 ∣∣ G M 3 ∣ n ⋅ n ⋅ ∣ G M 3 ∣ T ( M 1 , M 2 , M 3 ) = ∣ O 1 ∣∣ O 2 ∣ T ( M 1 , M 2 , M 3 ) , ( M 3 ) χ 3
since n / ∣ G M i ∣ = ∣ O i ∣ n/|G_{M_{i}}|=|O_{i}| n /∣ G M i ∣ = ∣ O i ∣ with one factor n n n coming from ∑ t \sum_{t} ∑ t , the other from S U ‾ , W t U = n − 1 S V ‾ 0 , W S^{U}_{\overline{U},W_{t}}=n^{-1}S_{\overline{V}_{0},W} S U , W t U = n − 1 S V 0 , W . Together with (2.4) this gives (i).
3 ) 3) 3 ) Individual coefficients: if M 1 , M 2 M_{1},M_{2} M 1 , M 2 are stable, ∣ O 1 ∣ = ∣ O 2 ∣ = 1 |O_{1}|=|O_{2}|=1 ∣ O 1 ∣ = ∣ O 2 ∣ = 1 , and (i) holds. If M 2 , M 3 M_{2},M_{3} M 2 , M 3 are stable, then for a ∈ G a\in G a ∈ G , by Proposition 2.4 (d) applied with h = a − 1 h=a^{-1} h = a − 1 , N M 1 ∘ a , M 2 M 3 = N M 1 , M 2 ∘ a − 1 M 3 ∘ a − 1 = N M 1 , M 2 M 3 N^{M_{3}}_{M_{1}\circ a,M_{2}}=N^{M_{3}\circ a^{-1}}_{M_{1},M_{2}\circ a^{-1}}=N^{M_{3}}_{M_{1},M_{2}} N M 1 ∘ a , M 2 M 3 = N M 1 , M 2 ∘ a − 1 M 3 ∘ a − 1 = N M 1 , M 2 M 3 since M 2 ∘ a − 1 ≃ M 2 M_{2}\circ a^{-1}\simeq M_{2} M 2 ∘ a − 1 ≃ M 2 and M 3 ∘ a − 1 ≃ M 3 M_{3}\circ a^{-1}\simeq M_{3} M 3 ∘ a − 1 ≃ M 3 . Thus, the left hand side of (2.4) equals ∣ O 1 ∣ N M 1 , M 2 M 3 |O_{1}|N^{M_{3}}_{M_{1},M_{2}} ∣ O 1 ∣ N M 1 , M 2 M 3 , with ∣ O 2 ∣ = 1 |O_{2}|=1 ∣ O 2 ∣ = 1 . The case when M 1 , M 3 M_{1},M_{3} M 1 , M 3 are stable is symmetric, i.e., N M 1 , M 2 ∘ b M 3 = N M 1 ∘ b − 1 , M 2 M 3 ∘ b − 1 = N M 1 , M 2 M 3 N^{M_{3}}_{M_{1},M_{2}\circ b}=N^{M_{3}\circ b^{-1}}_{M_{1}\circ b^{-1},M_{2}}=N^{M_{3}}_{M_{1},M_{2}} N M 1 , M 2 ∘ b M 3 = N M 1 ∘ b − 1 , M 2 M 3 ∘ b − 1 = N M 1 , M 2 M 3 .
Finally, let n = p a n=p^{a} n = p a . The residues of g 1 , g 2 , g 3 = g 1 g 2 g_{1},g_{2},g_{3}=g_{1}g_{2} g 1 , g 2 , g 3 = g 1 g 2 modulo p p p are ( r 1 , r 2 , r 1 + r 2 ) (r_{1},r_{2},r_{1}+r_{2}) ( r 1 , r 2 , r 1 + r 2 ) with ( r 1 , r 2 ) ≠ ( 0 , 0 ) (r_{1},r_{2})\ne(0,0) ( r 1 , r 2 ) = ( 0 , 0 ) , as ⟨ g 1 , g 2 ⟩ = G \langle g_{1},g_{2}\rangle=G ⟨ g 1 , g 2 ⟩ = G . If r 1 = 0 r_{1}=0 r 1 = 0 then g 2 , g 3 g_{2},g_{3} g 2 , g 3 have residue r 2 ≠ 0 r_{2}\ne0 r 2 = 0 . If r 2 = 0 r_{2}=0 r 2 = 0 then g 1 , g 3 g_{1},g_{3} g 1 , g 3 do. If r 1 , r 2 ≠ 0 r_{1},r_{2}\ne0 r 1 , r 2 = 0 and r 1 + r 2 = 0 r_{1}+r_{2}=0 r 1 + r 2 = 0 then g 1 , g 2 g_{1},g_{2} g 1 , g 2 are generators. In the remaining case all three are. In each case at least two of g 1 , g 2 , g 3 g_{1},g_{2},g_{3} g 1 , g 2 , g 3 generate G = Z / p a G=\mathbb{Z}/p^{a} G = Z / p a , and a module twisted by a generator g g g is stable since g ∈ G M g\in G_{M} g ∈ G M , [12 ] . □ \square □
Remark 2.12. Take n = 4 n=4 n = 4 and ( g 1 , g 2 , g 3 ) = ( σ 2 , σ , σ 3 ) (g_{1},g_{2},g_{3})=(\sigma^{2},\sigma,\sigma^{3}) ( g 1 , g 2 , g 3 ) = ( σ 2 , σ , σ 3 ) , with M 1 ∈ M ( σ 2 ) M_{1}\in\mathcal{M}(\sigma^{2}) M 1 ∈ M ( σ 2 ) non-stable, ( G M 1 = ⟨ σ 2 ⟩ , ∣ O 1 ∣ = 2 ) (G_{M_{1}}=\langle\sigma^{2}\rangle, |O_{1}|=2) ( G M 1 = ⟨ σ 2 ⟩ , ∣ O 1 ∣ = 2 ) and M 2 , M 3 M_{2},M_{3} M 2 , M 3 twisted by generators and, hence, stable. In 2), h h h runs over G M 1 ∩ G M 2 ∩ G M 3 = ⟨ σ 2 ⟩ G_{M_{1}}\cap G_{M_{2}}\cap G_{M_{3}}=\langle\sigma^{2}\rangle G M 1 ∩ G M 2 ∩ G M 3 = ⟨ σ 2 ⟩ . χ 3 \chi_{3} χ 3 runs over the four characters of G M 3 = G G_{M_{3}}=G G M 3 = G and ∑ χ 3 χ 3 ( h ) = 4 δ h , 1 \sum_{\chi_{3}}\chi_{3}(h)=4\delta_{h,1} ∑ χ 3 χ 3 ( h ) = 4 δ h , 1 . The prefactor in (iii) is n 2 8 = 2 = ∣ O 1 ∣ ∣ O 2 ∣ \frac{n^{2}}{8}=2=|O_{1}||O_{2}| 8 n 2 = 2 = ∣ O 1 ∣∣ O 2 ∣ . In 3), the case when M 2 , M 3 M_{2},M_{3} M 2 , M 3 are stable, the left side of (2.4) is N M 1 , M 2 M 3 + N M 1 ∘ σ , M 2 M 3 = 2 N M 1 , M 2 M 3 N^{M_{3}}_{M_{1},M_{2}}+N^{M_{3}}_{M_{1}\circ\sigma,M_{2}}=2N^{M_{3}}_{M_{1},M_{2}} N M 1 , M 2 M 3 + N M 1 ∘ σ , M 2 M 3 = 2 N M 1 , M 2 M 3 . Hence, N M 1 , M 2 M 3 = T ( M 1 , M 2 , M 3 ) N^{M_{3}}_{M_{1},M_{2}}=T(M_{1},M_{2},M_{3}) N M 1 , M 2 M 3 = T ( M 1 , M 2 , M 3 ) .
Remark 2.13. The remining open quesion: for n n n divisible by two distinct primes there are triples with at most one stable module (e.g., n = 6 n=6 n = 6 , ( g 1 , g 2 , g 3 ) = ( σ 2 , σ 3 , σ 5 ) (g_{1},g_{2},g_{3})=(\sigma^{2},\sigma^{3},\sigma^{5}) ( g 1 , g 2 , g 3 ) = ( σ 2 , σ 3 , σ 5 ) , M 1 , M 2 M_{1},M_{2} M 1 , M 2 are unstable). There (i) gives only the orbit average over O 1 × O 2 O_{1}\times O_{2} O 1 × O 2 . Whether the individual coefficients are given by T is not decided here (by Lemma 2.9 (a) the summand T is constant on orbits, thus the claim would imply that N M 1 , M 2 ∘ b M 3 N^{M_{3}}_{M_{1},M_{2}\circ b} N M 1 , M 2 ∘ b M 3 does not depend on $b). For non-cyclic abelian groups, projective representations of stabilizers appear and the argument must be modified.
Next proposition, the second main result of this paper, describes unitarity for generating sectors.
Proposition 2.14. Adopt both assumptions of Subsection 2.1 and let g g g generate G G G . Then ∣ M ( g ) ∣ = ∣ Q ∣ |\mathcal{M}(g)|=|\mathcal{Q}| ∣ M ( g ) ∣ = ∣ Q ∣ and the square matrix ( S X , Q ) X ∈ M ( g ) , Q ∈ Q (S_{X,Q})_{X\in\mathcal{M}(g),Q\in\mathcal{Q}} ( S X , Q ) X ∈ M ( g ) , Q ∈ Q is unitary. In particular, for X , X ′ ∈ M ( g ) X,X'\in\mathcal{M}(g) X , X ′ ∈ M ( g ) and Q , Q ′ ∈ Q Q,Q'\in\mathcal{Q} Q , Q ′ ∈ Q ,
∑ Q ∈ Q S X , Q S X ′ , Q ‾ = δ X , X ′ , ∑ X ∈ M ( g ) S X , Q ‾ S X , Q ′ = δ Q , Q ′ . \sum_{Q\in\mathcal{Q}}S_{X,Q}\overline{S_{X',Q}}=\delta_{X,X'},\qquad\sum_{X\in\mathcal{M}(g)}\overline{S_{X,Q}}S_{X,Q'}=\delta_{Q,Q'}. Q ∈ Q ∑ S X , Q S X ′ , Q = δ X , X ′ , X ∈ M ( g ) ∑ S X , Q S X , Q ′ = δ Q , Q ′ . More generally ∑ N ∈ M ( h ) , s t a b l e S X , N [ h ] S X ′ , N [ h ] ‾ = δ X , X ′ \sum_{N\in\mathcal{M}(h),\,\mathrm{stable}}S^{[h]}_{X,N}\overline{S^{[h]}_{X',N}}=\delta_{X,X'} ∑ N ∈ M ( h ) , stable S X , N [ h ] S X ′ , N [ h ] = δ X , X ′ for every h ∈ G h\in G h ∈ G .
Proof. Orthonormality of rows: let X , X ′ ∈ M ( g ) X,X'\in\mathcal{M}(g) X , X ′ ∈ M ( g ) . They are stable, g ∈ G X g\in G_{X} g ∈ G X , ⟨ g ⟩ = G \langle g\rangle=G ⟨ g ⟩ = G . Thus, all n n n components X s X_{s} X s , X s ′ ′ X'_{s'} X s ′ ′ , exist. Unitarity of S U S^{U} S U gives ∑ Y S X s , Y U S X s ′ ′ , Y U ‾ = δ X , X ′ δ s , s ′ \sum_{Y}S^{U}_{X_{s},Y}\overline{S^{U}_{X'_{s'},Y}}=\delta_{X,X'}\delta_{s,s'} ∑ Y S X s , Y U S X s ′ ′ , Y U = δ X , X ′ δ s , s ′ since distinct stable modules in M ( g ) \mathcal{M}(g) M ( g ) give inequivalent U U U -modules, see Proposition 2.6 (iii). By Lemma 2.8 (a) only Y Y Y with g ∈ G Y g \in G_Y g ∈ G Y , i.e., G Y = G G_Y = G G Y = G , contribute: Y = N t Y = N_t Y = N t , N ∈ M ( h ) N \in\mathcal{M}(h) N ∈ M ( h ) stable, h ∈ G h \in G h ∈ G , t ∈ Z / n t \in\mathbb{Z}/n t ∈ Z / n , each such Y Y Y once. By Lemma 2.8 (b) with G X = G X ′ = G G_X = G_{X'} = G G X = G X ′ = G , S X , N t U = n − 1 ψ s ( h ) ‾ ψ t ( g − 1 ) S X , N [ h ] S^{U}_{X,N_t} = n^{-1}\overline{\psi_s(h)}\psi_t(g^{-1})S^{[h]}_{X,N} S X , N t U = n − 1 ψ s ( h ) ψ t ( g − 1 ) S X , N [ h ] . Thus, the left hand side equals
n − 2 ∑ h ∈ G ψ s ( h ) ‾ ψ s ′ ( h ) A h ∑ t ∈ Z / n ∣ ψ t ( g − 1 ) ∣ 2 = n − 1 ∑ h ∈ G ψ s ′ − s ( h ) A h , A h : = ∑ N ∈ M ( h ) s t a b l e S X , N [ h ] S X ′ , N [ h ] ‾ . n^{-2}\sum_{h\in G}\overline{\psi_s(h)}\psi_{s'}(h)A_h\sum_{t\in\mathbb{Z}/n}\left|\psi_t(g^{-1})\right|^2=n^{-1}\sum_{h\in G}\psi_{s'-s}(h)A_h,\qquad A_h:=\sum_{N\in\mathcal{M}(h)\ \mathrm{stable}}S^{[h]}_{X,N}\overline{S^{[h]}_{X',N}}. n − 2 h ∈ G ∑ ψ s ( h ) ψ s ′ ( h ) A h t ∈ Z / n ∑ ψ t ( g − 1 ) 2 = n − 1 h ∈ G ∑ ψ s ′ − s ( h ) A h , A h := N ∈ M ( h ) stable ∑ S X , N [ h ] S X ′ , N [ h ] .
where complex conjugation acts on the second factor of each product. For X = X ′ X = X' X = X ′ the left side is δ s s ′ \delta_{ss'} δ s s ′ and for X ≠ X ′ X \ne X' X = X ′ it is zero. With a = s ′ − s a = s' - s a = s ′ − s ranging over Z / n \mathbb{Z}/n Z / n this gives ∑ h ψ a ( h ) A h = n δ X X ′ δ a , 0 \sum_h \psi_a(h)A_h = n\delta_{XX'}\delta_{a,0} ∑ h ψ a ( h ) A h = n δ X X ′ δ a , 0 for all a a a , hence, A h = δ X X ′ A_h = \delta_{XX'} A h = δ X X ′ for all h ∈ G h \in G h ∈ G by Fourier inversion on G G G , i.e., A h = n − 1 ∑ a ψ a ( h ) n δ X X ′ δ a , 0 = δ X X ′ A_h = n^{-1}\sum_a \psi_a(h)n\delta_{XX'}\delta_{a,0} = \delta_{XX'} A h = n − 1 ∑ a ψ a ( h ) n δ X X ′ δ a , 0 = δ X X ′ . For h = 1 h = 1 h = 1 , A 1 = ∑ Q ∈ Q S X , Q S X ′ , Q A_1 = \sum_{Q\in\mathcal{Q}}S_{X,Q}S_{X',Q} A 1 = ∑ Q ∈ Q S X , Q S X ′ , Q .
Denote by W ( g , h ) W_{(g,h)} W ( g , h ) the span of the functions Z M ( v , ( g , h ) , τ ) Z_M(v,(g,h),\tau) Z M ( v , ( g , h ) , τ ) , M ∈ M ( g , h ) M \in\mathcal{M}(g,h) M ∈ M ( g , h ) with v ∈ V ‾ 0 v \in\overline{V}_0 v ∈ V 0 , as in (2.2). By [16 , 12 ] , the functions of all sectors together are linearly independent. The dimension of their span is ∑ ( g , h ) ∣ M ( g , h ) ∣ \sum_{(g,h)}|\mathcal{M}(g,h)| ∑ ( g , h ) ∣ M ( g , h ) ∣ . In particular, dim W ( g , h ) = ∣ M ( g , h ) ∣ \dim W_{(g,h)} = |\mathcal{M}(g,h)| dim W ( g , h ) = ∣ M ( g , h ) ∣ for each commuting pair, with M ( g , h ) \mathcal{M}(g,h) M ( g , h ) the set of isomorphism classes of individual modules, so that, for sectors where some module is not h h h -stable or has non-trivial orbit, the count is that of [12 ] . By the transformation law, ρ ( S ) \rho(S) ρ ( S ) maps W ( g , h ) W_{(g,h)} W ( g , h ) into W ( h , g − 1 ) W_{(h,g^{-1})} W ( h , g − 1 ) , and ρ ( S ) − 1 = ρ ( S − 1 ) \rho(S)^{-1} = \rho(S^{-1}) ρ ( S ) − 1 = ρ ( S − 1 ) maps W ( h , g − 1 ) W_{(h,g^{-1})} W ( h , g − 1 ) into W ( g , h ) W_{(g,h)} W ( g , h ) . Indeed, we apply the law to S − 1 = ( 0 1 − 1 0 ) S^{-1} = \begin{pmatrix}0&1\\-1&0\end{pmatrix} S − 1 = ( 0 − 1 1 0 ) , which sends ( h , g − 1 ) (h,g^{-1}) ( h , g − 1 ) to ( g , h ) (g,h) ( g , h ) . Thus, ρ ( S ) \rho(S) ρ ( S ) restricts to a linear isomorphism W ( g , h ) → W ( h , g − 1 ) W_{(g,h)} \to W_{(h,g^{-1})} W ( g , h ) → W ( h , g − 1 ) , and ∣ M ( g , h ) ∣ = ∣ M ( h , g − 1 ) ∣ |\mathcal{M}(g,h)| = |\mathcal{M}(h,g^{-1})| ∣ M ( g , h ) ∣ = ∣ M ( h , g − 1 ) ∣ . For h = 1 h = 1 h = 1 and g g g a generator this reads ∣ M ( g ) ∣ = ∣ M ( 1 , g ) ∣ |\mathcal{M}(g)| = |\mathcal{M}(1,g)| ∣ M ( g ) ∣ = ∣ M ( 1 , g ) ∣ . Here M ( g , 1 ) = M ( g ) \mathcal{M}(g,1) = \mathcal{M}(g) M ( g , 1 ) = M ( g ) (all modules are stable, thus orbits are singletons) and M ( 1 , g ) = Q \mathcal{M}(1,g) = \mathcal{Q} M ( 1 , g ) = Q since g g g generates G G G . A square matrix with orthonormal rows is unitary, thus its columns are orthonormal. □
Let A = V ‾ 0 ⊕ J A = \overline{V}_0 \oplus J A = V 0 ⊕ J be the superalgebra in Rep ( V ‾ 0 ) \operatorname{Rep}(\overline{V}_0) Rep ( V 0 ) . For a simple M ∈ Rep ( V ‾ 0 ) M \in\operatorname{Rep}(\overline{V}_0) M ∈ Rep ( V 0 ) set F V ( M ) = M ⊕ J ⊠ M F_V(M) = M \oplus J \boxtimes M F V ( M ) = M ⊕ J ⊠ M with M M M even and J ⊠ M J \boxtimes M J ⊠ M odd. If { X , J ⊠ X } \{X,J \boxtimes X\} { X , J ⊠ X } is an orbit we call X 0 : = X X^0 := X X 0 := X an even representative. Then F V ( J ⊠ X ) = Π F V ( X ) F_V(J \boxtimes X) = \Pi F_V(X) F V ( J ⊠ X ) = Π F V ( X ) , the same module with opposite grading. In this paper, N X 1 , X 2 X 3 N^{X_3}_{X_1,X_2} N X 1 , X 2 X 3 always denotes a fusion coefficient of Rep ( V ‾ 0 ) \operatorname{Rep}(\overline{V}_0) Rep ( V 0 ) (see Definition 2.5). Quantities for V V V -modules appear under the assumptions below.
Minimal assumptions (Theorem 4.1 (b))
First, for every simple M ∈ Rep ( V ‾ 0 ) M \in\operatorname{Rep}(\overline{V}_0) M ∈ Rep ( V 0 ) of twist g g g , F V ( M ) F_V(M) F V ( M ) is a Z / 2 Z \mathbb{Z}/2\mathbb{Z} Z /2 Z -graded V V V -module, g g g -twisted if ϵ M = 1 \epsilon_M = 1 ϵ M = 1 and g P gP g P -twisted if ϵ M = − 1 \epsilon_M = -1 ϵ M = − 1 (see Definition 3.3), this includes J J J -fixed M M M .
Second, for simple X 1 X_1 X 1 , X 2 X_2 X 2 , X 3 X_3 X 3 with g 3 = g 1 g 2 g_3 = g_1g_2 g 3 = g 1 g 2 and even representatives X ‾ i = X i \overline{X}_i = X_i X i = X i , the space I ( F V ( X 3 ) ; F V ( X 1 ) , F V ( X 2 ) ) I(F_V(X_3);F_V(X_1),F_V(X_2)) I ( F V ( X 3 ) ; F V ( X 1 ) , F V ( X 2 )) of intertwining operators is Z / 2 Z \mathbb{Z}/2\mathbb{Z} Z /2 Z -graded (an operator Y \mathcal{Y} Y has parity a a a if Y ( u , z ) w ∈ F V ( X 3 ) ∣ u ∣ + ∣ w ∣ + a \mathcal{Y}(u,z)w \in F_V(X_3)^{|u|+|w|+a} Y ( u , z ) w ∈ F V ( X 3 ) ∣ u ∣ + ∣ w ∣ + a ), and its even part has dimension N X 1 , X 2 X 3 N^{X_3}_{X_1,X_2} N X 1 , X 2 X 3 , its odd part N X 1 , X 2 J ⊠ X 3 N^{J\boxtimes X_3}_{X_1,X_2} N X 1 , X 2 J ⊠ X 3 .
