Introduction
The paper [26 ] determines when the category C A l o c \mathcal{C}^{\mathrm{loc}}_A C A loc of local modules over a commutative exact algebra A A A in a braided finite tensor category C \mathcal{C} C is rigid, pivotal, ribbon, or a (possibly non-semisimple) modular tensor category. The motivating example is a vertex operator algebra (VOA) extension U ⊂ W U\subset W U ⊂ W , with C = Rep ( U ) \mathcal{C}=\operatorname{Rep}(U) C = Rep ( U ) , A = W A=W A = W and C A l o c ≃ Rep ( W ) \mathcal{C}^{\mathrm{loc}}_A\simeq\operatorname{Rep}(W) C A loc ≃ Rep ( W ) [18 , 8 ] . For vertex operator superalgebras (VOSAs) a VOSA extension of U U U is a supercommutative superalgebra in a category of U U U -modules, and the category of its (untwisted) modules is the category of local modules [7 , 8 ] . The semisimple theory is given in [21 , 8 ] . Rigidity for non-semisimple categories is the subject of [9 ] . According to its introduction, Section 4.3 there treats VOSA extensions.
Let us describe our approach. Following [8 ] , a superalgebra in C \mathcal{C} C is an algebra in the category of graded objects. The Koszul signs are placed in the braiding of that category (our c s c^{\mathrm{s}} c s , § 2.3 ). With this convention a superalgebra is an ordinary commutative algebra A ˉ \bar{A} A ˉ in an ordinary braided category B \mathcal{B} B , thus [26 ] applies to A ˉ \bar{A} A ˉ . But modularity needs extra care. B \mathcal{B} B contains the transparent fermion 1 ⊠ f \mathbb{1}\boxtimes f 1 ⊠ f , thus non-degeneracy is replaced by the condition that the Müger center is s V e c t \mathrm{sVect} sVect . We prove this by embedding B \mathcal{B} B in a modular category constructed with the toric code.
Main results of this paper
All notation is introduced in § 2 . B = C g r \mathcal{B}=\mathcal{C}^{\mathrm{gr}} B = C gr is the graded category, A ˉ = ( A 0 , A 1 ) \bar{A}=(A_0,A_1) A ˉ = ( A 0 , A 1 ) the algebra in B \mathcal{B} B attached to a superalgebra A A A , ε ∈ { ± 1 } \varepsilon\in\{\pm1\} ε ∈ { ± 1 } the twist of the fermion, and C A s − l o c = B A ˉ l o c \mathcal{C}^{\mathrm{s-loc}}_A=\mathcal{B}^{\mathrm{loc}}_{\bar{A}} C A s − loc = B A ˉ loc . For an algebra A ˉ \bar{A} A ˉ in B \mathcal{B} B we write A ˉ B A ˉ \bar{A}\mathcal{B}_{\bar{A}} A ˉ B A ˉ for the category of A ˉ \bar{A} A ˉ -bimodules in B \mathcal{B} B , and E ′ \mathcal{E}' E ′ for the Müger center of a braided category E \mathcal{E} E , as in [26 ] .
Theorem 1.1. Let C \mathcal{C} C be an additive rigid braided category with equalizers and coequalizers, and A A A a supercommutative superalgebra in C \mathcal{C} C (Definition 2.5).
(a) If the category A ˉ B A ˉ \bar{A}\mathcal{B}_{\bar{A}} A ˉ B A ˉ of A ˉ \bar{A} A ˉ -bimodules in B \mathcal{B} B is rigid, then C A s − l o c \mathcal{C}^{\mathrm{s-loc}}_A C A s − loc is rigid.
(b) If C \mathcal{C} C is pivotal, A ˉ B A ˉ \bar{A}\mathcal{B}_{\bar{A}} A ˉ B A ˉ is rigid, and u A 0 = p A 0 u_{A_0}=p_{A_0} u A 0 = p A 0 , u A 1 = ε p A 1 u_{A_1}=\varepsilon p_{A_1} u A 1 = ε p A 1 (Drinfeld isomorphism u u u and pivotal structure p p p of C \mathcal{C} C ), then C A s − l o c \mathcal{C}^{\mathrm{s-loc}}_A C A s − loc is pivotal.
(c) If C \mathcal{C} C is ribbon, A ˉ B A ˉ \bar{A}\mathcal{B}_{\bar{A}} A ˉ B A ˉ is rigid, θ A 0 = i d \theta_{A_0}=\mathrm{id} θ A 0 = id , θ A 1 = ε i d \theta_{A_1}=\varepsilon\mathrm{id} θ A 1 = ε id and A ˉ ∗ ∈ ( C A s − l o c ) ′ \bar{A}^{*}\in(\mathcal{C}^{\mathrm{s-loc}}_A)' A ˉ ∗ ∈ ( C A s − loc ) ′ , then C A s − l o c \mathcal{C}^{\mathrm{s-loc}}_A C A s − loc is ribbon.
2020 Mathematics Subject Classification. Primary 18M20, 18M40; Secondary 17B69, 16W55.
Key words and phrases. Commutative exact superalgebras, Super-modular tensor categories, Braided finite tensor categories, Vertex operator superalgebras, Super-local modules.
Remark 1.2. In (c), ( C A s -loc ) ′ (\mathcal{C}^{s\text{-loc}}_{A})' ( C A s -loc ) ′ is the Müger center, thus the condition says that the dual A ˉ ∗ \bar{A}^{*} A ˉ ∗ , equipped with the right A ˉ \bar{A} A ˉ -module structure of [26 ] , is transparent in C A s -loc \mathcal{C}^{s\text{-loc}}_{A} C A s -loc . If θ A ˉ s = id \theta^{s}_{\bar{A}}=\operatorname{id} θ A ˉ s = id (which is the assumption θ A 0 = id \theta_{A_{0}}=\operatorname{id} θ A 0 = id , θ A 1 = ε id \theta_{A_{1}}=\varepsilon\operatorname{id} θ A 1 = ε id , Proposition 2.6 (c)) this module is automatically local, and θ s \theta^{s} θ s is a twist of C A s -loc \mathcal{C}^{s\text{-loc}}_{A} C A s -loc [26 ] . The extra condition is then equivalent to θ s \theta^{s} θ s being a ribbon structure, i.e., compatible with duality. It holds, for instance, if A ˉ ≅ A ˉ ∗ \bar{A}\cong\bar{A}^{*} A ˉ ≅ A ˉ ∗ as right A ˉ \bar{A} A ˉ -modules (e.g., A ˉ \bar{A} A ˉ Frobenius, since the Frobenius isomorphism is one of right modules), since then A ˉ ∗ ≅ A ˉ \bar{A}^{*}\cong\bar{A} A ˉ ∗ ≅ A ˉ is the unit of C A s -loc \mathcal{C}^{s\text{-loc}}_{A} C A s -loc . In Theorem 1.5 (2) below the condition is not assumed. Ribbon property of C A s -loc \mathcal{C}^{s\text{-loc}}_{A} C A s -loc is obtained from an ambient modular category.
Definition 1.3. A braided finite tensor category D \mathcal{D} D with a distinguished invertible object f f f , f ⊗ f ≅ 1 f\otimes f\cong1 f ⊗ f ≅ 1 , c f , f = − id c_{f,f}=-\operatorname{id} c f , f = − id , is super-non-degenerate if its Müger center D ′ \mathcal{D}' D ′ equals ⟨ 1 , f ⟩ ≃ s V e c t \langle1,f\rangle\simeq\mathrm{sVect} ⟨ 1 , f ⟩ ≃ sVect (a property of the braided category with the object f f f ). If D \mathcal{D} D is moreover ribbon with θ f = ε id \theta_{f}=\varepsilon\operatorname{id} θ f = ε id , it is called ε \varepsilon ε -super-modular . This includes the choice of the ribbon structure.
Remark 1.4. Since θ f = q ( f ) dim f = − dim f \theta_{f}=q(f)\dim f=-\dim f θ f = q ( f ) dim f = − dim f (Remark 2.1), dim f = − ε \dim f=-\varepsilon dim f = − ε . Thus, ( − 1 ) (-1) ( − 1 ) -super-modular categories have dim f = 1 \dim f=1 dim f = 1 and, when spherical (and pseudo-unitary), are the super-modular categories of [5 ] . ( + 1 ) (+1) ( + 1 ) -super-modular categories have dim f = − 1 \dim f=-1 dim f = − 1 , so they are not pseudo-unitary (not all dimensions are positive). The sign of dim f \dim f dim f is not important on sphericality: for any pivotal structure one has dim r X = dim l X ∗ \dim_{r}X=\dim_{l}X^{*} dim r X = dim l X ∗ , [13 ] . Hence, every pivotal structure on a category all of whose simple objects are self-dual (e.g., s V e c t ε \mathrm{sVect}^{\varepsilon} sVect ε and the toric code T ε \mathcal{T}^{\varepsilon} T ε of § 4.1 , for both signs of ε \varepsilon ε ) is spherical.
Theorem 1.5. Let C \mathcal{C} C be a braided finite tensor category over an algebraically closed field of characteristic 0 0 0 and A A A a haploid supercommutative superalgebra with A ˉ \bar{A} A ˉ exact. Then:
(0) C A s -loc \mathcal{C}^{s\text{-loc}}_{A} C A s -loc is a braided finite tensor category. Moreover, A ˉ B ˉ A ˉ \bar{A}\bar{B}\bar{A} A ˉ B ˉ A ˉ is rigid, so that the rigidity assumption in Theorem 1.1 (a)-(c) is automatic, while the other assumptions of Theorem 1.1 (b), (c) (on u u u , p p p , θ \theta θ , A ˉ ∗ \bar{A}^{*} A ˉ ∗ , and the assumption that C \mathcal{C} C be pivotal, resp. ribbon) are still needed for pivotality, resp. ribbonness of C A s -loc \mathcal{C}^{s\text{-loc}}_{A} C A s -loc .
(1) If C \mathcal{C} C is non-degenerate (braided), then C A s -loc \mathcal{C}^{s\text{-loc}}_{A} C A s -loc is super-non-degenerate with fermion f ~ : = Π A ˉ = ( A 1 , A 0 ) \tilde{f}:=\Pi\bar{A}=(A_{1},A_{0}) f ~ := Π A ˉ = ( A 1 , A 0 ) . More precisely, there are a non-degenerate braided finite tensor category N \mathcal{N} N and an invertible f ~ ∈ N \tilde{f}\in\mathcal{N} f ~ ∈ N with C A s -loc = Z ( 2 ) ( ⟨ f ~ ⟩ ⊂ N ) \mathcal{C}^{s\text{-loc}}_{A}=\mathcal{Z}_{(2)}(\langle\tilde{f}\rangle\subset\mathcal{N}) C A s -loc = Z ( 2 ) (⟨ f ~ ⟩ ⊂ N ) and FPdim N = 2 FPdim C A s -loc \operatorname{FPdim}\mathcal{N}=2\operatorname{FPdim}\mathcal{C}^{s\text{-loc}}_{A} FPdim N = 2 FPdim C A s -loc .
(2) If in addition C \mathcal{C} C is ribbon (hence an MTC) and A ˉ \bar{A} A ˉ is symmetric Frobenius, i.e., Frobenius with θ A 0 = id \theta_{A_{0}}=\operatorname{id} θ A 0 = id , θ A 1 = ε id \theta_{A_{1}}=\varepsilon\operatorname{id} θ A 1 = ε id (Definition 2.7), then N \mathcal{N} N is a modular tensor category and C A s -loc \mathcal{C}^{s\text{-loc}}_{A} C A s -loc is ε \varepsilon ε -super-modular.
Corollary 1.6. Under (2), C A s -loc \mathcal{C}^{s\text{-loc}}_{A} C A s -loc has a modular extension N \mathcal{N} N of twice its Frobenius-Perron dimension.
We provide also graded simple currents (Proposition 5.1), examples (§ 6 ), the VOSA notions (§ 7 ), Witt classes (§ 8 ), a comparison with [8 , 9 ] (§ 9 ).
Theorems 1.1 and 1.5 (0) require the material of [26 , 9 ] applied to ( B , A ˉ ) (\mathcal{B},\bar{A}) ( B , A ˉ ) . Proposition 2.6 and Lemma 2.4 use the Koszul sign rule. Lemma 4.3 reflects induction/monodromy statements [24 , 21 ] . The new content of the paper is: the embedding B ↪ C ⊠ T ε \mathcal{B}\hookrightarrow\mathcal{C}\boxtimes\mathcal{T}^{\varepsilon} B ↪ C ⊠ T ε (Lemma 4.1) together with the proof that the image of an exact algebra is exact (Lemma 4.2 ); the resulting description of C A s -loc \mathcal{C}^{s\text{-loc}}_{A} C A s -loc as a centralizer, with its Müger center and its modular extension (Theorem 1.5 (1), (2)); the self-contained proof of Proposition 5.1; and the verification of signs, twists and dimensions in the examples. We do not treat the Ramond sector, other modular completions of s V e c t \mathrm{sVect} sVect , or a super Witt group (Remark 4.4, § 8 ).
Foundations
In this paper, finite tensor categories (FTCs) are over an algebraically closed field k k k of characteristic 0 0 0 . This assumption is used in Theorem 1.5 , Lemma 4.2 , Proposition 5.1 and § 7 : algebraic closedness for the results of [25 , 26 ] on finite tensor categories, characteristic 0 0 0 for the separability of group algebras and for the existence of the toric code. Theorem 1.1 and Sections 2.3.3 hold over an arbitrary field, since they use [26 ] , which is formulated over an arbitrary field. We follow [26 ] for the notions of closed, rigid, pivotal, ribbon categories (p = u θ p=u\theta p = u θ ), the Drinfeld isomorphism u u u , central algebras, C A \mathcal{C}_{A} C A , A C \mathcal{A}^{\mathcal{C}} A C , C A \mathcal{C}^{\mathcal{A}} C A , Frobenius and symmetric Frobenius algebras, exact algebras, haploid algebras, Müger center C ′ \mathcal{C}' C ′ , centralizers Z ( 2 ) ( D ⊂ C ) = { X ∣ c Y , X c X , Y = id ∀ Y ∈ D } \mathcal{Z}_{(2)}(\mathcal{D}\subset\mathcal{C})=\{X\mid c_{Y,X}c_{X,Y}=\operatorname{id}\ \forall Y\in\mathcal{D}\} Z ( 2 ) ( D ⊂ C ) = { X ∣ c Y , X c X , Y = id ∀ Y ∈ D } , non-degenerate braided FTCs (those with C ′ ≃ V e c t \mathcal{C}'\simeq\mathrm{Vect} C ′ ≃ Vect ), and MTCs (i.e., non-degenerate ribbon FTCs). We write mon X , Y = c Y , X ∘ c X , Y \operatorname{mon}_{X,Y}=c_{Y,X}\circ c_{X,Y} mon X , Y = c Y , X ∘ c X , Y . We use right modules. For local modules the left and right conventions agree (§ 4.3 and § 9 ). A tensor full subcategory of an FTC is a full subcategory closed under subquotients, tensor products and duals. In this paper, C A s − l o c \mathcal{C}^{\mathrm{s-loc}}_{\mathcal{A}} C A s − loc denotes the category of super-local modules, B ‾ A l o c \overline{\mathcal{B}}^{\mathrm{loc}}_{\mathcal{A}} B A loc the category of local modules in the ordinary braided category B \mathcal{B} B (these are the same category, Proposition 2.6), and D s \mathcal{D}^{\mathrm{s}} D s (see Remark 4.4) the enriched supercategory constructed on D \mathcal{D} D .
Quadratic data of invertible objects
For an invertible object X X X of a braided category, X ⊗ X X \otimes X X ⊗ X is simple (if 1 \mathbb{1} 1 is) and c X , X = q ( X ) id X ⊗ X c_{X,X}=q(X)\operatorname{id}_{X\otimes X} c X , X = q ( X ) id X ⊗ X for a scalar q ( X ) q(X) q ( X ) .
Remark 2.1. In a ribbon category, θ X = ptr ( c X , X ) \theta_X=\operatorname{ptr}(c_{X,X}) θ X = ptr ( c X , X ) , the partial trace taken with the pivotal structure p = u θ \mathfrak{p}=u\theta p = u θ (the convention of [26 ] ). Hence θ X = q ( X ) dim X \theta_X=q(X)\dim X θ X = q ( X ) dim X for invertible X X X . With the opposite convention θ − 1 \theta^{-1} θ − 1 in place of θ \theta θ nothing changes below, since only θ = ± 1 \theta=\pm1 θ = ± 1 occurs. Thus, c X , X c_{X,X} c X , X and θ X \theta_X θ X are independent data unless dim X \dim X dim X is fixed. Balancing θ X ⊗ Y = ( θ X ⊗ θ Y ) mon X , Y \theta_{X\otimes Y}=(\theta_X\otimes\theta_Y)\operatorname{mon}_{X,Y} θ X ⊗ Y = ( θ X ⊗ θ Y ) mon X , Y gives only θ X ⊗ X = θ X 2 q ( X ) 2 \theta_{X\otimes X}=\theta_X^2q(X)^2 θ X ⊗ X = θ X 2 q ( X ) 2 .
The fermion sVect-epsilon
sVect \operatorname{sVect} sVect is the category of finite-dimensional Z / 2 \mathbb{Z}/2 Z /2 -graded vector spaces with braiding c ( v ⊗ w ) = ( − 1 ) ∣ v ∣ ∣ w ∣ w ⊗ v c(v\otimes w)=(-1)^{|v||w|}w\otimes v c ( v ⊗ w ) = ( − 1 ) ∣ v ∣∣ w ∣ w ⊗ v ; f f f is the odd line, f ⊗ f ≅ 1 f\otimes f\cong\mathbb{1} f ⊗ f ≅ 1 , c f , f = − id c_{f,f}=-\operatorname{id} c f , f = − id , f f f transparent. For ε ∈ { ± 1 } \varepsilon\in\{\pm1\} ε ∈ { ± 1 } let sVect ε \operatorname{sVect}_{\varepsilon} sVect ε be sVect \operatorname{sVect} sVect with the ribbon structure θ 1 = id \theta_{\mathbb{1}}=\operatorname{id} θ 1 = id , θ f = ε id \theta_f=\varepsilon\operatorname{id} θ f = ε id (balancing holds since mon f , f = id \operatorname{mon}_{f,f}=\operatorname{id} mon f , f = id ). By Remark 2.1, dim f = − ε \dim f=-\varepsilon dim f = − ε .
The graded category
Let C \mathcal{C} C be an additive braided monoidal category (finite direct sums exist and ⊗ \otimes ⊗ is additive in each variable; if C \mathcal{C} C is rigid the latter is automatic since X ⊗ − X\otimes- X ⊗ − has adjoints). Define B = C g r \mathcal{B}=\mathcal{C}^{\mathrm{gr}} B = C gr by the following:
objects are pairs X = ( X 0 , X 1 ) X=(X_0,X_1) X = ( X 0 , X 1 ) of objects of C \mathcal{C} C ; ∣ x ∣ ∈ Z / 2 |x|\in\mathbb{Z}/2 ∣ x ∣ ∈ Z /2 denotes the degree of a component;
morphisms X → Y X\to Y X → Y are pairs ( f 0 , f 1 ) (f_0,f_1) ( f 0 , f 1 ) of morphisms of C \mathcal{C} C ;
( X ⊗ Y ) k = ⨁ i + j = k X i ⊗ Y j (X\otimes Y)_k=\bigoplus_{i+j=k}X_i\otimes Y_j ( X ⊗ Y ) k = ⨁ i + j = k X i ⊗ Y j , unit ( 1 , 0 ) (\mathbb{1},0) ( 1 , 0 ) , associator and unitors induced from C \mathcal{C} C (they involve no signs);
c X , Y s = ( − 1 ) i j c X i , Y j c^{\mathrm{s}}_{X,Y}=(-1)^{ij}c_{X_i,Y_j} c X , Y s = ( − 1 ) ij c X i , Y j on X i ⊗ Y j → Y j ⊗ X i X_i\otimes Y_j\to Y_j\otimes X_i X i ⊗ Y j → Y j ⊗ X i ;
if C \mathcal{C} C has a twist θ \theta θ and ε ∈ { ± 1 } \varepsilon\in\{\pm1\} ε ∈ { ± 1 } is fixed: θ s = ε i θ X i \theta^{\mathrm{s}}=\varepsilon^i\theta_{X_i} θ s = ε i θ X i on the component X i X_i X i ;
if C \mathcal{C} C is rigid, X ∗ = ( X 0 ∗ , X 1 ∗ ) X^*=(X_0^*,X_1^*) X ∗ = ( X 0 ∗ , X 1 ∗ ) with ev X = ev X 0 ⊕ ev X 1 \operatorname{ev}_X=\operatorname{ev}_{X_0}\oplus\operatorname{ev}_{X_1} ev X = ev X 0 ⊕ ev X 1 and coev X = coev X 0 ⊕ coev X 1 \operatorname{coev}_X=\operatorname{coev}_{X_0}\oplus\operatorname{coev}_{X_1} coev X = coev X 0 ⊕ coev X 1 (with no signs; these are morphisms of degree 0 0 0 ).
Then C g r \mathcal{C}^{\mathrm{gr}} C gr is C ⊠ sVect \mathcal{C}\boxtimes\operatorname{sVect} C ⊠ sVect whenever the Deligne product exists (this is automatic for finite abelian categories). It has equalizers and coequalizers if C \mathcal{C} C does (for each component), and is rigid if C \mathcal{C} C is (Lemma 2.2 ). The hexagons for c s c^{\mathrm{s}} c s follow from those of c c c since ( − 1 ) i j (-1)^{ij} ( − 1 ) ij is bilinear in the degrees; θ s \theta^{\mathrm{s}} θ s satisfies balancing since ε i + j = ε i ε j \varepsilon^{i+j}=\varepsilon^i\varepsilon^j ε i + j = ε i ε j and Lemma 2.3 . Thus, B \mathcal{B} B is braided (ribbon if C \mathcal{C} C is), and ( 0 , 1 ) (0,1) ( 0 , 1 ) is a transparent invertible object f f f with f ⊗ f ≅ 1 f\otimes f\cong\mathbb{1} f ⊗ f ≅ 1 , c f , f s = − id c^{\mathrm{s}}_{f,f}=-\operatorname{id} c f , f s = − id , θ f s = ε \theta^{\mathrm{s}}_f=\varepsilon θ f s = ε . If C \mathcal{C} C is an FTC, so is B \mathcal{B} B (its unit ( 1 , 0 ) (1,0) ( 1 , 0 ) has endomorphism algebra k k k ).
Lemma 2.2. Let C \mathcal{C} C be rigid. Then ev X \operatorname{ev}_X ev X and coev X \operatorname{coev}_X coev X of item 6. satisfy the zig-zag identities in B \mathcal{B} B , thus X ∗ X^* X ∗ is a left dual of X X X in B \mathcal{B} B . Right duals are obtained in the same way.
Proof. The monoidal structure of B \mathcal{B} B (items 1–3. above) is that of C ⊠ Vect Z / 2 \mathcal{C}\boxtimes\operatorname{Vect}_{\mathbb{Z}/2} C ⊠ Vect Z /2 , thus it involves no signs. Signs enter only through c s c^{\mathrm{s}} c s . Now ( X ∗ ⊗ X ) 0 = X 0 ∗ ⊗ X 0 ⊕ X 1 ∗ ⊗ X 1 (X^*\otimes X)_0=X_0^*\otimes X_0\oplus X_1^*\otimes X_1 ( X ∗ ⊗ X ) 0 = X 0 ∗ ⊗ X 0 ⊕ X 1 ∗ ⊗ X 1 , so ev X \operatorname{ev}_X ev X is well defined of degree 0 0 0 (and zero on the mixed summands X i ∗ ⊗ X j ∗ X_i^*\otimes X_j^* X i ∗ ⊗ X j ∗ , i ≠ j i\ne j i = j ), and likewise coev X : 1 → X 0 ⊗ X 0 ∗ ⊕ X 1 ⊗ X 1 ∗ \operatorname{coev}_X:\mathbb{1}\to X_0\otimes X_0^*\oplus X_1\otimes X_1^* coev X : 1 → X 0 ⊗ X 0 ∗ ⊕ X 1 ⊗ X 1 ∗ . The zig-zag composite ( id X ⊗ ev X ) ( coev X ⊗ id X ) (\operatorname{id}_X\otimes\operatorname{ev}_X)(\operatorname{coev}_X\otimes\operatorname{id}_X) ( id X ⊗ ev X ) ( coev X ⊗ id X ) , restricted to the summand X i X_i X i , passes only through the summands X i ⊗ X i ∗ ⊗ X i X_i\otimes X_i^*\otimes X_i X i ⊗ X i ∗ ⊗ X i and is therefore the zig-zag of X i X_i X i in C \mathcal{C} C , i.e., id X i \operatorname{id}_{X_i} id X i . The other identity is analogous. □
Lemma 2.3. mon X , Y s = mon X , Y \operatorname{mon}^{\mathrm{s}}_{X,Y}=\operatorname{mon}_{X,Y} mon X , Y s = mon X , Y componentwise.
Proof. mon s \operatorname{mon}^{\mathrm{s}} mon s is the composite of c X i , Y j s c^{\mathrm{s}}_{X_i,Y_j} c X i , Y j s and c Y j , X i s c^{\mathrm{s}}_{Y_j,X_i} c Y j , X i s . The two signs ( − 1 ) i j (-1)^{ij} ( − 1 ) ij cancel. □
The following lemma concerns the duality data of B \mathcal{B} B .
Lemma 2.4. Let C \mathcal{C} C be rigid braided. Then:
(i) The Drinfeld isomorphism of B \mathcal{B} B is u X s = ( − 1 ) i u X i u^{\mathrm{s}}_X=(-1)^i u_{X_i} u X s = ( − 1 ) i u X i on X i X_i X i .
(ii) If p \mathfrak{p} p is a pivotal structure of C \mathcal{C} C , then p X ε = ( − 1 ) i ε p X i \mathfrak{p}^{\varepsilon}_X=(-1)^{i\varepsilon}\mathfrak{p}_{X_i} p X ε = ( − 1 ) i ε p X i is a pivotal structure of B \mathcal{B} B (p ε = p \mathfrak{p}^{\varepsilon}=\mathfrak{p} p ε = p for ε = − 1 \varepsilon=-1 ε = − 1 ).
(iii) If C \mathcal{C} C is ribbon and p = u θ \mathfrak{p}=u\theta p = u θ , then p ε = u s θ s \mathfrak{p}^{\varepsilon}=u^{\mathrm{s}}\theta^{\mathrm{s}} p ε = u s θ s .
(iv) dim f = − ε \dim f=-\varepsilon dim f = − ε .
Proof. (i) With the conventions ev X : X ∗ ⊗ X → 1 \operatorname{ev}_X:X^*\otimes X\to\mathbb{1} ev X : X ∗ ⊗ X → 1 , coev X : 1 → X ⊗ X ∗ \operatorname{coev}_X:\mathbb{1}\to X\otimes X^* coev X : 1 → X ⊗ X ∗ ,
u X = ( ev X ⊗ id X ∗ ∗ ) ∘ ( c X , X ∗ ⊗ id X ∗ ∗ ) ∘ ( id X ⊗ coev X ∗ ) : X → X ∗ ∗ . u_X=(\operatorname{ev}_X\otimes\operatorname{id}_{X^{**}})\circ(c_{X,X^*}\otimes\operatorname{id}_{X^{**}})\circ(\operatorname{id}_X\otimes\operatorname{coev}_{X^*}):X\to X^{**}. u X = ( ev X ⊗ id X ∗∗ ) ∘ ( c X , X ∗ ⊗ id X ∗∗ ) ∘ ( id X ⊗ coev X ∗ ) : X → X ∗∗ .
In B \mathcal{B} B the evaluations and coevaluations are those of C \mathcal{C} C componentwise (Lemma 2.2 ), and u X S u^{\mathcal{S}}_{X} u X S involves one braiding c X , X ∗ S c^{\mathcal{S}}_{X,X^{*}} c X , X ∗ S . On the degree-i i i component, X i X_i X i and X i ∗ X_i^{*} X i ∗ both have degree i i i , thus c X i , X i ∗ S = ( − 1 ) i ⋅ i c X i , X i ∗ = ( − 1 ) i c X i , X i ∗ c^{\mathcal{S}}_{X_i,X_i^{*}}=(-1)^{i\cdot i}c_{X_i,X_i^{*}}=(-1)^ic_{X_i,X_i^{*}} c X i , X i ∗ S = ( − 1 ) i ⋅ i c X i , X i ∗ = ( − 1 ) i c X i , X i ∗ .
(ii) A pivotal structure is a monoidal isomorphism id ⇒ ( − ) ∗ ∗ \operatorname{id}\Rightarrow(-)^{**} id ⇒ ( − ) ∗∗ . Naturality for degree-preserving morphisms is clear. The canonical isomorphisms ( X ⊗ Y ) ∗ ∗ ≅ X ∗ ∗ ⊗ Y ∗ ∗ (X\otimes Y)^{**}\cong X^{**}\otimes Y^{**} ( X ⊗ Y ) ∗∗ ≅ X ∗∗ ⊗ Y ∗∗ are constructed from ev \operatorname{ev} ev , coev \operatorname{coev} coev and associators, with no braiding and hence no signs. Thus, monoidality means p X ⊗ Y ε = p X ε ⊗ p Y ε p^{\varepsilon}_{X\otimes Y}=p^{\varepsilon}_{X}\otimes p^{\varepsilon}_{Y} p X ⊗ Y ε = p X ε ⊗ p Y ε on X i ⊗ Y j X_i\otimes Y_j X i ⊗ Y j , i.e., ( − ε ) i + j = ( − ε ) i ( − ε ) j (-\varepsilon)^{i+j}=(-\varepsilon)^i(-\varepsilon)^j ( − ε ) i + j = ( − ε ) i ( − ε ) j , since ( − ε ) 2 = 1 (-\varepsilon)^2=1 ( − ε ) 2 = 1 . (iii) u S s S = ( − 1 ) i u θ = ( − ε ) i p u^{\mathcal{S}}s^{\mathcal{S}}=(-1)^iu\theta=(-\varepsilon)^ip u S s S = ( − 1 ) i u θ = ( − ε ) i p . (iv) dim B f = ev f ∗ ∘ ( p f ε ⊗ id f ∗ ) ∘ coev f \dim_{\mathcal{B}}f=\operatorname{ev}_{f^{*}}\circ(p^{\varepsilon}_{f}\otimes\operatorname{id}_{f^{*}})\circ\operatorname{coev}_{f} dim B f = ev f ∗ ∘ ( p f ε ⊗ id f ∗ ) ∘ coev f equals p f ε = − ε p^{\varepsilon}_{f}=-\varepsilon p f ε = − ε times dim C 1 = 1 \dim_{\mathcal{C}}1=1 dim C 1 = 1 .
The pivotal structure p ε p^{\varepsilon} p ε of B \mathcal{B} B is not the componentwise pivotal structure of C \mathcal{C} C . It differs from it by the sign ( − 1 ) i (-1)^i ( − 1 ) i on the degree-i i i component (they agree for ε = − 1 \varepsilon=-1 ε = − 1 only). The sign is chosen by u S = ( − 1 ) i u u^{\mathcal{S}}=(-1)^iu u S = ( − 1 ) i u and by the requirement p ε = u S θ S p^{\varepsilon}=u^{\mathcal{S}}\theta^{\mathcal{S}} p ε = u S θ S in the ribbon case. Therefore, we have dim B f = − ε \dim_{\mathcal{B}}f=-\varepsilon dim B f = − ε instead of + 1 +1 + 1 .
Superalgebras
Definition 2.5. A superalgebra in C \mathcal{C} C is an algebra ( A , μ , η ) (A,\mu,\eta) ( A , μ , η ) in C \mathcal{C} C with a decomposition A = A 0 ⊕ A 1 A=A_0\oplus A_1 A = A 0 ⊕ A 1 (the grading is part of the data) such that η \eta η factors through A 0 A_0 A 0 and μ ( A i ⊗ A j ) ⊂ A i + j \mu(A_i\otimes A_j)\subset A_{i+j} μ ( A i ⊗ A j ) ⊂ A i + j . It is supercommutative if μ ∘ c A , A = μ ∘ ς \mu\circ c_{A,A}=\mu\circ\varsigma μ ∘ c A , A = μ ∘ ς , where ς ∈ End ( A ⊗ A ) \varsigma\in\operatorname{End}(A\otimes A) ς ∈ End ( A ⊗ A ) acts as ( − 1 ) i j (-1)^{ij} ( − 1 ) ij on A i ⊗ A j A_i\otimes A_j A i ⊗ A j .
Proposition 2.6. Set A ‾ : = ( A 0 , A 1 ) ∈ B \overline{A}:=(A_0,A_1)\in\mathcal{B} A := ( A 0 , A 1 ) ∈ B . Then:
(a) A ↦ A ‾ A\mapsto\overline{A} A ↦ A is a bijection (on isomorphism classes) between superalgebras in C \mathcal{C} C and algebras in B \mathcal{B} B , and A A A is supercommutative iff A ‾ \overline{A} A is commutative in B \mathcal{B} B .
(b) Right A ‾ \overline{A} A -modules in B \mathcal{B} B are Z / 2 \mathbb{Z}/2 Z /2 -graded right A A A -modules M = ( M 0 , M 1 ) M=(M_0,M_1) M = ( M 0 , M 1 ) in C \mathcal{C} C with degree-preserving morphisms. M M M is local iff, with the right action a M : M ⊗ A → M a_M:M\otimes A\to M a M : M ⊗ A → M and the endomorphism mon : = c A , M ∘ c M , A \operatorname{mon}:=c_{A,M}\circ c_{M,A} mon := c A , M ∘ c M , A of M ⊗ A M\otimes A M ⊗ A , one has a M ∘ mon = a M a_M\circ\operatorname{mon}=a_M a M ∘ mon = a M .
(c) If C \mathcal{C} C is ribbon: θ A ‾ = id \theta_{\overline{A}}=\operatorname{id} θ A = id iff θ A 0 = id \theta_{A_0}=\operatorname{id} θ A 0 = id and θ A 1 = ε id \theta_{A_1}=\varepsilon\operatorname{id} θ A 1 = ε id . If C \mathcal{C} C is pivotal: u A ‾ S = p A ‾ ε u^{\mathcal{S}}_{\overline{A}}=p^{\varepsilon}_{\overline{A}} u A S = p A ε iff u A 0 = p A 0 u_{A_0}=p_{A_0} u A 0 = p A 0 and u A 1 = ε p A 1 u_{A_1}=\varepsilon p_{A_1} u A 1 = ε p A 1 .
Note that in (a) superalgebras are algebras with a given Z / 2 \mathbb{Z}/2 Z /2 -grading and graded isomorphisms, i.e., the statement is not about arbitrary ungraded algebra structures on A A A . In (b) we first have c M , A : M ⊗ A → A ⊗ M c_{M,A}:M\otimes A\to A\otimes M c M , A : M ⊗ A → A ⊗ M , then c A , M : A ⊗ M → M ⊗ A c_{A,M}:A\otimes M\to M\otimes A c A , M : A ⊗ M → M ⊗ A , both braidings of C \mathcal{C} C .
Proof. (a) An algebra structure on an object of B \mathcal{B} B is a graded algebra structure on the pair. On A i ⊗ A j A_i\otimes A_j A i ⊗ A j the equation μ ∘ c A ‾ , A ‾ S = μ \mu\circ c^{\mathcal{S}}_{\overline{A},\overline{A}}=\mu μ ∘ c A , A S = μ reads ( − 1 ) i j μ c A i , A j = μ (-1)^{ij}\mu c_{A_i,A_j}=\mu ( − 1 ) ij μ c A i , A j = μ , i.e., μ c = μ ς \mu c=\mu\varsigma μ c = μ ς . (b) A right A ‾ \overline{A} A -module is a graded object with a M : M ⊗ A ‾ → M a_M:M\otimes\overline{A}\to M a M : M ⊗ A → M , i.e., M i ⊗ A j → M i + j M_i\otimes A_j\to M_{i+j} M i ⊗ A j → M i + j . The locality equation in B \mathcal{B} B is a M ∘ c A , M S ∘ c M , A S = a M a_M\circ c^{\mathcal{S}}_{A,M}\circ c^{\mathcal{S}}_{M,A}=a_M a M ∘ c A , M S ∘ c M , A S = a M on M ⊗ A ‾ M\otimes\overline{A} M ⊗ A , and by Lemma 2.3 it is the stated equation. (c) θ S \theta^{\mathcal{S}} θ S acts by θ A 0 \theta_{A_0} θ A 0 on A 0 A_0 A 0 and ε θ A 1 \varepsilon\theta_{A_1} ε θ A 1 on A 1 A_1 A 1 ; for u S = p ε u^{\mathcal{S}}=p^{\varepsilon} u S = p ε use Lemma 2.4 : ( − 1 ) i u A i = ( − ε ) i p A i (-1)^iu_{A_i}=(-\varepsilon)^ip_{A_i} ( − 1 ) i u A i = ( − ε ) i p A i .
Definition 2.7
A A A is haploid ( Hom C ( 1 , A 0 ) ≅ k ) (\operatorname{Hom}_{\mathcal{C}}(1,A_0)\cong\mathbb{k}) ( Hom C ( 1 , A 0 ) ≅ k ) , exact , Frobenius , central , if A ‾ \overline{A} A is such as an algebra in B \mathcal{B} B (thus “A A A exact” always means exactness of A ‾ \overline{A} A in B \mathcal{B} B ). A Frobenius algebra in a pivotal FTC is symmetric if its Nakayama automorphism is the identity [26 ] . For a commutative Frobenius algebra in a braided category with twist, symmetric is equivalent to θ A = id \theta_A=\operatorname{id} θ A = id [26 ] , which is [15 ] (Proposition 2.10 (iii) below). Applied to A ‾ ∈ B \overline{A}\in\mathcal{B} A ∈ B with the twist θ S \theta^{\mathcal{S}} θ S : a commutative Frobenius algebra A A A is symmetric iff θ A ‾ S = id \theta^{\mathcal{S}}_{\overline{A}}=\operatorname{id} θ A S = id , i.e., (Proposition 2.6 (c)) iff θ A 0 = id \theta_{A_0}=\operatorname{id} θ A 0 = id and θ A 1 = ε id \theta_{A_1}=\varepsilon\operatorname{id} θ A 1 = ε id . We use this equivalence as the criterion; the definition is the Nakayama one.
The sign ε \varepsilon ε is part of the data. If A 1 ≠ 0 A_1\ne0 A 1 = 0 and θ A 1 \theta_{A_1} θ A 1 is a scalar ± 1 \pm1 ± 1 , then ε : = θ A 1 \varepsilon:=\theta_{A_1} ε := θ A 1 follows; if A 1 = 0 A_1=0 A 1 = 0 any ε \varepsilon ε works and C A s -loc = C A l o c ⊂ C V e c t ε \mathcal{C}^{s\text{-loc}}_{A}=\mathcal{C}^{\mathrm{loc}}_{A}\subset\mathcal{C}^{\mathrm{Vect}^{\varepsilon}} C A s -loc = C A loc ⊂ C Vect ε .
Preliminaries
For the convenience we list here preliminary results from [26 , 9 , 25 , 10 ] . D \mathcal{D} D denotes a braided monoidal category and A A A a commutative algebra in D \mathcal{D} D . The following proposition is due to [26 ] .
Proposition 2.8. (i) Rigidity: if D \mathcal{D} D is rigid with equalizers and coequalizers (more generally closed monoidal) and D \mathcal{D} D is rigid, then D A \mathcal{D}_A D A and D A loc \mathcal{D}^{\operatorname{loc}}_A D A loc are rigid.
(ii) Pivotal: if in addition D \mathcal{D} D is rigid with a pivotal structure p p p and u A = p A u_A=p_A u A = p A , then D A \mathcal{D}_A D A and D A loc \mathcal{D}^{\operatorname{loc}}_A D A loc are pivotal.
(iii) Ribbon: let D \mathcal{D} D be ribbon (twist θ \theta θ , p = u θ \mathbf{p}=u\theta p = u θ ) with equalizers and coequalizers, rigid, and let A D A {}_A\mathcal{D}_A A D A be rigid and θ A = id \theta_A=\operatorname{id} θ A = id . Then θ \theta θ is a twist of D A loc \mathcal{D}^{\operatorname{loc}}_A D A loc (Theorem 3.15, [21 ] ). The object D ( A ) = A ∗ D(A)=A^{*} D ( A ) = A ∗ with the right action of [26 ] belongs to D A l o c D^{\mathrm{loc}}_{A} D A loc (Prop. 3.18, which states that this holds iff θ A 2 = id \theta_{A}^{2} = \operatorname{id} θ A 2 = id ). D A l o c D^{\mathrm{loc}}_{A} D A loc is ribbon with twist θ \theta θ iff D ( A ) ∈ ( D A l o c ) ′ D(A) \in(D^{\mathrm{loc}}_{A})' D ( A ) ∈ ( D A loc ) ′ (Theorem 3.21).
Next lemma was taken from [14 ] (see also [26 ] ).
Lemma 2.9. Exactness and rigidity: if D \mathcal{D} D is an FTC, A D A _{A}\mathcal{D}_{A} A D A is rigid iff A A A is exact.
In [26 ] one finds
Proposition 2.10. (i) Local modules over exact algebras: if D \mathcal{D} D is a braided FTC and A A A is an exact, commutative, haploid algebra in D \mathcal{D} D , then D A \mathcal{D}_{A} D A is an FTC and D A l o c \mathcal{D}^{\mathrm{loc}}_{A} D A loc is a braided FTC. If D \mathcal{D} D is non-degenerate, so is D A l o c \mathcal{D}^{\mathrm{loc}}_{A} D A loc (see also [9 ] ).
(ii) Modularity: if D \mathcal{D} D is an MTC and A A A is a haploid, commutative, exact, symmetric Frobenius algebra in D \mathcal{D} D , then D A l o c \mathcal{D}^{\mathrm{loc}}_{A} D A loc is an MTC (a non-degenerate ribbon FTC), with the twist inherited from D \mathcal{D} D .
(iii) Symmetry criterion: Let D \mathcal{D} D be braided with a twist θ \theta θ . A commutative Frobenius algebra A A A in D \mathcal{D} D is symmetric iff θ A = id A \theta_{A} = \operatorname{id}_{A} θ A = id A (see also [15 ] ).
The next lemma describes centralizers [25 ] .
Lemma 2.11. For a braided FTC D \mathcal{D} D : D \mathcal{D} D is non-degenerate iff it is factorizable iff D ′ ≃ Vect \mathcal{D}' \simeq\operatorname{Vect} D ′ ≃ Vect (Theorem 1.1 ). For every tensor full subcategory S ⊂ D \mathcal{S} \subset\mathcal{D} S ⊂ D one has FPdim ( S ) FPdim ( S ′ ) = FPdim ( D ) FPdim ( S ∩ D ′ ) \operatorname{FPdim}(\mathcal{S})\operatorname{FPdim}(\mathcal{S}') = \operatorname{FPdim}(\mathcal{D})\operatorname{FPdim}(\mathcal{S} \cap\mathcal{D}') FPdim ( S ) FPdim ( S ′ ) = FPdim ( D ) FPdim ( S ∩ D ′ ) and S ′ ′ = S ∨ D ′ \mathcal{S}'' = \mathcal{S} \vee\mathcal{D}' S ′′ = S ∨ D ′ ; in particular, if D \mathcal{D} D is non-degenerate, FPdim ( S ) FPdim ( S ′ ) = FPdim ( D ) \operatorname{FPdim}(\mathcal{S})\operatorname{FPdim}(\mathcal{S}') = \operatorname{FPdim}(\mathcal{D}) FPdim ( S ) FPdim ( S ′ ) = FPdim ( D ) and S ′ ′ = S \mathcal{S}'' = \mathcal{S} S ′′ = S (Theorem 4.9). Here S ′ = Z ( 2 ) ( S ⊂ D ) \mathcal{S}' = \mathcal{Z}_{(2)}(\mathcal{S} \subset\mathcal{D}) S ′ = Z ( 2 ) ( S ⊂ D ) .