Classification and monoidality (Corollary 4.2)
First: if J ⊠ M ≇ M J \boxtimes M \not\cong M J ⊠ M ≅ M , F V ( M ) F_V(M) F V ( M ) is simple as a graded module, and every simple graded g g g - or g P gP g P -twisted V V V -module N N N with N ~ ≇ J ⊠ N ~ \widetilde{N} \not\cong J \boxtimes\widetilde{N} N ≅ J ⊠ N is isomorphic to some Π i F V ( M ) \Pi^iF_V(M) Π i F V ( M ) for a unique free orbit.
Second: on the full subcategory Rep ( V ) f r e e \operatorname{Rep}(V)^{\mathrm{free}} Rep ( V ) free generated by the Π i F V ( M ) \Pi^iF_V(M) Π i F V ( M ) , M M M free, a tensor product ⊠ V \boxtimes_V ⊠ V exists, and F V ( M ) ⊠ V F V ( M ′ ) ≃ F V ( M ⊠ M ′ ) F_V(M)\boxtimes_V F_V(M') \simeq F_V(M\boxtimes M') F V ( M ) ⊠ V F V ( M ′ ) ≃ F V ( M ⊠ M ′ ) whenever every simple part of M ⊠ M ′ M\boxtimes M' M ⊠ M ′ is J J J -free.
Li transport for superalgebras (Section 6)
Let h ∈ V ‾ 0 h \in\overline{V}_0 h ∈ V 0 be as in Section 6.
First: for every irreducible (parity-untwisted, parity-twisted, or twisted) V V V -module M M M , M ( h ) = ( M , Y M ( Δ ( h , z ) , z ) ) M^{(h)} = (M,Y_M(\Delta(h,z),z)) M ( h ) = ( M , Y M ( Δ ( h , z ) , z )) is a module twisted by ϕ ( h ) = e − 2 π i h 0 \phi(h) = e^{-2\pi i h_0} ϕ ( h ) = e − 2 π i h 0 (composed with the same parity twist), and Δ ( h 1 , z ) Δ ( h 2 , z ) = Δ ( h 1 + h 2 , z ) \Delta(h_1,z)\Delta(h_2,z) = \Delta(h_1+h_2,z) Δ ( h 1 , z ) Δ ( h 2 , z ) = Δ ( h 1 + h 2 , z ) .
Second: an intertwining operator Y \mathcal{Y} Y of type ( M 3 M 1 M 2 ) \binom{M_3}{M_1\ M_2} ( M 1 M 2 M 3 ) gives an intertwining operator V ( Δ ( h , z ) , z ) \mathcal{V}(\Delta(h,z),z) V ( Δ ( h , z ) , z ) of type ( M 3 ( h ) M 1 ( h ) M 2 ( h ) ) \binom{M_3^{(h)}}{M_1^{(h)}\ M_2^{(h)}} ( M 1 ( h ) M 2 ( h ) M 3 ( h ) ) .
Third: (Two twisted arguments.) N M 1 ( a h ) , M 2 ( b h ) M 3 ( a + b ) h = N M 1 , M 2 M 3 N^{M_3^{(a+b)h}}_{M_1^{(ah)},M_2^{(bh)}} = N^{M_3}_{M_1,M_2} N M 1 ( ah ) , M 2 ( bh ) M 3 ( a + b ) h = N M 1 , M 2 M 3 , where M ( n h ) M^{(nh)} M ( nh ) denotes the module obtained by the full spectral flow, an untwisted module. The corresponding identification of spaces of intertwining operators respects the Z / 2 Z \mathbb{Z}/2\mathbb{Z} Z /2 Z -gradings.
Note that conditions 2.5.3, 2.5.2, 2.5.1 are additional conditions and are not consequences of two assumptions of Subsection 2.1 or of the cited results. For g = 1 g = 1 g = 1 , conditions 2.5.1-2.5.2 are established in [30 , 8 ] . For vertex operator algebras, the first and second conditions of 2.5.3 are theorems [33 , 14 ] . Assumptions 2.5.3 state them for superalgebras. The third condition of 2.5.3 is not covered by those results (where only two of the three arguments are twisted). For inner automorphisms, conditions 2.5.1–2.5.2 reduce to the untwisted theory under assumptions 2.5.3 (see Proposition 6.1 ).
Sign data and equivariance of the orbifold S S S -matrix
First, let us list conventions for this section:
(i) For W ∈ { V ‾ 0 , U } W \in\{\overline{V}_0,U\} W ∈ { V 0 , U } , S W S^W S W is the S S S -matrix of C W \mathcal{C}_W C W defined by modular transformation (for W = U W=U W = U : of the functions Z Y ( v , τ ) Z_Y(v,\tau) Z Y ( v , τ ) , v ∈ U v \in U v ∈ U ). An S S S without superscript is the untwisted S S S -matrix of V ‾ 0 \overline{V}_0 V 0 or an orbifold entry S X , Q S_{X,Q} S X , Q (known from the indices).
(ii) θ A = e 2 π i h A \theta_A=e^{2\pi i h_A} θ A = e 2 π i h A .
(iii) For simple Z Z Z , Y Y Y in a modular category C W \mathcal{C}_W C W set χ Y ( Z ) : = S Z , Y / S 0 , Y \chi_Y(Z):=S_{Z,Y}/S_{0,Y} χ Y ( Z ) := S Z , Y / S 0 , Y . By the Verlinde formula this is a character of the fusion ring, thus S A ⊠ B , Y W = χ Y ( A ) S B , Y W S^W_{A\boxtimes B,Y}=\chi_Y(A)S^W_{B,Y} S A ⊠ B , Y W = χ Y ( A ) S B , Y W .
(iv) Symmetry is used only for S U S^U S U and for the untwisted S S S of V ‾ 0 \overline{V}_0 V 0 .
(v) The quantities d d d , κ \kappa κ , ϵ \epsilon ϵ , η \eta η , π \pi π of Definition 3.3 are ratios of S S S -matrix entries. Their interpretation in terms of V V V -modules requires conditions 2.5.1.
The lemma below provides the Gauss identity.
Lemma 3.1. Let W W W be a simple, rational, C 2 C_2 C 2 -cofinite, self-dual vertex operator algebra of CFT type (we use W = U W=U W = U and W = V ‾ 0 W=\overline{V}_0 W = V 0 ) with irreducible modules a a a , b b b , … \ldots … , vacuum module 0 = W \mathbf{0}=W 0 = W , conformal weights h a h_a h a , twists θ a = e 2 π i h a \theta_a=e^{2\pi i h_a} θ a = e 2 π i h a , S = S W S=S^W S = S W and fusion coefficients N a b f N^f_{ab} N ab f . Then
S a b = θ a θ b ∑ f N a b f θ f − 1 S f 0 S_{ab}=\theta_a\theta_b\sum_f N^f_{ab}\theta_f^{-1}S_{f0} S ab = θ a θ b f ∑ N ab f θ f − 1 S f 0 for all a a a , b b b .
With the opposite convention for T T T one replaces θ \theta θ by θ − 1 \theta^{-1} θ − 1 , and the central charge drops out.
Proof. Let T ^ = diag ( γ θ a ) \widehat{T}=\operatorname{diag}(\gamma\theta_a) T = diag ( γ θ a ) , γ = e − 2 π i c / 24 \gamma=e^{-2\pi i c/24} γ = e − 2 π i c /24 . By Zhu theorem the characters of the irreducible W W W -modules are linearly independent and span an S L 2 ( Z ) \mathrm{SL}_2(\mathbb{Z}) SL 2 ( Z ) -representation in which S S S , T ^ \widehat{T} T are the matrices of τ ↦ − 1 / τ \tau\mapsto-1/\tau τ ↦ − 1/ τ , τ ↦ τ + 1 \tau\mapsto\tau+1 τ ↦ τ + 1 . Hence, ( S T ^ ) 3 = S 2 (S\widehat{T})^3=S^2 ( S T ) 3 = S 2 or ( T ^ S ) 3 = S 2 (\widehat{T}S)^3=S^2 ( T S ) 3 = S 2 , according to the order of composition. In the first case S T ^ S T ^ S T ^ = S S S\widehat{T}S\widehat{T}S\widehat{T}=SS S T S T S T = S S , and cancelling S S S on the left gives T ^ S T ^ S T ^ = S \widehat{T}S\widehat{T}S\widehat{T}=S T S T S T = S . In the second, T ^ S T ^ S T ^ S = S S \widehat{T}S\widehat{T}S\widehat{T}S=SS T S T S T S = S S , and cancelling S S S on the right gives the same relation. Multiplying T ^ S T ^ S T ^ = S \widehat{T}S\widehat{T}S\widehat{T}=S T S T S T = S on the left and on the right by T ^ − 1 \widehat{T}^{-1} T − 1 gives S T ^ S = T ^ − 1 S T ^ − 1 S\widehat{T}S=\widehat{T}^{-1}S\widehat{T}^{-1} S T S = T − 1 S T − 1 . Evaluate the ( a , b ) (a,b) ( a , b ) entry: since T ^ e = γ θ e \widehat{T}_e=\gamma\theta_e T e = γ θ e , we have
∑ e S a e S b e θ e = γ − 3 θ a − 1 θ b − 1 S a b . (6) \sum_e S_{ae}S_{be}\theta_e=\gamma^{-3}\theta_a^{-1}\theta_b^{-1}S_{ab}. \tag*{(6)} e ∑ S a e S b e θ e = γ − 3 θ a − 1 θ b − 1 S ab . (6) By the Verlinde formula, e ↦ S a e / S 0 e e\mapsto S_{ae}/S_{0e} e ↦ S a e / S 0 e is a character of the fusion ring, thus S a e S b e / S 0 e = ∑ f N a b f S f e S_{ae}S_{be}/S_{0e}=\sum_f N^f_{ab}S_{fe} S a e S b e / S 0 e = ∑ f N ab f S f e . Hence, using (6) for the pair ( f , 0 ) (f,0) ( f , 0 ) and θ 0 = 1 \theta_0=1 θ 0 = 1 , we get
∑ e S a e S b e θ e = ∑ f N a b f ∑ e S f e S 0 e θ e = γ − 3 ∑ f N a b f θ f − 1 S f 0 . \sum_e S_{ae}S_{be}\theta_e=\sum_f N^f_{ab}\sum_e S_{fe}S_{0e}\theta_e=\gamma^{-3}\sum_f N^f_{ab}\theta_f^{-1}S_{f0}. e ∑ S a e S b e θ e = f ∑ N ab f e ∑ S f e S 0 e θ e = γ − 3 f ∑ N ab f θ f − 1 S f 0 . Comparing with (6) (the factor γ − 3 \gamma^{-3} γ − 3 cancels) gives the statement of Lemma. □ \square □
Lemma 3.2. S X , V ‾ 0 ≠ 0 S_{X,\overline{V}_0}\ne0 S X , V 0 = 0 for every X ∈ R X\in\mathcal{R} X ∈ R , and S V ‾ 0 , Q ≠ 0 S_{\overline{V}_0,Q}\ne0 S V 0 , Q = 0 for every Q ∈ Q Q\in\mathcal{Q} Q ∈ Q .
Proof. In the modular category C U \mathcal{C}_U C U , S U , Y U = S U , U U dim q Y S^U_{U,Y}=S^U_{U,U}\dim_qY S U , Y U = S U , U U dim q Y with S U , U U ≠ 0 S^U_{U,U}\ne0 S U , U U = 0 and dim q Y ≠ 0 \dim_qY\ne0 dim q Y = 0 for every simple Y Y Y , since dimensions of simple objects of a fusion category are nonzero, and S U S^U S U is symmetric. Thus, S Y , U U = S U , Y U ≠ 0 S^U_{Y,U}=S^U_{U,Y}\ne0 S Y , U U = S U , Y U = 0 . The column V ‾ 0 ∈ Q \overline{V}_0\in\mathcal{Q} V 0 ∈ Q with t = 0 t=0 t = 0 is U U U (with normalization φ V ‾ 0 = σ \varphi_{\overline{V}_0}=\sigma φ V 0 = σ ). Lemma 2.8 (b) with t = 0 t=0 t = 0 gives, for every X ∈ M ( g ) X\in\mathcal{M}(g) X ∈ M ( g ) and any χ \chi χ , S X , U U = ∣ G X ∣ − 1 χ ( 1 ) ψ 0 ( g − 1 ) S X , V ‾ 0 = ∣ G X ∣ − 1 S X , V ‾ 0 S^U_{X,U}=|G_X|^{-1}\chi(1)\psi_0(g^{-1})S_{X,\overline{V}_0}=|G_X|^{-1}S_{X,\overline{V}_0} S X , U U = ∣ G X ∣ − 1 χ ( 1 ) ψ 0 ( g − 1 ) S X , V 0 = ∣ G X ∣ − 1 S X , V 0 , hence, S X , V ‾ 0 ≠ 0 S_{X,\overline{V}_0}\ne0 S X , V 0 = 0 . With X = V ‾ 0 X=\overline{V}_0 X = V 0 (g = 1 g=1 g = 1 , χ = ψ 0 \chi=\psi_0 χ = ψ 0 , X χ = U X_\chi=U X χ = U ) and column Q Q Q : S U , Q t U = n − 1 S V ‾ 0 , Q S^U_{U,Q_t}=n^{-1}S_{\overline{V}_0,Q} S U , Q t U = n − 1 S V 0 , Q , which is nonzero. □ \square □
Definition 3.3. With the normalizations φ J = σ ∣ J \varphi_J=\sigma|_J φ J = σ ∣ J , φ V ‾ 0 = σ ∣ V ‾ 0 \varphi_{\overline{V}_0}=\sigma|_{\overline{V}_0} φ V 0 = σ ∣ V 0 on the columns J , V ‾ 0 ∈ Q J,\overline{V}_0\in\mathcal{Q} J , V 0 ∈ Q , set
d : = S J , V ‾ 0 S V ‾ 0 , V ‾ 0 , κ X : = S X , J S X , V ‾ 0 ( X ∈ R ) , ϵ X : = κ X d , η Q : = κ Q ( Q ∈ Q ) , d:=\frac{S_{J,\overline{V}_0}}{S_{\overline{V}_0,\overline{V}_0}},\qquad
\kappa_X:=\frac{S_{X,J}}{S_{X,\overline{V}_0}}\ (X\in\mathcal{R}),\qquad
\epsilon_X:=\frac{\kappa_X}{d},\qquad
\eta_Q:=\kappa_Q\ (Q\in\mathcal{Q}), d := S V 0 , V 0 S J , V 0 , κ X := S X , V 0 S X , J ( X ∈ R ) , ϵ X := d κ X , η Q := κ Q ( Q ∈ Q ) ,
and the parity type π ( X ) ∈ Z / 2 Z \pi(X) \in\mathbb{Z}/2\mathbb{Z} π ( X ) ∈ Z /2 Z by ( − 1 ) π ( X ) = ϵ X (-1)^{\pi(X)} = \epsilon_X ( − 1 ) π ( X ) = ϵ X . All equalities and sums of values of π \pi π are taken in Z / 2 Z \mathbb{Z}/2\mathbb{Z} Z /2 Z . Thus, π ( X 1 ) + π ( X 2 ) = π ( X 3 ) \pi(X_1) + \pi(X_2) = \pi(X_3) π ( X 1 ) + π ( X 2 ) = π ( X 3 ) iff ϵ X 1 ϵ X 2 ϵ X 3 = 1 \epsilon_{X_1}\epsilon_{X_2}\epsilon_{X_3} = 1 ϵ X 1 ϵ X 2 ϵ X 3 = 1 . These are ratios of S S S -matrix entries (well defined by Lemma 3.2). Lemma 3.5 (2) identifies them with values of characters of C U C_U C U , and conditions 2.5.1 gives ϵ X \epsilon_X ϵ X a module-theoretic meaning (parity type).
Induced normalization and uniform shift are given in Lemma 3.4.
Lemma 3.4. Let X ∈ M ( g ) X \in\mathcal{M}(g) X ∈ M ( g ) with stabilizer G X = ⟨ σ X ⟩ G_X = \langle\sigma_X \rangle G X = ⟨ σ X ⟩ of order m m m and fixed φ X \varphi_X φ X . Then J ⊠ X ∈ M ( g ) J \boxtimes X \in\mathcal{M}(g) J ⊠ X ∈ M ( g ) has G J ⊠ X = G X G_{J \boxtimes X} = G_X G J ⊠ X = G X , and:
(i) Let 0 ≠ Y ∈ I ( J ⊠ X ; J , X ) 0 \ne\mathcal{Y} \in I(J \boxtimes X; J, X) 0 = Y ∈ I ( J ⊠ X ; J , X ) (a one-dimensional space). There is a unique isomorphism θ : J ⊠ X → ( J ⊠ X ) ∘ σ X \theta: J \boxtimes X \to(J \boxtimes X) \circ\sigma_X θ : J ⊠ X → ( J ⊠ X ) ∘ σ X with
θ Y ( u , z ) w = Y ( σ X u , z ) φ X ( σ X ) w , u ∈ J , w ∈ X , (7) \theta\mathcal{Y}(u,z)w = \mathcal{Y}(\sigma_Xu,z)\varphi_X(\sigma_X)w,\qquad u \in J,\ w \in X, \tag*{(7)} θ Y ( u , z ) w = Y ( σ X u , z ) φ X ( σ X ) w , u ∈ J , w ∈ X , (7) and θ m = 1 \theta^m = 1 θ m = 1 . Hence, φ i n d ( σ X k ) : = θ k \varphi^{\mathrm{ind}}(\sigma_X^k) := \theta^k φ ind ( σ X k ) := θ k is a representation of G X G_X G X on J ⊠ X J \boxtimes X J ⊠ X (the induced normalization), satisfying (3.2) for all h ∈ G X h \in G_X h ∈ G X .
(ii) With respect to φ X \varphi_X φ X and φ i n d \varphi^{\mathrm{ind}} φ ind : J 0 ⊠ U X χ ≃ ( J ⊠ X ) χ J_0 \boxtimes_U X_\chi\simeq(J \boxtimes X)_\chi J 0 ⊠ U X χ ≃ ( J ⊠ X ) χ for every χ ∈ G ^ X \chi\in\widehat{G}_X χ ∈ G X (the label is not shifted).
(iii) If J ⊠ X J \boxtimes X J ⊠ X is given another normalization λ φ i n d \lambda\varphi^{\mathrm{ind}} λ φ ind , λ ∈ G ^ X \lambda\in\widehat{G}_X λ ∈ G X , then J 0 ⊠ U X χ ≃ ( J ⊠ X ) χ λ J_0 \boxtimes_U X_\chi\simeq(J \boxtimes X)_{\chi\lambda} J 0 ⊠ U X χ ≃ ( J ⊠ X ) χ λ for all χ \chi χ : the shift is the same character λ \lambda λ for all χ \chi χ .
Proof. Stabilizer: by Proposition 2.4 (d), ( J ⊠ X ) ∘ a ≃ ( J ∘ a ) ⊠ ( X ∘ a ) ≃ J ⊠ ( X ∘ a ) (J \boxtimes X) \circ a \simeq(J \circ a) \boxtimes(X \circ a) \simeq J \boxtimes(X \circ a) ( J ⊠ X ) ∘ a ≃ ( J ∘ a ) ⊠ ( X ∘ a ) ≃ J ⊠ ( X ∘ a ) as J J J is stable, and J ⊠ ( ⋅ ) J \boxtimes(\cdot) J ⊠ ( ⋅ ) is an equivalence. Thus, J ⊠ ( X ∘ a ) ≃ J ⊠ X J \boxtimes(X \circ a) \simeq J \boxtimes X J ⊠ ( X ∘ a ) ≃ J ⊠ X iff X ∘ a ≃ X X \circ a \simeq X X ∘ a ≃ X .
One-dimensionality: J J J is invertible in Rep ( V ‾ 0 ) \operatorname{Rep}(\overline{V}_0) Rep ( V 0 ) , thus J ⊠ X J \boxtimes X J ⊠ X is simple and N J , X J ⊠ X = 1 N^{J\boxtimes X}_{J,X} = 1 N J , X J ⊠ X = 1 . By Proposition 2.4 (b), dim I ( J ⊠ X ; J , X ) = 1 \dim I(J \boxtimes X; J, X) = 1 dim I ( J ⊠ X ; J , X ) = 1 .