The simplicity implies exactness [10 ] (see also [26 ] ) due to
Lemma 2.12. An algebra in a finite tensor category which is simple (has no two-sided ideals other than 0 and itself), or a direct product of simple algebras, is exact.
In the fusion case [9 ] one has
Lemma 2.13. If A A A is a simple commutative algebra in a braided fusion category D \mathcal{D} D , then D A \mathcal{D}_{A} D A is a fusion category and D A l o c \mathcal{D}^{\mathrm{loc}}_{A} D A loc is a braided fusion category.
For pointed braided categories one obtains [20 ] (see also [13 ] )
Lemma 2.14. Braided pointed fusion categories with group of invertibles H H H are classified, up to braided monoidal equivalence, by the quadratic form q ( h ) = c h , h q(h) = c_{h,h} q ( h ) = c h , h on H H H . The form q ≡ 1 q \equiv1 q ≡ 1 corresponds to Vect H \operatorname{Vect}_{H} Vect H with trivial associator and trivial braiding; for H = Z / 2 H = \mathbb{Z}/2 H = Z /2 , q ( h ) = ± 1 q(h) = \pm1 q ( h ) = ± 1 corresponds to trivial associator, and q = − 1 q = -1 q = − 1 is sVect \operatorname{sVect} sVect .
Witt equivalence [26 ] ([11 ] for the semisimple case) is given by
Lemma 2.15. Non-degenerate braided FTCs C , D \mathcal{C}, \mathcal{D} C , D are Witt equivalent if C ⊠ Z ( A ) ≃ D ⊠ Z ( B ) \mathcal{C} \boxtimes\mathcal{Z}(A) \simeq\mathcal{D} \boxtimes\mathcal{Z}(B) C ⊠ Z ( A ) ≃ D ⊠ Z ( B ) for FTCs A , B A, B A , B . The set W n s W_{\mathrm{ns}} W ns of Witt classes is an abelian group under ⊠ \boxtimes ⊠ , written additively, with [ C ] + [ D ] = [ C ⊠ D ] [\mathcal{C}] + [\mathcal{D}] = [\mathcal{C} \boxtimes\mathcal{D}] [ C ] + [ D ] = [ C ⊠ D ] , zero [ Vect ] [\operatorname{Vect}] [ Vect ] and inverse − [ C ] = [ C r e v ] -[\mathcal{C}] = [\mathcal{C}^{\mathrm{rev}}] − [ C ] = [ C rev ] (since C ⊠ C r e v ≃ Z ( C ) \mathcal{C} \boxtimes\mathcal{C}^{\mathrm{rev}} \simeq\mathcal{Z}(\mathcal{C}) C ⊠ C rev ≃ Z ( C ) for non-degenerate C \mathcal{C} C , [25 ] ). If C \mathcal{C} C is non-degenerate and A ∈ C A \in\mathcal{C} A ∈ C is a haploid, commutative, exact algebra, then C A l o c \mathcal{C}^{\mathrm{loc}}_{A} C A loc is non-degenerate and [ C A l o c ] = [ C ] [\mathcal{C}^{\mathrm{loc}}_{A}] = [\mathcal{C}] [ C A loc ] = [ C ] , [26 ] .
Lemma 2.16. Let N \mathcal{N} N be a non-degenerate braided FTC and S ⊂ N \mathcal{S} \subset\mathcal{N} S ⊂ N a tensor full subcategory. Then:
(i) FPdim S ⋅ FPdim S ′ = FPdim N \operatorname{FPdim}\mathcal{S} \cdot\operatorname{FPdim}\mathcal{S}' = \operatorname{FPdim}\mathcal{N} FPdim S ⋅ FPdim S ′ = FPdim N ;
(ii) S ′ ′ = S \mathcal{S}'' = \mathcal{S} S ′′ = S ;
(iii) the Müger center of S ′ \mathcal{S}' S ′ equals the Müger center S ∩ S ′ \mathcal{S} \cap\mathcal{S}' S ∩ S ′ of S \mathcal{S} S . In particular, if S = ⟨ 1 , f ~ ⟩ ≃ sVect \mathcal{S} = \langle1,\widetilde{f}\rangle\simeq\operatorname{sVect} S = ⟨ 1 , f ⟩ ≃ sVect then S \mathcal{S} S is symmetric, S ⊂ S ′ \mathcal{S} \subset\mathcal{S}' S ⊂ S ′ , and the Müger center of S ′ \mathcal{S}' S ′ is S \mathcal{S} S .
Proof. (i), (ii) are Lemma 2.11 with N ′ = Vect \mathcal{N}' = \operatorname{Vect} N ′ = Vect . (iii) The Müger center of a tensor full subcategory E \mathcal{E} E is E ∩ E ′ \mathcal{E} \cap\mathcal{E}' E ∩ E ′ . For E = S ′ \mathcal{E} = \mathcal{S}' E = S ′ this is S ′ ∩ S ′ ′ = S ′ ∩ S \mathcal{S}' \cap\mathcal{S}'' = \mathcal{S}' \cap\mathcal{S} S ′ ∩ S ′′ = S ′ ∩ S by (ii). If S \mathcal{S} S is symmetric then S ⊂ S ′ \mathcal{S} \subset\mathcal{S}' S ⊂ S ′ , so S ′ ∩ S = S \mathcal{S}' \cap\mathcal{S} = \mathcal{S} S ′ ∩ S = S .
Rigidity, pivotality and ribbon structure
Proof of Theorem 1.1. B \mathcal{B} B is an additive rigid braided category (§ 2.3 and Lemma 2.2 ), hence closed monoidal, and it has equalizers and coequalizers (they are computed componentwise, § 2.3 ). A ‾ \overline{\mathcal{A}} A is commutative algebra in B \mathcal{B} B according to Proposition 2.6 (a). Next, C A ‾ s -loc = B A ‾ l o c \mathcal{C}^{s\text{-loc}}_{\overline{A}}=\mathcal{B}^{\mathrm{loc}}_{\overline{A}} C A s -loc = B A loc due to Proposition 2.6 (b). We apply the results of [26 ] mentioned in § 2.6 .
Proposition 2.8 (i) applied to ( B , A ‾ ) (\mathcal{B},\overline{A}) ( B , A ) : if A ‾ B A ‾ {}_{\overline{A}}\mathcal{B}_{\overline{A}} A B A is rigid, then B A ‾ l o c \mathcal{B}^{\mathrm{loc}}_{\overline{A}} B A loc is rigid.
The pivotal structure on B \mathcal{B} B is p ε p^{\varepsilon} p ε (Lemma 2.4 (ii)), the Drinfeld isomorphism of B \mathcal{B} B is u s u^{s} u s (Lemma 2.4 (i)), and the condition u A ‾ s = p A ‾ ε u^{s}_{\overline{A}}=p^{\varepsilon}_{\overline{A}} u A s = p A ε of Proposition 2.8 (ii) is equivalent to u A 0 = p A 0 u_{A_0}=p_{A_0} u A 0 = p A 0 , u A 1 = ε p A 1 u_{A_1}=\varepsilon p_{A_1} u A 1 = ε p A 1 by Proposition 2.6 (c).
B \mathcal{B} B is ribbon with twist θ s \theta^{s} θ s (§ 2.3 ), and p ε = u s θ s p^{\varepsilon}=u^{s}\theta^{s} p ε = u s θ s (Lemma 2.4 (iii)). The condition θ A ‾ s = id \theta^{s}_{\overline{A}}=\operatorname{id} θ A s = id of Proposition 2.8 (iii) is equivalent to θ A 0 = id \theta_{A_0}=\operatorname{id} θ A 0 = id , θ A 1 = ε id \theta_{A_1}=\varepsilon\operatorname{id} θ A 1 = ε id (Proposition 2.6 (c)). By Proposition 2.8 (iii), θ s \theta^{s} θ s is then a twist of B A ‾ l o c \mathcal{B}^{\mathrm{loc}}_{\overline{A}} B A loc , and B A ‾ l o c \mathcal{B}^{\mathrm{loc}}_{\overline{A}} B A loc is ribbon iff A ‾ ∗ ∈ ( B A ‾ l o c ) ′ \overline{A}^{*}\in(\mathcal{B}^{\mathrm{loc}}_{\overline{A}})^{\prime} A ∗ ∈ ( B A loc ) ′ , which is the remaining assumption. □
Note that we use only that A ‾ \overline{A} A is an ordinary commutative algebra of B \mathcal{B} B . The condition θ 2 = id \theta^{2}=\operatorname{id} θ 2 = id of [26 ] , resp. [9 ] , determines only whether A ‾ ∗ \overline{A}^{*} A ∗ (resp. the dualizing object D A ‾ D\overline{A} D A ) is local. The transparency of A ‾ ∗ \overline{A}^{*} A ∗ required in (c) is a stronger condition, equivalent to the compatibility of θ s \theta^{s} θ s with duality (Remark 1.2). In our setting ( θ A ‾ s ) 2 = id (\theta^{s}_{\overline{A}})^{2}=\operatorname{id} ( θ A s ) 2 = id means θ A 0 2 = id \theta^{2}_{A_0}=\operatorname{id} θ A 0 2 = id and θ A 1 2 = id \theta^{2}_{A_1}=\operatorname{id} θ A 1 2 = id .
Finite tensor categories and super-modularity
The toric code
Let T = Z ( Vect Z / 2 ) \mathcal{T}=\mathcal{Z}(\operatorname{Vect}_{\mathbb{Z}/2}) T = Z ( Vect Z /2 ) be the toric code, regarded as a braided fusion category: Inv ( T ) = { 1 , e , m , ψ } \operatorname{Inv}(\mathcal{T})=\{1,e,m,\psi\} Inv ( T ) = { 1 , e , m , ψ } with ψ = e ⊗ m \psi=e\otimes m ψ = e ⊗ m and group Z / 2 × Z / 2 \mathbb{Z}/2\times\mathbb{Z}/2 Z /2 × Z /2 , trivial associator (its underlying fusion category is Vect Z / 2 × Z / 2 \operatorname{Vect}_{\mathbb{Z}/2\times\mathbb{Z}/2} Vect Z /2 × Z /2 with trivial 3 3 3 -cocycle), c e , e = c m , m = id c_{e,e}=c_{m,m}=\operatorname{id} c e , e = c m , m = id , mon e , m = mon ψ , e = mon ψ , m = − 1 \operatorname{mon}_{e,m}=\operatorname{mon}_{\psi,e}=\operatorname{mon}_{\psi,m}=-1 mon e , m = mon ψ , e = mon ψ , m = − 1 , mon e , e = mon m , m = 1 \operatorname{mon}_{e,e}=\operatorname{mon}_{m,m}=1 mon e , e = mon m , m = 1 ; hence c ψ , ψ = − id c_{\psi,\psi}=-\operatorname{id} c ψ , ψ = − id . It is semisimple and non-degenerate (T ′ = ⟨ 1 ⟩ \mathcal{T}^{\prime}=\langle1\rangle T ′ = ⟨ 1 ⟩ , since each of e , m , ψ e,m,\psi e , m , ψ has monodromy − 1 -1 − 1 with some simple object). Its standard ribbon structure is θ e = θ m = 1 \theta_e=\theta_m=1 θ e = θ m = 1 , θ ψ = − 1 \theta_\psi=-1 θ ψ = − 1 .
For ε ∈ { ± 1 } \varepsilon\in\{\pm1\} ε ∈ { ± 1 } define the ribbon category T ε \mathcal{T}^{\varepsilon} T ε as follows. For ε = − 1 \varepsilon=-1 ε = − 1 it is T \mathcal{T} T with the standard twist. For ε = + 1 \varepsilon=+1 ε = + 1 set θ X ′ : = θ X mon e , X \theta'_X:=\theta_X\operatorname{mon}_{e,X} θ X ′ := θ X mon e , X . Since monodromy is bimultiplicative on invertible objects, X ↦ mon e , X X\mapsto\operatorname{mon}_{e,X} X ↦ mon e , X is a ± 1 \pm1 ± 1 -valued character. Thus, θ X ⊗ Y ′ = θ X θ Y mon X , Y mon e , X mon e , Y = θ X ′ θ Y ′ mon X , Y \theta'_{X\otimes Y}=\theta_X\theta_Y\operatorname{mon}_{X,Y}\operatorname{mon}_{e,X}\operatorname{mon}_{e,Y}=\theta'_X\theta'_Y\operatorname{mon}_{X,Y} θ X ⊗ Y ′ = θ X θ Y mon X , Y mon e , X mon e , Y = θ X ′ θ Y ′ mon X , Y and θ X ∗ ′ = θ X ′ \theta'_{X^{*}}=\theta'_X θ X ∗ ′ = θ X ′ . Therefore, θ ′ \theta' θ ′ is a ribbon structure, with θ e ′ = 1 \theta'_e=1 θ e ′ = 1 , θ m ′ = − 1 \theta'_m=-1 θ m ′ = − 1 , θ ψ ′ = + 1 \theta'_\psi=+1 θ ψ ′ = + 1 and dim ψ = dim m = − 1 \dim\psi=\dim m=-1 dim ψ = dim m = − 1 (by Remark 2.1, since q ( m ) = 1 q(m)=1 q ( m ) = 1 and q ( ψ ) = − 1 q(\psi)=-1 q ( ψ ) = − 1 ). So T + \mathcal{T}^{+} T + is not pseudo-unitary (some dimensions are negative); it is nevertheless spherical, as all simple objects are self-dual (see Remark 1.4). The underlying braided category of T + \mathcal{T}^{+} T + and T − \mathcal{T}^{-} T − is the same, but the ribbon structure (hence pivotal structure and dimensions) differs (both are non-degenerate). In both, ⟨ 1 , ψ ⟩ ≃ sVect ε \langle1,\psi\rangle\simeq\operatorname{sVect}^{\varepsilon} ⟨ 1 , ψ ⟩ ≃ sVect ε is a full ribbon subcategory equal to its centralizer.
Lemma 4.1. Let C \mathcal{C} C be a braided FTC and B ~ : = C ⊠ T ε \widetilde{\mathcal{B}}:=\mathcal{C}\boxtimes\mathcal{T}^{\varepsilon} B := C ⊠ T ε . Choose an isomorphism m ψ : ψ ⊗ ψ → 1 m_{\psi}:\psi\otimes\psi\to1 m ψ : ψ ⊗ ψ → 1 with m ψ ⊗ id ψ = id ψ ⊗ m ψ m_{\psi}\otimes\operatorname{id}_{\psi}=\operatorname{id}_{\psi}\otimes m_{\psi} m ψ ⊗ id ψ = id ψ ⊗ m ψ (possible since the associator of ⟨ 1 , ψ ⟩ \langle1,\psi\rangle ⟨ 1 , ψ ⟩ is trivial; see below). Then ι 0 : sVect → T ε \iota_0:\operatorname{sVect}\to\mathcal{T}^{\varepsilon} ι 0 : sVect → T ε , 1 ↦ 1 1\mapsto1 1 ↦ 1 , f ↦ ψ f\mapsto\psi f ↦ ψ with tensor structure given by m ψ m_{\psi} m ψ on f ⊗ f f\otimes f f ⊗ f and by identities otherwise, is a braided tensor functor, ribbon for sVect ε \operatorname{sVect}^{\varepsilon} sVect ε , and
ι : = id C ⊠ ι 0 : B = C ⊠ sVect → B ~ , ( X 0 , X 1 ) ↦ X 0 ⊠ 1 ⊕ X 1 ⊠ ψ , \iota:=\operatorname{id}_{\mathcal{C}}\boxtimes\iota_0:\mathcal{B}=\mathcal{C}\boxtimes\operatorname{sVect}\to\widetilde{\mathcal{B}},\qquad(X_0,X_1)\mapsto X_0\boxtimes1\oplus X_1\boxtimes\psi, ι := id C ⊠ ι 0 : B = C ⊠ sVect → B , ( X 0 , X 1 ) ↦ X 0 ⊠ 1 ⊕ X 1 ⊠ ψ , is a fully faithful braided tensor functor, ribbon if C \mathcal{C} C is ribbon, and ι ( B ) = Z ( 2 ) ( 1 ⊠ ψ ⊂ B ~ ) \iota(\mathcal{B})=\mathcal{Z}_{(2)}(1\boxtimes\psi\subset\widetilde{\mathcal{B}}) ι ( B ) = Z ( 2 ) ( 1 ⊠ ψ ⊂ B ) . If C \mathcal{C} C is ribbon, ι \iota ι preserves the pivotal structures p ε = u s θ s p^{\varepsilon}=u^{s}\theta^{s} p ε = u s θ s and p ~ = u ~ θ ~ \widetilde{p}=\widetilde{u}\widetilde{\theta} p = u θ (with θ ~ ψ = − ε \widetilde{\theta}_{\psi}=-\varepsilon θ ψ = − ε ).
Proof. Tensor structure. For ι 0 \iota_0 ι 0 to be a tensor functor the only axiom with non-trivial content is the hexagon on f ⊗ f ⊗ f f\otimes f\otimes f f ⊗ f ⊗ f , which reads m ψ ⊗ id ψ = id ψ ⊗ m ψ m_{\psi}\otimes\operatorname{id}_{\psi}=\operatorname{id}_{\psi}\otimes m_{\psi} m ψ ⊗ id ψ = id ψ ⊗ m ψ (the associators of sVect \operatorname{sVect} sVect and of ⟨ 1 , ψ ⟩ \langle1,\psi\rangle ⟨ 1 , ψ ⟩ being trivial). Both sides belong to the one-dimensional space Hom ( ψ ⊗ 3 , ψ ) \operatorname{Hom}(\psi^{\otimes3},\psi) Hom ( ψ ⊗ 3 , ψ ) , so they differ by a scalar λ \lambda λ ; λ = ± 1 \lambda=\pm1 λ = ± 1 is the value of the associator on ( ψ , ψ , ψ ) (\psi,\psi,\psi) ( ψ , ψ , ψ ) , which does not change under rescaling m ψ m_{\psi} m ψ and is 1 1 1 for the trivial associator of T \mathcal{T} T . Hence, ι 0 \iota_0 ι 0 is, as required, an isomorphism, and ι 0 \iota_0 ι 0 is a tensor functor. It is braided: the non-trivial condition is ι 0 ( c f , f ) m ψ = m ψ c ψ , ψ \iota_0(c_{f,f})m_{\psi}=m_{\psi}c_{\psi,\psi} ι 0 ( c f , f ) m ψ = m ψ c ψ , ψ , i.e., ( − 1 ) m ψ = m ψ ( − 1 ) (-1)m_{\psi}=m_{\psi}(-1) ( − 1 ) m ψ = m ψ ( − 1 ) . It is ribbon for sVect ε \operatorname{sVect}^{\varepsilon} sVect ε since θ ψ = ε \theta_{\psi}=\varepsilon θ ψ = ε . Hence, ι = id C ⊠ ι 0 \iota=\operatorname{id}_{\mathcal{C}}\boxtimes\iota_0 ι = id C ⊠ ι 0 is a braided (ribbon) tensor functor, and its coherence constraints are those of ι 0 \iota_0 ι 0 with id C \operatorname{id}_{\mathcal{C}} id C : for X , Y ∈ B X,Y\in\mathcal{B} X , Y ∈ B the tensor ι ( X ) ⊗ ι ( Y ) → ι ( X ⊗ Y ) \iota(X)\otimes\iota(Y)\to\iota(X\otimes Y) ι ( X ) ⊗ ι ( Y ) → ι ( X ⊗ Y ) is id X i ⊗ Y j ⊠ ι i , j \operatorname{id}_{X_i\otimes Y_j}\boxtimes\iota_{i,j} id X i ⊗ Y j ⊠ ι i , j on the summand X i ⊠ ψ i ⊗ Y j ⊠ ψ j X_i\boxtimes\psi^i\otimes Y_j\boxtimes\psi^j X i ⊠ ψ i ⊗ Y j ⊠ ψ j , where ι 1 , 1 = m ψ \iota_{1,1}=m_{\psi} ι 1 , 1 = m ψ and ι i , j = id \iota_{i,j}=\operatorname{id} ι i , j = id otherwise. Different choices of m ψ m_{\psi} m ψ differ by a non-zero scalar and give isomorphic functors.
Fully faithful. Hom C ⊠ T ( X 0 ⊠ 1 ⊕ X 1 ⊠ ψ , Y 0 ⊠ 1 ⊕ Y 1 ⊠ ψ ) = Hom ( X 0 , Y 0 ) ⊕ Hom ( X 1 , Y 1 ) \operatorname{Hom}_{\mathcal{C}\boxtimes\mathcal{T}}(X_0\boxtimes1\oplus X_1\boxtimes\psi,Y_0\boxtimes1\oplus Y_1\boxtimes\psi)=\operatorname{Hom}(X_0,Y_0)\oplus\operatorname{Hom}(X_1,Y_1) Hom C ⊠ T ( X 0 ⊠ 1 ⊕ X 1 ⊠ ψ , Y 0 ⊠ 1 ⊕ Y 1 ⊠ ψ ) = Hom ( X 0 , Y 0 ) ⊕ Hom ( X 1 , Y 1 ) since Hom T ( 1 , ψ ) = 0 \operatorname{Hom}_{\mathcal{T}}(1,\psi)=0 Hom T ( 1 , ψ ) = 0 .
Braided and ribbon. Explicitly, c X ⊠ ψ , Y ⊠ ψ = c X , Y ⊠ c ψ , ψ = − c X , Y = ε s c_{X\boxtimes\psi,Y\boxtimes\psi}=c_{X,Y}\boxtimes c_{\psi,\psi}=-c_{X,Y}=\varepsilon^{s} c X ⊠ ψ , Y ⊠ ψ = c X , Y ⊠ c ψ , ψ = − c X , Y = ε s on odd-odd, and c X ⊠ ψ , Y ⊠ 1 = c X , Y ⊠ c ψ , 1 = c X , Y c_{X\boxtimes\psi,Y\boxtimes1}=c_{X,Y}\boxtimes c_{\psi,1}=c_{X,Y} c X ⊠ ψ , Y ⊠ 1 = c X , Y ⊠ c ψ , 1 = c X , Y . It is ribbon since θ X ⊠ ψ = θ X ⊠ ε = θ s \theta_{X\boxtimes\psi}=\theta_X\boxtimes\varepsilon=\theta^{s} θ X ⊠ ψ = θ X ⊠ ε = θ s . As ι \iota ι is braided and compatible with duals, it preserves the Drinfeld isomorphisms, and together with the twists it preserves p = u θ p=u\theta p = u θ . On 1 ⊠ ψ 1\boxtimes\psi 1 ⊠ ψ this gives p ~ ψ = u ~ ψ θ ψ = ( − 1 ) ε \widetilde{p}_{\psi}=\widetilde{u}_{\psi}\theta_{\psi}=(-1)^{\varepsilon} p ψ = u ψ θ ψ = ( − 1 ) ε .