(i) Existence and uniqueness: let Θ \Theta Θ be the set of isomorphisms θ 0 : J ⊠ X → ( J ⊠ X ) ∘ σ X \theta_0 : J \boxtimes X \to(J \boxtimes X) \circ\sigma_X θ 0 : J ⊠ X → ( J ⊠ X ) ∘ σ X . It is nonempty since J ⊠ X J \boxtimes X J ⊠ X is G X G_X G X -stable, and is a C × \mathbb{C}^{\times} C × -torsor. The same map Y \mathcal{Y} Y is an intertwining operator of type ( ( J ⊠ X ) ∘ σ X J ∘ σ X X ∘ σ X ) \left(\begin{smallmatrix}(J\boxtimes X)\circ\sigma_X\\ J\circ\sigma_X\quad X\circ\sigma_X\end{smallmatrix}\right) ( ( J ⊠ X ) ∘ σ X J ∘ σ X X ∘ σ X ) , [12 ] . Explicitly, M ∘ h M \circ h M ∘ h has vertex operators Y M ∘ h ( v , z ) = Y M ( h v , z ) Y_{M\circ h}(v,z) = Y_M(hv,z) Y M ∘ h ( v , z ) = Y M ( h v , z ) . The Jacobi identity for Y \mathcal{Y} Y is, for each v ∈ V ‾ 0 v \in\overline{V}_0 v ∈ V 0 homogeneous for the automorphisms occurring in it (a joint eigenvector), an identity between Y M 3 ( v , ⋅ ) Y_{M_3}(v,\cdot) Y M 3 ( v , ⋅ ) , Y M 2 ( v , ⋅ ) Y_{M_2}(v,\cdot) Y M 2 ( v , ⋅ ) , Y M 1 ( v , ⋅ ) Y_{M_1}(v,\cdot) Y M 1 ( v , ⋅ ) and Y \mathcal{Y} Y . Since G G G is abelian, h v hv h v belongs to the same joint eigenspace as v v v . Thus, the identity for h v hv h v , valid for the original modules, is the identity for v v v in the h h h -conjugated modules. Composing with the isomorphisms φ J ( σ X ) = σ X ∣ J : J → J ∘ σ X \varphi_J(\sigma_X) = \sigma_X|_J : J \to J \circ\sigma_X φ J ( σ X ) = σ X ∣ J : J → J ∘ σ X and φ X ( σ X ) : X → X ∘ σ X \varphi_X(\sigma_X) : X \to X \circ\sigma_X φ X ( σ X ) : X → X ∘ σ X , and then composing with θ 0 − 1 \theta_0^{-1} θ 0 − 1 gives Y θ 0 ( u , z ) w : = θ 0 − 1 Y ( σ X u , z ) φ X ( σ X ) w ∈ I ( J ⊠ X ; J , X ) \mathcal{Y}^{\theta_0}(u,z)w := \theta_0^{-1}\mathcal{Y}(\sigma_Xu,z)\varphi_X(\sigma_X)w \in I(J \boxtimes X; J, X) Y θ 0 ( u , z ) w := θ 0 − 1 Y ( σ X u , z ) φ X ( σ X ) w ∈ I ( J ⊠ X ; J , X ) which is nonzero. Hence, Y θ 0 = c ( θ 0 ) Y \mathcal{Y}^{\theta_0} = c(\theta_0)\mathcal{Y} Y θ 0 = c ( θ 0 ) Y with c ( θ 0 ) ∈ C × c(\theta_0) \in\mathbb{C}^{\times} c ( θ 0 ) ∈ C × . For μ ∈ C × \mu\in\mathbb{C}^{\times} μ ∈ C × , c ( μ θ 0 ) = μ − 1 c ( θ 0 ) c(\mu\theta_0) = \mu^{-1}c(\theta_0) c ( μ θ 0 ) = μ − 1 c ( θ 0 ) . Thus θ : = c ( θ 0 ) θ 0 \theta:= c(\theta_0)\theta_0 θ := c ( θ 0 ) θ 0 is the unique element of Θ \Theta Θ with c ( θ ) = 1 c(\theta) = 1 c ( θ ) = 1 , which is (3.2). Uniqueness can also be seen: two solutions differ by a scalar μ \mu μ with ( μ − 1 ) Y = 0 (\mu- 1)\mathcal{Y} = 0 ( μ − 1 ) Y = 0 .
The relation θ m = 1 \theta^m = 1 θ m = 1 : iterate (3.2) by applying it to ( σ X u , φ X ( σ X ) w ) (\sigma_Xu,\varphi_X(\sigma_X)w) ( σ X u , φ X ( σ X ) w ) . The functor ∘ σ X \circ\sigma_X ∘ σ X does not change underlying linear maps. Then it gives θ k Y ( u , z ) w = Y ( σ X k u , z ) φ X ( σ X ) k w \theta^k\mathcal{Y}(u,z)w = \mathcal{Y}(\sigma_X^ku,z)\varphi_X(\sigma_X)^kw θ k Y ( u , z ) w = Y ( σ X k u , z ) φ X ( σ X ) k w for k ≥ 1 k \ge1 k ≥ 1 , where θ k \theta^k θ k is the composite of k k k copies of θ \theta θ , a linear isomorphism J ⊠ X → ( J ⊠ X ) ∘ σ X k J \boxtimes X \to(J \boxtimes X)\circ\sigma_X^k J ⊠ X → ( J ⊠ X ) ∘ σ X k compatible with the module structures. For k = m k=m k = m we have σ X m = 1 \sigma_X^m = 1 σ X m = 1 . Thus, θ m \theta^m θ m is an endomorphism of the simple V ‾ 0 \overline{V}_0 V 0 -module J ⊠ X J \boxtimes X J ⊠ X , hence a scalar μ \mu μ by Schur lemma. Since σ X m = 1 \sigma_X^m = 1 σ X m = 1 on J J J and φ X ( σ X ) m = φ X ( 1 ) = 1 \varphi_X(\sigma_X)^m = \varphi_X(1) = 1 φ X ( σ X ) m = φ X ( 1 ) = 1 (φ X \varphi_X φ X is a representation of G X G_X G X ), θ m \theta^m θ m fixes coefficients of Y ( u , z ) w \mathcal{Y}(u,z)w Y ( u , z ) w which are not all zero, because Y ( u , z ) w ≠ 0 \mathcal{Y}(u,z)w \ne0 Y ( u , z ) w = 0 for u ≠ 0 ≠ w u \ne0 \ne w u = 0 = w by [12 ] . Hence, μ = 1 \mu= 1 μ = 1 and θ m = i d \theta^m = \mathrm{id} θ m = id .
(ii) For h ∈ G X h \in G_X h ∈ G X , u ∈ J 0 u \in J_0 u ∈ J 0 (h u = u hu = u h u = u ) and w ∈ X χ w \in X_\chi w ∈ X χ , (3.2) gives φ i n d ( h ) Y ( u , z ) w = χ ( h ) Y ( u , z ) w \varphi^{\mathrm{ind}}(h)\mathcal{Y}(u,z)w = \chi(h)\mathcal{Y}(u,z)w φ ind ( h ) Y ( u , z ) w = χ ( h ) Y ( u , z ) w . Thus, Y ∣ J 0 × X χ \mathcal{Y}|_{J_0\times X_\chi} Y ∣ J 0 × X χ is a nonzero U U U -intertwining operator with coefficients in ( J ⊠ X ) χ (J \boxtimes X)_\chi ( J ⊠ X ) χ , an irreducible U U U -module (see Proposition 2.6 (i)). Thus, N J 0 , X χ ( J ⊠ X ) χ ≥ 1 N^{(J\boxtimes X)_\chi}_{J_0,X_\chi} \ge1 N J 0 , X χ ( J ⊠ X ) χ ≥ 1 . Since J 0 J_0 J 0 is a simple current of C U C_U C U , J 0 ⊠ U X χ J_0 \boxtimes_U X_\chi J 0 ⊠ U X χ is simple and J 0 ⊠ U X χ ≃ ( J ⊠ X ) χ J_0 \boxtimes_U X_\chi\simeq(J \boxtimes X)_\chi J 0 ⊠ U X χ ≃ ( J ⊠ X ) χ .
(iii) Replacing φ i n d ( h ) \varphi^{\mathrm{ind}}(h) φ ind ( h ) by λ ( h ) \lambda(h) λ ( h ) multiplies the eigenvalue on every ( J ⊠ X ) χ (J \boxtimes X)_\chi ( J ⊠ X ) χ by λ ( h ) \lambda(h) λ ( h ) , i.e., relabels simultaneously χ ↦ χ λ \chi\mapsto\chi\lambda χ ↦ χ λ for all χ \chi χ .
Lemma 3.5. (1) d = dim q J = dim q J 0 ∈ { ± 1 } d = \dim_q J = \dim_q J_0 \in\{\pm1\} d = dim q J = dim q J 0 ∈ { ± 1 } .
(2) κ X = χ X ( J 0 ) ∈ { ± 1 } \kappa_X = \chi_X(J_0) \in\{\pm1\} κ X = χ X ( J 0 ) ∈ { ± 1 } for every X ∈ R X \in\mathcal{R} X ∈ R and every χ ∈ G ^ X \chi\in\widehat{G}_X χ ∈ G X . Hence ϵ X , η Q ∈ { ± 1 } \epsilon_X,\eta_Q \in\{\pm1\} ϵ X , η Q ∈ { ± 1 } .
(3) χ J 0 ( J 0 ) = d \chi_{J_0}(J_0) = d χ J 0 ( J 0 ) = d .
(4) κ J ⊠ X = κ X \kappa_{J\boxtimes X} = \kappa_X κ J ⊠ X = κ X . Hence, ϵ J ⊠ X = ϵ X \epsilon_{J\boxtimes X} = \epsilon_X ϵ J ⊠ X = ϵ X , π ( J ⊠ X ) = π ( X ) \pi(J \boxtimes X) = \pi(X) π ( J ⊠ X ) = π ( X ) , η J ⊠ Q = η Q \eta_{J\boxtimes Q} = \eta_Q η J ⊠ Q = η Q .
Proof. (1) Lemma 2.8 (b) with rows J J J (g = 1 g=1 g = 1 , χ = ψ 0 \chi=\psi_0 χ = ψ 0 , thus X χ = J 0 X_\chi=J_0 X χ = J 0 ) and column V ‾ 0 \overline{V}_0 V 0 (t = 0 t=0 t = 0 ): S J 0 , U U = n − 1 S J , V ‾ 0 S^U_{J_0,U}=n^{-1}S_{J,\overline{V}_0} S J 0 , U U = n − 1 S J , V 0 , S U , U U = n − 1 S V ‾ 0 , V ‾ 0 U S^U_{U,U}=n^{-1}S^U_{\overline{V}_0,\overline{V}_0} S U , U U = n − 1 S V 0 , V 0 U . Thus, d = S J 0 , U U / S U , U U = dim q J 0 d=S^U_{J_0,U}/S^U_{U,U}=\dim_q J_0 d = S J 0 , U U / S U , U U = dim q J 0 . Also, d = S J , V ‾ 0 / S V ‾ 0 , V ‾ 0 = dim q J d=S_{J,\overline{V}_0}/S_{\overline{V}_0,\overline{V}_0}=\dim_q J d = S J , V 0 / S V 0 , V 0 = dim q J in C V ‾ 0 C_{\overline{V}_0} C V 0 . As J 0 ⊠ U J 0 ≃ U J_0\boxtimes_U J_0\simeq U J 0 ⊠ U J 0 ≃ U , d 2 = 1 d^2=1 d 2 = 1 .
(2) Lemma 2.8 (b) with columns J J J and V ‾ 0 \overline{V}_0 V 0 at t = 0 t=0 t = 0 (where J 0 J_0 J 0 and U U U are the t = 0 t=0 t = 0 components) and the symmetry of S U S^U S U : κ X = S X χ , J 0 U S X χ , U U = S J 0 , X χ U S U , X χ U = χ X χ ( J 0 ) \kappa_X=\frac{S^U_{X_\chi,J_0}}{S^U_{X_\chi,U}}=\frac{S^U_{J_0,X_\chi}}{S^U_{U,X_\chi}}=\chi_{X_\chi}(J_0) κ X = S X χ , U U S X χ , J 0 U = S U , X χ U S J 0 , X χ U = χ X χ ( J 0 ) . Then χ ( J 0 ) 2 = χ ( J 0 ⊠ U J 0 ) = χ ( U ) = 1 \chi(J_0)^2=\chi(J_0\boxtimes_U J_0)=\chi(U)=1 χ ( J 0 ) 2 = χ ( J 0 ⊠ U J 0 ) = χ ( U ) = 1 .
(3) Apply Lemma 3.1 to W = U W=U W = U , a = b = J 0 a=b=J_0 a = b = J 0 : since J 0 ⊠ U J 0 ≃ U J_0\boxtimes_U J_0\simeq U J 0 ⊠ U J 0 ≃ U and θ J 0 = e 4 π i h J = 1 \theta_{J_0}=e^{4\pi i h_J}=1 θ J 0 = e 4 π i h J = 1 (2 h J ∈ Z 2h_J\in\mathbb{Z} 2 h J ∈ Z because V V V is 1 2 Z \frac{1}{2}\mathbb{Z} 2 1 Z -graded), S J 0 , J 0 U = S U , U U S^U_{J_0,J_0}=S^U_{U,U} S J 0 , J 0 U = S U , U U . By (1), S U , J 0 U = S J 0 , U U = d S U , U U S^U_{U,J_0}=S^U_{J_0,U}=dS^U_{U,U} S U , J 0 U = S J 0 , U U = d S U , U U . Hence, χ J 0 ( J 0 ) = S J 0 , J 0 U / S U , J 0 U = 1 / d = d \chi_{J_0}(J_0)=S^U_{J_0,J_0}/S^U_{U,J_0}=1/d=d χ J 0 ( J 0 ) = S J 0 , J 0 U / S U , J 0 U = 1/ d = d .
(4) By Lemma 3.4 (ii), ( J ⊠ X ) χ = J 0 ⊠ U X χ (J\boxtimes X)_\chi=J_0\boxtimes_U X_\chi ( J ⊠ X ) χ = J 0 ⊠ U X χ (with induced normalization). By (2) κ J ⊠ X = χ J 0 ⊠ U X χ ( J 0 ) = S J 0 , J 0 ⊠ U X χ U / S U , J 0 ⊠ U X χ U \kappa_{J\boxtimes X}=\chi_{J_0\boxtimes_U X_\chi}(J_0)=S^U_{J_0,J_0\boxtimes_U X_\chi}/S^U_{U,J_0\boxtimes_U X_\chi} κ J ⊠ X = χ J 0 ⊠ U X χ ( J 0 ) = S J 0 , J 0 ⊠ U X χ U / S U , J 0 ⊠ U X χ U . By symmetry, multiplicativity and (3): S J 0 , J 0 ⊠ X χ U = S J 0 ⊠ X χ , J 0 U = χ J 0 ( J 0 ) S X χ , J 0 U = d S J 0 , X χ U S^U_{J_0,J_0\boxtimes X_\chi}=S^U_{J_0\boxtimes X_\chi,J_0}=\chi_{J_0}(J_0)S^U_{X_\chi,J_0}=dS^U_{J_0,X_\chi} S J 0 , J 0 ⊠ X χ U = S J 0 ⊠ X χ , J 0 U = χ J 0 ( J 0 ) S X χ , J 0 U = d S J 0 , X χ U and S U , J 0 ⊠ X χ U = S J 0 ⊠ X χ , U U = χ U ( J 0 ) S X χ , U U = d S U , X χ U S^U_{U,J_0\boxtimes X_\chi}=S^U_{J_0\boxtimes X_\chi,U}=\chi_U(J_0)S^U_{X_\chi,U}=dS^U_{U,X_\chi} S U , J 0 ⊠ X χ U = S J 0 ⊠ X χ , U U = χ U ( J 0 ) S X χ , U U = d S U , X χ U . Dividing gives κ X \kappa_X κ X .
Equivariance in the row index is given in
Lemma 3.6. For all X ∈ R X\in\mathcal{R} X ∈ R and Q ∈ Q Q\in\mathcal{Q} Q ∈ Q : S J ⊠ X , Q = η Q S X , Q S_{J\boxtimes X,Q}=\eta_QS_{X,Q} S J ⊠ X , Q = η Q S X , Q .
Proof. Let g = g X g=g_X g = g X ; J ⊠ X J\boxtimes X J ⊠ X has the same twist and stabilizer. Use the induced normalization on J ⊠ X J\boxtimes X J ⊠ X , the row entries do not depend on it. By Lemma 3.4 (ii), ( J ⊠ X ) χ = J 0 ⊠ U X χ (J\boxtimes X)_\chi=J_0\boxtimes_U X_\chi ( J ⊠ X ) χ = J 0 ⊠ U X χ , thus S ( J ⊠ X ) χ , Q t U = χ Q t ( J 0 ) S X χ , Q t U = η Q S X χ , Q t U S^U_{(J\boxtimes X)_\chi,Q_t}=\chi_{Q_t}(J_0)S^U_{X_\chi,Q_t}=\eta_QS^U_{X_\chi,Q_t} S ( J ⊠ X ) χ , Q t U = χ Q t ( J 0 ) S X χ , Q t U = η Q S X χ , Q t U (Lemma 3.5 (2) for Q Q Q ). Lemma 2.8 (b) on both sides (column Q Q Q untwisted and the factors ∣ G X ∣ − 1 ψ t ( g − 1 ) |G_X|^{-1}\psi_t(g^{-1}) ∣ G X ∣ − 1 ψ t ( g − 1 ) are the same) gives S J ⊠ X , Q = η Q S X , Q S_{J\boxtimes X,Q}=\eta_QS_{X,Q} S J ⊠ X , Q = η Q S X , Q . There is no distinction between stable and unstable X X X since the stabilizer enters only through ∣ G X ∣ |G_X| ∣ G X ∣ .
Equivariance in the column index is given by
Lemma 3.7. Let Q ∈ Q Q\in\mathcal{Q} Q ∈ Q , and give J ⊠ Q ∈ Q J\boxtimes Q\in\mathcal{Q} J ⊠ Q ∈ Q the induced normalization. Then for every X ∈ R X\in\mathcal{R} X ∈ R (any stabilizer)
S X , J ⊠ Q = κ X S X , Q , in particular, S V ‾ 0 , J ⊠ Q = d S V ‾ 0 , Q . S_{X,J\boxtimes Q}=\kappa_XS_{X,Q},\qquad\text{in particular, }S_{\overline{V}_0,J\boxtimes Q}=dS_{\overline{V}_0,Q}. S X , J ⊠ Q = κ X S X , Q , in particular, S V 0 , J ⊠ Q = d S V 0 , Q .
If J ⊠ Q J\boxtimes Q J ⊠ Q carries the normalization λ φ i n d \lambda\varphi^{\mathrm{ind}} λ φ ind , λ ∈ G ^ \lambda\in\widehat{G} λ ∈ G , then S X , J ⊠ Q = λ ( g X ) κ X S X , Q S_{X,J\boxtimes Q}=\lambda(g_X)\kappa_XS_{X,Q} S X , J ⊠ Q = λ ( g X ) κ X S X , Q .
Proof. Symmetry is applied here only to S U S^U S U . By Lemma 3.4 (ii), ( J ⊠ Q ) t = J 0 ⊠ U Q t (J\boxtimes Q)_t=J_0\boxtimes_U Q_t ( J ⊠ Q ) t = J 0 ⊠ U Q t for all t t t . For X ∈ M ( g ) X\in\mathcal{M}(g) X ∈ M ( g ) and χ ∈ G ^ d X \chi\in\widehat{G}_{d_X} χ ∈ G d X ,
S X χ , ( J ⊠ Q ) t U = S J 0 ⊠ U Q t , X χ U = χ X χ ( J 0 ) S Q t , X χ U = κ X S X χ , Q t U S^U_{X_\chi,(J\boxtimes Q)_t}=S^U_{J_0\boxtimes_U Q_t,X_\chi}=\chi_{X_\chi}(J_0)S^U_{Q_t,X_\chi}=\kappa_XS^U_{X_\chi,Q_t} S X χ , ( J ⊠ Q ) t U = S J 0 ⊠ U Q t , X χ U = χ X χ ( J 0 ) S Q t , X χ U = κ X S X χ , Q t U
(by Lemma 3.5 (2)). Lemma 2.8 (b) with the column J ⊠ Q J\boxtimes Q J ⊠ Q , respectively, Q Q Q , gives ∣ G X ∣ − 1 ψ t ( g − 1 ) S X , J ⊠ Q = κ X ∣ G X ∣ − 1 ψ t ( g − 1 ) S X , Q |G_X|^{-1}\psi_t(g^{-1})S_{X,J\boxtimes Q}=\kappa_X|G_X|^{-1}\psi_t(g^{-1})S_{X,Q} ∣ G X ∣ − 1 ψ t ( g − 1 ) S X , J ⊠ Q = κ X ∣ G X ∣ − 1 ψ t ( g − 1 ) S X , Q . For a different normalization λ φ i n d \lambda\varphi^{\mathrm{ind}} λ φ ind of J ⊠ Q J\boxtimes Q J ⊠ Q , (2.3) multiplies the left side by λ ( g X ) \lambda(g_X) λ ( g X ) . Here λ \lambda λ depends only on the normalization of the column J ⊠ Q J\boxtimes Q J ⊠ Q and does not depend on X X X . For three rows with twists g 1 g_1 g 1 , g 2 g_2 g 2 , g 3 = g 1 g 2 g_3=g_1g_2 g 3 = g 1 g 2 the resulting phases λ ( g 1 ) λ ( g 2 ) λ ( g 3 ) = λ ( g 1 g 2 g 3 − 1 ) = 1 \lambda(g_1)\lambda(g_2)\lambda(g_3)=\lambda(g_1g_2g_3^{-1})=1 λ ( g 1 ) λ ( g 2 ) λ ( g 3 ) = λ ( g 1 g 2 g 3 − 1 ) = 1 cancel in f 0 f_0 f 0 (Theorem 4.1 , Part 2).
For g = 1 g=1 g = 1 these are [8 ] for C V ‾ 0 C_{\overline{V}_0} C V 0 . For g ≠ 1 g\ne1 g = 1 , once normalizations are compatible no phase appears. With an arbitrary normalization the phase is the character λ ( g X ) \lambda(g_X) λ ( g X ) , which is the same for all X X X and cancels in the fusion rules since g 3 = g 1 g 2 g_3=g_1g_2 g 3 = g 1 g 2 (see proof of Theorem 4.1 ).
Lemma 3.8. If d = − 1 d=-1 d = − 1 then J ⊠ X ≄ X J\boxtimes X\not\simeq X J ⊠ X ≃ X for every X ∈ R X\in\mathcal{R} X ∈ R .
Proof. If J ⊠ X ≃ X J\boxtimes X\simeq X J ⊠ X ≃ X , Lemma 3.6 with Q = V ‾ 0 Q=\overline{V}_0 Q = V 0 (η V ‾ 0 = κ V ‾ 0 = d \eta_{\overline{V}_0}=\kappa_{\overline{V}_0}=d η V 0 = κ V 0 = d ) gives S X , V ‾ 0 = d S X , V ‾ 0 = − S X , V ‾ 0 S_{X,\overline{V}_0}=dS_{X,\overline{V}_0}=-S_{X,\overline{V}_0} S X , V 0 = d S X , V 0 = − S X , V 0 , contradicting Lemma 3.2.
Concerning fixed points we state the following
Lemma 3.9. (i) If Q ∈ Q Q\in\mathcal{Q} Q ∈ Q and J ⊠ Q ≃ Q J\boxtimes Q\simeq Q J ⊠ Q ≃ Q , then d = 1 d=1 d = 1 and η Q = − 1 \eta_Q=-1 η Q = − 1 .
(ii) If X ∈ R X\in\mathcal{R} X ∈ R and J ⊠ X ≃ X J\boxtimes X\simeq X J ⊠ X ≃ X , then S X , Q = 0 S_{X,Q}=0 S X , Q = 0 for every Q ∈ Q Q\in\mathcal{Q} Q ∈ Q with η Q = − 1 \eta_Q=-1 η Q = − 1 .
(iii) If g X g_X g X generates G G G , then J ⊠ X ≃ X J\boxtimes X\simeq X J ⊠ X ≃ X if and only if S X , Q = 0 S_{X,Q}=0 S X , Q = 0 for all Q ∈ Q Q\in\mathcal{Q} Q ∈ Q with η Q = − 1 \eta_Q=-1 η Q = − 1 , i.e., iff ∑ Q η Q ∣ S X , Q ∣ 2 = 1 \sum_{Q}\eta_Q|S_{X,Q}|^2=1 ∑ Q η Q ∣ S X , Q ∣ 2 = 1 .
Proof. (i) d = 1 d=1 d = 1 by Lemma 3.8, thus θ J = − 1 \theta_J=-1 θ J = − 1 (Proposition 2.3 (ii)). Apply Lemma 3.1 to W = V ‾ 0 W=\overline{V}_0 W = V 0 , a = J a=J a = J , b = Q b=Q b = Q (an untwisted module). Then N J Q f = δ f , J ⊠ Q = δ f , Q N^f_{JQ}=\delta_{f,J\boxtimes Q}=\delta_{f,Q} N J Q f = δ f , J ⊠ Q = δ f , Q , thus S J , Q = θ J θ Q θ Q − 1 S Q , V ‾ 0 = θ J S Q , V ‾ 0 S_{J,Q}=\theta_J\theta_Q\theta_Q^{-1}S_{Q,\overline{V}_0}=\theta_JS_{Q,\overline{V}_0} S J , Q = θ J θ Q θ Q − 1 S Q , V 0 = θ J S Q , V 0 . By symmetry of the untwisted S S S of V ‾ 0 \overline{V}_0 V 0 , η Q = S Q , J / S Q , V ‾ 0 = θ J = − 1 \eta_Q=S_{Q,J}/S_{Q,\overline{V}_0}=\theta_J=-1 η Q = S Q , J / S Q , V 0 = θ J = − 1 .
(ii) By Lemma 3.6, S X , Q = S J ⊠ X , Q = η Q S X , Q S_{X,Q}=S_{J\boxtimes X,Q}=\eta_QS_{X,Q} S X , Q = S J ⊠ X , Q = η Q S X , Q .
(iii) X X X and J ⊠ X J\boxtimes X J ⊠ X both belong to M ( g X ) \mathcal{M}(g_X) M ( g X ) (the same twist), and g X g_X g X generates G G G . Proposition 2.14 (for rows) gives
∑ Q ∈ Q S J ⊠ X , Q S X , Q ‾ = δ J ⊠ X , X , ∑ Q ∈ Q ∣ S X , Q ∣ 2 = 1. \sum_{Q\in\mathcal{Q}}S_{J\boxtimes X,Q}\overline{S_{X,Q}}=\delta_{J\boxtimes X,X},\qquad\sum_{Q\in\mathcal{Q}}|S_{X,Q}|^2=1. Q ∈ Q ∑ S J ⊠ X , Q S X , Q = δ J ⊠ X , X , Q ∈ Q ∑ ∣ S X , Q ∣ 2 = 1.
By Lemma 3.6, S J ⊠ X , Q = η Q S X , Q S_{J\boxtimes X,Q}=\eta_QS_{X,Q} S J ⊠ X , Q = η Q S X , Q , thus the first sum equals to ∑ Q η Q ∣ S X , Q ∣ 2 \sum_Q\eta_Q|S_{X,Q}|^2 ∑ Q η Q ∣ S X , Q ∣ 2 . If J ⊠ X ≅ X J\boxtimes X\cong X J ⊠ X ≅ X , both sums equal one. By subtracting, ∑ Q ( 1 − η Q ) ∣ S X , Q ∣ 2 = 0 \sum_Q(1-\eta_Q)|S_{X,Q}|^2=0 ∑ Q ( 1 − η Q ) ∣ S X , Q ∣ 2 = 0 , i.e., 2 ∑ η Q = − 1 ∣ S X , Q ∣ 2 = 0 2\sum_{\eta_Q=-1}|S_{X,Q}|^2=0 2 ∑ η Q = − 1 ∣ S X , Q ∣ 2 = 0 . Thus, S X , Q = 0 S_{X,Q}=0 S X , Q = 0 whenever η Q = − 1 \eta_Q=-1 η Q = − 1 . Conversely, if S X , Q = 0 S_{X,Q}=0 S X , Q = 0 whenever η Q = − 1 \eta_Q=-1 η Q = − 1 then ∑ Q η Q ∣ S X , Q ∣ 2 = ∑ Q ∣ S X , Q ∣ 2 = 1 \sum_Q\eta_Q|S_{X,Q}|^2=\sum_Q|S_{X,Q}|^2=1 ∑ Q η Q ∣ S X , Q ∣ 2 = ∑ Q ∣ S X , Q ∣ 2 = 1 , thus δ J ⊠ X , X = 1 \delta_{J\boxtimes X,X}=1 δ J ⊠ X , X = 1 and J ⊠ X ≅ X J\boxtimes X\cong X J ⊠ X ≅ X . Note that (i) concerns untwisted fixed modules (Q ∈ Q Q\in\mathcal{Q} Q ∈ Q ), and (iii) concerns twisted and untwisted X X X with generating twist. □
Conformal weight criterion in untwisted sector is given in
Lemma 3.10. For every untwisted X ∈ M ( 1 ) X\in\mathcal{M}(1) X ∈ M ( 1 ) (stable or not) θ J ⊠ X / θ X = − d ϵ X \theta_{J\boxtimes X}/\theta_X=-d\epsilon_X θ J ⊠ X / θ X = − d ϵ X , i.e., e 2 π i ( h J ⊠ X − h X ) = − d ϵ X e^{2\pi i(h_{J\boxtimes X}-h_X)}=-d\epsilon_X e 2 π i ( h J ⊠ X − h X ) = − d ϵ X for the conformal weights.
Proof. Apply Lemma 3.1 to W = V ‾ 0 W=\overline{V}_0 W = V 0 , a = J a=J a = J , b = X b=X b = X : N J X f = δ f , J ⊠ X N^f_{JX}=\delta_{f,J\boxtimes X} N J X f = δ f , J ⊠ X , thus S J , X = θ J θ X θ J ⊠ X − 1 S J ⊠ X , V ‾ 0 S_{J,X}=\theta_J\theta_X\theta_{J\boxtimes X}^{-1}S_{J\boxtimes X,\overline{V}_0} S J , X = θ J θ X θ J ⊠ X − 1 S J ⊠ X , V 0 . Here S J ⊠ X , V ‾ 0 = χ V ‾ 0 ( J ) S X , V ‾ 0 = d S X , V ‾ 0 S_{J\boxtimes X,\overline{V}_0}=\chi_{\overline{V}_0}(J)S_{X,\overline{V}_0}=dS_{X,\overline{V}_0} S J ⊠ X , V 0 = χ V 0 ( J ) S X , V 0 = d S X , V 0 (multiplicativity of the character χ V ‾ 0 \chi_{\overline{V}_0} χ V 0 and d = S J , V ‾ 0 / S V ‾ 0 , V ‾ 0 d=S_{J,\overline{V}_0}/S_{\overline{V}_0,\overline{V}_0} d = S J , V 0 / S V 0 , V 0 ). By symmetry of the untwisted S S S , κ X = S X , J / S X , V ‾ 0 = S J , X / S V ‾ 0 , X \kappa_X=S_{X,J}/S_{X,\overline{V}_0}=S_{J,X}/S_{\overline{V}_0,X} κ X = S X , J / S X , V 0 = S J , X / S V 0 , X , and, therefore, κ X = d θ J θ X / θ J ⊠ X \kappa_X=d\theta_J\theta_X/\theta_{J\boxtimes X} κ X = d θ J θ X / θ J ⊠ X , i.e., θ J ⊠ X / θ X = d θ J / κ X = θ J ϵ X = − d ϵ X \theta_{J\boxtimes X}/\theta_X=d\theta_J/\kappa_X=\theta_J\epsilon_X=-d\epsilon_X θ J ⊠ X / θ X = d θ J / κ X = θ J ϵ X = − d ϵ X (using d 2 = ϵ X 2 = 1 d^2=\epsilon_X^2=1 d 2 = ϵ X 2 = 1 , θ J = − d \theta_J=-d θ J = − d ). All entries S X , V ‾ 0 S_{X,\overline{V}_0} S X , V 0 have cancelled, thus the result does not depend on the normalization of S S S . With the opposite convention for T T T the computation gives the same result since ϵ X θ J = ± 1 \epsilon_X\theta_J=\pm1 ϵ X θ J = ± 1 is real. □
Remark 3.11. Let us make a remark about parity type and weights for fixed twisted modules. For d = 1 d=1 d = 1 (J J J has weights in Z + 1 2 \mathbb{Z}+\frac{1}{2} Z + 2 1 ) Lemma 3.10 gives ϵ X = 1 \epsilon_X=1 ϵ X = 1 iff the weights of J ⊠ X J\boxtimes X J ⊠ X and X X X differ by 1 2 m o d Z \frac{1}{2}\bmod\mathbb{Z} 2 1 mod Z (NS-type); ϵ X = − 1 \epsilon_X=-1 ϵ X = − 1 iff they agree m o d Z \bmod\mathbb{Z} mod Z (Ramond-type). For d = − 1 d=-1 d = − 1 the roles are reversed. For twisted X X X and n n n odd, the analogous statement e 2 π i n ( h J ⊠ X − h X ) = − d ϵ X e^{2\pi in(h_{J\boxtimes X}-h_X)}=-d\epsilon_X e 2 π in ( h J ⊠ X − h X ) = − d ϵ X holds in Examples 7.1 and 7.3. The first condition of 2.5.1 requires that for the sign ϵ X \epsilon_X ϵ X , and it is an independent check of the labelling ϵ ↔ \epsilon\leftrightarrow ϵ ↔ parity type that does not use S S S -matrices. Note that an untwisted fixed module is always parity-twisted (ϵ = − 1 \epsilon=-1 ϵ = − 1 , Lemma 3.9 (i)), whereas in Example 7.2 (n = 2 n=2 n = 2 ) the fixed twisted module T 1 T_1 T 1 has ϵ = 1 \epsilon=1 ϵ = 1 and T 2 T_2 T 2 has ϵ = − 1 \epsilon=-1 ϵ = − 1 .
The twisted Verlinde formula for V
Let Q f i x = { Q ∈ Q : J ⊠ Q ≅ Q } \mathcal{Q}^{\mathrm{fix}}=\{Q\in\mathcal{Q}:J\boxtimes Q\cong Q\} Q fix = { Q ∈ Q : J ⊠ Q ≅ Q } . It is empty if d = − 1 d=-1 d = − 1 by Lemma 3.8). Let Q f r e e / J \mathcal{Q}^{\mathrm{free}}/J Q free / J be the set of two-element orbits [ Q ] = { Q , J ⊠ Q } [Q]=\{Q,J\boxtimes Q\} [ Q ] = { Q , J ⊠ Q } on Q \mathcal{Q} Q . By Lemma 3.5 (4) η \eta η is constant on orbits. Set Q ± f r e e = { [ Q ] : η Q = ± 1 } \mathcal{Q}^{\mathrm{free}}_{\pm}=\{[Q]:\eta_Q=\pm1\} Q ± free = {[ Q ] : η Q = ± 1 } and Q ± f i x = { Q ∈ Q f i x : η Q = ± 1 } \mathcal{Q}^{\mathrm{fix}}_{\pm}=\{Q\in\mathcal{Q}^{\mathrm{fix}}:\eta_Q=\pm1\} Q ± fix = { Q ∈ Q fix : η Q = ± 1 } . By Lemma 3.9 (i), Q + f i x = ∅ \mathcal{Q}^{\mathrm{fix}}_+=\varnothing Q + fix = ∅ . The next main result of this paper, the twisted Verlinde formula for V V V , is the content of
Theorem 4.1. Adopt assumptions of Subsection 2.1. Let g 1 , g 2 ∈ G g_1,g_2\in G g 1 , g 2 ∈ G with ⟨ g 1 , g 2 ⟩ = G \langle g_1,g_2\rangle=G ⟨ g 1 , g 2 ⟩ = G , g 3 = g 1 g 2 g_3=g_1g_2 g 3 = g 1 g 2 , and let X i ∈ M ( g i ) X_i\in\mathcal{M}(g_i) X i ∈ M ( g i ) be simple objects of Rep ( V ‾ 0 ) \operatorname{Rep}(\overline{V}_0) Rep ( V 0 ) (free or J J J -fixed). Set
N ± : = N X 1 , X 2 X 3 ± N X 1 , X 2 J ⊠ X 3 . N^{\pm}:=N^{X_3}_{X_1,X_2}\pm N^{J\boxtimes X_3}_{X_1,X_2}. N ± := N X 1 , X 2 X 3 ± N X 1 , X 2 J ⊠ X 3 . Let f 0 ( Q ) = S X 1 , Q S X 2 , Q S X 3 , Q ‾ / S V ‾ 0 , Q f_0(Q)=S_{X_1,Q}S_{X_2,Q}\overline{S_{X_3,Q}}/S_{\overline{V}_0,Q} f 0 ( Q ) = S X 1 , Q S X 2 , Q S X 3 , Q / S V 0 , Q . Then
(a) N ± = δ π ( X 1 ) + π ( X 2 ) , π ( X 3 ) ( 4 ∑ [ Q ] ∈ Q ± f r e e f 0 ( Q ) + 2 ∑ Q ∈ Q ± f i x f 0 ( Q ) ) , N^{\pm}=\delta_{\pi(X_1)+\pi(X_2),\pi(X_3)}\left(4\sum_{[Q]\in\mathcal{Q}^{\mathrm{free}}_{\pm}}f_0(Q)+2\sum_{Q\in\mathcal{Q}^{\mathrm{fix}}_{\pm}}f_0(Q)\right), N ± = δ π ( X 1 ) + π ( X 2 ) , π ( X 3 ) ( 4 ∑ [ Q ] ∈ Q ± free f 0 ( Q ) + 2 ∑ Q ∈ Q ± fix f 0 ( Q ) ) ,
where the Kronecker delta is taken in Z / 2 Z \mathbb{Z}/2\mathbb{Z} Z /2 Z (equivalently, it is one iff ϵ X 1 ϵ X 2 ϵ X 3 = 1 \epsilon_{X_1}\epsilon_{X_2}\epsilon_{X_3}=1 ϵ X 1 ϵ X 2 ϵ X 3 = 1 ), and Q Q Q is any representative of [ Q ] [Q] [ Q ] .
(b) in conditions 2.5.1 N + N^+ N + is the dimension and N − N^- N − the superdimension of the graded space ( F V ( X 3 ) ; F V ( X 1 ) , F V ( X 2 ) ) (F_V(X_3);F_V(X_1),F_V(X_2)) ( F V ( X 3 ) ; F V ( X 1 ) , F V ( X 2 )) of intertwining operators (gradings by the even representatives X i X_i X i ). F V ( X i ) F_V(X_i) F V ( X i ) is g i g_i g i -twisted for π ( X i ) = 0 \pi(X_i)=0 π ( X i ) = 0 and g i P g_iP g i P -twisted for π ( X i ) = 1 \pi(X_i)=1 π ( X i ) = 1 .
Note that all fusion coefficients and the tensor product ⊠ \boxtimes ⊠ here are those of Rep ( V ‾ 0 ) \operatorname{Rep}(\overline{V}_0) Rep ( V 0 ) , Definition 2.5 . N X 1 , X 2 X 3 N^{X_3}_{X_1,X_2} N X 1 , X 2 X 3 is the individual coefficient if at least two of X 1 X_1 X 1 , X 2 X_2 X 2 , X 3 X_3 X 3 are G G G -stable, and otherwise the orbit average ∣ O X 1 ∣ − 1 ∣ O X 2 ∣ − 1 ∑ N i ∈ O X i N N 1 , N 2 X 3 |O_{X_1}|^{-1}|O_{X_2}|^{-1}\sum_{N_i\in O_{X_i}}N^{X_3}_{N_1,N_2} ∣ O X 1 ∣ − 1 ∣ O X 2 ∣ − 1 ∑ N i ∈ O X i N N 1 , N 2 X 3 of Theorem 2.11 (i). The same convention applies to N X 1 , X 2 J ⊠ X 3 N^{J\boxtimes X_3}_{X_1,X_2} N X 1 , X 2 J ⊠ X 3 , and J ⊠ X 3 J\boxtimes X_3 J ⊠ X 3 has the same stabilizer as X 3 X_3 X 3 by Lemma 3.4 . The set Q \mathcal{Q} Q refers to G = ⟨ g 1 , g 2 ⟩ G=\langle g_1,g_2\rangle G = ⟨ g 1 , g 2 ⟩ , while the entries S X , Q S_{X,\mathcal{Q}} S X , Q depend only on g X g_X g X and on the normalization of Q \mathcal{Q} Q , Lemma 2.9 (b).
When in (4.1) the delta equals one the summand f 0 f_0 f 0 is constant on each free orbit. Moreover, N X 1 , X 2 X 3 = 1 2 ( N + + N − ) N^{X_3}_{X_1,X_2}=\frac{1}{2}(N^+ +N^-) N X 1 , X 2 X 3 = 2 1 ( N + + N − ) , N X 1 , X 2 J ⊠ X 3 = 1 2 ( N + − N − ) N^{J\boxtimes X_3}_{X_1,X_2}=\frac{1}{2}(N^+ -N^-) N X 1 , X 2 J ⊠ X 3 = 2 1 ( N + − N − ) . Both vanish if π ( X 3 ) ≠ π ( X 1 ) + π ( X 2 ) \pi(X_3)\ne\pi(X_1)+\pi(X_2) π ( X 3 ) = π ( X 1 ) + π ( X 2 ) in Z / 2 Z \mathbb{Z}/2\mathbb{Z} Z /2 Z . For d = − 1 d=-1 d = − 1 no fixed points occur and (4.1) is 4 ∑ [ Q ] ∈ Q f r e e f 0 ( Q ) 4\sum_{[\mathcal{Q}]\in\mathcal{Q}^{\mathrm{free}}}f_0(\mathcal{Q}) 4 ∑ [ Q ] ∈ Q free f 0 ( Q ) . For d = 1 d=1 d = 1 the fixed points contribute only to N − N^- N − .
Proof. (a) 1. By Theorem 2.11 applied to ( X 1 , X 2 , X 3 ) (X_1,X_2,X_3) ( X 1 , X 2 , X 3 ) and to ( X 1 , X 2 , J ⊠ X 3 ) (X_1,X_2,J\boxtimes X_3) ( X 1 , X 2 , J ⊠ X 3 ) (the same stability pattern), and by Lemma 3.6 (η Q \eta_{\mathcal{Q}} η Q is real),
N X 1 , X 2 X 3 = ∑ Q ∈ Q f 0 ( Q ) , N X 1 , X 2 J ⊠ X 3 = ∑ Q ∈ Q η Q f 0 ( Q ) . N^{X_3}_{X_1,X_2}=\sum_{\mathcal{Q}\in\mathcal{Q}}f_0(\mathcal{Q}),\qquad N^{J\boxtimes X_3}_{X_1,X_2}=\sum_{\mathcal{Q}\in\mathcal{Q}}\eta_{\mathcal{Q}}f_0(\mathcal{Q}). N X 1 , X 2 X 3 = Q ∈ Q ∑ f 0 ( Q ) , N X 1 , X 2 J ⊠ X 3 = Q ∈ Q ∑ η Q f 0 ( Q ) .
thus N ± = ∑ Q ( 1 ± η Q ) f 0 ( Q ) = 2 ∑ Q : η Q = ± 1 f 0 ( Q ) N^\pm=\sum_{\mathcal{Q}}(1\pm\eta_{\mathcal{Q}})f_0(\mathcal{Q})=2\sum_{\mathcal{Q}:\eta_{\mathcal{Q}}=\pm1}f_0(\mathcal{Q}) N ± = ∑ Q ( 1 ± η Q ) f 0 ( Q ) = 2 ∑ Q : η Q = ± 1 f 0 ( Q ) , since 1 ± η Q ∈ { 0 , 2 } 1\pm\eta_{\mathcal{Q}}\in\{0,2\} 1 ± η Q ∈ { 0 , 2 } . This is the origin of the outer factor 2.
Let Q ∈ Q \mathcal{Q}\in\mathcal{Q} Q ∈ Q and give J ⊠ Q J\boxtimes\mathcal{Q} J ⊠ Q the induced normalization since f 0 f_0 f 0 does not depend on normalizations, Theorem 2.11. By Lemma 3.7 and κ = d ϵ \kappa=d\epsilon κ = d ϵ , d 2 = 1 d^2=1 d 2 = 1 , and
f 0 ( J ⊠ Q ) = κ X 1 κ X 2 κ X 3 d f 0 ( Q ) = ϵ X 1 ϵ X 2 ϵ X 3 f 0 ( Q ) . f_0(J\boxtimes\mathcal{Q})=\frac{\kappa_{X_1}\kappa_{X_2}\kappa_{X_3}}{d}f_0(\mathcal{Q})=\epsilon_{X_1}\epsilon_{X_2}\epsilon_{X_3}f_0(\mathcal{Q}). f 0 ( J ⊠ Q ) = d κ X 1 κ X 2 κ X 3 f 0 ( Q ) = ϵ X 1 ϵ X 2 ϵ X 3 f 0 ( Q ) .