Centralizer. T \mathcal{T} T is pointed, thus every object of B ~ \widetilde{\mathcal{B}} B is canonically ⨁ t ∈ Inv ( T ) X t ⊠ t \bigoplus_{t\in\operatorname{Inv}(\mathcal{T})}X_t\boxtimes t ⨁ t ∈ Inv ( T ) X t ⊠ t . The monodromy of X t ⊠ t X_t\boxtimes t X t ⊠ t with 1 ⊠ ψ 1\boxtimes\psi 1 ⊠ ψ is mon t , ψ ⋅ id \operatorname{mon}_{t,\psi}\cdot\operatorname{id} mon t , ψ ⋅ id , with mon t , ψ = − 1 \operatorname{mon}_{t,\psi}=-1 mon t , ψ = − 1 for t ∈ { e , m } t\in\{e,m\} t ∈ { e , m } . It is the identity iff X e = X m = 0 X_e=X_m=0 X e = X m = 0 . □
Exactness of the image of A ˉ \bar{A} A ˉ
The following lemma supplies the step on which Theorem 1.5 (1), (2) depend. Exactness of an algebra is not preserved by arbitrary tensor functors, and ι \iota ι is far from surjective, thus the statement is proven using the structure of T \mathcal{T} T .
Lemma 4.2. Let C \mathcal{C} C be a braided FTC and let A ˉ \bar{A} A ˉ be a haploid commutative exact algebra in B = C g r \mathcal{B}=\mathcal{C}^{\mathrm{gr}} B = C gr . Then A ′ : = ι ( A ˉ ) A':=\iota(\bar{A}) A ′ := ι ( A ˉ ) is a haploid commutative exact algebra in B ~ = C ⊠ T e \widetilde{\mathcal{B}}=\mathcal{C}\boxtimes\mathcal{T}^{e} B = C ⊠ T e .
Proof. Commutative: ι \iota ι is a braided tensor functor, thus μ ′ ∘ c A ′ , A ′ = ι ( μ ∘ c A ˉ , A ˉ ) = μ ′ \mu'\circ c_{A',A'}=\iota(\mu\circ c_{\bar{A},\bar{A}})=\mu' μ ′ ∘ c A ′ , A ′ = ι ( μ ∘ c A ˉ , A ˉ ) = μ ′ .
Haploid: Hom B ~ ( 1 , A ′ ) = Hom C ( 1 , A 0 ) ⊗ Hom T ( 1 , 1 ) ⊕ Hom C ( 1 , A 1 ) ⊗ Hom T ( 1 , ψ ) = Hom C ( 1 , A 0 ) ≅ k \operatorname{Hom}_{\widetilde{\mathcal{B}}}(1,A')=\operatorname{Hom}_{\mathcal{C}}(1,A_0)\otimes\operatorname{Hom}_{\mathcal{T}}(1,1)\oplus\operatorname{Hom}_{\mathcal{C}}(1,A_1)\otimes\operatorname{Hom}_{\mathcal{T}}(1,\psi)=\operatorname{Hom}_{\mathcal{C}}(1,A_0)\cong k Hom B ( 1 , A ′ ) = Hom C ( 1 , A 0 ) ⊗ Hom T ( 1 , 1 ) ⊕ Hom C ( 1 , A 1 ) ⊗ Hom T ( 1 , ψ ) = Hom C ( 1 , A 0 ) ≅ k .
Exact: By [14 ] , A ′ A' A ′ is exact iff the category B ~ A ′ \widetilde{\mathcal{B}}_{A'} B A ′ of right A ′ A' A ′ -modules is an exact left B ~ \widetilde{\mathcal{B}} B -module category, i.e., iff P ~ ⊗ M \widetilde{P}\otimes M P ⊗ M is projective in B ~ A ′ \widetilde{\mathcal{B}}_{A'} B A ′ for every projective P ~ ∈ B ~ \widetilde{P}\in\widetilde{\mathcal{B}} P ∈ B and every M ∈ B ~ A ′ M\in\widetilde{\mathcal{B}}_{A'} M ∈ B A ′ [13 ] . Likewise for A ˉ \bar{A} A ˉ and B A ˉ \mathcal{B}_{\bar{A}} B A ˉ . As T \mathcal{T} T is semisimple, B ~ = ⨁ t ∈ Inv ( T ) C ⊠ t \widetilde{\mathcal{B}}=\bigoplus_{t\in\operatorname{Inv}(\mathcal{T})}\mathcal{C}\boxtimes t B = ⨁ t ∈ Inv ( T ) C ⊠ t as an abelian category, and the projective objects are the sums of objects P ⊠ t P\boxtimes t P ⊠ t with P P P projective in C \mathcal{C} C .
Set T 1 = { 1 , ψ } \mathcal{T}_1=\{1,\psi\} T 1 = { 1 , ψ } and T e = { e , m } \mathcal{T}_e=\{e,m\} T e = { e , m } . The algebra A ′ A' A ′ belongs to C ⊠ T 1 \mathcal{C}\boxtimes\mathcal{T}_1 C ⊠ T 1 , and tensoring on the right with 1 1 1 or ψ \psi ψ preserves T 1 \mathcal{T}_1 T 1 and T e \mathcal{T}_e T e (since e ⊗ ψ = m e\otimes\psi=m e ⊗ ψ = m , m ⊗ ψ = e m\otimes\psi=e m ⊗ ψ = e ). Hence, every right A ′ A' A ′ -module decomposes canonically as M = M 1 ⊕ M e M=M_1\oplus M_e M = M 1 ⊕ M e (M 1 M_1 M 1 supported on C ⊠ T 1 \mathcal{C}\boxtimes\mathcal{T}_1 C ⊠ T 1 , M e M_e M e on C ⊠ T e \mathcal{C}\boxtimes\mathcal{T}_e C ⊠ T e ), and B ~ A ′ = ( B ~ A ′ ) 1 ⊕ ( B ~ A ′ ) e \widetilde{\mathcal{B}}_{A'}=(\widetilde{\mathcal{B}}_{A'})_1\oplus(\widetilde{\mathcal{B}}_{A'})_e B A ′ = ( B A ′ ) 1 ⊕ ( B A ′ ) e as abelian categories. The functor ι \iota ι identifies B \mathcal{B} B with the full subcategory C ⊠ T 1 \mathcal{C}\boxtimes\mathcal{T}_1 C ⊠ T 1 and, being fully faithful and monoidal, identifies ( B ~ A ′ ) 1 (\widetilde{\mathcal{B}}_{A'})_1 ( B A ′ ) 1 with B A ˉ \mathcal{B}_{\bar{A}} B A ˉ . For the second summand consider E : B → C ⊠ T e E:\mathcal{B}\to\mathcal{C}\boxtimes\mathcal{T}_e E : B → C ⊠ T e , E ( Y 0 , Y 1 ) = Y 0 ⊠ e ⊕ Y 1 ⊠ m E(Y_0,Y_1)=Y_0\boxtimes e\oplus Y_1\boxtimes m E ( Y 0 , Y 1 ) = Y 0 ⊠ e ⊕ Y 1 ⊠ m . Using e ⊗ 1 = e e\otimes1=e e ⊗ 1 = e , e ⊗ ψ = m e\otimes\psi=m e ⊗ ψ = m , m ⊗ ψ = e m\otimes\psi=e m ⊗ ψ = e one has E ( Y ) ⊗ ι ( Z ) ≅ E ( Y ⊗ Z ) E(Y)\otimes\iota(Z)\cong E(Y\otimes Z) E ( Y ) ⊗ ι ( Z ) ≅ E ( Y ⊗ Z ) , and since the associator of T \mathcal{T} T is trivial these isomorphisms can be chosen coherently. Thus, E E E is an equivalence of right B \mathcal{B} B -module categories. T e \mathcal{T}_e T e is a free orbit of ⟨ 1 , ψ ⟩ ≃ Vect Z / 2 \langle1,\psi\rangle\simeq\operatorname{Vect}_{\mathbb{Z}/2} ⟨ 1 , ψ ⟩ ≃ Vect Z /2 , so C ⊠ T e \mathcal{C}\boxtimes\mathcal{T}_e C ⊠ T e is the regular module category B \mathcal{B} B . Consequently, ( B ~ A ′ ) e ≃ B A ˉ (\widetilde{\mathcal{B}}_{A'})_e\simeq\mathcal{B}_{\bar{A}} ( B A ′ ) e ≃ B A ˉ as well, and B ~ A ′ ≃ B A ˉ ⊕ B A ˉ \widetilde{\mathcal{B}}_{A'}\simeq\mathcal{B}_{\bar{A}}\oplus\mathcal{B}_{\bar{A}} B A ′ ≃ B A ˉ ⊕ B A ˉ , an object being projective iff both components are.
Let P ~ = P ⊠ t \widetilde{P}=P\boxtimes t P = P ⊠ t with P P P projective in C \mathcal{C} C and t ∈ { 1 , e , m , ψ } t\in\{1,e,m,\psi\} t ∈ { 1 , e , m , ψ } . Then ( P , 0 ) ∈ B (P,0)\in\mathcal{B} ( P , 0 ) ∈ B is projective. A computation with the fusion rules of T \mathcal{T} T shows that under the identifications above, left multiplication by P ~ \widetilde{P} P is, on each of the two copies of B A ˉ \mathcal{B}_{\bar{A}} B A ˉ , isomorphic to N ↦ Π a ( ( P , 0 ) ⊗ N ) N\mapsto\Pi^{a}((P,0)\otimes N) N ↦ Π a (( P , 0 ) ⊗ N ) followed by a possible interchange of the two copies. Explicitly, for Y = ( Y 0 , Y 1 ) Y=(Y_0,Y_1) Y = ( Y 0 , Y 1 ) :
( P ⊠ 1 ) ⊗ ι ( Y ) = ι ( ( P , 0 ) ⊗ Y ) , ( P ⊠ ψ ) ⊗ ι ( Y ) = ι ( ( 0 , P ) ⊗ Y ) = ι ( Π ( ( P , 0 ) ⊗ Y ) ) , (P\boxtimes1)\otimes\iota(Y)=\iota((P,0)\otimes Y),\qquad
(P\boxtimes\psi)\otimes\iota(Y)=\iota((0,P)\otimes Y)=\iota(\Pi((P,0)\otimes Y)), ( P ⊠ 1 ) ⊗ ι ( Y ) = ι (( P , 0 ) ⊗ Y ) , ( P ⊠ ψ ) ⊗ ι ( Y ) = ι (( 0 , P ) ⊗ Y ) = ι ( Π (( P , 0 ) ⊗ Y )) , ( P ⊠ e ) ⊗ ι ( Y ) = E ( ( P , 0 ) ⊗ Y ) , ( P ⊠ m ) ⊗ ι ( Y ) = E ( Π ( ( P , 0 ) ⊗ Y ) ) , (P\boxtimes e)\otimes\iota(Y)=E((P,0)\otimes Y),\qquad
(P\boxtimes m)\otimes\iota(Y)=E(\Pi((P,0)\otimes Y)), ( P ⊠ e ) ⊗ ι ( Y ) = E (( P , 0 ) ⊗ Y ) , ( P ⊠ m ) ⊗ ι ( Y ) = E ( Π (( P , 0 ) ⊗ Y )) , ( P ⊠ e ) ⊗ E ( Y ) = ι ( ( P , 0 ) ⊗ Y ) , ( P ⊠ m ) ⊗ E ( Y ) = ι ( Π ( ( P , 0 ) ⊗ Y ) ) , (P\boxtimes e)\otimes E(Y)=\iota((P,0)\otimes Y),\qquad
(P\boxtimes m)\otimes E(Y)=\iota(\Pi((P,0)\otimes Y)), ( P ⊠ e ) ⊗ E ( Y ) = ι (( P , 0 ) ⊗ Y ) , ( P ⊠ m ) ⊗ E ( Y ) = ι ( Π (( P , 0 ) ⊗ Y )) , and ( P ⊠ 1 ) ⊗ E ( Y ) = E ( ( P , 0 ) ⊗ Y ) (P\boxtimes1)\otimes E(Y)=E((P,0)\otimes Y) ( P ⊠ 1 ) ⊗ E ( Y ) = E (( P , 0 ) ⊗ Y ) , ( P ⊠ ψ ) ⊗ E ( Y ) = E ( Π ( ( P , 0 ) ⊗ Y ) ) (P\boxtimes\psi)\otimes E(Y)=E(\Pi((P,0)\otimes Y)) ( P ⊠ ψ ) ⊗ E ( Y ) = E ( Π (( P , 0 ) ⊗ Y )) , where Π = f ⊠ − \Pi=f\boxtimes- Π = f ⊠ − is the parity shift. The identities hold as identities of A ˉ \bar{A} A ˉ -modules, the action being id P ⊗ α \operatorname{id}_P\otimes\alpha id P ⊗ α . Since A ˉ \bar{A} A ˉ is exact in B \mathcal{B} B and ( P , 0 ) (P,0) ( P , 0 ) is projective in B \mathcal{B} B , ( P , 0 ) ⊗ N (P,0)\otimes N ( P , 0 ) ⊗ N is projective in B A ˉ \mathcal{B}_{\bar{A}} B A ˉ for all N ∈ B A ˉ N\in\mathcal{B}_{\bar{A}} N ∈ B A ˉ , and Π \Pi Π is an autoequivalence of B A ˉ \mathcal{B}_{\bar{A}} B A ˉ . Hence P ~ ⊗ M \widetilde{P}\otimes M P ⊗ M is projective in B ~ A ′ \widetilde{\mathcal{B}}_{A'} B A ′ for all M M M .
The induction lemma
Let D \mathcal{D} D be a braided monoidal category with coequalizers preserved by ⊗ \otimes ⊗ in each variable (e.g., a finite tensor category), and A A A a commutative algebra in D \mathcal{D} D . For a right module N N N set λ N : = a N ∘ c A , N : A ⊗ N → N \lambda_N:=a_{N}\circ c_{A,N}:A\otimes N\to N λ N := a N ∘ c A , N : A ⊗ N → N (a left action). For a right module M M M and a right module N N N with λ N \lambda_N λ N , M ⊗ A N M\otimes_A N M ⊗ A N is the coequalizer of a M ⊗ id N a_{M}\otimes\operatorname{id}_N a M ⊗ id N and id M ⊗ λ N \operatorname{id}_M\otimes\lambda_N id M ⊗ λ N on M ⊗ A ⊗ N ⇒ M ⊗ N M\otimes A\otimes N\Rightarrow M\otimes N M ⊗ A ⊗ N ⇒ M ⊗ N . Denote by π M , N \pi_{M,N} π M , N the projection. For local N N N , a N c A , N = a N c N , A − 1 a_Nc_{A,N}=a_Nc^{-1}_{N,A} a N c A , N = a N c N , A − 1 (apply c N , A − 1 c^{-1}_{N,A} c N , A − 1 to a N mon = a N a_N\operatorname{mon}=a_N a N mon = a N ). Thus, the left action does not depend on the choice between c c c and c − 1 c^{-1} c − 1 and D A l o c \mathcal{D}^{\mathrm{loc}}_A D A loc (with ⊗ A \otimes_A ⊗ A ) is the same as in [24 , 21 , 9 ] (where left modules are used; see § 9 for the correspondence). For local M , N M,N M , N the braiding c M , N A : M ⊗ A N → N ⊗ A M c^A_{M,N}:M\otimes_A N\to N\otimes_A M c M , N A : M ⊗ A N → N ⊗ A M is the map induced by c M , N c_{M,N} c M , N , i.e., c M , N A π M , N = π N , M c M , N c^A_{M,N}\pi_{M,N}=\pi_{N,M}c_{M,N} c M , N A π M , N = π N , M c M , N [24 ] .
Let Z A : = { X ∈ D ∣ c A , X ∘ c X , A = id X ⊗ A } Z_A:=\{X\in\mathcal{D}\mid c_{A,X}\circ c_{X,A}=\operatorname{id}_{X\otimes A}\} Z A := { X ∈ D ∣ c A , X ∘ c X , A = id X ⊗ A } (equivalently c X , A c A , X = id A ⊗ X c_{X,A}c_{A,X}=\operatorname{id}_{A\otimes X} c X , A c A , X = id A ⊗ X ), F ( X ) : = X ⊗ A F(X):=X\otimes A F ( X ) := X ⊗ A with right action id X ⊗ μ \operatorname{id}_X\otimes\mu id X ⊗ μ .
Lemma 4.3. Let X , Y ∈ Z A X,Y\in Z_A X , Y ∈ Z A and let M M M be a local right A A A -module. Then:
1 ∈ Z A 1\in Z_A 1 ∈ Z A , X ⊗ Y ∈ Z A X\otimes Y\in Z_A X ⊗ Y ∈ Z A , and F ( X ) F(X) F ( X ) is local. ::
(ii) There are isomorphisms ρ X , M : F ( X ) ⊗ A M → X ⊗ M \rho_{X,M}: F(X) \otimes_A M \to X \otimes M ρ X , M : F ( X ) ⊗ A M → X ⊗ M (valid for every right module M M M ; an isomorphism of right A A A -modules, X ⊗ M X \otimes M X ⊗ M carrying the action id X ⊗ a M \operatorname{id}_X \otimes a_M id X ⊗ a M ) and σ M , X : M ⊗ A F ( X ) → M ⊗ X \sigma_{M,X}: M \otimes_A F(X) \to M \otimes X σ M , X : M ⊗ A F ( X ) → M ⊗ X (an isomorphism in D D D ), given by ρ π = id X ⊗ λ M \rho\pi=\operatorname{id}_X \otimes\lambda_M ρ π = id X ⊗ λ M and σ π = ( a M ⊗ id X ) ( id M ⊗ c X , A ) \sigma\pi=(a_M \otimes\operatorname{id}_X)(\operatorname{id}_M \otimes c_{X,A}) σ π = ( a M ⊗ id X ) ( id M ⊗ c X , A ) . Moreover ρ X , F ( Y ) : F ( X ) ⊗ A F ( Y ) → X ⊗ Y ⊗ A = F ( X ⊗ Y ) \rho_{X,F(Y)}: F(X) \otimes_A F(Y) \to X \otimes Y \otimes A=F(X \otimes Y) ρ X , F ( Y ) : F ( X ) ⊗ A F ( Y ) → X ⊗ Y ⊗ A = F ( X ⊗ Y ) is an isomorphism of right modules.
(iii) σ M , X c F ( X ) , M A ρ X , M − 1 = c X , M \sigma_{M,X}c^A_{F(X),M}\rho^{-1}_{X,M}=c_{X,M} σ M , X c F ( X ) , M A ρ X , M − 1 = c X , M and ρ X , M c M , F ( X ) A σ M , X − 1 = c M , X \rho_{X,M}c^A_{M,F(X)}\sigma^{-1}_{M,X}=c_{M,X} ρ X , M c M , F ( X ) A σ M , X − 1 = c M , X . Consequently, under σ \sigma σ the monodromy c F ( X ) , M A c M , F ( X ) A c^A_{F(X),M}c^A_{M,F(X)} c F ( X ) , M A c M , F ( X ) A of M M M and F ( X ) F(X) F ( X ) in D A l o c D^{{\rm loc}}_A D A loc is the monodromy c X , M c M , X c_{X,M}c_{M,X} c X , M c M , X of M M M and X X X in D D D (as an endomorphism of M ⊗ X M \otimes X M ⊗ X ). In particular, it is the identity iff mon M , X = id M ⊗ X \operatorname{mon}_{M,X}=\operatorname{id}_{M\otimes X} mon M , X = id M ⊗ X .
(iv) The composite
F ( X ⊗ Y ) = X ⊗ Y ⊗ A → ρ X , F ( Y ) − 1 F ( X ) ⊗ A F ( Y ) → c F ( X ) , F ( Y ) A F ( Y ) ⊗ A F ( X ) → ρ Y , F ( X ) Y ⊗ X ⊗ A = F ( Y ⊗ X ) F(X\otimes Y)=X\otimes Y\otimes A \xrightarrow{\rho^{-1}_{X,F(Y)}} F(X)\otimes_A F(Y) \xrightarrow{c^A_{F(X),F(Y)}} F(Y)\otimes_A F(X) \xrightarrow{\rho_{Y,F(X)}} Y\otimes X\otimes A=F(Y\otimes X) F ( X ⊗ Y ) = X ⊗ Y ⊗ A ρ X , F ( Y ) − 1 F ( X ) ⊗ A F ( Y ) c F ( X ) , F ( Y ) A F ( Y ) ⊗ A F ( X ) ρ Y , F ( X ) Y ⊗ X ⊗ A = F ( Y ⊗ X )
equals c X , Y ⊗ id A c_{X,Y}\otimes\operatorname{id}_A c X , Y ⊗ id A .
(v) If D D D is ribbon and θ A = id \theta_A=\operatorname{id} θ A = id , then θ F ( X ) = θ X ⊗ id A \theta_{F(X)}=\theta_X\otimes\operatorname{id}_A θ F ( X ) = θ X ⊗ id A .
The isomorphisms ρ X , F ( Y ) \rho_{X,F(Y)} ρ X , F ( Y ) in (ii) are the components of the well-known monoidal structure of the induction functor F F F [24 , 21 ] . The proof of Theorem 1.5 uses that ρ ψ , F ( ψ ) \rho_{\psi,F(\psi)} ρ ψ , F ( ψ ) is an isomorphism and the identities (iii)-(v), which are proved below from the definitions.
Proof. We suppress associators and unitors.