In an arbitrary normalization of J ⊠ Q J\boxtimes\mathcal{Q} J ⊠ Q the extra factor is λ ( g 1 ) λ ( g 2 ) λ ( g 3 ) = λ ( g 1 g 2 g 3 − 1 ) = 1 \lambda(g_1)\lambda(g_2)\lambda(g_3)=\lambda(g_1g_2g_3^{-1})=1 λ ( g 1 ) λ ( g 2 ) λ ( g 3 ) = λ ( g 1 g 2 g 3 − 1 ) = 1 . If Q \mathcal{Q} Q is fixed, J ⊠ Q ≃ Q J\boxtimes\mathcal{Q}\simeq\mathcal{Q} J ⊠ Q ≃ Q and f 0 ( J ⊠ Q ) = f 0 ( Q ) f_0(J\boxtimes\mathcal{Q})=f_0(\mathcal{Q}) f 0 ( J ⊠ Q ) = f 0 ( Q ) , thus f 0 ( Q ) = 0 f_0(\mathcal{Q})=0 f 0 ( Q ) = 0 unless ϵ X 1 ϵ X 2 ϵ X 3 = 1 \epsilon_{X_1}\epsilon_{X_2}\epsilon_{X_3}=1 ϵ X 1 ϵ X 2 ϵ X 3 = 1 .
Orbit decomposition: Q \mathcal{Q} Q is the disjoint union of Q f i x \mathcal{Q}^{\mathrm{fix}} Q fix and the free orbits [ Q ] = { Q , J ⊠ Q } [\mathcal{Q}]=\{\mathcal{Q},J\boxtimes\mathcal{Q}\} [ Q ] = { Q , J ⊠ Q } of size two (Q f i x = { ∅ } \mathcal{Q}^{\mathrm{fix}}=\{\varnothing\} Q fix = { ∅ } if d = − 1 d=-1 d = − 1 , so then all orbits are free). For a free orbit, η J ⊠ Q = η Q \eta_{J\boxtimes\mathcal{Q}}=\eta_{\mathcal{Q}} η J ⊠ Q = η Q (see Lemma 3.5 (4)). Thus, both members belong to the same set { η = 1 } \{\eta=1\} { η = 1 } or { η = − 1 } \{\eta=-1\} { η = − 1 } , and, according to part 2. above, their contributions are f 0 ( Q ) f_0(\mathcal{Q}) f 0 ( Q ) and ϵ X 1 ϵ X 2 ϵ X 3 f 0 ( Q ) \epsilon_{X_1}\epsilon_{X_2}\epsilon_{X_3}f_0(\mathcal{Q}) ϵ X 1 ϵ X 2 ϵ X 3 f 0 ( Q ) . Let δ = 1 \delta=1 δ = 1 if ϵ X 1 ϵ X 2 ϵ X 3 = 1 \epsilon_{X_1}\epsilon_{X_2}\epsilon_{X_3}=1 ϵ X 1 ϵ X 2 ϵ X 3 = 1 and δ = 0 \delta=0 δ = 0 otherwise. Then
∑ η Q = ± 1 f 0 ( Q ) = ∑ [ Q ] ∈ Q ± f r e e ( f 0 ( Q ) + f 0 ( J ⊠ Q ) ) + ∑ Q ∈ Q ± f i x f 0 ( Q ) = δ ( 2 ∑ Q ± f r e e f 0 + ∑ Q ± f i x f 0 ) . \sum_{\eta_{\mathcal{Q}}=\pm1}f_0(\mathcal{Q})=\sum_{[\mathcal{Q}]\in\mathcal{Q}^{\mathrm{free}}_\pm}\bigl(f_0(\mathcal{Q})+f_0(J\boxtimes\mathcal{Q})\bigr)+\sum_{\mathcal{Q}\in\mathcal{Q}^{\mathrm{fix}}_\pm}f_0(\mathcal{Q})=\delta\left(2\sum_{\mathcal{Q}^{\mathrm{free}}_\pm}f_0+\sum_{\mathcal{Q}^{\mathrm{fix}}_\pm}f_0\right). η Q = ± 1 ∑ f 0 ( Q ) = [ Q ] ∈ Q ± free ∑ ( f 0 ( Q ) + f 0 ( J ⊠ Q ) ) + Q ∈ Q ± fix ∑ f 0 ( Q ) = δ 2 Q ± free ∑ f 0 + Q ± fix ∑ f 0 .
since f 0 ( Q ) + f 0 ( J ⊠ Q ) = ( 1 + ϵ X 1 ϵ X 2 ϵ X 3 ) f 0 ( Q ) = 2 δ f 0 ( Q ) f_0(\mathcal{Q})+f_0(J\boxtimes\mathcal{Q})=(1+\epsilon_{X_1}\epsilon_{X_2}\epsilon_{X_3})f_0(\mathcal{Q})=2\delta f_0(\mathcal{Q}) f 0 ( Q ) + f 0 ( J ⊠ Q ) = ( 1 + ϵ X 1 ϵ X 2 ϵ X 3 ) f 0 ( Q ) = 2 δ f 0 ( Q ) , and f 0 ( Q ) = 0 f_0(\mathcal{Q})=0 f 0 ( Q ) = 0 at a fixed point when δ = 0 \delta=0 δ = 0 (see part 2). The inner 2 is the size of a free J J J -orbit (two equal terms), a fixed point is a single term. Multiplying by the outer 2 of part 1. above gives the coefficients 4 (free orbits) and 2 (fixed points) in (4.1). Since ϵ X 1 ϵ X 2 ϵ X 3 = 1 \epsilon_{X_1}\epsilon_{X_2}\epsilon_{X_3}=1 ϵ X 1 ϵ X 2 ϵ X 3 = 1 iff π ( X 1 ) + π ( X 2 ) = π ( X 3 ) \pi(X_1)+\pi(X_2)=\pi(X_3) π ( X 1 ) + π ( X 2 ) = π ( X 3 ) in Z / 2 Z \mathbb{Z}/2\mathbb{Z} Z /2 Z , this is the delta of (4.1).
(b): By the second condition of 2.5.1 the even and odd parts of I ( F V ( X 3 ) ; F V ( X 1 ) , F V ( X 2 ) ) I(F_V(X_3);F_V(X_1),F_V(X_2)) I ( F V ( X 3 ) ; F V ( X 1 ) , F V ( X 2 )) have dimensions N X 1 , X 2 X 3 N^{X_3}_{X_1,X_2} N X 1 , X 2 X 3 and N X 1 , X 2 J ⊠ X 3 N^{J\boxtimes X_3}_{X_1,X_2} N X 1 , X 2 J ⊠ X 3 . Their sum and difference are N ± N^\pm N ± . The twist statement is given by the first condition of 2.5.1 and Lemma 2.2.
We next adopt the following parity conventions and dependence on representatives. Replacing the even representative X 1 X_1 X 1 by J ⊠ X 1 J\boxtimes X_1 J ⊠ X 1 (i.e., passing from F V ( X 1 ) F_V(X_1) F V ( X 1 ) to Π F V ( X 1 ) \Pi F_V(X_1) Π F V ( X 1 ) ) gives, using J ⊠ J ≃ V ‾ 0 J\boxtimes J\simeq\overline{V}_0 J ⊠ J ≃ V 0 , N J ⊠ X 1 , X 2 J ⊠ X 1 , X 3 = N X 1 , X 2 J ⊠ X 3 N^{J\boxtimes X_1,X_3}_{J\boxtimes X_1,X_2}=N^{J\boxtimes X_3}_{X_1,X_2} N J ⊠ X 1 , X 2 J ⊠ X 1 , X 3 = N X 1 , X 2 J ⊠ X 3 and N J ⊠ X 1 , X 2 J ⊠ X 1 , X 3 = N X 1 , X 2 X 3 N^{J\boxtimes X_1,X_3}_{J\boxtimes X_1,X_2}=N^{X_3}_{X_1,X_2} N J ⊠ X 1 , X 2 J ⊠ X 1 , X 3 = N X 1 , X 2 X 3 . Hence, N + ↦ N + N^+\mapsto N^+ N + ↦ N + and N − ↦ − N − N^-\mapsto-N^- N − ↦ − N − . This agrees with (4.1), since f 0 f_0 f 0 changes by the factor η Q \eta_{\mathcal{Q}} η Q (Lemma 3.6), which is 1 on the domain of N + N^+ N + and − 1 -1 − 1 on that of N − N^- N − . Thus, the dimension is independent of the parity conventions and the superdimension changes sign with the parity of an odd number of inputs/outputs. The next of this paper main results concerns the fusion product. Let us work in conditions 2.5.1 and 2.5.2.
Corollary 4.2. Let X 1 X_1 X 1 , X 2 X_2 X 2 be J J J -free with ⟨ g 1 , g 2 ⟩ = G \langle g_1,g_2\rangle=G ⟨ g 1 , g 2 ⟩ = G , and assume that every simple constituent of X 1 ⊠ X 2 X_1\boxtimes X_2 X 1 ⊠ X 2 is J J J -free (if g 3 g_3 g 3 generates G G G this can be tested on the S S S -matrix by Lemma 3.9 (iii)). Then in Rep ( V ) f r e e \operatorname{Rep}(V)^{\mathrm{free}} Rep ( V ) free
F V ( X 1 ) ⊠ V F V ( X 2 ) ≃ ⨁ [ X 3 ] ( N X 1 , X 2 X 3 F V ( X 3 ) ⊕ N X 1 , X 2 J ⊠ X 3 Π F V ( X 3 ) ) , F_V(X_1)\boxtimes_V F_V(X_2)\simeq\bigoplus_{[X_3]}\left(N^{X_3}_{X_1,X_2}F_V(X_3)\oplus N^{J\boxtimes X_3}_{X_1,X_2}\Pi F_V(X_3)\right), F V ( X 1 ) ⊠ V F V ( X 2 ) ≃ [ X 3 ] ⨁ ( N X 1 , X 2 X 3 F V ( X 3 ) ⊕ N X 1 , X 2 J ⊠ X 3 Π F V ( X 3 ) ) ,
the sum over the J J J -orbits [ X 3 ] = { X 3 , J ⊠ X 3 } [X_3]=\{X_3,J\boxtimes X_3\} [ X 3 ] = { X 3 , J ⊠ X 3 } of simple constituents with g 3 = g 1 g 2 g_3=g_1g_2 g 3 = g 1 g 2 , π ( X 3 ) = π ( X 1 ) + π ( X 2 ) \pi(X_3)=\pi(X_1)+\pi(X_2) π ( X 3 ) = π ( X 1 ) + π ( X 2 ) (in Z / 2 Z \mathbb{Z}/2\mathbb{Z} Z /2 Z ), and multiplicities are given by (4.1).
Proof. The second conditions of 2.5.2 gives F V ( X 1 ) ⊠ V F V ( X 2 ) ≅ F V ( X 1 ⊠ X 2 ) = ⨁ X 3 N X 1 , X 2 X 3 F V ( X 3 ) F_V(X_1)\boxtimes_V F_V(X_2)\cong F_V(X_1\boxtimes X_2)=\bigoplus_{X_3}N_{X_1,X_2}^{X_3}F_V(X_3) F V ( X 1 ) ⊠ V F V ( X 2 ) ≅ F V ( X 1 ⊠ X 2 ) = ⨁ X 3 N X 1 , X 2 X 3 F V ( X 3 ) , the sum over all simple X 3 X_3 X 3 (both members of each J J J -orbit). Let us group the terms by orbits. Since F V ( J ⊠ X 3 ) = Π F V ( X 3 ) F_V(J\boxtimes X_3)=\Pi F_V(X_3) F V ( J ⊠ X 3 ) = Π F V ( X 3 ) and N X 1 , X 2 J ⊠ X 3 N_{X_1,X_2}^{J\boxtimes X_3} N X 1 , X 2 J ⊠ X 3 is the multiplicity of the second member, the orbit [ X 3 ] [X_3] [ X 3 ] contributes N X 1 , X 2 X 3 F V ( X 3 ) ⊕ N X 1 , X 2 J ⊠ X 3 Π F V ( X 3 ) N_{X_1,X_2}^{X_3}F_V(X_3)\oplus N_{X_1,X_2}^{J\boxtimes X_3}\Pi F_V(X_3) N X 1 , X 2 X 3 F V ( X 3 ) ⊕ N X 1 , X 2 J ⊠ X 3 Π F V ( X 3 ) . Replacing the chosen representative X 3 X_3 X 3 by J ⊠ X 3 J\boxtimes X_3 J ⊠ X 3 interchanges the two summands. I.e., using J ⊠ J ≅ V ‾ 0 J\boxtimes J\cong\overline{V}_0 J ⊠ J ≅ V 0 , N X 1 , X 2 J ⊠ X 3 N_{X_1,X_2}^{J\boxtimes X_3} N X 1 , X 2 J ⊠ X 3 and N X 1 , X 2 X 3 N_{X_1,X_2}^{X_3} N X 1 , X 2 X 3 swap and F V ( J ⊠ X 3 ) = Π F V ( X 3 ) F_V(J\boxtimes X_3)=\Pi F_V(X_3) F V ( J ⊠ X 3 ) = Π F V ( X 3 ) , Π Π = i d \Pi\Pi=\mathrm{id} ΠΠ = id . Thus, the displayed sum does not depend on the representatives. Vanishing for π ( X 3 ) ≠ π ( X 1 ) + π ( X 2 ) \pi(X_3)\ne\pi(X_1)+\pi(X_2) π ( X 3 ) = π ( X 1 ) + π ( X 2 ) is given by Theorem 4.1 (a).
Consistency checks
In this Section we do some consistency checks.
Proposition 5.1. Let g 2 g_2 g 2 generate G G G and X 2 X_2 X 2 , X 3 ∈ M ( g 2 ) X_3\in\mathcal{M}(g_2) X 3 ∈ M ( g 2 ) . For X 1 = V ‾ 0 X_1=\overline{V}_0 X 1 = V 0 : N + = δ X 3 , X 2 + δ X 3 , J ⊠ X 2 N^+=\delta_{X_3,X_2}+\delta_{X_3,J\boxtimes X_2} N + = δ X 3 , X 2 + δ X 3 , J ⊠ X 2 and N − = δ X 3 , X 2 − δ X 3 , J ⊠ X 2 N^-=\delta_{X_3,X_2}-\delta_{X_3,J\boxtimes X_2} N − = δ X 3 , X 2 − δ X 3 , J ⊠ X 2 , as required by V ⊠ V F V ( X 2 ) ≅ F V ( X 2 ) V\boxtimes_V F_V(X_2)\cong F_V(X_2) V ⊠ V F V ( X 2 ) ≅ F V ( X 2 ) .
Proof. Here f 0 ( Q ) = S X 2 , Q S ‾ X 3 , Q f_0(Q)=S_{X_2,Q}\overline{S}_{X_3,Q} f 0 ( Q ) = S X 2 , Q S X 3 , Q and part 1 of the proof of Theorem 4.1 gives N ± = ∑ Q ( 1 ± η Q ) S X 2 , Q S ‾ X 3 , Q N^\pm=\sum_Q(1\pm\eta_Q)S_{X_2,Q}\overline{S}_{X_3,Q} N ± = ∑ Q ( 1 ± η Q ) S X 2 , Q S X 3 , Q . By Proposition 2.14, ∑ Q S X 2 , Q S ‾ X 3 , Q = δ X 2 , X 3 \sum_Q S_{X_2,Q}\overline{S}_{X_3,Q}=\delta_{X_2,X_3} ∑ Q S X 2 , Q S X 3 , Q = δ X 2 , X 3 , and by Lemma 3.6 and Proposition 2.14 (note that J ⊠ X 2 ∈ M ( g 2 ) J\boxtimes X_2\in\mathcal{M}(g_2) J ⊠ X 2 ∈ M ( g 2 ) ), ∑ Q η Q S X 2 , Q S ‾ X 3 , Q = ∑ Q S J ⊠ X 2 , Q S ‾ X 3 , Q = δ J ⊠ X 2 , X 3 \sum_Q\eta_QS_{X_2,Q}\overline{S}_{X_3,Q}=\sum_QS_{J\boxtimes X_2,Q}\overline{S}_{X_3,Q}=\delta_{J\boxtimes X_2,X_3} ∑ Q η Q S X 2 , Q S X 3 , Q = ∑ Q S J ⊠ X 2 , Q S X 3 , Q = δ J ⊠ X 2 , X 3 .
The selection rules are decribed by
Corollary 5.2. Under the assumptions of Theorem 4.1 (a):
N X 1 , X 2 X 3 = N X 1 , X 2 J ⊠ X 3 = 0 N_{X_1,X_2}^{X_3}=N_{X_1,X_2}^{J\boxtimes X_3}=0 N X 1 , X 2 X 3 = N X 1 , X 2 J ⊠ X 3 = 0 if π ( X 3 ) ≠ π ( X 1 ) + π ( X 2 ) \pi(X_3)\ne\pi(X_1)+\pi(X_2) π ( X 3 ) = π ( X 1 ) + π ( X 2 ) in Z / 2 Z \mathbb{Z}/2\mathbb{Z} Z /2 Z ;
otherwise N + ≥ ∣ N − ∣ N^+\geq\lvert N^-\rvert N + ≥ ∣ N − ∣ and N + ≡ N − N^+\equiv N^- N + ≡ N − (mod 2), for individual coefficients.
The dimension sum rule is given by
Proposition 5.3. Let g 1 g_1 g 1 , g 2 ∈ G g_2\in G g 2 ∈ G and assume that g 3 = g 1 g 2 g_3=g_1g_2 g 3 = g 1 g 2 generates G G G (then ⟨ g 1 , g 2 ⟩ = G \langle g_1,g_2\rangle=G ⟨ g 1 , g 2 ⟩ = G , thus Theorem 2.11 applies). Set δ ( X ) = S X , V ‾ 0 / S V ‾ 0 , V ‾ 0 \delta(X)=S_{X,\overline{V}_0}/S_{\overline{V}_0,\overline{V}_0} δ ( X ) = S X , V 0 / S V 0 , V 0 . For X 1 ∈ M ( g 1 ) X_1\in\mathcal{M}(g_1) X 1 ∈ M ( g 1 ) , X 2 ∈ M ( g 2 ) X_2\in\mathcal{M}(g_2) X 2 ∈ M ( g 2 ) ,
∑ X 3 ∈ M ( g 3 ) N X 1 , X 2 X 3 δ ( X 3 ) = δ ( X 1 ) δ ( X 2 ) , \sum_{X_3\in\mathcal{M}(g_3)}N_{X_1,X_2}^{X_3}\delta(X_3)=\delta(X_1)\delta(X_2), X 3 ∈ M ( g 3 ) ∑ N X 1 , X 2 X 3 δ ( X 3 ) = δ ( X 1 ) δ ( X 2 ) , where N N N is replaced by the orbit average of Theorem 2.11 (i) if neither X 1 X_1 X 1 nor X 2 X_2 X 2 is stable.
Proof. All X 3 ∈ M ( g 3 ) X_3\in\mathcal{M}(g_3) X 3 ∈ M ( g 3 ) are stable, thus by Theorem 2.11 the left side equals to ∑ X 3 T ( X 1 , X 2 , X 3 ) δ ( X 3 ) = ∑ Q S X 1 , Q S X 2 , Q S V ‾ 0 , Q 1 S V ‾ 0 , V ‾ 0 ∑ X 3 S ‾ X 3 , Q S X 3 , V ‾ 0 \sum_{X_3}\mathrm{T}(X_1,X_2,X_3)\delta(X_3)=\sum_Q\frac{S_{X_1,Q}S_{X_2,Q}}{S_{\overline{V}_0,Q}}\frac{1}{S_{\overline{V}_0,\overline{V}_0}}\sum_{X_3}\overline{S}_{X_3,Q}S_{X_3,\overline{V}_0} ∑ X 3 T ( X 1 , X 2 , X 3 ) δ ( X 3 ) = ∑ Q S V 0 , Q S X 1 , Q S X 2 , Q S V 0 , V 0 1 ∑ X 3 S X 3 , Q S X 3 , V 0 . By Proposition 2.14 (columns, for the generating sector g 3 g_3 g 3 ) the inner sum is δ Q , V ‾ 0 \delta_{Q,\overline{V}_0} δ Q , V 0 , which leaves S X 1 , V ‾ 0 S X 2 , V ‾ 0 / S V ‾ 0 , V ‾ 0 2 S_{X_1,\overline{V}_0}S_{X_2,\overline{V}_0}/S_{\overline{V}_0,\overline{V}_0}^2 S X 1 , V 0 S X 2 , V 0 / S V 0 , V 0 2 .
Since δ ( X ) ⋅ ∣ O X ∣ \delta(X)\cdot\lvert O_X\rvert δ ( X ) ⋅ ∣ O X ∣ is the dimension of the U U U -module X χ X_\chi X χ (Lemma 2.8 (b)), this is the usual dimension count of the fusion rules. It is an independent consistency check of statements from Lemma 2.8 to Theorem 2.11, and it is verified in all examples (see § 7 ).
Inner automorphisms
Let h ∈ V ‾ 0 h\in\overline{V}_0 h ∈ V 0 be of weight one with L ( m ) h = δ m , 0 h L(m)h=\delta_{m,0}h L ( m ) h = δ m , 0 h and h m h = δ m , 1 ⟨ h , h ⟩ 1 h_mh=\delta_{m,1}\langle h,h\rangle\mathbf{1} h m h = δ m , 1 ⟨ h , h ⟩ 1 (m ≥ 0 m\geq0 m ≥ 0 ), such that h 0 h_0 h 0 acts semisimply on V V V with spectrum in 1 n Z \frac{1}{n}\mathbb{Z} n 1 Z . Let σ = e 2 π i h 0 \sigma=e^{2\pi i h_0} σ = e 2 π i h 0 have order n n n and act faithfully on V ‾ 0 \overline{V}_0 V 0 as in [12 ] . Li operator Δ ( h , z ) = z h 0 exp ( ∑ m ≥ 1 h m − m ( − z ) − m ) \Delta(h,z)=z^{h_0}\exp\left(\sum_{m\geq1}\frac{h_m}{-m}(-z)^{-m}\right) Δ ( h , z ) = z h 0 exp ( ∑ m ≥ 1 − m h m ( − z ) − m ) turns a module M M M into M ( h ) = ( M , Y M ( Δ ( h , z ) ⋅ , z ) ) M^{(h)}=(M,Y_M(\Delta(h,z)\cdot,z)) M ( h ) = ( M , Y M ( Δ ( h , z ) ⋅ , z )) [33 ] . We consider such h 0 h_0 h 0 that has the following property: it acts semisimply on every irreducible untwisted V ‾ 0 \overline{V}_0 V 0 -module (we call this property in what follows the property of § 6 ).