(i) The hexagon axioms give c X ⊗ A , A = ( c X , A ⊗ id A ) ( id X ⊗ c A , A ) c_{X\otimes A,A}=(c_{X,A}\otimes\operatorname{id}_A)(\operatorname{id}_X\otimes c_{A,A}) c X ⊗ A , A = ( c X , A ⊗ id A ) ( id X ⊗ c A , A ) and c A , X ⊗ A = ( id X ⊗ c A , A ) ( c A , X ⊗ id A ) c_{A,X\otimes A}=(\operatorname{id}_X\otimes c_{A,A})(c_{A,X}\otimes\operatorname{id}_A) c A , X ⊗ A = ( id X ⊗ c A , A ) ( c A , X ⊗ id A ) , hence the endomorphism mon F ( X ) , A = c A , F ( X ) c F ( X ) , A \operatorname{mon}_{F(X),A}=c_{A,F(X)}c_{F(X),A} mon F ( X ) , A = c A , F ( X ) c F ( X ) , A of F ( X ) ⊗ A F(X)\otimes A F ( X ) ⊗ A is
mon F ( X ) , A = c A , X ⊗ A c X ⊗ A , A = ( id X ⊗ c A , A ) ( ( c A , X c X , A ) ⊗ id A ) ( id X ⊗ c A , A ) . \operatorname{mon}_{F(X),A}=c_{A,X\otimes A}c_{X\otimes A,A}=(\operatorname{id}_X\otimes c_{A,A})((c_{A,X}c_{X,A})\otimes\operatorname{id}_A)(\operatorname{id}_X\otimes c_{A,A}). mon F ( X ) , A = c A , X ⊗ A c X ⊗ A , A = ( id X ⊗ c A , A ) (( c A , X c X , A ) ⊗ id A ) ( id X ⊗ c A , A ) .
For X ∈ Z A X\in Z_A X ∈ Z A the middle factor is the identity, thus mon F ( X ) , A = id X ⊗ c A , A 2 \operatorname{mon}_{F(X),A}=\operatorname{id}_X\otimes c^2_{A,A} mon F ( X ) , A = id X ⊗ c A , A 2 , and a F ( X ) mon F ( X ) , A = id X ⊗ μ c A , A c A , A = id X ⊗ μ = a F ( X ) a_{F(X)}\operatorname{mon}_{F(X),A}=\operatorname{id}_X\otimes\mu c_{A,A}c_{A,A}=\operatorname{id}_X\otimes\mu=a_{F(X)} a F ( X ) mon F ( X ) , A = id X ⊗ μ c A , A c A , A = id X ⊗ μ = a F ( X ) by commutativity applied twice. The same hexagon computation shows Z A Z_A Z A is closed under ⊗ \otimes ⊗ and contains 1 1 1 .
(ii) For ρ \rho ρ , apply X ⊗ − X\otimes- X ⊗ − to the diagram A ⊗ A ⊗ M ⇉ A ⊗ M → λ M M A\otimes A\otimes M \rightrightarrows A\otimes M \xrightarrow{\lambda_M} M A ⊗ A ⊗ M ⇉ A ⊗ M λ M M with the two maps μ ⊗ id M \mu\otimes\operatorname{id}_M μ ⊗ id M and id A ⊗ λ M \operatorname{id}_A\otimes\lambda_M id A ⊗ λ M . This is a split coequalizer, with sections t = η ⊗ id M : M → A ⊗ M t=\eta\otimes\operatorname{id}_M:M\to A\otimes M t = η ⊗ id M : M → A ⊗ M and u = η ⊗ id A ⊗ id M : A ⊗ M → A ⊗ A ⊗ M u=\eta\otimes\operatorname{id}_A\otimes\operatorname{id}_M:A\otimes M\to A\otimes A\otimes M u = η ⊗ id A ⊗ id M : A ⊗ M → A ⊗ A ⊗ M : indeed λ M t = id \lambda_Mt=\operatorname{id} λ M t = id , ( μ ⊗ id ) u = id (\mu\otimes\operatorname{id})u=\operatorname{id} ( μ ⊗ id ) u = id , ( id A ⊗ λ M ) u = t λ M (\operatorname{id}_A\otimes\lambda_M)u=t\lambda_M ( id A ⊗ λ M ) u = t λ M , and λ M ( μ ⊗ id M ) = λ M ( id A ⊗ λ M ) \lambda_M(\mu\otimes\operatorname{id}_M)=\lambda_M(\operatorname{id}_A\otimes\lambda_M) λ M ( μ ⊗ id M ) = λ M ( id A ⊗ λ M ) by the module axiom. Split coequalizers are absolute, thus X ⊗ − X\otimes- X ⊗ − preserves it, and X ⊗ M X\otimes M X ⊗ M with projection id X ⊗ λ M \operatorname{id}_X\otimes\lambda_M id X ⊗ λ M is the coequalizer defining F ( X ) ⊗ A M F(X)\otimes_A M F ( X ) ⊗ A M . ρ \rho ρ is a right module map for the action id X ⊗ a M \operatorname{id}_X\otimes a_M id X ⊗ a M since λ M ( id A ⊗ a M ) = a M ( λ M ⊗ id A ) \lambda_M(\operatorname{id}_A\otimes a_M)=a_M(\lambda_M\otimes\operatorname{id}_A) λ M ( id A ⊗ a M ) = a M ( λ M ⊗ id A ) ; explicitly, by naturality of c A , − c_{A,-} c A , − , associativity of the module and commutativity of A A A ,
λ M ( id A ⊗ a M ) = a M c A , M ( id A ⊗ a M ) = a M ( a M ⊗ id A ) c A , M ⊗ A = a M ( id M ⊗ μ ) ( id M ⊗ c A , A ) ( c A , M ⊗ id A ) = a M ( id M ⊗ μ ) ( c A , M ⊗ id A ) = a M ( a M ⊗ id A ) ( c A , M ⊗ id A ) = a M ( λ M ⊗ id A ) . \begin{aligned}
\lambda_M(\operatorname{id}_A\otimes a_M)=a_Mc_{A,M}(\operatorname{id}_A\otimes a_M)=a_M(a_M\otimes\operatorname{id}_A)c_{A,M\otimes A}=a_M(\operatorname{id}_M\otimes\mu)(\operatorname{id}_M\otimes c_{A,A})(c_{A,M}\otimes\operatorname{id}_A) \\
&=a_M(\operatorname{id}_M\otimes\mu)(c_{A,M}\otimes\operatorname{id}_A)=a_M(a_M\otimes\operatorname{id}_A)(c_{A,M}\otimes\operatorname{id}_A)=a_M(\lambda_M\otimes\operatorname{id}_A).
\end{aligned} λ M ( id A ⊗ a M ) = a M c A , M ( id A ⊗ a M ) = a M ( a M ⊗ id A ) c A , M ⊗ A = a M ( id M ⊗ μ ) ( id M ⊗ c A , A ) ( c A , M ⊗ id A ) = a M ( id M ⊗ μ ) ( c A , M ⊗ id A ) = a M ( a M ⊗ id A ) ( c A , M ⊗ id A ) = a M ( λ M ⊗ id A ) .
For σ \sigma σ , the left action on F ( X ) F(X) F ( X ) is λ F ( X ) = a F ( X ) c A , F ( X ) = ( id X ⊗ μ ) ( c A , X ⊗ id A ) \lambda_{F(X)}=a_{F(X)}c_{A,F(X)}=(\operatorname{id}_X\otimes\mu)(c_{A,X}\otimes\operatorname{id}_A) λ F ( X ) = a F ( X ) c A , F ( X ) = ( id X ⊗ μ ) ( c A , X ⊗ id A ) (using μ c A , A = μ \mu c_{A,A}=\mu μ c A , A = μ ). For X ∈ Z A X\in Z_A X ∈ Z A one has c X , A A = c A , X c^A_{X,A}=c_{A,X} c X , A A = c A , X , and then by naturality of c X , − c_{X,-} c X , − with respect to μ \mu μ and the hexagon, c X , A λ F ( X ) = ( id X ⊗ μ ) c X , A ⊗ A = ( id X ⊗ μ ) ( id A ⊗ c X , A ) ( c X , A ⊗ id A ) c_{X,A}\lambda_{F(X)}=(\operatorname{id}_X\otimes\mu)c_{X,A\otimes A}=(\operatorname{id}_X\otimes\mu)(\operatorname{id}_A\otimes c_{X,A})(c_{X,A}\otimes\operatorname{id}_A) c X , A λ F ( X ) = ( id X ⊗ μ ) c X , A ⊗ A = ( id X ⊗ μ ) ( id A ⊗ c X , A ) ( c X , A ⊗ id A ) . Thus c X , A c A , X = id c_{X,A}c_{A,X}=\operatorname{id} c X , A c A , X = id . Thus c X , A : F ( X ) → ( A ⊗ X , μ ⊗ id X ) c_{X,A}:F(X)\to(A\otimes X,\mu\otimes\operatorname{id}_X) c X , A : F ( X ) → ( A ⊗ X , μ ⊗ id X ) transports λ F ( X ) \lambda_{F(X)} λ F ( X ) to the free module structure, and M ⊗ A F ( X ) ≅ M ⊗ A ( A ⊗ X ) ≅ M ⊗ X M\otimes_A F(X)\cong M\otimes_A(A\otimes X)\cong M\otimes X M ⊗ A F ( X ) ≅ M ⊗ A ( A ⊗ X ) ≅ M ⊗ X via the split coequalizer M ⊗ A ⊗ A ⊗ X ⇉ M ⊗ A ⊗ X → a M ⊗ id X M ⊗ X M\otimes A\otimes A\otimes X\rightrightarrows M\otimes A\otimes X\xrightarrow{a_M\otimes\operatorname{id}_X}M\otimes X M ⊗ A ⊗ A ⊗ X ⇉ M ⊗ A ⊗ X a M ⊗ id X M ⊗ X (maps id M ⊗ μ ⊗ id X \operatorname{id}_M\otimes\mu\otimes\operatorname{id}_X id M ⊗ μ ⊗ id X and a M ⊗ id A ⊗ id X a_M\otimes\operatorname{id}_A\otimes\operatorname{id}_X a M ⊗ id A ⊗ id X ; sections id M ⊗ η ⊗ id X \operatorname{id}_M\otimes\eta\otimes\operatorname{id}_X id M ⊗ η ⊗ id X and id M ⊗ A ⊗ η ⊗ id X \operatorname{id}_{M\otimes A}\otimes\eta\otimes\operatorname{id}_X id M ⊗ A ⊗ η ⊗ id X ). The composite is σ \sigma σ , with σ π = ( a M ⊗ id X ) ( id M ⊗ c X , A ) \sigma\pi=(a_M\otimes\operatorname{id}_X)(\operatorname{id}_M\otimes c_{X,A}) σ π = ( a M ⊗ id X ) ( id M ⊗ c X , A ) . The statement about F ( X ) ⊗ A F ( Y ) F(X)\otimes_A F(Y) F ( X ) ⊗ A F ( Y ) is ρ X , F ( Y ) \rho_{X,F(Y)} ρ X , F ( Y ) (take M = F ( Y ) M=F(Y) M = F ( Y ) in the first part, with X ⊗ F ( Y ) = X ⊗ Y ⊗ A X\otimes F(Y)=X\otimes Y\otimes A X ⊗ F ( Y ) = X ⊗ Y ⊗ A ).
(iii) For inverses, ρ − 1 π = π X ( id X ⊗ η ⊗ id M ) = π \rho^{-1}\pi=\pi_X(\operatorname{id}_X\otimes\eta\otimes\operatorname{id}_M)=\pi ρ − 1 π = π X ( id X ⊗ η ⊗ id M ) = π , σ − 1 π = π M ( id M ⊗ id X ⊗ η ) \sigma^{-1}\pi=\pi_M(\operatorname{id}_M\otimes\operatorname{id}_X\otimes\eta) σ − 1 π = π M ( id M ⊗ id X ⊗ η ) . Indeed σ π ( id ⊗ η ⊗ id ) = id X ⊗ λ M ( η ⊗ id M ) = id \sigma\pi(\operatorname{id}\otimes\eta\otimes\operatorname{id})=\operatorname{id}_X\otimes\lambda_M(\eta\otimes\operatorname{id}_M)=\operatorname{id} σ π ( id ⊗ η ⊗ id ) = id X ⊗ λ M ( η ⊗ id M ) = id and σ π ( id M ⊗ η ⊗ id ) = ( a M ⊗ id X ) ( id M ⊗ c X , A ) ( id M ⊗ η ⊗ id X ) = id \sigma\pi(\operatorname{id}_M\otimes\eta\otimes\operatorname{id})=(a_M\otimes\operatorname{id}_X)(\operatorname{id}_M\otimes c_{X,A})(\operatorname{id}_M\otimes\eta\otimes\operatorname{id}_X)=\operatorname{id} σ π ( id M ⊗ η ⊗ id ) = ( a M ⊗ id X ) ( id M ⊗ c X , A ) ( id M ⊗ η ⊗ id X ) = id (naturality of c X , − c_{X,-} c X , − with respect to η \eta η ). Now compute, using c A π = π c c^A\pi=\pi c c A π = π c and c A , M ( η ⊗ id M ) = id M ⊗ η c_{A,M}(\eta\otimes\operatorname{id}_M)=\operatorname{id}_M\otimes\eta c A , M ( η ⊗ id M ) = id M ⊗ η :
c X ⊗ A , M ( id X ⊗ η ⊗ id M ) = ( c X , M ⊗ id A ) ( id X ⊗ c A , M ) ( id X ⊗ η ⊗ id M ) = ( id M ⊗ X ⊗ η ) c X , M , c_{X\otimes A,M}(\operatorname{id}_X\otimes\eta\otimes\operatorname{id}_M)=(c_{X,M}\otimes\operatorname{id}_A)(\operatorname{id}_X\otimes c_{A,M})(\operatorname{id}_X\otimes\eta\otimes\operatorname{id}_M)=(\operatorname{id}_{M\otimes X}\otimes\eta)c_{X,M}, c X ⊗ A , M ( id X ⊗ η ⊗ id M ) = ( c X , M ⊗ id A ) ( id X ⊗ c A , M ) ( id X ⊗ η ⊗ id M ) = ( id M ⊗ X ⊗ η ) c X , M ,
thus c F ( X ) , M A ρ − 1 = σ π M , X ( id M ⊗ X ⊗ η ) c X , M = c X , M c^A_{F(X),M}\rho^{-1}=\sigma\pi_{M,X}(\operatorname{id}_{M\otimes X}\otimes\eta)c_{X,M}=c_{X,M} c F ( X ) , M A ρ − 1 = σ π M , X ( id M ⊗ X ⊗ η ) c X , M = c X , M . Similarly,
c M , X ⊗ A ( id M ⊗ X ⊗ η ) = ( id X ⊗ c M , A ) ( c M , X ⊗ η ) = ( id X ⊗ η ⊗ id M ) c M , X , c_{M,X\otimes A}(\operatorname{id}_{M\otimes X}\otimes\eta)=(\operatorname{id}_X\otimes c_{M,A})(c_{M,X}\otimes\eta)=(\operatorname{id}_X\otimes\eta\otimes\operatorname{id}_M)c_{M,X}, c M , X ⊗ A ( id M ⊗ X ⊗ η ) = ( id X ⊗ c M , A ) ( c M , X ⊗ η ) = ( id X ⊗ η ⊗ id M ) c M , X ,
thus ρ c M , F ( X ) A σ − 1 = ρ π X , M ( id ⊗ η ⊗ id ) c M , X = c M , X \rho c^A_{M,F(X)}\sigma^{-1}=\rho\pi_{X,M}(\operatorname{id}\otimes\eta\otimes\operatorname{id})c_{M,X}=c_{M,X} ρ c M , F ( X ) A σ − 1 = ρ π X , M ( id ⊗ η ⊗ id ) c M , X = c M , X . The monodromy statement follows by composing the two identities. Note that locality of M M M and F ( X ) F(X) F ( X ) enters through the existence of c A c^A c A . (iv) By (iii) with M = F ( Y ) M=F(Y) M = F ( Y ) : σ F ( Y ) , X c F ( X ) , F ( Y ) A ρ X , F ( Y ) − 1 = c X , F ( Y ) \sigma_{F(Y),X}c^{A}_{F(X),F(Y)}\rho^{-1}_{X,F(Y)}=c_{X,F(Y)} σ F ( Y ) , X c F ( X ) , F ( Y ) A ρ X , F ( Y ) − 1 = c X , F ( Y ) , a morphism X ⊗ F ( Y ) → F ( Y ) ⊗ X X\otimes F(Y)\to F(Y)\otimes X X ⊗ F ( Y ) → F ( Y ) ⊗ X (note F ( Y ) F(Y) F ( Y ) is local by (i)). Also, ρ Y , F ( X ) σ F ( Y ) , X − 1 = ( id Y ⊗ λ F ( X ) ) ( id Y ⊗ A ⊗ X ⊗ η ) = id Y ⊗ c A , X \rho_{Y,F(X)}\sigma^{-1}_{F(Y),X}=(\operatorname{id}_{Y}\otimes\lambda_{F(X)})(\operatorname{id}_{Y}\otimes A\otimes X\otimes\eta)=\operatorname{id}_{Y}\otimes c_{A,X} ρ Y , F ( X ) σ F ( Y ) , X − 1 = ( id Y ⊗ λ F ( X ) ) ( id Y ⊗ A ⊗ X ⊗ η ) = id Y ⊗ c A , X , a morphism Y ⊗ A ⊗ X → Y ⊗ X ⊗ A Y\otimes A\otimes X\to Y\otimes X\otimes A Y ⊗ A ⊗ X → Y ⊗ X ⊗ A . Hence, the composite in (iv) equals ( id Y ⊗ c A , X ) c X , Y ⊗ A = ( id Y ⊗ c A , X ) ( id Y ⊗ c X , A ) ( c X , Y ⊗ id A ) = c X , Y ⊗ id A (\operatorname{id}_{Y}\otimes c_{A,X})c_{X,Y\otimes A}=(\operatorname{id}_{Y}\otimes c_{A,X})(\operatorname{id}_{Y}\otimes c_{X,A})(c_{X,Y}\otimes\operatorname{id}_{A})=c_{X,Y}\otimes\operatorname{id}_{A} ( id Y ⊗ c A , X ) c X , Y ⊗ A = ( id Y ⊗ c A , X ) ( id Y ⊗ c X , A ) ( c X , Y ⊗ id A ) = c X , Y ⊗ id A , since c A , X c X , A = id c_{A,X}c_{X,A}=\operatorname{id} c A , X c X , A = id on X ⊗ A X\otimes A X ⊗ A .
(v) The twist of D A l o c \mathcal{D}^{\mathrm{loc}}_{A} D A loc is θ M \theta_{M} θ M when θ A = id \theta_{A}=\operatorname{id} θ A = id (Proposition 2.8 (iii)). Balancing: θ X ⊗ A = ( θ X ⊗ θ A ) mon X , A = θ X ⊗ id A \theta_{X\otimes A}=(\theta_{X}\otimes\theta_{A})\operatorname{mon}_{X,A}=\theta_{X}\otimes\operatorname{id}_{A} θ X ⊗ A = ( θ X ⊗ θ A ) mon X , A = θ X ⊗ id A for X ∈ Z A X\in\mathcal{Z}_{A} X ∈ Z A .
Proof of Theorem 1.5
Proof. 0) B = C g r \mathcal{B}=\mathcal{C}^{\mathrm{gr}} B = C gr is a braided FTC (see § 2.3 ), and A ˉ \bar{A} A ˉ is a haploid commutative exact algebra in B \mathcal{B} B (Proposition 2.6 and Definition 2.7). Hence, B A ˉ l o c = C A s − l o c \mathcal{B}_{\bar{A}}^{\mathrm{loc}}=\mathcal{C}_{A}^{\mathrm{s-loc}} B A ˉ loc = C A s − loc is a braided FTC by Proposition 2.10 (i). Moreover A ˉ B A ˉ {}_{\bar{A}}\mathcal{B}_{\bar{A}} A ˉ B A ˉ is rigid by Lemma 2.9, thus the rigidity assumption in Theorem 1.1 holds, and the remaining statements are those of Theorem 1.1.
The ambient category. Let B ~ = C ⊠ T ε \widetilde{\mathcal{B}}=\mathcal{C}\boxtimes\mathcal{T}^{\varepsilon} B = C ⊠ T ε . It is non-degenerate: since T \mathcal{T} T is semisimple and pointed, an object M = ⨁ t X t ⊠ t M=\bigoplus_{t}X_{t}\boxtimes t M = ⨁ t X t ⊠ t is transparent iff each non-zero summand is. The monodromy of X t ⊠ t X_{t}\boxtimes t X t ⊠ t with 1 ⊠ s 1\boxtimes s 1 ⊠ s is mon t , s \operatorname{mon}_{t,s} mon t , s , thus X t ≠ 0 X_{t}\ne0 X t = 0 leads to t ∈ T ′ = ⟨ 1 ⟩ t\in\mathcal{T}'=\langle1\rangle t ∈ T ′ = ⟨ 1 ⟩ , and then X 1 X_{1} X 1 must be transparent in C \mathcal{C} C , i.e., X 1 ∈ C ′ ≃ V e c t X_{1}\in\mathcal{C}'\simeq\mathrm{Vect} X 1 ∈ C ′ ≃ Vect . Thus, B ~ ′ ≃ V e c t \widetilde{\mathcal{B}}'\simeq\mathrm{Vect} B ′ ≃ Vect , and B ~ \widetilde{\mathcal{B}} B is non-degenerate (see Lemma 2.11).
The algebra. Set A ′ : = ι ( A ˉ ) ∈ B ~ A':=\iota(\bar{A})\in\widetilde{\mathcal{B}} A ′ := ι ( A ˉ ) ∈ B . By Lemma 4.2 it is haploid, commutative and exact. Hence, N : = B ~ A ′ l o c \mathcal{N}:=\widetilde{\mathcal{B}}_{A'}^{\mathrm{loc}} N := B A ′ loc is a non-degenerate braided FTC by Proposition 2.10 (i).
The fermion. A ′ ∈ ι ( B ) = Z ( 2 ) ( 1 ⊠ ψ ) A'\in\iota(\mathcal{B})=\mathcal{Z}_{(2)}(1\boxtimes\psi) A ′ ∈ ι ( B ) = Z ( 2 ) ( 1 ⊠ ψ ) (see Lemma 4.1). Thus, ψ : = 1 ⊠ ψ \psi:=1\boxtimes\psi ψ := 1 ⊠ ψ centralizes A ′ A' A ′ , i.e., ψ ∈ Z A ′ \psi\in\mathcal{Z}_{A'} ψ ∈ Z A ′ . Set f ~ : = F ( ψ ) = ψ ⊗ A ′ \widetilde{f}:=F(\psi)=\psi\otimes A' f := F ( ψ ) = ψ ⊗ A ′ . By Lemma 4.3 (i) f ~ \widetilde{f} f is local. By (ii) and ψ ⊗ ψ ≃ 1 \psi\otimes\psi\simeq1 ψ ⊗ ψ ≃ 1 , f ~ ⊗ A ′ f ~ ≃ F ( ψ ⊗ ψ ) ≃ A ′ = 1 N \widetilde{f}\otimes_{A'}\widetilde{f}\simeq F(\psi\otimes\psi)\simeq A'=1_{\mathcal{N}} f ⊗ A ′ f ≃ F ( ψ ⊗ ψ ) ≃ A ′ = 1 N . Thus, f ~ \widetilde{f} f is invertible, hence simple (A ′ A' A ′ is haploid, so the unit of N \mathcal{N} N is simple). By (iv) with X = Y = ψ X=Y=\psi X = Y = ψ , the braiding c f ~ , f ~ A ′ c^{A'}_{\widetilde{f},\widetilde{f}} c f , f A ′ corresponds to c ψ , ψ ⊗ id A ′ = − id c_{\psi,\psi}\otimes\operatorname{id}_{A'}=-\operatorname{id} c ψ , ψ ⊗ id A ′ = − id on f ~ ⊗ A ′ f ~ ≃ A ′ \widetilde{f}\otimes_{A'}\widetilde{f}\simeq A' f ⊗ A ′ f ≃ A ′ . Since the braiding of the unit object with itself is the identity, f ~ ≄ 1 N \widetilde{f}\not\simeq1_{\mathcal{N}} f ≃ 1 N , thus ⟨ 1 N , f ~ ⟩ ≃ s V e c t \langle1_{\mathcal{N}},\widetilde{f}\rangle\simeq\mathrm{sVect} ⟨ 1 N , f ⟩ ≃ sVect as braided categories (see Lemma 2.14, q ( f ~ ) = − 1 q(\widetilde{f})=-1 q ( f ) = − 1 ). As a right A ′ A' A ′ -module f ~ = ψ ⊗ A ′ \widetilde{f}=\psi\otimes A' f = ψ ⊗ A ′ with action id ψ ⊗ μ \operatorname{id}_{\psi}\otimes\mu id ψ ⊗ μ . Under ι \iota ι this is ι ( f ⊗ A ˉ ) \iota(f\otimes\bar{A}) ι ( f ⊗ A ˉ ) where f ⊗ A ˉ = ( A 1 , A 0 ) = : Π A ˉ f\otimes\bar{A}=(A_{1},A_{0})=:\Pi\bar{A} f ⊗ A ˉ = ( A 1 , A 0 ) =: Π A ˉ with action id f ⊗ μ A ˉ \operatorname{id}_{f}\otimes\mu_{\bar{A}} id f ⊗ μ A ˉ , i.e., the module with graded object ( A 1 , A 0 ) (A_{1},A_{0}) ( A 1 , A 0 ) and the action of A i A_{i} A i on A j A_{j} A j given by μ \mu μ . Since ι \iota ι is a tensor functor, ι ( Π A ˉ ) ≃ f ~ \iota(\Pi\bar{A})\simeq\widetilde{f} ι ( Π A ˉ ) ≃ f as right A ′ A' A ′ -modules.
Identification of C A s − l o c \mathcal{C}_{A}^{\mathrm{s-loc}} C A s − loc . ι \iota ι is fully faithful, thus right A ˉ \bar{A} A ˉ -modules in B \mathcal{B} B are the right A ′ A' A ′ -modules in B ~ \widetilde{\mathcal{B}} B whose underlying object belongs to ι ( B ) \iota(\mathcal{B}) ι ( B ) (the action map is a morphism between objects of ι ( B ) \iota(\mathcal{B}) ι ( B ) , hence in ι ( B ) \iota(\mathcal{B}) ι ( B ) ), with the same locality equation. Thus, B A ˉ l o c \mathcal{B}_{\bar{A}}^{\mathrm{loc}} B A ˉ loc is the full subcategory of N \mathcal{N} N of local modules with underlying object in ι ( B ) \iota(\mathcal{B}) ι ( B ) . Let M ∈ N M\in\mathcal{N} M ∈ N . By Lemma 4.3 (iii) (with D = B ~ , X = ψ D=\widetilde{\mathcal{B}},X=\psi D = B , X = ψ ) the monodromy of M M M and f ~ \widetilde{f} f in N \mathcal{N} N is conjugate, via the isomorphism mon M , ψ \operatorname{mon}_{M,\psi} mon M , ψ , to the monodromy mon M , ψ \operatorname{mon}_{M,\psi} mon M , ψ of the underlying object of M M M and ψ \psi ψ in B ~ \widetilde{\mathcal{B}} B . Therefore, M ∈ Z ( 2 ) ( ⟨ f ~ ⟩ ⊂ N ) M\in\mathcal{Z}_{(2)}(\langle\widetilde{f}\rangle\subset\mathcal{N}) M ∈ Z ( 2 ) (⟨ f ⟩ ⊂ N ) , which is equivalent to mon M , f ~ = id \operatorname{mon}_{M,\widetilde{f}}=\operatorname{id} mon M , f = id , equivalently to mon M , ψ = id M ⊠ ψ \operatorname{mon}_{M,\psi}=\operatorname{id}_{M\boxtimes\psi} mon M , ψ = id M ⊠ ψ , equivalently to (Lemma 4.1) the underlying object of M M M belonging to ι ( B ) \iota(\mathcal{B}) ι ( B ) , which is equivalent to M ∈ B A ˉ l o c M\in\mathcal{B}_{\bar{A}}^{\mathrm{loc}} M ∈ B A ˉ loc . Note that each step is an equivalence: σ \sigma σ is an isomorphism, and since ⟨ f ~ ⟩ \langle\widetilde{f}\rangle ⟨ f ⟩ is generated by f ~ \widetilde{f} f the centralizer condition is the single condition on f ~ \widetilde{f} f . Thus, C A s − l o c = D : = Z ( 2 ) ( ⟨ f ~ ⟩ ⊂ N ) \mathcal{C}_{A}^{\mathrm{s-loc}}=\mathcal{D}:=\mathcal{Z}_{(2)}(\langle\widetilde{f}\rangle\subset\mathcal{N}) C A s − loc = D := Z ( 2 ) (⟨ f ⟩ ⊂ N ) .
Müger center and dimension. S : = ⟨ 1 N , f ~ ⟩ \mathcal{S}:=\langle1_{\mathcal{N}},\widetilde{f}\rangle S := ⟨ 1 N , f ⟩ (objects isomorphic to sums of 1 N 1_{\mathcal{N}} 1 N and f ~ \widetilde{f} f ) is a symmetric tensor full subcategory of N \mathcal{N} N , equivalent to s V e c t \mathrm{sVect} sVect , with FPdim S = 2 \operatorname{FPdim}\mathcal{S}=2 FPdim S = 2 . By Lemma 2.16, the Müger center of D = S ′ \mathcal{D}=\mathcal{S}' D = S ′ is S \mathcal{S} S , and FPdim D = FPdim N / 2 \operatorname{FPdim}\mathcal{D}=\operatorname{FPdim}\mathcal{N}/2 FPdim D = FPdim N /2 . Thus, D = C A s − l o c \mathcal{D}=\mathcal{C}_{A}^{\mathrm{s-loc}} D = C A s − loc is super-non-degenerate with fermion f ~ \widetilde{f} f (see Definition 1.3).
Let C \mathcal{C} C be ribbon and A ˉ \bar{A} A ˉ symmetric Frobenius. Then B \mathcal{B} B (with θ S \theta^{\mathcal{S}} θ S ), T ε \mathcal{T}^{\varepsilon} T ε and B ~ \widetilde{\mathcal{B}} B are ribbon, B ~ \widetilde{\mathcal{B}} B is an MTC, and ι \iota ι is a ribbon functor preserving the pivotal structures (Lemma 4.1). Then A ′ A' A ′ is Frobenius (a strong monoidal functor preserves the algebra, coalgebra and Frobenius relations), commutative, and θ A ′ = ι ( θ A ˉ S ) = id \theta_{A'}=\iota(\theta_{\bar{A}}^{\mathcal{S}})=\operatorname{id} θ A ′ = ι ( θ A ˉ S ) = id . Hence A ′ A' A ′ is symmetric Frobenius (Proposition 2.10 (iii)), and, being haploid, commutative and exact (Lemma 4.2), N \mathcal{N} N is an MTC (Proposition 2.10 (ii)).
We distinguish three ribbon structures:
(a) N \mathcal{N} N carries the twist θ M N = θ ~ M \theta_{M}^{\mathcal{N}}=\widetilde{\theta}_{M} θ M N = θ M of the underlying object (Proposition 2.8 (iii)), for which N \mathcal{N} N is an MTC.
(b) D ⊂ N \mathcal{D}\subset\mathcal{N} D ⊂ N is a full subcategory closed under tensor products (it is a centralizer) and duals, hence inherits the ribbon structure θ N ∣ D \theta^{\mathcal{N}}|_{\mathcal{D}} θ N ∣ D . (c) Under the identification D = B A ‾ l o c \mathcal{D}=\mathcal{B}_{\overline{A}}^{\mathrm{loc}} D = B A loc of part (1), θ M D = θ ~ l ( M ′ ) = ι ( θ M ′ S ) \theta_{M}^{\mathcal{D}}=\widetilde{\theta}_{l(M')}=\iota(\theta_{M'}^{\mathcal{S}}) θ M D = θ l ( M ′ ) = ι ( θ M ′ S ) for M = ι ( M ′ ) M=\iota(M') M = ι ( M ′ ) with M ′ ∈ B A ‾ l o c M'\in\mathcal{B}_{\overline{A}}^{\mathrm{loc}} M ′ ∈ B A loc (Lemma 4.1). Thus, the ribbon structure on D \mathcal{D} D of (b) coincides with the twist θ S \theta^{\mathcal{S}} θ S that Proposition 2.8 (iii) induces on B A ‾ l o c = C A s -loc \mathcal{B}_{\overline{A}}^{\mathrm{loc}}=\mathcal{C}_{A}^{s\text{-loc}} B A loc = C A s -loc . Finally, θ f ~ D = θ ψ ⊗ id A ′ = ε id \theta_{\widetilde{f}}^{\mathcal{D}}=\theta_{\psi}\otimes\operatorname{id}_{A'}=\varepsilon\operatorname{id} θ f D = θ ψ ⊗ id A ′ = ε id by Lemma 4.3 (v) (applicable since θ ~ A ′ = id \widetilde{\theta}_{A'}=\operatorname{id} θ A ′ = id ), and θ ψ = ε \theta_{\psi}=\varepsilon θ ψ = ε in T ε \mathcal{T}^{\varepsilon} T ε (both for T − \mathcal{T}^{-} T − , ε = − 1 \varepsilon=-1 ε = − 1 , and T + \mathcal{T}^{+} T + , ε = + 1 \varepsilon=+1 ε = + 1 ). Thus, D \mathcal{D} D is ε \varepsilon ε -super-modular. □
Remark 4.4. Let us discuss the enriched supercategory. Theorem 1.5 (1) concerns the underlying category D = C A s -loc \mathcal{D}=\mathcal{C}_{A}^{s\text{-loc}} D = C A s -loc (even morphisms). Let Π : = f ~ ⊗ A ′ − \Pi:=\widetilde{f}\otimes_{A'}- Π := f ⊗ A ′ − be the parity shift. On D = B A ‾ l o c \mathcal{D}=\mathcal{B}_{\overline{A}}^{\mathrm{loc}} D = B A loc this is f ⊗ − f\otimes- f ⊗ − . Define the supercategory D s \mathcal{D}^{s} D s with the same objects and Hom D s ( M , N ) = Hom D ( M , N ) ⊕ Hom D ( M , Π N ) \operatorname{Hom}_{\mathcal{D}^{s}}(M,N)=\operatorname{Hom}_{\mathcal{D}}(M,N)\oplus\operatorname{Hom}_{\mathcal{D}}(M,\Pi N) Hom D s ( M , N ) = Hom D ( M , N ) ⊕ Hom D ( M , Π N ) (even ⊕ \oplus ⊕ odd), composition through Π 2 ≅ id \Pi^{2}\cong\operatorname{id} Π 2 ≅ id , induced by f ~ ⊗ A ′ f ~ ≅ 1 N \widetilde{f}\otimes_{A'}\widetilde{f}\cong\mathbf{1}_{N} f ⊗ A ′ f ≅ 1 N . This is the correspondence of [8 ] between even morphisms into Π N \Pi N Π N and odd morphisms into N N N . We use it as a definition of the objects and morphisms of D s \mathcal{D}^{s} D s . In D s \mathcal{D}^{s} D s the odd isomorphism 1 → Π 1 = f ~ \mathbf{1}\to\Pi\mathbf{1}=\widetilde{f} 1 → Π 1 = f identifies f ~ \widetilde{f} f with 1 \mathbf{1} 1 . Call an object of D s \mathcal{D}^{s} D s transparent if it is transparent in D \mathcal{D} D (an object-level notion, involving only even morphisms). We say that D s \mathcal{D}^{s} D s is non-degenerate in the super sense if every transparent object is isomorphic in D s \mathcal{D}^{s} D s to 1 ⊕ n \mathbf{1}^{\oplus n} 1 ⊕ n for some n n n . Then Theorem 1.5 (1) implies that D s \mathcal{D}^{s} D s is non-degenerate in the super sense. By (1) the transparent objects of D \mathcal{D} D are the sums 1 a ⊕ f ~ b \mathbf{1}^{a}\oplus\widetilde{f}^{b} 1 a ⊕ f b , and f ~ ≅ 1 \widetilde{f}\cong\mathbf{1} f ≅ 1 in D s \mathcal{D}^{s} D s . In § 7 the identification of the even part is the assumption (v2).
Remark 4.5. The functor ι \iota ι , and the proof of Theorem 1.5 , depend on the choice of T ε \mathcal{T}^{\varepsilon} T ε . Namely, N \mathcal{N} N depends on it while D = C A s -loc \mathcal{D}=\mathcal{C}_{A}^{s\text{-loc}} D = C A s -loc does not. There are sixteen minimal modular extensions of sVect (Kitaev 16-fold way, [5 ] ). In this paper we use the toric code. Lemmas 4.1 and 4.2 use that T \mathcal{T} T is pointed with ⟨ 1 , ψ ⟩ \langle1,\psi\rangle ⟨ 1 , ψ ⟩ acting freely on the other simple objects. By Proposition 8.1 , [ N ] = [ C ] [\mathcal{N}]=[\mathcal{C}] [ N ] = [ C ] in the Witt group of non-degenerate braided FTCs for the toric code. Corollary 1.6 is formulated in this case. For nice VOSAs with U = V ‾ 0 U=\overline{V}_{0} U = V 0 , [12 ] shows that Rep ( V ‾ 0 ) \operatorname{Rep}(\overline{V}_{0}) Rep ( V 0 ) is a modular category containing the NS category as the centralizer of a fermion. This is consistent with Example 6.1 where N \mathcal{N} N is again an Ising-type category.
Constructions
Graded simple currents
Let H ≤ Inv ( C ) H\leq\operatorname{Inv}(\mathcal{C}) H ≤ Inv ( C ) be a subgroup of the (finite) group of isomorphism classes of invertible objects of a braided FTC and χ : H → Z / 2 \chi:H\to\mathbb{Z}/2 χ : H → Z /2 a character (the grading is part of the data). Let X h X_h X h be an invertible object of class h h h , q ( h ) : = q ( X h ) q(h):=q(X_h) q ( h ) := q ( X h ) , and A H = ⨁ h X h A_H=\bigoplus_h X_h A H = ⨁ h X h with deg X h = χ ( h ) \deg X_h=\chi(h) deg X h = χ ( h ) . Then we have
Proposition 5.1. (a) Let us assume: q ( h ) = ( − 1 ) χ ( h ) q(h)=(-1)^{\chi(h)} q ( h ) = ( − 1 ) χ ( h ) for all h ∈ H h\in H h ∈ H . Then A H A_H A H (with the multiplication constructed in the proof) is a haploid supercommutative superalgebra, A ‾ H \overline{A}_H A H is exact (indeed separable), and C A ‾ H s -loc \mathcal{C}_{\overline{A}_H}^{s\text{-loc}} C A H s -loc is a braided FTC.
(b) Assume in addition: C \mathcal{C} C is ribbon and θ X h = ε χ ( h ) id \theta_{X_h}=\varepsilon^{\chi(h)}\operatorname{id} θ X h = ε χ ( h ) id for all h ∈ H h\in H h ∈ H . Then A ‾ H \overline{A}_H A H is symmetric Frobenius and C A ‾ H s -loc \mathcal{C}_{\overline{A}_H}^{s\text{-loc}} C A H s -loc is ribbon.
(c) With assumptions of (a), (b) and that C \mathcal{C} C is an MTC, then C A ‾ H s -loc \mathcal{C}_{\overline{A}_H}^{s\text{-loc}} C A H s -loc is ε \varepsilon ε -super-modular.
Proof. 1): Triviality monodromy on H H H . The monodromy bicharacter is mon g , h = q ( g h ) q ( g ) − 1 q ( h ) − 1 \operatorname{mon}_{g,h}=q(gh)q(g)^{-1}q(h)^{-1} mon g , h = q ( g h ) q ( g ) − 1 q ( h ) − 1 (polarization of the quadratic form of the pointed braided subcategory, [13 ] ). By assumption of Proposition 5.1 (a) q = ( − 1 ) χ q=(-1)^{\chi} q = ( − 1 ) χ is a character, thus mon g , h = 1 \operatorname{mon}_{g,h}=1 mon g , h = 1 .
2): A braided model. Let Y h ∈ B Y_h\in\mathcal{B} Y h ∈ B be X h X_h X h placed in degree χ ( h ) \chi(h) χ ( h ) . As χ \chi χ is a character, Y g Y h ≅ Y g h Y_gY_h\cong Y_{gh} Y g Y h ≅ Y g h and h ↦ Y h h\mapsto Y_h h ↦ Y h is an injective homomorphism H → Inv ( B ) H\to\operatorname{Inv}(\mathcal{B}) H → Inv ( B ) . The braided subcategory ⟨ Y h ⟩ ⊂ B \langle Y_h\rangle\subset\mathcal{B} ⟨ Y h ⟩ ⊂ B has quadratic form q B ( h ) = ε Y h , Y h s = ( − 1 ) χ ( h ) q ( h ) = 1 q^{\mathcal{B}}(h)=\varepsilon_{Y_h,Y_h}^{s}=(-1)^{\chi(h)}q(h)=1 q B ( h ) = ε Y h , Y h s = ( − 1 ) χ ( h ) q ( h ) = 1 by assumption of Proposition 5.1 (a). By Lemma 2.14 there is a braided monoidal equivalence G : Vect H → ⟨ Y h ⟩ G:\operatorname{Vect}_{H}\to\langle Y_h\rangle G : Vect H → ⟨ Y h ⟩ with G ( δ h ) ≅ Y h G(\delta_h)\cong Y_h G ( δ h ) ≅ Y h , where Vect H \operatorname{Vect}_{H} Vect H has trivial associator and trivial braiding.
3): The multiplication. The group algebra k [ H ] = ⨁ h δ h \mathbb{k}[H]=\bigoplus_h\delta_h k [ H ] = ⨁ h δ h (multiplication δ g ⊗ δ h → δ g h \delta_g\otimes\delta_h\to\delta_{gh} δ g ⊗ δ h → δ g h ) is a commutative algebra in Vect H \operatorname{Vect}_{H} Vect H , since the braiding is trivial. Set A ‾ H : = G ( k [ H ] ) \overline{A}_H:=G(\mathbb{k}[H]) A H := G ( k [ H ]) . As G G G is braided monoidal, A ‾ H \overline{A}_H A H is a commutative algebra in B \mathcal{B} B with underlying object ⨁ h Y h = ( A 0 , A 1 ) \bigoplus_hY_h=(A_0,A_1) ⨁ h Y h = ( A 0 , A 1 ) , A i = ⨁ χ ( h ) = i X h A_i=\bigoplus_{\chi(h)=i}X_h A i = ⨁ χ ( h ) = i X h , and its unit and multiplication preserve the degree since χ ( g ) + χ ( h ) = χ ( g h ) \chi(g)+\chi(h)=\chi(gh) χ ( g ) + χ ( h ) = χ ( g h ) . By Proposition 2.6 (a), A H A_H A H is a supercommutative superalgebra in C \mathcal{C} C . Explicitly, the multiplication consists of isomorphisms m g , h : X g ⊗ X h → X g h m_{g,h}:X_g\otimes X_h\to X_{gh} m g , h : X g ⊗ X h → X g h (the components of the tensor structure of G G G ).
We separate three points: (i) Triviality of the associativity class. That the m g , h m_{g,h} m g , h satisfy the associativity axiom says that they trivialize the associator of the pointed category ⟨ X h ⟩ ⊂ C \langle X_h\rangle\subset\mathcal{C} ⟨ X h ⟩ ⊂ C , whose class belongs to H a b 3 ( H , k × ) ≅ Quad ( H ) H^3_{\mathrm{ab}}(H,\mathbb{k}^{\times})\cong\operatorname{Quad}(H) H ab 3 ( H , k × ) ≅ Quad ( H ) . This class is trivial since the assumption of Proposition 5.1 (a) makes the quadratic form of ⟨ X h ⟩ \langle X_h\rangle ⟨ X h ⟩ equal to ( − 1 ) χ (-1)^\chi ( − 1 ) χ , that of Vect H \operatorname{Vect}_H Vect H with trivial associator and braiding β ( g , h ) = ( − 1 ) χ ( g ) χ ( h ) \beta(g,h)=(-1)^{\chi(g)\chi(h)} β ( g , h ) = ( − 1 ) χ ( g ) χ ( h ) [20 ] .