Then we have the following statement about inner automorphisms.
Proposition 6.1. (i) Every irreducible σ a \sigma^a σ a -twisted V ‾ 0 \overline{V}_0 V 0 -module is M ( a h ) M^{(ah)} M ( ah ) for an irreducible V ‾ 0 \overline{V}_0 V 0 -module M M M [33 , 14 ] . Moreover, when the first argument is untwisted, N M 1 , M 2 M 3 ( a h ) = N M 1 , M 2 M 3 N_{M_1,M_2}^{M_3^{(ah)}}=N_{M_1,M_2}^{M_3} N M 1 , M 2 M 3 ( ah ) = N M 1 , M 2 M 3 for untwisted M 1 M_1 M 1 , M 2 M_2 M 2 , M 3 M_3 M 3 [14 ] .
(ii) Every irreducible twisted V ‾ 0 \overline{V}_0 V 0 -module is G G G -stable. Hence, Q = M ( 1 ) \mathcal{Q}=\mathcal{M}(1) Q = M ( 1 ) , and Theorem 2.11 (ii) holds for all g 1 g_1 g 1 , g 2 g_2 g 2 with ⟨ g 1 , g 2 ⟩ = G \langle g_1,g_2\rangle=G ⟨ g 1 , g 2 ⟩ = G and all n n n (including composite and even n n n ), and Theorem 4.1 (a) holds for all simple X i X_i X i .
(iii) Adopt assumptions 2.5.3. Then conditions 2.5.1 and 2.5.2 follow from the untwisted statements of [8 , 30 ] by transport of structure.
Proof. (i) is given by the corresponding citation results mentioned above with proofs using the Jacobi identity and Δ ( h , z ) \Delta(h,z) Δ ( h , z ) .
(ii) By (i) the module is M ( a h ) M^{(ah)} M ( ah ) . The zero mode of h h h on M ( a h ) M^{(ah)} M ( ah ) is h 0 M ( a h ) + α ⟨ h , h ⟩ h_0^{M^{(ah)}}+\alpha\langle h,h\rangle h 0 M ( ah ) + α ⟨ h , h ⟩ [33 ] , semisimple by the condition of this Section). It commutes with L ( 0 ) L(0) L ( 0 ) , and satisfies [ h 0 M ( a h ) , Y ( v , z ) ] = Y ( h 0 v , z ) [h_0^{M^{(ah)}},Y(v,z)]=Y(h_0v,z) [ h 0 M ( ah ) , Y ( v , z )] = Y ( h 0 v , z ) since h h h is fixed by σ a \sigma^a σ a and has integer modes. Hence, e 2 π i h 0 M ( a h ) : M ( a h ) → M ( a h ) ∘ σ e^{2\pi i h_0^{M^{(ah)}}}:M^{(ah)}\to M^{(ah)}\circ\sigma e 2 π i h 0 M ( ah ) : M ( ah ) → M ( ah ) ∘ σ is an isomorphism (cf. [12 ] ).
(iii) is a transport of structure through Δ \Delta Δ , with the following dependence. The first condition of 2.5.1: by the first condition of 2.5.3 F V ( M ( a h ) ) = F V ( M ) ( a h ) F_V(M^{(ah)})=F_V(M)^{(ah)} F V ( M ( ah ) ) = F V ( M ) ( ah ) is a module twisted by ϕ ( a h ) \phi(ah) ϕ ( ah ) with the same parity twist as F V ( M ) F_V(M) F V ( M ) , and F V ( M ) F_V(M) F V ( M ) is an (untwisted or parity-twisted) V V V -module by the untwisted theory [8 ] . The second condition of 2.5.1: for X 3 X_3 X 3 twisted and one of X 1 , X 2 X_1,X_2 X 1 , X 2 untwisted, the second conditions of 2.5.3 transports the graded space of intertwining operators of the untwisted case. For X 1 , X 2 X_1,X_2 X 1 , X 2 both twisted one requires the third condition of 2.5.3) (including its compatibility with the Z / 2 Z \mathbb{Z}/2\mathbb{Z} Z /2 Z -gradings, which is part of the formulation of the third condition of 2.5.3). Conditions 2.5.2: the classification and the tensor product on Rep ( V ) f r e e \operatorname{Rep}(V)^{\mathrm{free}} Rep ( V ) free for twisted sectors are transported from the untwisted ones by conditions 2.5.3 if one additionally assumes that the tensor product of twisted V V V -modules exists and that M ↦ M ( a h ) M\mapsto M^{(ah)} M ↦ M ( ah ) is compatible with it.
Remark 6.2. Let us discuss (i). The results [33 , 14 ] twist the second and third arguments of an intertwining operator and leave the first untwisted. A fusion coefficient of two twisted modules, N M 1 ( a h ) , M 2 ( b h ) M 3 ( ( a + b ) h ) N_{M_1^{(ah)},M_2^{(bh)}}^{M_3^{((a+b)h)}} N M 1 ( ah ) , M 2 ( bh ) M 3 (( a + b ) h ) , is not of this form. The statement that it equals N M 1 , M 2 M 3 N_{M_1,M_2}^{M_3} N M 1 , M 2 M 3 is the third condition of 2.5.3, which we do not prove. A proof would use that V 0 ˉ ( a h ) V_{\bar{0}}^{(ah)} V 0 ˉ ( ah ) is an invertible object for a tensor product of twisted modules and the associativity of intertwining operators, i.e., a piece of the G G G -crossed theory. The numerical evidence below (all examples with inner automorphisms, in particular, the agreement of the formulas with the closed-form rule in verify_grouplaw.py, and Example 7.6) is consistent with the third condition of 2.5.3. By (ii) all modules are stable, thus Theorem 2.11 (ii) is available for every generating triple and all coefficients are computed from U U U by part 0.
In Examples 7.1-7.5 the automorphism is of this type and the condition of this Section holds (h 0 h_0 h 0 is a lattice or Cartan zero mode).
Examples
Computations in examples
In computations related to our examples below we used Plain Python 3. The code is contained in python3 verify_all.py etc. Outputs are given in verification_log.txt.
Examples 7.1-7.5 is in verify_all.py. Independent test of Lemma 2.8 in Example 7.1 from twisted traces (see Remark 2.10) is in verify_twisted_traces.py.
( S T ) 3 = S 2 (ST)^3=S^2 ( S T ) 3 = S 2 and the Gauss identity of Lemma 3.1 on the modular data used here (V Z β V_{\mathbb{Z}\beta} V Z β with β 2 = 4 , 6 , 18 \beta^2=4,6,18 β 2 = 4 , 6 , 18 ; Virasoro M ( 3 , 5 ) M(3,5) M ( 3 , 5 ) ; Ising Z / 4 \mathbb{Z}/4 Z /4 ); residuals ≤ 4 ⋅ 10 − 15 \le4\cdot10^{-15} ≤ 4 ⋅ 1 0 − 15 are in verify_gauss.py.
Comparison of the formulas (Theorem 2.11, (4.1)) with fusion coefficients given by the closed form spectral flow rule in Examples 7.1, 7.3, 7.4 (see Remark 6.2) is in verify_grouplaw.py.
Example 7.6 is in verify_product.py.
The barred/unbarred illustration in Remark 2.10 is in verify_conv.py.
Each script header lists its inputs and its normalization conventions. These are the conventions of Section 2.4 (in particular φ Q k = e − 2 π i k / 2 n σ \varphi_{Q_k}=e^{-2\pi i k/2n}\sigma φ Q k = e − 2 π ik /2 n σ in Example 7.1, e − 2 π i k / n σ e^{-2\pi i k/n}\sigma e − 2 π ik / n σ in Example 7.3). The scripts are supplementary material and are not reproduced here.
Inputs and checks: the inputs to verify_all.py are
(i) the modular data of U U U (S U S^U S U and the fusion rules of C U \mathcal{C}_U C U );
(ii) for every V 0 ˉ V_{\bar{0}} V 0 ˉ -module the list of its U U U -components (Proposition 2.6, each list is justified in the example by counting). Other subjects are computed:
(R) Lemma 2.8 (b) is checked for all components of all modules. The orbifold S S S -matrix is computed as S X , Q = ∣ G X ∣ S X 0 , Q ϵ 0 U S_{X,Q}=|G_X|S^U_{X_0,Q\epsilon_0} S X , Q = ∣ G X ∣ S X 0 , Q ϵ 0 U .
(U) Proposition 2.14: the block ( S X , Q ) X ∈ M ( g ) , Q ∈ Q (S_{X,Q})_{X\in\mathcal{M}_{(g)},Q\in\mathcal{Q}} ( S X , Q ) X ∈ M ( g ) , Q ∈ Q is square and unitary for every generating sector g g g .
(F) Fusion coefficients are computed by Part 0 of Theorem 2.11 from the fusion rules of U U U (U U U -components, divided by ∣ O 1 ∣ ∣ O 2 ∣ |O_1||O_2| ∣ O 1 ∣∣ O 2 ∣ ).
(TV, 4.1) Theorem 2.11 and formula (4.1) (with the fixed-point weights) are verified on all in-scope triples (⟨ g 1 , g 2 ⟩ = G \langle g_1,g_2\rangle= G ⟨ g 1 , g 2 ⟩ = G , g 3 = g 1 g 2 g_3 = g_1g_2 g 3 = g 1 g 2 ). Where all modules are stable, the triples with ⟨ g 1 , g 2 ⟩ ≠ G \langle g_1,g_2\rangle\ne G ⟨ g 1 , g 2 ⟩ = G (including g 1 = g 2 = 1 g_1 = g_2 = 1 g 1 = g 2 = 1 ), which lie outside Theorems 2.11 and 4.1 as stated, were checked too and are counted separately (“extra”).
(SR, W, 3.9) the sum rule (Proposition 5.3 ); the weight criterion of Lemma 3.10 and Remark 3.11; Lemma 3.9 including the criterion (iii).
Table 7.
∗ ^{*} ∗ untwisted modules only (for even n n n the twisted weight test is uninformative). “Extra” triples are not covered by the theorems as stated; in Example 7.2 there are unstable modules and no extra triples are checked. Example 7.6 is run by a separate script from products of the data of its factors, thus (R), W and 3.9(iii) do not apply to it.
The content of tests for each example
Stable-column formula (Lemma 2.8 (b)) was tested independently of S U S^{U} S U in Example 7.1 with twisted traces, n = 3 , 9 n = 3, 9 n = 3 , 9 . In Examples 7.1 and 7.3 composite n n n are 9 and 15. In Examples 7.4 and 7.6 composite n n n is 4, and no composite n n n in Examples 7.2 and 7.5. Non-trivial stabilizers are testes in Example 7.2 with n = 2 n = 2 n = 2 , and only by factorization in Example 7.6. In every example the checks (R), (U), (F), (TV, 4.1), (SR) test the algebra of the proofs and the stated conventions given the inputs. They do not test Propositions 2.4 and 2.6 .
Example 7.1. In this example we consider complex fermion with inner automorphism of odd order n n n . Let α \alpha α have ( α , α ) = 1 (\alpha,\alpha) = 1 ( α , α ) = 1 , V = V Z α V = V_{\mathbb{Z}\alpha} V = V Z α , V ‾ 0 = V L \overline{V}_0 = V_L V 0 = V L with L = 2 Z α L = 2\mathbb{Z}\alpha L = 2 Z α , J = V α + L J = V_{\alpha+L} J = V α + L (weights in Z + 1 2 \mathbb{Z} + \frac{1}{2} Z + 2 1 , thus d = 1 d = 1 d = 1 ), σ = e 2 π i α ( 0 ) / n \sigma= e^{2\pi i\alpha(0)/n} σ = e 2 π i α ( 0 ) / n , n ∈ { 3 , 9 , 15 } n \in\{3,9,15\} n ∈ { 3 , 9 , 15 } (prime, respectively, composite).
Untwisted data: Q k = V L + k α / 2 Q_k = V_{L+k\alpha/2} Q k = V L + k α /2 , k ∈ Z / 4 k \in\mathbb{Z}/4 k ∈ Z /4 , S Q k , Q l = 1 2 e − 2 π i k l / 4 S_{Q_k,Q_l} = \frac{1}{2}e^{-2\pi ikl/4} S Q k , Q l = 2 1 e − 2 π ik l /4 , J = Q 2 J = Q_2 J = Q 2 , η Q k = ( − 1 ) k \eta_{Q_k} = (-1)^k η Q k = ( − 1 ) k . J J J freely acts by k ↦ k + 2 k \mapsto k+2 k ↦ k + 2 . All Q k Q_k Q k are stable: Q = { Q 0 , … , Q 3 } Q = \{Q_0,\ldots,Q_3\} Q = { Q 0 , … , Q 3 } , Q + f r e e = { [ Q 0 ] } Q^{\mathrm{free}}_+ = \{[Q_0]\} Q + free = {[ Q 0 ]} , Q − f r e e = { [ Q 1 ] } Q^{\mathrm{free}}_- = \{[Q_1]\} Q − free = {[ Q 1 ]} .
The fixed-point algebra: U = V L ′ U = V_{L'} U = V L ′ , L ′ = 2 n Z α L' = 2n\mathbb{Z}\alpha L ′ = 2 n Z α (norm 4 n 2 4n^2 4 n 2 ), C U C_U C U is pointed, Z / 4 n 2 \mathbb{Z}/4n^2 Z /4 n 2 , modules V L ′ + j α / ( 2 n ) V_{L'+j\alpha/(2n)} V L ′ + j α / ( 2 n ) , S j j ′ U = ( 2 n ) − 1 e − 2 π i j j ′ / 4 n 2 S^U_{jj'} = (2n)^{-1}e^{-2\pi ijj'/4n^2} S j j ′ U = ( 2 n ) − 1 e − 2 π ij j ′ /4 n 2 . The module Q k Q_k Q k consists of the U U U -modules j = n k + 4 n s j = nk+4ns j = nk + 4 n s , the σ a \sigma^a σ a -twisted module T m ( a ) = V L + ( m / 2 + a / n ) α T_m^{(a)} = V_{L_+ (m/2+a/n)\alpha} T m ( a ) = V L + ( m /2 + a / n ) α (a ∈ Z / n a \in\mathbb{Z}/n a ∈ Z / n , m ∈ Z / 4 m \in\mathbb{Z}/4 m ∈ Z /4 ) of j = n m + 2 a + 4 n s j = nm+2a+4ns j = nm + 2 a + 4 n s , s ∈ Z / n s \in\mathbb{Z}/n s ∈ Z / n , with T m ( a + n ) = T m + 2 ( a ) T_m^{(a+n)} = T_{m+2}^{(a)} T m ( a + n ) = T m + 2 ( a ) . The count 4 n 2 = n ⋅ ( n ⋅ 4 ) 4n^2 = n\cdot(n\cdot4) 4 n 2 = n ⋅ ( n ⋅ 4 ) confirms Proposition 2.6 . V G = V n Z α V^{G} = V_{n\mathbb{Z}\alpha} V G = V n Z α is a simple lattice VOSA.
Orbifold data: with φ Q k = e − 2 π i k / 2 n σ a \varphi_{Q_k} = e^{-2\pi ik/2n}\sigma^a φ Q k = e − 2 π ik /2 n σ a (a representation, since σ n = ( − 1 ) k \sigma^n = (-1)^k σ n = ( − 1 ) k on Q k Q_k Q k ). The label of the component j = n k + 4 n s − 2 s j = nk+4ns-2s j = nk + 4 n s − 2 s , Lemma 2.8 gives
S T m ( a ) , Q k = 1 2 e − 2 π i m k / 4 e − 2 π i a k / 2 n , κ T m ( a ) = ( − 1 ) m = ϵ T m ( a ) . S_{T_m^{(a)},Q_k} = \frac{1}{2}e^{-2\pi imk/4}e^{-2\pi iak/2n}, \qquad\kappa_{T_m^{(a)}} = (-1)^m = \epsilon_{T_m^{(a)}}. S T m ( a ) , Q k = 2 1 e − 2 π imk /4 e − 2 π iak /2 n , κ T m ( a ) = ( − 1 ) m = ϵ T m ( a ) .
even m m m corresponds to the σ a \sigma^a σ a -twisted modules V Z α + a α / n V_{\mathbb{Z}\alpha+a\alpha/n} V Z α + a α / n (NS), odd m m m to the σ a P \sigma^aP σ a P -twisted ones (Ramond). This labelling is confirmed independently by the weight test of Remark 3.11, e 2 π i n ( h J x − h X ) = − e X e^{2\pi in(h_Jx-h_X)} = -e_X e 2 π in ( h J x − h X ) = − e X , valid for all modules and sectors. J J J acts by m ↦ m + 2 m \mapsto m+2 m ↦ m + 2 on all sectors.
Independent test of the orbifold S-matrix: the entries above were derived from S U S^U S U by Lemma 2.8 . For this family the twisted characters are theta functions and the law (2.2) can be tested: for n = 3 , 9 n = 3, 9 n = 3 , 9 , all sectors and three values of τ \tau τ , F a , m ( − 1 / τ ) = ∑ k S T m ( a ) , Q k G a , k ( τ ) F_{a,m}(-1/\tau) = \sum_k S_{T_m^{(a)},Q_k}G_{a,k}(\tau) F a , m ( − 1/ τ ) = ∑ k S T m ( a ) , Q k G a , k ( τ ) holds with residual ≤ 1.2 ⋅ 10 − 15 \le1.2\cdot10^{-15} ≤ 1.2 ⋅ 1 0 − 15 , where F a , m F_{a,m} F a , m is the character of T m ( a ) T_m^{(a)} T m ( a ) and G a , k G_{a,k} G a , k the σ − a \sigma^{-a} σ − a -trace on Q k Q_k Q k in the normalization φ Q k = e − 2 π i k / 2 n σ \varphi_{Q_k} = e^{-2\pi ik/2n}\sigma φ Q k = e − 2 π ik /2 n σ (verify_twisted_traces.py). Changing the sign of a a a or conjugating S S S gives residuals 0.68–1.22. This does not use the modular data of U U U , and checks Lemma 2.8 and the conventions of Remark 2.10 for this family (it tests characters, i.e., v = 1 v = 1 v = 1 , and one family only). Checks: the twisted Verlinde formula reproduces the fusion rules T m ( a ) ⊠ T m ′ ( b ) = T m + m ′ ( a + b ) T_m^{(a)} \boxtimes T_{m'}^{(b)} = T_{m+m'}^{(a+b)} T m ( a ) ⊠ T m ′ ( b ) = T m + m ′ ( a + b ) obtained from part 0: for a + b < n a+b<n a + b < n the phases e − 2 π i k ( a + b − c ) / 2 n e^{-2\pi ik(a+b-c)/2n} e − 2 π ik ( a + b − c ) /2 n cancel. For a + b ≥ n a+b\ge n a + b ≥ n they produce ( − 1 ) k (-1)^k ( − 1 ) k , i.e., the shift m ↦ m + 2 m\mapsto m+2 m ↦ m + 2 . (4.1) gives N + = 4 f 0 ( Q 0 ) = 1 N^+ = 4f_0(Q_0) = 1 N + = 4 f 0 ( Q 0 ) = 1 and N − = 4 f 0 ( Q 1 ) = ± 1 N^- = 4f_0(Q_1) = \pm1 N − = 4 f 0 ( Q 1 ) = ± 1 if π ( X 1 ) + π ( X 2 ) = π ( X 3 ) \pi(X_1)+\pi(X_2)=\pi(X_3) π ( X 1 ) + π ( X 2 ) = π ( X 3 ) (and zero otherwise).
Counts: for n = 9 n=9 n = 9 the sectors a ∈ { 3 , 6 } a\in\{3,6\} a ∈ { 3 , 6 } generate a subgroup H H H of order three. There the theorem is to be applied with H H H , and by Lemma 2.9 (b) the entries are those of the case n = 3 n=3 n = 3 checked for each entry by the script. Here Q H = Q G Q_H=Q_G Q H = Q G since all modules are stable, thus the formulas agree.
Summarising the results we have the following. With the first assumption of Subsection 2.1, V Z α V_{\mathbb{Z}\alpha} V Z α is a simple lattice VOSA [26 ] . V L V_L V L , L L L even positive definite is simple, rational, C 2 C_2 C 2 -cofinite, self-dual [9 ] . With the second assumption of Subsection 2.1, σ \sigma σ acts on e 2 m α e^{2m\alpha} e 2 m α by e 4 π i m / n e^{4\pi im/n} e 4 π im / n : order n n n on V ‾ 0 \overline{V}_0 V 0 (n n n odd). For V G V^G V G simple, then one has Lemma 2.1 , in fact V G = V n Z α V^G=V_{n\mathbb{Z}\alpha} V G = V n Z α . With Theorem 2.11 (ii), n = 3 n=3 n = 3 prime, n = 9 , 15 n=9,15 n = 9 , 15 : all modules stable, With Proposition 6.1 (ii), the property of § 6 holds for lattice zero modes. For fixed points, one has that J J J acts freely on all of R \mathcal{R} R . Propositions 2.14, 5.3 and Lemma 3.10 are checked.