(ii) Choice of the multiplication. Triviality of the class does not specify the m g , h m_{g,h} m g , h : the tensor structures with the required property form a torsor under the normalized 2-cocycles Z 2 ( H , k × ) Z^2(H,\Bbbk^\times) Z 2 ( H , k × ) (rescaling each m g , h m_{g,h} m g , h ).
(iii) Commutativity. The condition μ c s = μ \mu c^s=\mu μ c s = μ in B \mathcal{B} B makes the ratio of two admissible choices to be a symmetric 2-cocycle. Since k × \Bbbk^\times k × is divisible, Ext Z 1 ( H , k × ) = 0 \operatorname{Ext}^1_{\mathbb{Z}}(H,\Bbbk^\times)=0 Ext Z 1 ( H , k × ) = 0 , thus every symmetric 2-cocycle on the finite abelian group H H H is a coboundary. Hence, any two commutative multiplications are isomorphic (by rescaling the identifications of X h X_h X h ), and A H A_H A H is well defined up to isomorphism. The unit and associativity axioms hold since G \mathcal{G} G is monoidal.
4): Haploid, separable, exact. Hom B ( 1 , A ‾ H ) = ⨁ h Hom ( 1 , Y h ) = k \operatorname{Hom}_{\mathcal{B}}(\mathbf{1},\overline{A}_H)=\bigoplus_h\operatorname{Hom}(\mathbf{1},Y_h)=\Bbbk Hom B ( 1 , A H ) = ⨁ h Hom ( 1 , Y h ) = k , since X h ≇ 1 X_h\not\cong\mathbf{1} X h ≅ 1 for h ≠ 1 h\ne1 h = 1 and χ ( 1 ) = 0 \chi(1)=0 χ ( 1 ) = 0 . Next k [ H ] \Bbbk[H] k [ H ] is separable in Vect H \operatorname{Vect}_H Vect H : Δ ( δ h ) = ∣ H ∣ − 1 ∑ g ∈ H δ g ⊗ δ g − 1 h \Delta(\delta_h)=|H|^{-1}\sum_{g\in H}\delta_g\otimes\delta_{g^{-1}h} Δ ( δ h ) = ∣ H ∣ − 1 ∑ g ∈ H δ g ⊗ δ g − 1 h is a bimodule map with μ Δ = id \mu\Delta=\operatorname{id} μ Δ = id (∣ H ∣ |H| ∣ H ∣ is invertible as char k = 0 \operatorname{char}\Bbbk=0 char k = 0 ). Hence A ‾ H = G ( k [ H ] ) \overline{A}_H=\mathcal{G}(\Bbbk[H]) A H = G ( k [ H ]) is separable in B \mathcal{B} B . A separable algebra A ‾ \overline{A} A in a finite tensor category is exact: the multiplication M ⊗ A ‾ → M M\otimes\overline{A}\to M M ⊗ A → M of any right module M M M splits as a map of right modules, so for P P P projective in B \mathcal{B} B the module P ⊗ M P\otimes M P ⊗ M is a direct summand of the free module ( P ⊗ M ) ⊗ A ‾ (P\otimes M)\otimes\overline{A} ( P ⊗ M ) ⊗ A , which is projective in B A ‾ \mathcal{B}_{\overline{A}} B A since P ⊗ M P\otimes M P ⊗ M is projective in B \mathcal{B} B [13 ] and the free functor is left adjoint to the exact forgetful functor. Thus, B A ‾ \mathcal{B}_{\overline{A}} B A is an exact B \mathcal{B} B -module category, i.e., A ‾ H \overline{A}_H A H is exact. By Theorem 1.5 (0), C A ‾ H s -loc \mathcal{C}^{s\text{-loc}}_{\overline{A}_H} C A H s -loc is a braided FTC. This proves (a).
5): Frobenius and ribbon. k [ H ] \Bbbk[H] k [ H ] is a Frobenius algebra in Vect H \operatorname{Vect}_H Vect H (comultiplication Δ \Delta Δ above, counit δ h ↦ δ h , 1 \delta_h\mapsto\delta_{h,1} δ h ↦ δ h , 1 ), thus A ‾ H = G ( k [ H ] ) \overline{A}_H=\mathcal{G}(\Bbbk[H]) A H = G ( k [ H ]) is a Frobenius algebra (G \mathcal{G} G is strong monoidal). Under assumption of Proposition 5.1 (b), θ Y h = ε χ ( h ) θ X h = ε 2 χ ( h ) = id \theta_{Y_h}=\varepsilon^{\chi(h)}\theta_{X_h}=\varepsilon^{2\chi(h)}=\operatorname{id} θ Y h = ε χ ( h ) θ X h = ε 2 χ ( h ) = id , hence θ A ‾ H s = id \theta^s_{\overline{A}_H}=\operatorname{id} θ A H s = id and, A ‾ H \overline{A}_H A H being commutative Frobenius, it is symmetric by Proposition 2.10 (iii) (applied in the ribbon category B \mathcal{B} B ). Moreover, the Frobenius isomorphism is an isomorphism of right A ‾ H \overline{A}_H A H -modules A ‾ H ≅ A ‾ H ∗ \overline{A}_H\cong\overline{A}_H^* A H ≅ A H ∗ [26 ] , thus A ‾ H ∗ \overline{A}_H^* A H ∗ is the unit of B A ‾ H loc \mathcal{B}^{\operatorname{loc}}_{\overline{A}_H} B A H loc and belongs to the Müger center. Theorem 1.1 (c) (the rigidity assumption being automatic by Theorem 1.5 (0)) shows that C A ‾ H s -loc \mathcal{C}^{s\text{-loc}}_{\overline{A}_H} C A H s -loc is ribbon. This proves (b).
(c) is Theorem 1.5 (2): A ‾ H \overline{A}_H A H is haploid, commutative, exact and symmetric Frobenius.
Necessity of assumption of Proposition 5.1 (a). If A ‾ H \overline{A}_H A H is commutative and the multiplication maps m h , h m_{h,h} m h , h (isomorphisms onto the summands X h 2 X_{h^2} X h 2 , the nonzero components of μ \mu μ on Y h ⊗ Y h Y_h\otimes Y_h Y h ⊗ Y h ) are nonzero, then μ c s = μ \mu c^s=\mu μ c s = μ on the simple object Y h ⊗ Y h Y_h\otimes Y_h Y h ⊗ Y h leads to c Y h , Y h s = id c^s_{Y_h,Y_h}=\operatorname{id} c Y h , Y h s = id , i.e., assumption of Proposition 5.1 (a).
Remark 5.2. For χ = 0 \chi=0 χ = 0 , assumption of Proposition 5.1 (a) reads q = 1 q=1 q = 1 , and Proposition 5.1 (a) recovers the simple current algebra construction of [26 ] (their §6.1). If moreover θ X h ≠ 1 \theta_{X_h}\ne1 θ X h = 1 for some h h h , then assumption of Proposition 5.1 (b) fails (it requires θ X h = ε 0 = 1 \theta_{X_h}=\varepsilon^0=1 θ X h = ε 0 = 1 ), so that parts (b), (c) do not apply, although part (a) does: this is the situation of a 1 2 Z \frac{1}{2}\mathbb{Z} 2 1 Z -graded vertex operator algebra extension (case (2) of § 7.1 ), for which θ A ≠ id \theta_A\ne\operatorname{id} θ A = id and C A loc \mathcal{C}^{\operatorname{loc}}_A C A loc has no ribbon structure induced from θ \theta θ by Proposition 2.8 (iii) (consistent with [8 ] , where e 2 π i L 0 e^{2\pi iL_0} e 2 π i L 0 is not a module map). Thus, (a) holds in all three cases of § 7.1 under assumption of Proposition 5.1 (a), while (b), (c) hold in cases (1) and (3) only.
Examples
Example 6.1. We discuss here the free fermion case. Let C \mathcal{C} C be the Ising MTC (the category of L ( 1 2 , 0 ) = V 0 ‾ L(\frac{1}{2},0)=V_{\overline{0}} L ( 2 1 , 0 ) = V 0 , simples 1 , ψ , σ \mathbf{1},\psi,\sigma 1 , ψ , σ ), ε = − 1 \varepsilon=-1 ε = − 1 , H = { 1 , ψ } H=\{1,\psi\} H = { 1 , ψ } , χ = id \chi=\operatorname{id} χ = id . Then q ( ψ ) = c ψ , ψ = − 1 = ( − 1 ) χ ( ψ ) q(\psi)=c_{\psi,\psi}=-1=(-1)^{\chi(\psi)} q ( ψ ) = c ψ , ψ = − 1 = ( − 1 ) χ ( ψ ) and θ ψ = − 1 = ε \theta_\psi=-1=\varepsilon θ ψ = − 1 = ε , thus Proposition 5.1 (a)-(c) apply: A = 1 ⊕ ψ A=\mathbf{1}\oplus\psi A = 1 ⊕ ψ (the free Majorana fermion VOA) is a haploid supercommutative symmetric Frobenius algebra. The Frobenius-Perron dimensions are FPdim C = 4 \operatorname{FPdim}\mathcal{C}=4 FPdim C = 4 , FPdim B = 8 \operatorname{FPdim}\mathcal{B}=8 FPdim B = 8 , FPdim B ~ = 16 \operatorname{FPdim}\widetilde{\mathcal{B}}=16 FPdim B = 16 , FPdim A ‾ = 2 \operatorname{FPdim}\overline{A}=2 FPdim A = 2 , FPdim N = FPdim B ~ / FPdim ( A ′ ) 2 = 16 / 4 = 4 \operatorname{FPdim}\mathcal{N}=\operatorname{FPdim}\widetilde{\mathcal{B}}/\operatorname{FPdim}(A')^2=16/4=4 FPdim N = FPdim B / FPdim ( A ′ ) 2 = 16/4 = 4 , FPdim C A s -loc = FPdim N / 2 = 2 \operatorname{FPdim}\mathcal{C}^{s\text{-loc}}_A=\operatorname{FPdim}\mathcal{N}/2=2 FPdim C A s -loc = FPdim N /2 = 2 (the formula for N \mathcal{N} N is by [11 ] , applicable since A ′ = ( 1 , 1 ) ⊕ ( ψ , ψ T ) A'=(\mathbf{1},\mathbf{1})\oplus(\psi,\psi_T) A ′ = ( 1 , 1 ) ⊕ ( ψ , ψ T ) is a connected étale algebra in the non-degenerate fusion category B ~ = Ising ⊠ T \widetilde{\mathcal{B}}=\operatorname{Ising}\boxtimes T B = Ising ⊠ T ). Explicitly, the objects X ⊠ t X\boxtimes t X ⊠ t of Ising ⊠ T \operatorname{Ising}\boxtimes T Ising ⊠ T whose monodromy with ψ ⊠ ψ T \psi\boxtimes\psi_T ψ ⊠ ψ T is trivial are ( 1 , 1 ) , ( ψ , 1 ) , ( 1 , ψ T ) , ( ψ , ψ T ) (\mathbf{1},\mathbf{1}),(\psi,\mathbf{1}),(\mathbf{1},\psi_T),(\psi,\psi_T) ( 1 , 1 ) , ( ψ , 1 ) , ( 1 , ψ T ) , ( ψ , ψ T ) and ( σ , e ) , ( σ , m ) (\sigma,e),(\sigma,m) ( σ , e ) , ( σ , m ) (use m ψ , ψ = − 1 m_{\psi,\psi}=-1 m ψ , ψ = − 1 , m ψ , ψ T = 1 m_{\psi,\psi_T}=1 m ψ , ψ T = 1 , m 1 , ψ T = − 1 m_{\mathbf{1},\psi_T}=-1 m 1 , ψ T = − 1 ). The three free modules F ( 1 , 1 ) ≅ F ( ψ , ψ T ) = 1 N F(\mathbf{1},\mathbf{1})\cong F(\psi,\psi_T)=\mathbf{1}_{\mathcal{N}} F ( 1 , 1 ) ≅ F ( ψ , ψ T ) = 1 N , F ( ψ , 1 ) ≅ F ( 1 , ψ T ) = f ~ F(\psi,\mathbf{1})\cong F(\mathbf{1},\psi_T)=\widetilde{f} F ( ψ , 1 ) ≅ F ( 1 , ψ T ) = f (both of FP dimension 1), F ( σ , e ) ≅ F ( σ , m ) = : s F(\sigma,e)\cong F(\sigma,m)=:s F ( σ , e ) ≅ F ( σ , m ) =: s . The module s s s has underlying object ( σ , e ) ⊕ ( σ , m ) (\sigma,e)\oplus(\sigma,m) ( σ , e ) ⊕ ( σ , m ) of FP dimension 2 2 2\sqrt{2} 2 2 , hence FP dimension 2 2 / FPdim A ′ = 2 2\sqrt{2}/\operatorname{FPdim}A'=\sqrt{2} 2 2 / FPdim A ′ = 2 in N \mathcal{N} N . It is simple, since a non-zero object of N \mathcal{N} N has FP dimension at least 1, so 2 < 2 \sqrt{2}<2 2 < 2 excludes a decomposition into two summands. As N \mathcal{N} N is semisimple, 1 2 + 1 2 + ( 2 ) 2 = 4 = FPdim N 1^2+1^2+(\sqrt{2})^2=4=\operatorname{FPdim}\mathcal{N} 1 2 + 1 2 + ( 2 ) 2 = 4 = FPdim N shows that 1 N , f ~ , s \mathbf{1}_{\mathcal{N}},\widetilde{f},s 1 N , f , s are all the simple objects of N \mathcal{N} N . Since s s s is the only simple object of dimension 2 \sqrt{2} 2 , it is self-dual, and s ⊗ s s\otimes s s ⊗ s has FP dimension 2 and contains 1 N 1_{\mathcal{N}} 1 N once, so s ⊗ s = 1 N ⊕ f ~ s \otimes s = 1_{\mathcal{N}} \oplus\tilde{f} s ⊗ s = 1 N ⊕ f ~ : N \mathcal{N} N has the fusion rules of the Ising category (we call such a braided fusion category Ising-type). The centralizer of f ~ \tilde{f} f ~ is { 1 , f ~ } \{1,\tilde{f}\} { 1 , f ~ } since mon s , f ~ = − 1 \operatorname{mon}_{s,\tilde{f}}=-1 mon s , f ~ = − 1 . Thus, C A s -loc = ⟨ f ~ ⟩ ≃ s V e c t \mathcal{C}^{s\text{-loc}}_{A}=\langle\tilde{f}\rangle\simeq\mathrm{sVect} C A s -loc = ⟨ f ~ ⟩ ≃ sVect : the NS sector of the free fermion has one simple module up to parity. The Ramond sector arises from σ \sigma σ : mon σ , ψ = − 1 \operatorname{mon}_{\sigma,\psi}=-1 mon σ , ψ = − 1 , thus σ ⊗ A \sigma\otimes A σ ⊗ A is a graded A A A -module that is not super-local. Note that ψ \psi ψ is not transparent in C \mathcal{C} C .
Example 6.2. Here we discuss the symplectic fermions, d = 1 d=1 d = 1 . First, recall the following
Proposition 6.3. (1) C = Rep ( W ( 2 ) ) \mathcal{C}=\operatorname{Rep}(W(2)) C = Rep ( W ( 2 )) (W ( 2 ) W(2) W ( 2 ) the triplet algebra at c = − 2 c=-2 c = − 2 , equal to the even part S F + \mathrm{SF}^{+} SF + of the symplectic fermion VOSA V = S F 1 V=\mathrm{SF}_{1} V = SF 1 [1 ] ) is a non-degenerate braided FTC with ribbon twist e 2 π i L 0 e^{2\pi iL_{0}} e 2 π i L 0 ;
(2) the odd part V ‾ 1 = W ( 1 ) \overline{V}_{1}=W(1) V 1 = W ( 1 ) (the irreducible module of lowest weight 1 ) is invertible, W ( 1 ) ⊠ W ( 1 ) ≅ W ( 0 ) = 1 W(1)\boxtimes W(1)\cong W(0)=1 W ( 1 ) ⊠ W ( 1 ) ≅ W ( 0 ) = 1 [27 ] ;
(3) V V V is a supercommutative superalgebra in C \mathcal{C} C with A 0 = 1 A_{0}=1 A 0 = 1 , A 1 = W ( 1 ) A_{1}=W(1) A 1 = W ( 1 ) , whose local modules are the NS V V V -supermodules, i.e., ( v 1 ) , ( v 2 ) (v1),(v2) ( v 1 ) , ( v 2 ) of § 7 , with non-zero product V ‾ 1 ⊗ V ‾ 1 → V ‾ 0 \overline{V}_{1}\otimes\overline{V}_{1}\to\overline{V}_{0} V 1 ⊗ V 1 → V 0 .
Categorical consequences of Proposition 6.3 (1)-(3). θ W ( 1 ) = e 2 π i ⋅ 1 = + 1 \theta_{W(1)}=e^{2\pi i\cdot1}=+1 θ W ( 1 ) = e 2 π i ⋅ 1 = + 1 , thus ε = + 1 \varepsilon=+1 ε = + 1 (Z-graded VOSA; “incorrect statistics” in the sense of [8 ] ). Since Hom ( W ( 1 ) ⊗ W ( 1 ) , 1 ) \operatorname{Hom}(W(1)\otimes W(1),1) Hom ( W ( 1 ) ⊗ W ( 1 ) , 1 ) is one-dimensional, c W ( 1 ) , W ( 1 ) = q id c_{W(1),W(1)}=q\,\operatorname{id} c W ( 1 ) , W ( 1 ) = q id , and μ c = − μ \mu c=-\mu μ c = − μ with μ ≠ 0 \mu\ne0 μ = 0 leads to q ( W ( 1 ) ) = − 1 q(W(1))=-1 q ( W ( 1 )) = − 1 . Then dim W ( 1 ) = θ / q = − 1 \dim W(1)=\theta/q=-1 dim W ( 1 ) = θ / q = − 1 , matching the value of [8 ] . Thus, assumptions of Proposition 5.1 (a) and (b) hold for H = { 1 , W ( 1 ) } H=\{1,W(1)\} H = { 1 , W ( 1 )} , χ = id \chi=\operatorname{id} χ = id , ε = + 1 \varepsilon=+1 ε = + 1 , and Proposition 5.1 (c) shows that C A s -loc \mathcal{C}^{s\text{-loc}}_{A} C A s -loc is a ( + 1 ) (+1) ( + 1 ) -super-modular category (in particular not pseudo-unitary, Remark 1.4). The dimension formula FPdim N = FPdim B ~ / FPdim ( A ′ ) 2 \operatorname{FPdim}\mathcal{N}=\operatorname{FPdim}\widetilde{\mathcal{B}}/\operatorname{FPdim}(A')^{2} FPdim N = FPdim B / FPdim ( A ′ ) 2 of [23 ] (applicable since A ′ A' A ′ is a special Frobenius algebra) gives, with FPdim A ‾ = 2 \operatorname{FPdim}\overline{A}=2 FPdim A = 2 , FPdim N = 4 FPdim C / 2 2 = FPdim C \operatorname{FPdim}\mathcal{N}=4\operatorname{FPdim}\mathcal{C}/2^{2}=\operatorname{FPdim}\mathcal{C} FPdim N = 4 FPdim C / 2 2 = FPdim C and FPdim C A s -loc = FPdim C / 2 \operatorname{FPdim}\mathcal{C}^{s\text{-loc}}_{A}=\operatorname{FPdim}\mathcal{C}/2 FPdim C A s -loc = FPdim C /2 . By Proposition 6.3 (3) it is Rep N S ( S F 1 ) \operatorname{Rep}^{\mathrm{NS}}(\mathrm{SF}_{1}) Rep NS ( SF 1 ) . Exactness and the Frobenius property of A ‾ \overline{A} A come from Proposition 5.1.
The VOSA notions
In this section we recall material from [18 , 7 , 8 , 19 ] . Let V = V ‾ 0 ⊕ V ‾ 1 V=\overline{V}_{0}\oplus\overline{V}_{1} V = V 0 ⊕ V 1 be a VOSA and U ⊆ V ‾ 0 U\subseteq\overline{V}_{0} U ⊆ V 0 a vertex subalgebra with the same conformal vector. C \mathcal{C} C is a category of U U U -modules containing V ‾ 0 , V ‾ 1 \overline{V}_{0},\overline{V}_{1} V 0 , V 1 . Concerning a VOSA we state the following assumptions:
(v1) [7 , 8 ] : V V V is a supercommutative superalgebra in C \mathcal{C} C (Definition 2.5) with A i = V ‾ i A_{i}=\overline{V}_{i} A i = V i . This is an algebra in the supercategory S C \mathcal{SC} SC of [8 ] with even structure maps. The braiding of S C \mathcal{SC} SC is ⨁ ( − 1 ) i 1 i 2 R W i 1 , W i 2 \bigoplus(-1)^{i_{1}i_{2}}R^{W_{i_{1}},W_{i_{2}}} ⨁ ( − 1 ) i 1 i 2 R W i 1 , W i 2 , i.e., our c s c^{s} c s .
(v2) [8 ] : the category of NS (untwisted) V V V -supermodules in C \mathcal{C} C with even morphisms is equivalent, as a braided monoidal category, to C A s -loc = B A l o c \mathcal{C}^{s\text{-loc}}_{A}=\mathcal{B}^{\mathrm{loc}}_{A} C A s -loc = B A loc .
(v3) Three separate facts on C \mathcal{C} C , which are not to be conflated:
(v3a) Tensor structure: C \mathcal{C} C carries the braided tensor structure of [19 ] (its existence is an assumption on U U U , e.g., cofiniteness conditions as in [17 ] );
(v3b) Finiteness and rigidity: C \mathcal{C} C is a braided finite tensor category, i.e., finite abelian and rigid (for strongly rational U U U see [16 ] ; for strongly finite U U U with rigid Rep ( U ) \operatorname{Rep}(U) Rep ( U ) see [17 ] for the braided monoidal structure and [9 ] for rigidity; otherwise it is an assumption);
(v3c) Twist: the ribbon twist of C \mathcal{C} C (when C \mathcal{C} C is ribbon) is θ = e 2 π i L 0 \theta=e^{2\pi iL_{0}} θ = e 2 π i L 0 .