Example 7.2. In this example we consider two Majorana fermions, p = 2 p=2 p = 2 : non-inner, unstable modules, and fixed twisted modules. Let V = F ⊗ F V=F\otimes F V = F ⊗ F (F ‾ 0 = I 0 \overline{F}_0=I_0 F 0 = I 0 , F ‾ 1 = I 1 / 2 \overline{F}_1=I_{1/2} F 1 = I 1/2 Ising modules), V ‾ 0 = I 0 ⊗ I 0 ⊕ I 1 / 2 ⊗ I 1 / 2 \overline{V}_0=I_0\otimes I_0\oplus I_{1/2}\otimes I_{1/2} V 0 = I 0 ⊗ I 0 ⊕ I 1/2 ⊗ I 1/2 , J = I 0 ⊗ I 1 / 2 ⊕ I 1 / 2 ⊗ I 0 J=I_0\otimes I_{1/2}\oplus I_{1/2}\otimes I_0 J = I 0 ⊗ I 1/2 ⊕ I 1/2 ⊗ I 0 , σ = 1 ⊗ P \sigma=1\otimes P σ = 1 ⊗ P , order two, nontrivial on V ‾ 0 \overline{V}_0 V 0 . Then U = I 0 ⊗ I 0 U=I_0\otimes I_0 U = I 0 ⊗ I 0 , C U \mathcal{C}_U C U the product of two Ising categories (modules ( a , b ) (a,b) ( a , b ) , a , b ∈ { 1 , σ , ψ } a,b\in\{1,\sigma,\psi\} a , b ∈ { 1 , σ , ψ } ), J 0 = ( ψ , 1 ) J_0=(\psi,1) J 0 = ( ψ , 1 ) , d = 1 d=1 d = 1 . The U U U -components are: V ‾ 0 = ( 1 , 1 ) + ( ψ , ψ ) \overline{V}_0=(1,1)+(\psi,\psi) V 0 = ( 1 , 1 ) + ( ψ , ψ ) , J = ( ψ , 1 ) + ( 1 , ψ ) J=(\psi,1)+(1,\psi) J = ( ψ , 1 ) + ( 1 , ψ ) , T 1 = ( 1 , σ ) + ( ψ , σ ) T_1=(1,\sigma)+(\psi,\sigma) T 1 = ( 1 , σ ) + ( ψ , σ ) , T 2 = ( σ , 1 ) + ( σ , ψ ) T_2=(\sigma,1)+(\sigma,\psi) T 2 = ( σ , 1 ) + ( σ , ψ ) , and the unstable pair Q [ 1 ] , Q [ 3 ] Q_{[1]},Q_{[3]} Q [ 1 ] , Q [ 3 ] with the single U U U -module ( σ , σ ) (\sigma,\sigma) ( σ , σ ) (count 9 = 2 ⋅ 4 + 1 9=2\cdot4+1 9 = 2 ⋅ 4 + 1 ). Thus, Q = { V ‾ 0 , J } Q=\{\overline{V}_0,J\} Q = { V 0 , J } , M ( σ ) = { T 1 , T 2 } \mathcal{M}(\sigma)=\{T_1,T_2\} M ( σ ) = { T 1 , T 2 } . Lemma 2.8 gives, in the columns ( V ‾ 0 , J ) (\overline{V}_0,J) ( V 0 , J ) ,
S T 1 , ⋅ = ( 2 2 , 2 2 ) , S T 2 , ⋅ = ( 2 2 , − 2 2 ) , S Q [ 1 ] , ⋅ = ( 1 2 , − 1 2 ) . S_{T_1,\cdot}=\left(\frac{\sqrt{2}}{2},\frac{\sqrt{2}}{2}\right),\qquad S_{T_2,\cdot}=\left(\frac{\sqrt{2}}{2},-\frac{\sqrt{2}}{2}\right),\qquad S_{Q_{[1]},\cdot}=\left(\frac{1}{2},-\frac{1}{2}\right). S T 1 , ⋅ = ( 2 2 , 2 2 ) , S T 2 , ⋅ = ( 2 2 , − 2 2 ) , S Q [ 1 ] , ⋅ = ( 2 1 , − 2 1 ) .
Part 0 gives T 1 ⊠ T 1 = T 2 ⊠ T 2 = V ‾ 0 ⊕ J T_1\boxtimes T_1=T_2\boxtimes T_2=\overline{V}_0\oplus J T 1 ⊠ T 1 = T 2 ⊠ T 2 = V 0 ⊕ J , T 1 ⊠ T 2 = Q [ 1 ] ⊕ Q [ 3 ] T_1\boxtimes T_2=Q_{[1]}\oplus Q_{[3]} T 1 ⊠ T 2 = Q [ 1 ] ⊕ Q [ 3 ] (an unstable output) and Q [ 1 ] ⊠ T 1 = T 2 Q_{[1]}\boxtimes T_1=T_2 Q [ 1 ] ⊠ T 1 = T 2 (an unstable input). Theorem 2.11 reproduces all of them (48 in-scope triples, including these).
Signs: κ T 1 = 1 \kappa_{T_1}=1 κ T 1 = 1 , κ T 2 = − 1 \kappa_{T_2}=-1 κ T 2 = − 1 , κ Q [ 1 ] = − 1 \kappa_{Q_{[1]}}=-1 κ Q [ 1 ] = − 1 . T 1 T_1 T 1 is the σ \sigma σ -twisted NS⊗ \otimes ⊗ R module and T 2 T_2 T 2 the σ P \sigma P σ P -twisted R⊗ \otimes ⊗ NS module. By the physical identification, the first conditions of 2.5.1 is the statement that this labelling is correct. The weight test holds for the four untwisted modules. (4.1) holds, e.g., X 1 = T 1 X_1=T_1 X 1 = T 1 , X 2 = T 2 X_2=T_2 X 2 = T 2 , X 3 = Q [ 1 ] X_3=Q_{[1]} X 3 = Q [ 1 ] gives N + = 4 f 0 ( V ‾ 0 ) = 2 N^+=4f_0(\overline{V}_0)=2 N + = 4 f 0 ( V 0 ) = 2 , N − = 0 N^-=0 N − = 0 .
J ⊠ T i ≃ T i J\boxtimes T_i\simeq T_i J ⊠ T i ≃ T i : every twisted module is a fixed point (Lemma 3.9 (iii) detects this on the S S S -matrix), thus no triple with g i = σ g_i=\sigma g i = σ consists of free objects and Corollary 4.2 does not apply. Conditions 2.5.1 are claimed for fixed objects (conditions 2.5.1) but nothing is known about the structure of F V ( T i ) = T i ⊕ T i F_{V}(T_i)=T_i\oplus T_i F V ( T i ) = T i ⊕ T i . Theorem 4.1 (a) is unaffected. Note that T 1 T_1 T 1 is parity-un-twisted although fixed (see Remark 3.11).
Higher rank (not verified): for V = F 0 ‾ ⊗ 2 N V=F_{\overline{0}}^{\otimes2N} V = F 0 ⊗ 2 N with σ \sigma σ flipping one fermion one has V ‾ 0 = s o ( 2 N ) 1 \overline{V}_0=\mathfrak{so}(2N)_1 V 0 = so ( 2 N ) 1 and U = F 0 ‾ ⊗ ( 2 N − 1 ) ⊗ F 0 U=F_{\overline{0}}^{\otimes(2N-1)}\otimes F_0 U = F 0 ⊗ ( 2 N − 1 ) ⊗ F 0 . By the standard description of s o ( m ) 1 \mathfrak{so}(m)_1 so ( m ) 1 (m m m odd) its modular category has the Ising S S S -matrix. Thus, the S S S -data of this example would be the same for every N N N . We verified only N = 1 N=1 N = 1 by script. The identification for N ≥ 2 N\ge2 N ≥ 2 is the standard.
Summarizing result, we have the following. With both assumptions of Subsection 2.1, V = F ⊗ 2 V=F^{\otimes2} V = F ⊗ 2 is simple, V ‾ 0 \overline{V}_0 V 0 is regular, σ \sigma σ is nontrivial on V ‾ 0 \overline{V}_0 V 0 , and V G = F ⊗ I 0 V^G=F\otimes I_0 V G = F ⊗ I 0 . With G ∩ ⟨ P ⟩ = 1 G\cap\langle P\rangle=1 G ∩ ⟨ P ⟩ = 1 , σ ≠ P \sigma\ne P σ = P and V = F ⊗ P V=F\otimes P V = F ⊗ P . With Theorem 2.11, n = 2 n=2 n = 2 prime [12 ] , unstable Q [ 1 ] Q_{[1]} Q [ 1 ] as input and output (part 0 is checked). Propositions 2.14, 5.3 are checked.
Example 7.3. In this example we consider L 1 ( o s p ( 1 ∣ 2 ) ) L_1(\mathfrak{osp}(1|2)) L 1 ( osp ( 1∣2 )) : wrong statistics, non-pointed, orders 3, 9, 15.
Known from the literature:
(a) V = L 1 ( o s p ( 1 ∣ 2 ) ) V=L_1(\mathfrak{osp}(1|2)) V = L 1 ( osp ( 1∣2 )) is a simple VOSA with V ‾ 0 = V Z α ⊗ M ( 3 , 5 ) \overline{V}_0=V_{\mathbb{Z}\alpha}\otimes M(3,5) V 0 = V Z α ⊗ M ( 3 , 5 ) , ( α , α ) = 2 (\alpha,\alpha)=2 ( α , α ) = 2 (V Z α = L 1 ( s l 2 ) V_{\mathbb{Z}\alpha}=L_1(\mathfrak{sl}_2) V Z α = L 1 ( sl 2 ) ) and J = V Z α + α / 2 ⊗ ϕ 1 , 4 J=V_{\mathbb{Z}\alpha+\alpha/2}\otimes\phi_{1,4} J = V Z α + α /2 ⊗ ϕ 1 , 4 , the coset construction of [6 ] . One requires that V Z α + α / 2 ⊗ ϕ 1 , 4 V_{\mathbb{Z}\alpha+\alpha/2}\otimes\phi_{1,4} V Z α + α /2 ⊗ ϕ 1 , 4 is a simple current of order two for V ‾ 0 \overline{V}_0 V 0 whose superalgebra extension V V V is simple.
(b) The Virasoro minimal model M ( 3 , 5 ) M(3,5) M ( 3 , 5 ) is rational and C 2 C_2 C 2 -cofinite, [39 ] .
We checked here: the central charge 2 5 = 1 + ( − 3 5 ) \frac{2}{5}=1+\left(-\frac{3}{5}\right) 5 2 = 1 + ( − 5 3 ) and the weight 1 4 + 3 4 = 1 \frac{1}{4}+\frac{3}{4}=1 4 1 + 4 3 = 1 of J J J . The S S S -matrix S μ ν = 2 5 ( − 1 ) μ + ν sin ( 3 π μ ν / 5 ) S_{\mu\nu}=\sqrt{\frac{2}{5}}(-1)^{\mu+\nu}\sin(3\pi\mu\nu/5) S μν = 5 2 ( − 1 ) μ + ν sin ( 3 π μν /5 ) of M ( 3 , 5 ) M(3,5) M ( 3 , 5 ) is the modular transformation of the Rocha-Caridi characters χ 1 , μ \chi_{1,\mu} χ 1 , μ (residual < 10 − 15 <10^{-15} < 1 0 − 15 at four values of τ \tau τ ), is symmetric and orthogonal, and gives non-negative integer fusion coefficients with ϕ 1 , 4 ⊠ ϕ 1 , 4 = ϕ 1 , 1 \phi_{1,4}\boxtimes\phi_{1,4}=\phi_{1,1} ϕ 1 , 4 ⊠ ϕ 1 , 4 = ϕ 1 , 1 , ϕ 1 , 2 ⊠ ϕ 1 , 2 = ϕ 1 , 1 ⊕ ϕ 1 , 3 \phi_{1,2}\boxtimes\phi_{1,2}=\phi_{1,1}\oplus\phi_{1,3} ϕ 1 , 2 ⊠ ϕ 1 , 2 = ϕ 1 , 1 ⊕ ϕ 1 , 3 , dim q ϕ 1 , μ = ( 1 , 0.618 , − 0.618 , − 1 ) \dim_q\phi_{1,\mu}= (1,0.618,-0.618,-1) dim q ϕ 1 , μ = ( 1 , 0.618 , − 0.618 , − 1 ) . Hence, dim q J = 1 ⋅ ( − 1 ) \dim_qJ=1\cdot(-1) dim q J = 1 ⋅ ( − 1 ) , θ J = e 2 π i = 1 \theta_J=e^{2\pi i}=1 θ J = e 2 π i = 1 : d = − 1 d=-1 d = − 1 and θ J = − d \theta_J=-d θ J = − d (Proposition 2.3 (ii)).
Everything below in this example, in particular the value d = − 1 d=-1 d = − 1 , the sign data and counts depend on input (a). Take σ = σ n = e 2 π i α ( 0 ) / n ⊗ 1 \sigma=\sigma_n=e^{2\pi i\alpha(0)/n}\otimes1 σ = σ n = e 2 π i α ( 0 ) / n ⊗ 1 , n ∈ { 3 , 9 , 15 } n\in\{3,9,15\} n ∈ { 3 , 9 , 15 } : order n n n on V ‾ 0 \overline{V}_0 V 0 and on J J J , trivial on the Virasoro factor.
Data: modules of V 0 V_0 V 0 : ( ℓ , μ ) (\ell,\mu) ( ℓ , μ ) , ℓ ∈ Z / 2 \ell\in\mathbb{Z}/2 ℓ ∈ Z /2 . J = ( 1 , 4 ) J=(1,4) J = ( 1 , 4 ) freely acts by ( ℓ , μ ) ↦ ( ℓ + 1 , 5 − μ ) (\ell,\mu)\mapsto(\ell+1,5-\mu) ( ℓ , μ ) ↦ ( ℓ + 1 , 5 − μ ) , Lemma 3.8. Twisted modules: ( T ℓ ( a ) , μ ) (T_\ell^{(a)},\mu) ( T ℓ ( a ) , μ ) , T ℓ ( a ) = V Z α + ( ℓ / 2 + a / n ) α T_\ell^{(a)}=V_{\mathbb{Z}\alpha+(\ell/2+a/n)\alpha} T ℓ ( a ) = V Z α + ( ℓ /2 + a / n ) α . All modules are stable (inner). The lattice factor has U l a t = V n Z α U_{\mathrm{lat}}=V_{n\mathbb{Z}\alpha} U lat = V n Z α , C U C_U C U of rank 2 n 2 2n^2 2 n 2 , S j j ′ U = 1 n 2 e − 2 π i j j ′ / 2 n S^U_{jj'}=\frac{1}{n\sqrt{2}}e^{-2\pi ijj'/2n} S j j ′ U = n 2 1 e − 2 π ij j ′ /2 n . The U U U -components of ( T ℓ ( a ) , μ ) (T_\ell^{(a)},\mu) ( T ℓ ( a ) , μ ) are ( j = n ℓ + 2 a + 2 n s , μ ) (j=n\ell+2a+2ns,\mu) ( j = n ℓ + 2 a + 2 n s , μ ) , s ∈ Z / n s\in\mathbb{Z}/n s ∈ Z / n . In the normalization φ = σ \varphi=\sigma φ = σ the orbifold entries are the sector independent numbers
S ( T ℓ ( a ) , μ ) , ( Q k , ν ) = 1 2 ( − 1 ) ℓ k S μ ν . S_{(T_\ell^{(a)},\mu),(Q_k,\nu)}=\frac{1}{\sqrt{2}}(-1)^{\ell k}S_{\mu\nu}. S ( T ℓ ( a ) , μ ) , ( Q k , ν ) = 2 1 ( − 1 ) ℓ k S μν .
Here the script uses the normalization with φ Q k = e − 2 π i k / n σ \varphi_{Q_k}=e^{-2\pi ik/n}\sigma φ Q k = e − 2 π ik / n σ , which multiplies this by e − 2 π i a k / n e^{-2\pi iak/n} e − 2 π iak / n in the sector a a a by (2.3).
Sign data: η ( k , ν ) = ( − 1 ) k + ν \eta_{(k,\nu)}=(-1)^{k+\nu} η ( k , ν ) = ( − 1 ) k + ν , d = η ( 0 , 1 ) = − 1 d=\eta_{(0,1)}=-1 d = η ( 0 , 1 ) = − 1 , ϵ ( ℓ , μ ) = − ( − 1 ) ℓ + μ \epsilon_{(\ell,\mu)}=-(-1)^{\ell+\mu} ϵ ( ℓ , μ ) = − ( − 1 ) ℓ + μ . The NS (ϵ = 1 \epsilon=1 ϵ = 1 ) J J J -orbits are { ( 0 , 1 ) , ( 1 , 4 ) } \{(0,1),(1,4)\} {( 0 , 1 ) , ( 1 , 4 )} , { ( 0 , 3 ) , ( 1 , 2 ) } \{(0,3),(1,2)\} {( 0 , 3 ) , ( 1 , 2 )} ; the Ramond ones { ( 0 , 2 ) , ( 1 , 3 ) } \{(0,2),(1,3)\} {( 0 , 2 ) , ( 1 , 3 )} , { ( 0 , 4 ) , ( 1 , 1 ) } \{(0,4),(1,1)\} {( 0 , 4 ) , ( 1 , 1 )} . As d = − 1 d=-1 d = − 1 , η = 1 \eta=1 η = 1 on the Ramond orbits: N + N^+ N + is a sum over parity-twisted and N − N^- N − over parity-untwisted stable modules, as in [8 ] . The weight test of Remark 3.11 holds for all modules and sectors.
Checks: the fusion coefficients computed by Part 0 are N ( a , ℓ 1 , μ 1 ) , ( b , ℓ 2 , μ 2 ) ( c , ℓ 3 , μ 3 ) = δ c , a + b δ ℓ 3 , ℓ 1 + ℓ 2 N μ 1 μ 2 μ 3 N^{(c,\ell_3,\mu_3)}_{(a,\ell_1,\mu_1),(b,\ell_2,\mu_2)}=\delta_{c,a+b}\delta_{\ell_3,\ell_1+\ell_2}N^{\mu_3}_{\mu_1\mu_2} N ( a , ℓ 1 , μ 1 ) , ( b , ℓ 2 , μ 2 ) ( c , ℓ 3 , μ 3 ) = δ c , a + b δ ℓ 3 , ℓ 1 + ℓ 2 N μ 1 μ 2 μ 3 with the (non-pointed) Virasoro fusion coefficients; Theorem 2.11 and (4.1) reproduce them (counts in the table). Example: X 1 = X 2 = ( T 0 ( 1 ) , ϕ 1 , 2 ) X_1=X_2=(T_0^{(1)},\phi_{1,2}) X 1 = X 2 = ( T 0 ( 1 ) , ϕ 1 , 2 ) (both Ramond) gives X 3 = ( T 0 ( 2 ) , ϕ 1 , 1 ) X_3=(T_0^{(2)},\phi_{1,1}) X 3 = ( T 0 ( 2 ) , ϕ 1 , 1 ) or ( T 0 ( 2 ) , ϕ 1 , 3 ) (T_0^{(2)},\phi_{1,3}) ( T 0 ( 2 ) , ϕ 1 , 3 ) (NS), N ± = 1 N^\pm=1 N ± = 1 , consistent with ϵ X 1 ϵ X 2 ϵ X 3 = 1 \epsilon_{X_1}\epsilon_{X_2}\epsilon_{X_3}=1 ϵ X 1 ϵ X 2 ϵ X 3 = 1 .
Summarizing results we have the following. With the first assumption of subsection 2.1, inputs (a), (b) above and consistency are checked. With the second assumption of Subsection 2.1, σ \sigma σ is inner on the lattice factor, order n n n on V ‾ 0 \overline{V}_0 V 0 (n n n odd). V G V^G V G is simple by Lemma 2.1. With Theorem 2.11 (ii), n = 3 n=3 n = 3 prime; for n = 9 , 15 n=9,15 n = 9 , 15 : all modules are stable. Fixed points: none (d = − 1 d=-1 d = − 1 , Lemma 3.8. Propositions 2.14, 5.3 are checked with 8 × 8 8\times8 8 × 8 blocks. This is the only example with d = − 1 d=-1 d = − 1 and it is based on (a). It shows that (4.1) carries information beyond the twisted fusion coefficients, namely the sign restriction of the sums.
Example 7.4. In this example of rank two, even composite order n = 4 n=4 n = 4 . Let V = V Z 2 V=V_{\mathbb{Z}^2} V = V Z 2 (two complex fermions, ( α i , α j ) = δ i j (\alpha_i,\alpha_j)=\delta_{ij} ( α i , α j ) = δ ij ), σ = e 2 π i α 1 ( 0 ) / 4 \sigma=e^{2\pi i\alpha_1(0)/4} σ = e 2 π i α 1 ( 0 ) /4 . Then V ‾ 0 = V L 0 \overline{V}_0=V_{L_0} V 0 = V L 0 , L 0 = { ( k 1 , k 2 ) : k 1 + k 2 even } L_0=\{(k_1,k_2):k_1+k_2\text{ even}\} L 0 = {( k 1 , k 2 ) : k 1 + k 2 even } , J = V L 0 + ( 1 , 0 ) J=V_{L_0+(1,0)} J = V L 0 + ( 1 , 0 ) , d = 1 d=1 d = 1 ; σ \sigma σ has order 4 4 4 on V V V and on V ‾ 0 \overline{V}_0 V 0 , with σ 2 = ( − 1 ) k 1 ≠ P = ( − 1 ) k 1 + k 2 \sigma^2=(-1)^{k_1}\ne P=(-1)^{k_1+k_2} σ 2 = ( − 1 ) k 1 = P = ( − 1 ) k 1 + k 2 , thus G ∩ ⟨ P ⟩ = 1 G\cap\langle P\rangle=1 G ∩ ⟨ P ⟩ = 1 and G × ⟨ P ⟩ ≃ Z / 4 × Z / 2 G\times\langle P\rangle\simeq\mathbb{Z}/4\times\mathbb{Z}/2 G × ⟨ P ⟩ ≃ Z /4 × Z /2 is not cyclic. The category of V ‾ 0 \overline{V}_0 V 0 is pointed, Z / 2 × Z / 2 \mathbb{Z}/2\times\mathbb{Z}/2 Z /2 × Z /2 : V ‾ 0 \overline{V}_0 V 0 , J J J , A = ( 1 2 , 1 2 ) A=\left(\frac{1}{2},\frac{1}{2}\right) A = ( 2 1 , 2 1 ) , B = ( 1 2 , − 1 2 ) B=\left(\frac{1}{2},-\frac{1}{2}\right) B = ( 2 1 , − 2 1 ) , ( J ⊠ A = B ⊠ ( J ⊠ A ) = ( + , + , − , − ) ) (J\boxtimes A=B\boxtimes(J\boxtimes A)=(+,+,-,-)) ( J ⊠ A = B ⊠ ( J ⊠ A ) = ( + , + , − , − )) . U = V 4 Z α U=V_{4\mathbb{Z}\alpha} U = V 4 Z α , rank 64 = n 2 ⋅ 4 64=n^2\cdot4 64 = n 2 ⋅ 4 , modules U U U -modules are Z / 16 × Z / 4 \mathbb{Z}/16\times\mathbb{Z}/4 Z /16 × Z /4 with S U = 1 8 e − 2 π i ( j 1 j 1 ′ / 16 + j 2 j 2 ′ / 4 ) S^U=\frac{1}{8}e^{-2\pi i(j_1j'_1/16+j_2j'_2/4)} S U = 8 1 e − 2 π i ( j 1 j 1 ′ /16 + j 2 j 2 ′ /4 ) ; the U U U -components of T ( a ) = V L 0 + λ Q + a / 4 T^{(a)}=V_{L_0+\lambda_Q+a/4} T ( a ) = V L 0 + λ Q + a /4 are ( 4 λ Q , 1 + a + 4 t , 2 λ Q , 2 + 2 t ) (4\lambda_{Q,1}+a+4t,2\lambda_{Q,2}+2t) ( 4 λ Q , 1 + a + 4 t , 2 λ Q , 2 + 2 t ) , t ∈ Z / 4 t\in\mathbb{Z}/4 t ∈ Z /4 . All modules are stable (inner); Q + f r e e = { [ V ‾ 0 ] } Q^{{\rm free}}_+=\{[\overline{V}_0]\} Q + free = {[ V 0 ]} , Q − f r e e = { [ A ] } Q^{{\rm free}}_-=\{[A]\} Q − free = {[ A ]} . Theorem 2.11 reproduces the Part 0 fusion rules and (4.1) holds on all in-scope triples, ϵ \epsilon ϵ takes both values. This example has even n n n (thus σ n / 2 ≠ P \sigma^{n/2}\ne P σ n /2 = P is essential) and a non-cyclic group G × ⟨ P ⟩ G\times\langle P\rangle G × ⟨ P ⟩ . Two assumptions of Subsection 2.1, unitarity, sum rule and the untwisted weight test are as in Example 7.1/the table. The correspondences depend on conditions 2.5.3.
Example 7.5. In this example we consider three Majorana fermions: a J J J -fixed untwisted module. Let V = F ⊗ 3 = V Z α ⊗ F ( ( α , α ) = 1 ) V=F^{\otimes3}=V_{Z\alpha}\otimes F\ ((\alpha,\alpha)=1) V = F ⊗ 3 = V Z α ⊗ F (( α , α ) = 1 ) , V ‾ 0 = L 2 ( s l 2 ) \overline{V}_0=L_2(\mathfrak{sl}_2) V 0 = L 2 ( sl 2 ) (c = 3 / 2 c=3/2 c = 3/2 ), J J J the spin-one module (h = 1 2 h=\frac{1}{2} h = 2 1 , d = 1 d=1 d = 1 ), σ = e π i α ( 0 ) ⊗ 1 \sigma=e^{\pi i\alpha(0)}\otimes1 σ = e π i α ( 0 ) ⊗ 1 (a rotation by π \pi π ; order two, inner on V ‾ 0 \overline{V}_0 V 0 ). Untwisted modules (all stable): [ 0 ] = V ‾ 0 [0]=\overline{V}_0 [ 0 ] = V 0 , [ 1 ] = J [1]=J [ 1 ] = J , [ 1 2 ] [\frac{1}{2}] [ 2 1 ] (h = 3 / 16 h=3/16 h = 3/16 ); J J J swaps [ 0 ] ↔ [ 1 ] [0]\leftrightarrow[1] [ 0 ] ↔ [ 1 ] and fixes [ 1 2 ] [\frac{1}{2}] [ 2 1 ] . U = I 0 ⊗ V 2 Z α U=I_0\otimes V_{2\mathbb{Z}\alpha} U = I 0 ⊗ V 2 Z α (Ising ⊠ Z / 4 \boxtimes\mathbb{Z}/4 ⊠ Z /4 , 12 = 2 ⋅ 6 12=2\cdot6 12 = 2 ⋅ 6 modules); as U U U -modules [ 0 ] = ( 1 , 0 ) + ( ψ , 2 ) [0]=(1,0)+(\psi,2) [ 0 ] = ( 1 , 0 ) + ( ψ , 2 ) , [ 1 ] = ( ψ , 0 ) + ( 1 , 2 ) [1]=(\psi,0)+(1,2) [ 1 ] = ( ψ , 0 ) + ( 1 , 2 ) , [ 1 2 ] = ( σ , 1 ) + ( σ , 3 ) [\frac{1}{2}]=(\sigma,1)+(\sigma,3) [ 2 1 ] = ( σ , 1 ) + ( σ , 3 ) and the three σ \sigma σ -twisted modules are T A = ( 1 , 1 ) + ( ψ , 3 ) T_A=(1,1)+(\psi,3) T A = ( 1 , 1 ) + ( ψ , 3 ) , T B = ( 1 , 3 ) + ( ψ , 1 ) T_B=(1,3)+(\psi,1) T B = ( 1 , 3 ) + ( ψ , 1 ) , T C = ( σ , 0 ) + ( σ , 2 ) T_C=(\sigma,0)+(\sigma,2) T C = ( σ , 0 ) + ( σ , 2 ) (the components of each module are related by multiplication with ( ψ , 2 ) (\psi,2) ( ψ , 2 ) , the σ \sigma σ -odd part of V ‾ 0 \overline{V}_0 V 0 ). Lemma 2.8 gives, in the columns ( [ 0 ] , [ 1 ] , [ 1 2 ] ) ([0],[1],[\frac{1}{2}]) ([ 0 ] , [ 1 ] , [ 2 1 ]) ,
S T A , ⋅ = ( 1 2 , 1 2 , − i 2 2 ) , S T B , ⋅ = ( 1 2 , 1 2 , i 2 2 ) , S T C , ⋅ = ( 2 2 , − 2 2 , 0 ) , S_{T_A,\cdot}=\left(\frac{1}{2},\frac{1}{2},-\frac{i\sqrt{2}}{2}\right),\qquad
S_{T_B,\cdot}=\left(\frac{1}{2},\frac{1}{2},\frac{i\sqrt{2}}{2}\right),\qquad
S_{T_C,\cdot}=\left(\frac{\sqrt{2}}{2},-\frac{\sqrt{2}}{2},0\right), S T A , ⋅ = ( 2 1 , 2 1 , − 2 i 2 ) , S T B , ⋅ = ( 2 1 , 2 1 , 2 i 2 ) , S T C , ⋅ = ( 2 2 , − 2 2 , 0 ) ,
and S [ 1 2 ] , ⋅ = ( 2 2 , − 2 2 , 0 ) S_{[\frac{1}{2}],\cdot}=(\frac{\sqrt{2}}{2},-\frac{\sqrt{2}}{2},0) S [ 2 1 ] , ⋅ = ( 2 2 , − 2 2 , 0 ) , η = ( 1 , 1 , − 1 ) \eta=(1,1,-1) η = ( 1 , 1 , − 1 ) : the fixed point [ 1 2 ] [\frac{1}{2}] [ 2 1 ] has η = − 1 \eta=-1 η = − 1 (Lemma 3.9 (i)); by Lemma 3.9 (iii) the fixed twisted module T C T_C T C is detected by ∑ Q η Q ∣ S T C , Q ∣ 2 = 1 2 + 1 2 − 0 = 1 \sum_Q\eta_Q|S_{T_C,Q}|^2=\frac{1}{2}+\frac{1}{2}-0=1 ∑ Q η Q ∣ S T C , Q ∣ 2 = 2 1 + 2 1 − 0 = 1 , while ∑ Q η Q ∣ S T A , Q ∣ 2 = 1 4 + 1 4 − 1 2 = 0 ≠ 1 \sum_Q\eta_Q|S_{T_A,Q}|^2=\frac{1}{4}+\frac{1}{4}-\frac{1}{2}=0\ne1 ∑ Q η Q ∣ S T A , Q ∣ 2 = 4 1 + 4 1 − 2 1 = 0 = 1 (T A T_A T A is free). J ⊠ T A = T B J\boxtimes T_A=T_B J ⊠ T A = T B , J ⊠ T C = T C J\boxtimes T_C=T_C J ⊠ T C = T C ; ϵ = ( + , + , − ) \epsilon=(+,+,-) ϵ = ( + , + , − ) on ( T A , T B , T C ) (T_A,T_B,T_C) ( T A , T B , T C ) , ϵ [ 1 / 2 ] = − 1 \epsilon_{[1/2]}=-1 ϵ [ 1/2 ] = − 1 .
The fixed point matters: take ( X 1 , X 2 , X 3 ) = ( T A , T B , [ 0 ] ) (X_1,X_2,X_3)=(T_A,T_B,[0]) ( X 1 , X 2 , X 3 ) = ( T A , T B , [ 0 ]) : N T A , T B [ 0 ] = 1 N_{T_A,T_B}^{[0]}=1 N T A , T B [ 0 ] = 1 , N T A , T B [ 1 ] = 0 N_{T_A,T_B}^{[1]}=0 N T A , T B [ 1 ] = 0 , thus N + = N − = 1 N^+=N^-=1 N + = N − = 1 . By (4.1): N + = 4 f 0 ( [ 0 ] ) = 4 ⋅ 1 4 = 1 N^+=4f_0([0])=4\cdot\frac{1}{4}=1 N + = 4 f 0 ([ 0 ]) = 4 ⋅ 4 1 = 1 , and N − = 4 ∑ Q f r e e f 0 + 2 f 0 ( [ 1 2 ] ) = 0 + 2 ⋅ 1 2 = 1 N^-=4\sum_{Q_{\mathrm{free}}}f_0+2f_0([\frac{1}{2}])=0+2\cdot\frac{1}{2}=1 N − = 4 ∑ Q free f 0 + 2 f 0 ([ 2 1 ]) = 0 + 2 ⋅ 2 1 = 1 : the minus sign comes entirely from the fixed point, with weight two (weight 4 would give 2, weight 0 would give 0). Over the 81 in-scope triples, treating [ 1 2 ] [\frac{1}{2}] [ 2 1 ] as a free orbit gives a wrong N − N^- N − in 24 cases. Also T C ⊠ T C = [ 0 ] ⊕ [ 1 ] T_C\boxtimes T_C=[0]\oplus[1] T C ⊠ T C = [ 0 ] ⊕ [ 1 ] (N + = 2 N^+=2 N + = 2 , N − = 0 N^-=0 N − = 0 ) and T A ⊠ T C = [ 1 2 ] T_A\boxtimes T_C=[\frac{1}{2}] T A ⊠ T C = [ 2 1 ] (N + = 2 N^+=2 N + = 2 , N − = 0 N^-=0 N − = 0 since S T C , [ 1 / 2 ] = 0 S_{T_C,[1/2]}=0 S T C , [ 1/2 ] = 0 , in accordance with Lemma 3.9 (ii)).
Summarizing results, the script has checked that, with both assumptions of Subsection 2.1, F ⊗ 3 F^{\otimes3} F ⊗ 3 is simple, L 2 ( s l 2 ) L_2(\mathfrak{sl}_2) L 2 ( sl 2 ) is regular, σ \sigma σ is nontrivial on V ‾ 0 \overline{V}_0 V 0 , and V G = U ⊕ J 0 = F ⊗ V 2 Z α V^G=U\oplus J_0=F\otimes V_{2\mathbb{Z}\alpha} V G = U ⊕ J 0 = F ⊗ V 2 Z α . For Theorem 2.11 (ii), n = 2 n=2 n = 2 prime, all modules are stable. For fixed points, [ 1 2 ] ∈ Q f i x [\frac{1}{2}]\in Q^{\mathrm{fix}} [ 2 1 ] ∈ Q fix , T C T_C T C , Lemma 3.9 (i)-(iii) is checked. Conditions 2.5.1 are checked for fixed objects, depending on conditions 2.5.3.
Example 7.6. In this example we consider a product with a proper stabilizer, n = 4 n=4 n = 4 .
Let A = L 1 ( s l 2 ) = V Z β A=L_1(\mathfrak{sl}_2)=V_{\mathbb{Z}\beta} A = L 1 ( sl 2 ) = V Z β (( β , β ) = 2 ( \beta,\beta)=2 ( β , β ) = 2 ), σ A = e 2 π i β ( 0 ) / 8 \sigma_A=e^{2\pi i\beta(0)/8} σ A = e 2 π i β ( 0 ) /8 (order 4 on A A A ), and let V = A ⊗ F ⊗ F V=A\otimes F\otimes F V = A ⊗ F ⊗ F , thus V ‾ 0 = A ⊗ B 0 \overline{V}_0=A\otimes B_0 V 0 = A ⊗ B 0 with B 0 B_0 B 0 the even part of Example 7.2, J = A ⊗ J B J=A\otimes J_B J = A ⊗ J B , d = 1 d=1 d = 1 , and σ = σ A ⊗ ( 1 ⊗ P ) \sigma=\sigma_A\otimes(1\otimes P) σ = σ A ⊗ ( 1 ⊗ P ) , G = ⟨ σ ⟩ ≃ Z / 4 G=\langle\sigma\rangle\simeq\mathbb{Z}/4 G = ⟨ σ ⟩ ≃ Z /4 . The first assumption of Subsection 2.1 holds; σ 2 = σ A 2 ⊗ 1 ≠ 1 \sigma^2=\sigma_A^2\otimes1\ne1 σ 2 = σ A 2 ⊗ 1 = 1 on A A A and σ k \sigma^k σ k is nontrivial on B 0 B_0 B 0 for k k k odd, thus G G G is faithful on V ‾ 0 \overline{V}_0 V 0 the second assumption of Subsection 2.1, and σ k ≠ P \sigma^k\ne P σ k = P for all k k k (Lemma 2.2).
Structure: twisted modules of A ⊗ B 0 A\otimes B_0 A ⊗ B 0 are tensor products of twisted modules, and twisted traces and fusion coefficients factorize. All A A A -modules are stable. A pair ( X A , y ) (X_A,y) ( X A , y ) with y ∈ { Q [ 1 ] , Q [ 3 ] } y\in\{Q_{[1]},Q_{[3]}\} y ∈ { Q [ 1 ] , Q [ 3 ] } (the unstable pair of Example 7.2, in even sectors) has G G G -orbit of size two and stabilizer ⟨ σ 2 ⟩ \langle\sigma^2\rangle ⟨ σ 2 ⟩ . The twisted modules of odd sectors have B B B -part T 1 T_1 T 1 or T 2 T_2 T 2 and are J J J -fixed. In-scope triples include ( g 1 , g 2 , g 3 ) = ( σ 2 , σ , σ 3 ) (g_1,g_2,g_3)=(\sigma^2,\sigma,\sigma^3) ( g 1 , g 2 , g 3 ) = ( σ 2 , σ , σ 3 ) with M 1 M_1 M 1 non-stable and M 2 , M 3 M_2,M_3 M 2 , M 3 stable, the situation of Remark 2.12.
Check (verify_product.py): the S S S -entries are the products of those of the factors (the A A A -factor from U A = V 4 Z β U_A=V_{4\mathbb{Z}\beta} U A = V 4 Z β , the B B B -factor from Example 7.2, with the B B B -sector a a a mod 2, Lemma 2.9 (b)). The fusion coefficients are the products of Part 0 coefficients of the factors. For G = Z / 4 G=\mathbb{Z}/4 G = Z /4 with Q = Q A × Q B Q=Q_A\times Q_B Q = Q A × Q B (4 modules), the generating blocks (a = 1 , 3 a=1,3 a = 1 , 3 ) are 4 × 4 4\times4 4 × 4 unitary (Proposition 2.14). Theorem 2.11 and (4.1) hold on all 1536 in-scope triples (of which 768 have one non-stable module, of stabilizer order two, and two stable ones; all involve J J J -fixed twisted modules), and the sum rule holds (256 checks).
Limitation: due to factorization, the script does not compute U = V 0 G U=V_0^G U = V 0 G and test Lemma 2.8 for the diagonal group. The outcome is implied by the corresponding statements for the factors (A A A with Z / 4 \mathbb{Z}/4 Z /4 , all modules stable, and B B B with Z / 2 \mathbb{Z}/2 Z /2 ). The example tests the formulation for composite order with proper stabilizers, i.e., index set Q G Q_G Q G , orbit conventions, J J J -fixed twisted modules in the composite-order setting. A non-product example is listed in Section 8.
Affine even part and open questions
Affine even part
8.1. Affine even part. Let V ‾ 0 = L g ( k , 0 ) \overline{V}_0=L_g(k,0) V 0 = L g ( k , 0 ) be an affine vertex operator algebra at positive integer level and J = L g ( k ω J ) J=L_g(k\omega_J) J = L g ( k ω J ) a simple current of order two with θ J = − dim q J \theta_J=-\dim_q J θ J = − dim q J , so that V = V ‾ 0 ⊕ J V=\overline{V}_0\oplus J V = V 0 ⊕ J is a simple VOSA (e.g., s l 2 \mathfrak{sl}_2 sl 2 with k ≡ 2 ( m o d 4 ) k\equiv2\pmod4 k ≡ 2 ( mod 4 ) ; s l 2 m \mathfrak{sl}_{2m} sl 2 m , J = L ( k ω m ) J=L(k\omega_m) J = L ( k ω m ) with k m ≡ 2 ( m o d 4 ) ; E x a m p l e s 7.2 , 7.5 ) . I f km\equiv2\pmod4; Examples 7.2, 7.5). If k m ≡ 2 ( mod 4 ) ; E x am pl es 7.2 , 7.5 ) . I f \sigma$ is a diagram automorphism of order two preserving J J J , then by [12 ] the orbifold S S S -matrix is given by the Kac–Peterson coefficients and the fusion coefficients N λ 1 , λ 2 λ 3 N_{\lambda_1,\lambda_2}^{\lambda_3} N λ 1 , λ 2 λ 3 (λ 1 \lambda_1 λ 1 untwisted, λ 2 † , λ 3 † \lambda_2^\dagger,\lambda_3^\dagger λ 2 † , λ 3 † twisted) by the twisted Kac–Walton formula. Theorem 4.1 (a) then expresses N ± = N λ 1 , λ 2 λ 3 ± N λ 1 , λ 2 λ 3 † N^\pm=N_{\lambda_1,\lambda_2}^{\lambda_3}\pm N_{\lambda_1,\lambda_2}^{\lambda_3^\dagger} N ± = N λ 1 , λ 2 λ 3 ± N λ 1 , λ 2 λ 3 † through these coefficients, with the parity selection rule. In the affine cases of Examples 7.2, 7.3, 7.5 everything is computed from the modular data of U U U . For inner automorphisms of affine V ‾ 0 \overline{V}_0 V 0 all twisted modules are stable under the property of Section 6 (Proposition 6.1 (ii)). Thus Theorem 2.11 (ii) and Theorem 4.1 (a) apply to every generating triple. The resulting coefficients agree with the untwisted ones when the first argument is untwisted (Proposition 6.1 (ii)).
Open problems
8.2. Open problems. In this Subsection we list the remaining open problems related to the material of this paper.
Composite order with proper stabilizers, beyond products: example 7.6 is a product and tests only statements. A non-product test in which U = V 0 G U=V_0^G U = V 0 G is computed is open: e.g., V = F ⊗ 4 V=F^{\otimes4} V = F ⊗ 4 with a 4-cycle (thus V ‾ 0 = L 1 ( s l 2 ) ⊗ 2 \overline{V}_0=L_1(\mathfrak{sl}_2)^{\otimes2} V 0 = L 1 ( sl 2 ) ⊗ 2 , σ = s w a p ∘ ( inner involution ) \sigma=\mathrm{swap}\circ(\text{inner involution}) σ = swap ∘ ( inner involution ) , G = Z / 4 G=\mathbb{Z}/4 G = Z /4 , the orbit { ( 0 , 1 2 ) , ( 1 2 , 0 ) } \{(0,\frac{1}{2}),(\frac{1}{2},0)\} {( 0 , 2 1 ) , ( 2 1 , 0 )} with stabilizer
Z/2). Another open question is represented by two unstable modules in a composite-order triple beyond the orbit average (see Remark 2.13).
Non-cyclic abelian and non-abelian G G G (projective representations of stabilizers).
The correspondence for non-inner automorphisms (Example 7.2): existence of a tensor product of twisted V V V -modules and monoidality of F V F_V F V . For inner automorphisms it depends on conditions 2.5.3 (Li results for superalgebras).
Type-Q Q Q structure of F V ( X ) F_V(X) F V ( X ) for J J J -fixed X X X : Theorem 4.1 (a) counts correctly but nothing is known beyond conditions 2.5.1.
Parity type of fixed twisted modules: for n n n odd, is a J J J -fixed twisted module necessarily parity-twisted? See Remark 3.11.
Examples with d = − 1 d=-1 d = − 1 and non-inner automorphisms: Example 7.3 relies on quoted input (a). The Kac-Walton route for k > 1 k>1 k > 1 is untested.
Twisted×twisted transport of the third condition of 2.5.3 and the G G G -crossed identification: whether Theorem 2.11 is the cyclic case of the G G G -crossed Verlinde formula [37 , 10 ] , and a comparison of Lemma 2.8 with the relations of [21 ] for general stabilizers.
The material of this paper is also useful in other areas of mathematical physics [24 , 25 , 27 ] , in particular, in computations of topological invariants of condensed matter theory [42 , 43 , 44 , 45 , 32 ] , and foliation theory [46 , 47 , 48 ] .
Acknowledgments
The author is supported by the Institute of Mathematics, Academy of Sciences of the Czech Republic (RVO 67985840). The author would like to thank C. Dong and S. Kanade for previous discussions.
Data Availability. Data sharing is not applicable to this article as no datasets were generated or analysed during the current study.
Declarations
This paper arose in partial collaboration with ChatGPT, an AI assistant developed by OpenAI. The collaboration involved exploring proof strategies, refining arguments, locating references, Pyton computations, and drafting. Revision and writing of final text have been done by the authors. The authors take full responsibility for the mathematical content and the final text.
Conflict of interest. The authors have no conflicts of interest to declare that are relevant to the content of this article.
Institute of Mathematics, Czech Academy of Sciences, Žitná 25, Prague, Czech Republic
Email address: zuevsky@yahoo.com