(v4) L 0 L_{0} L 0 acts semisimply on V ‾ 1 \overline{V}_{1} V 1 (automatic for the VOSA V V V ).
(v5) All weights of V ‾ 0 \overline{V}_{0} V 0 are integral, i.e., θ V ‾ 0 = id \theta_{\overline{V}_{0}}=\operatorname{id} θ V 0 = id (used only for ribbon statements).
The three cases
[8 ] distinguishes: (1) Z \mathbb{Z} Z -graded VOSAs; (2) 1 2 Z \frac{1}{2}\mathbb{Z} 2 1 Z -graded VOAs; (3) 1 2 Z \frac{1}{2}\mathbb{Z} 2 1 Z -graded VOSAs whose odd part has half-integral weights (“correct statistics”). Assume V = V + ⊕ V − V=V_{+}\oplus V_{-} V = V + ⊕ V − with V − V_{-} V − a self-dual simple current, V − ⊗ V − ≅ V + V_{-}\otimes V_{-}\cong V_{+} V − ⊗ V − ≅ V + , and the product V − ⊗ V − → V + V_{-}\otimes V_{-}\to V_{+} V − ⊗ V − → V + non-zero, so that Hom ( V − ⊗ V − , V + ) \operatorname{Hom}(V_{-}\otimes V_{-},V_{+}) Hom ( V − ⊗ V − , V + ) is one-dimensional. In [4] this is derived when V + V_{+} V + is regular. In the non-semisimple case it is a starting point (in Example 6.2 it is Proposition 6.3 (2)). Then we have the following table:
Table 1.
Derivation of the table.
Let χ ∈ { 0 , 1 } \chi\in\{0,1\} χ ∈ { 0 , 1 } be the parity of V − V_{-} V − in the superalgebra V V V (χ = 1 \chi=1 χ = 1 in the VOSA cases (1), (3), χ = 0 \chi=0 χ = 0 in the VOA case (2)).
Twist. V − V_{-} V − is a simple current, thus L 0 L_{0} L 0 acts on it with a single class of weights modulo Z \mathbb{Z} Z (by (v4) it acts semisimply), and θ V − = e 2 π i L 0 = e 2 π i h \theta_{V_{-}}=e^{2\pi iL_{0}}=e^{2\pi ih} θ V − = e 2 π i L 0 = e 2 π ih where h h h is any weight of V − V_{-} V − : this is + 1 +1 + 1 if the weights are integral (case (1)) and − 1 -1 − 1 if they are half-integral (cases (2), (3)). In the two VOSA cases ε = θ V − \varepsilon=\theta_{V_{-}} ε = θ V − by definition of ε \varepsilon ε (see Definition 2.7).
Quadratic form. By (v1) the non-zero product μ : V − ⊗ V − → V + \mu:V_{-}\otimes V_{-}\to V_{+} μ : V − ⊗ V − → V + satisfies μ c V − , V − = ( − 1 ) χ χ μ = ( − 1 ) χ μ \mu c_{V_{-},V_{-}}=(-1)^{\chi}\chi\mu=(-1)^{\chi}\mu μ c V − , V − = ( − 1 ) χ χ μ = ( − 1 ) χ μ . Since c V − , V − = q id c_{V_{-},V_{-}}=q\operatorname{id} c V − , V − = q id on the simple object V − ⊗ V − ≅ V + V_{-}\otimes V_{-}\cong V_{+} V − ⊗ V − ≅ V + , this gives q = ( − 1 ) χ q=(-1)^{\chi} q = ( − 1 ) χ .
Dimension. dim V − = θ V − / q \dim V_{-}=\theta_{V_{-}}/q dim V − = θ V − / q by Remark 2.1. The last column contains the quantum dimension of V − V_{-} V − as given in [8 ] in each case. The entries of the first columns are given by (v1) and the weights. The agreement of the last column with θ / q \theta/q θ / q is a consistency check of our sign conventions c s , θ s c^{s},\theta^{s} c s , θ s against [8 ] .
In all three cases assumption of Proposition 5.1 (a) holds, i.e., q = ( − 1 ) χ q=(-1)^{\chi} q = ( − 1 ) χ . Hence, Proposition 5.1 (a) applies to H = { 1 , V − } , χ H=\{1,V_{-}\},\chi H = { 1 , V − } , χ . Assumption of Proposition 5.1 (b) holds in cases (1) and (3) (with ε = θ V − \varepsilon=\theta_{V_{-}} ε = θ V − ). Thus, Proposition 5.1 (b), (c) apply there. In case (2) χ = 0 \chi=0 χ = 0 and θ V − = − 1 ≠ ε \theta_{V_{-}}=-1\ne\varepsilon θ V − = − 1 = ε , thus assumption of Proposition 5.1 (b) fails: θ A ‾ ≠ id \theta_{\overline{A}}\ne\operatorname{id} θ A = id and C A ‾ l o c \mathcal{C}_{\overline{A}}^{\mathrm{loc}} C A loc has no ribbon structure induced from θ \theta θ (consistent with [8 ] : e 2 π i L 0 e^{2\pi iL_{0}} e 2 π i L 0 is not a module map there), see Remark 5.2.
The twist
For M = ( M 0 , M 1 ) ∈ C A s -loc M=(M_{0},M_{1})\in\mathcal{C}_{A}^{s\text{-loc}} M = ( M 0 , M 1 ) ∈ C A s -loc (an NS V V V -supermodule) the grading is the Z / 2 \mathbb{Z}/2 Z /2 -grading of the supermodule, P M P_{M} P M is the parity involution (+ 1 +1 + 1 on M 0 M_{0} M 0 , − 1 -1 − 1 on M 1 M_{1} M 1 ), and the vertex operators satisfy Y M ( v , x ) M i ⊂ M i + ∣ v ∣ ( ( x ) ) Y_{M}(v,x)M_{i}\subset M_{i+\lvert v\rvert}((x)) Y M ( v , x ) M i ⊂ M i + ∣ v ∣ (( x )) , so P M Y M ( v , x ) P M − 1 = ( − 1 ) ∣ v ∣ Y M ( v , x ) P_{M}Y_{M}(v,x)P_{M}^{-1}=(-1)^{\lvert v\rvert}Y_{M}(v,x) P M Y M ( v , x ) P M − 1 = ( − 1 ) ∣ v ∣ Y M ( v , x ) . In the NS sector the modes v n v_{n} v n of Y M ( v , x ) Y_{M}(v,x) Y M ( v , x ) are indexed by n ∈ Z n\in\mathbb{Z} n ∈ Z , so that for v v v of weight wt v \operatorname{wt}v wt v one has e 2 π i L 0 v n e − 2 π i L 0 = e 2 π i wt v v n e^{2\pi iL_{0}}v_{n}e^{-2\pi iL_{0}}=e^{2\pi i\operatorname{wt}v}v_{n} e 2 π i L 0 v n e − 2 π i L 0 = e 2 π i wt v v n .
On M = ( M 0 , M 1 ) ∈ C A s -loc M=(M_{0},M_{1})\in\mathcal{C}_{A}^{s\text{-loc}} M = ( M 0 , M 1 ) ∈ C A s -loc , θ M = θ M 0 ⊕ ε θ M 1 \theta_{M}=\theta_{M_{0}}\oplus\varepsilon\theta_{M_{1}} θ M = θ M 0 ⊕ ε θ M 1 . For ε = + 1 \varepsilon=+1 ε = + 1 this is e 2 π i L 0 e^{2\pi iL_{0}} e 2 π i L 0 . For ε = − 1 \varepsilon=-1 ε = − 1 it is e 2 π i L 0 P M e^{2\pi iL_{0}}P_{M} e 2 π i L 0 P M . In case (3) it is a V V V -module endomorphism: conjugation by e 2 π i L 0 e^{2\pi iL_{0}} e 2 π i L 0 multiplies Y ( v , x ) Y(v,x) Y ( v , x ) , v v v odd, by e 2 π i wt v = − 1 e^{2\pi i\operatorname{wt}v}=-1 e 2 π i wt v = − 1 , as does conjugation by P P P , thus the product commutes with Y ( v , x ) Y(v,x) Y ( v , x ) . On V V V , e 2 π i L 0 P V = id e^{2\pi iL_{0}}P_{V}=\operatorname{id} e 2 π i L 0 P V = id ([8 ] : θ V = P V \theta_{V}=P_{V} θ V = P V for correct statistics), as required for a ribbon structure on a category with unit V V V . Thus, the answer to the question “does θ s \theta^{s} θ s agree with the twist e 2 π i L 0 e^{2\pi iL_{0}} e 2 π i L 0 on NS supermodules?” is: for ε = + 1 \varepsilon=+1 ε = + 1 , and up to the parity operator for ε = − 1 \varepsilon=-1 ε = − 1 , where the correction is forced.
Applying the theorems
Corollary 7.1. Assume (v1)-(v4) (and (v5) for part (c)) and that A ‾ \overline{A} A is exact and haploid. Then:
(a) Braided conclusion: Rep N S ( V ) \operatorname{Rep}^{NS}(V) Rep N S ( V ) is a rigid braided finite tensor category.
(b) Super-non-degeneracy: if moreover C \mathcal{C} C is non-degenerate, then Rep N S ( V ) \operatorname{Rep}^{NS}(V) Rep N S ( V ) is super-non-degenerate and is the centralizer of a fermion in a non-degenerate braided FTC N \mathcal{N} N .
(c) Ribbon conclusion: if in addition C \mathcal{C} C is ribbon with twist θ = e 2 π i L 0 \theta=e^{2\pi iL_{0}} θ = e 2 π i L 0 (∀ 3 c \forall3c ∀3 c ), θ V 1 = ε id \theta_{V_{1}}=\varepsilon\operatorname{id} θ V 1 = ε id for a sign ε \varepsilon ε (all weights of V 1 V_{1} V 1 integral, resp. half-integral), C \mathcal{C} C is non-degenerate and A ‾ \overline{A} A is Frobenius, then A ‾ \overline{A} A is symmetric Frobenius, N \mathcal{N} N is an MTC, and Rep N S ( V ) \operatorname{Rep}^{NS}(V) Rep N S ( V ) is ribbon with the twist θ s \theta^{s} θ s and ε \varepsilon ε -super-modular.
Proof. (a) is Theorem 1.5 (0) with (v2). (b) is Theorem 1.5 (1). (c): by (v5) and the assumption on V 1 V_{1} V 1 , θ A ‾ = id \theta_{\overline{A}}=\operatorname{id} θ A = id and θ A 1 = ε id \theta_{A_{1}}=\varepsilon\operatorname{id} θ A 1 = ε id , i.e., θ A ‾ s = id \theta_{\overline{A}}^{s}=\operatorname{id} θ A s = id (Proposition 2.6 (c)). A Frobenius commutative algebra with θ A ‾ s = id \theta_{\overline{A}}^{s}=\operatorname{id} θ A s = id is symmetric by Proposition 2.10 (iii), and Theorem 1.5 (2) applies. □
Lemma 7.2. Assume (v1)-(v3), Hom ( 1 , V ‾ 0 ) = k \operatorname{Hom}(1,\overline{V}_{0})=\mathbb{k} Hom ( 1 , V 0 ) = k , and that V V V is simple as a V V V -supermodule (no proper nonzero graded submodule). Then A ‾ \overline{A} A is a simple algebra in B \mathcal{B} B , hence exact.
Proof. We use the definition of [5] (also [21]): A ‾ \overline{A} A is simple if its only two-sided ideals (sub-bimodules of A ‾ \overline{A} A ) are 0 0 0 and A ‾ \overline{A} A . Let I ⊂ A ‾ I\subset\overline{A} I ⊂ A be any right submodule, with inclusion j j j , j a I = μ ( j ⊗ id ) j a_{I}=\mu(j\otimes\operatorname{id}) j a I = μ ( j ⊗ id ) . We show that I I I is local: with mon I = c A , I c I , A \operatorname{mon}_{I}=c_{A,I}c_{I,A} mon I = c A , I c I , A , naturality of c c c gives j a I mon I = μ ( j ⊗ id ) mon I = μ mon A ‾ , A ( j ⊗ id ) = μ ( j ⊗ id ) = j a I j a_{I}\operatorname{mon}_{I}=\mu(j\otimes\operatorname{id})\operatorname{mon}_{I}=\mu\operatorname{mon}_{\overline{A},A}(j\otimes\operatorname{id})=\mu(j\otimes\operatorname{id})=j a_{I} j a I mon I = μ ( j ⊗ id ) mon I = μ mon A , A ( j ⊗ id ) = μ ( j ⊗ id ) = j a I (A ‾ \overline{A} A is local over itself, by commutativity: μ c 2 = μ \mu c^{2}=\mu μ c 2 = μ ), and j j j is a monomorphism, thus a I mon I = a I a_{I}\operatorname{mon}_{I}=a_{I} a I mon I = a I . By (v2) I I I is a graded V V V -submodule of V V V , thus I ∈ { 0 , V } I\in\{0,V\} I ∈ { 0 , V } . Hence, the only right submodules of A ‾ \overline{A} A are 0 , A ‾ 0,\overline{A} 0 , A . This is stronger than what is needed, since every two-sided ideal is in particular a right submodule, thus the only two-sided ideals are 0 , A ‾ 0,\overline{A} 0 , A and A ‾ \overline{A} A is simple. By Lemma 2.12 [6], valid for algebras in finite tensor categories. Here B \mathcal{B} B is an FTC (by (v3b)), A ‾ \overline{A} A is exact. □
The next corollary concerns the fusion case.
Corollary 7.3. Assume (v1)–(v3), C \mathcal{C} C a braided fusion category (e.g., U U U strongly rational), V V V simple, Hom C ( 1 , V ‾ 0 ) = k \operatorname{Hom}_{\mathcal{C}}(\mathbf{1},\overline{V}_{0})=\mathsf{k} Hom C ( 1 , V 0 ) = k . Then C A s -loc = B A ‾ l o c \mathcal{C}^{s\text{-loc}}_{A}=\mathcal{B}^{\mathrm{loc}}_{\overline{A}} C A s -loc = B A loc is a braided fusion category.
Proof. B = C ⊠ s V e c t \mathcal{B}=\mathcal{C}\boxtimes\mathrm{sVect} B = C ⊠ sVect is fusion and A ‾ \overline{A} A is a simple commutative algebra in it (first part of Lemma 7.2 ), thus [9 ] Lemma 2.13 applies. □ \square □
This is one categorical ingredient (for the even part V ‾ 0 \overline{V}_{0} V 0 strongly rational) of the strong-rationality statement for extensions announced in [9 ] ; we do not claim the strong-rationality statement.
Witt classes
Let W n s \mathcal{W}^{\mathrm{ns}} W ns be the Witt group of non-degenerate braided FTCs, with Witt equivalence as in Lemma 2.15 : C ∼ D \mathcal{C}\sim\mathcal{D} C ∼ D iff C ⊠ Z ( A ) ≃ D ⊠ Z ( B ) \mathcal{C}\boxtimes\mathcal{Z}(\mathcal{A})\simeq\mathcal{D}\boxtimes\mathcal{Z}(\mathcal{B}) C ⊠ Z ( A ) ≃ D ⊠ Z ( B ) for some FTCs A , B \mathcal{A},\mathcal{B} A , B . We write W n s \mathcal{W}^{\mathrm{ns}} W ns additively. We use only the following consequence of [26 ] (see Lemma 2.15 ): if C \mathcal{C} C is non-degenerate and A ∈ C A\in\mathcal{C} A ∈ C is a haploid commutative exact algebra, then C A l o c \mathcal{C}^{\mathrm{loc}}_{A} C A loc is non-degenerate and Witt equivalent to C \mathcal{C} C .
Proposition 8.1. Let C \mathcal{C} C be a non-degenerate braided FTC and A A A a haploid supercommutative superalgebra with A ‾ \overline{A} A exact. Then N = B ι ( A ‾ ) l o c \mathcal{N}=\mathcal{B}^{\mathrm{loc}}_{\iota(\overline{A})} N = B ι ( A ) loc is non-degenerate and [ C ] = [ N ] [\mathcal{C}]=[\mathcal{N}] [ C ] = [ N ] in W n s \mathcal{W}^{\mathrm{ns}} W ns .
Proof. B ~ = C ⊠ T ε \widetilde{\mathcal{B}}=\mathcal{C}\boxtimes\mathcal{T}^{\varepsilon} B = C ⊠ T ε is non-degenerate (§ 4.4 ) and ι ( A ‾ ) \iota(\overline{A}) ι ( A ) is haploid, commutative and exact by Lemma 4.2 . Thus, [ B ~ ] = [ N ] [\widetilde{\mathcal{B}}]=[\mathcal{N}] [ B ] = [ N ] by Lemma 2.15 . Next, T = Z ( Vect Z / 2 ) \mathcal{T}=\mathcal{Z}(\operatorname{Vect}_{\mathbb{Z}/2}) T = Z ( Vect Z /2 ) , thus T ⊠ Z ( Vect ) ≃ Vect ⊠ Z ( Vect Z / 2 ) \mathcal{T}\boxtimes\mathcal{Z}(\operatorname{Vect})\simeq\operatorname{Vect}\boxtimes\mathcal{Z}(\operatorname{Vect}_{\mathbb{Z}/2}) T ⊠ Z ( Vect ) ≃ Vect ⊠ Z ( Vect Z /2 ) and [ T ] = [ Vect ] = 0 [\mathcal{T}]=[\operatorname{Vect}]=0 [ T ] = [ Vect ] = 0 , hence [ B ~ ] = [ C ] + [ T ] = [ C ] [\widetilde{\mathcal{B}}]=[\mathcal{C}]+[\mathcal{T}]=[\mathcal{C}] [ B ] = [ C ] + [ T ] = [ C ] . The choice of ε \varepsilon ε is irrelevant: T + \mathcal{T}^{+} T + and T − \mathcal{T}^{-} T − have the same underlying braided category (§ 4.1 ), and W n s \mathcal{W}^{\mathrm{ns}} W ns is defined for braided categories. □ \square □
We do not define super Witt group here: it would require the fermionic product ( D 1 ⊠ D 2 ) 1 ⊕ f 1 ⊠ f 2 l o c (\mathcal{D}_{1}\boxtimes\mathcal{D}_{2})^{\mathrm{loc}}_{1\oplus f_{1}\boxtimes f_{2}} ( D 1 ⊠ D 2 ) 1 ⊕ f 1 ⊠ f 2 loc (condensation of the boson f 1 ⊠ f 2 f_{1}\boxtimes f_{2} f 1 ⊠ f 2 ), a proof that it is well defined on classes, and non-semisimple analogues of [11 ] .
Comparison with existing formulations
The remaining open questions are the super Witt group, other modular completions, the Ramond sector, odd morphisms for modules beyond Remark 4.4, and the material of § 4.3 of [9 ] . The material of this paper is also useful in computations of topological invariants in non-commutative geometry setup in condensed matter theory [6 , 28 , 29 , 30 , 22 ] , and computations of characteristic classes in foliation theory [2 , 3 , 4 , 31 , 32 ] . Next we compare our formulation with those of [8 ] and [9 ] .
Creutzig-Kanade-McRae
The comparison concerns the introduction and Sections 2.1–2.2 of [8 ] .
The category. [8 ] defines S C \mathcal{SC} SC with objects W = ( W 0 ‾ , W 1 ‾ ) W=(W^{\overline{0}},W^{\overline{1}}) W = ( W 0 , W 1 ) and Hom S C ( W 1 , W 2 ) = Hom C ( W 1 0 ‾ ⊕ W 1 1 ‾ , W 2 0 ‾ ⊕ W 2 1 ‾ ) \operatorname{Hom}_{\mathcal{SC}}(W_{1},W_{2})=\operatorname{Hom}_{\mathcal{C}}(W_{1}^{\overline{0}}\oplus W_{1}^{\overline{1}},W_{2}^{\overline{0}}\oplus W_{2}^{\overline{1}}) Hom SC ( W 1 , W 2 ) = Hom C ( W 1 0 ⊕ W 1 1 , W 2 0 ⊕ W 2 1 ) , graded into even and odd morphisms. Its underlying category has the even morphisms [8 ] , and an odd morphism W 1 → W 2 W_{1}\to W_{2} W 1 → W 2 is an even morphism W 1 → Π W 2 W_{1}\to\Pi W_{2} W 1 → Π W 2 . The tensor product is ( W 1 ⊠ W 2 ) i = ⨁ i 1 + i 2 = i W 1 i 1 ⊠ W 2 i 2 (W_{1}\boxtimes W_{2})^{i}=\bigoplus_{i_{1}+i_{2}=i}W_{1}^{i_{1}}\boxtimes W_{2}^{i_{2}} ( W 1 ⊠ W 2 ) i = ⨁ i 1 + i 2 = i W 1 i 1 ⊠ W 2 i 2 and the braiding is ⨁ ( − 1 ) i 1 i 2 R W 1 i 1 , W 2 i 2 \bigoplus(-1)^{i_{1}i_{2}}\mathcal{R}_{W_{1}^{i_{1}},W_{2}^{i_{2}}} ⨁ ( − 1 ) i 1 i 2 R W 1 i 1 , W 2 i 2 [8 ] . These are the definitions of § 2.3 : B \mathcal{B} B is the underlying category of S C \mathcal{SC} SC , with Π = f ⊗ − \Pi=f\otimes- Π = f ⊗ − .
Algebras and modules. A superalgebra in C \mathcal{C} C is an algebra in S C \mathcal{SC} SC with even structure maps, and Rep 0 A \operatorname{Rep}^{0}A Rep 0 A is the category of local modules with even structure maps [8 ] . Its underlying category is B A ‾ l o c \mathcal{B}^{\mathrm{loc}}_{\overline{A}} B A loc by Proposition 2.6.
Twist. [8 ] states that θ V = e 2 π i L 0 \theta_{V}=e^{2\pi iL_{0}} θ V = e 2 π i L 0 and the parity involution P V P_{V} P V coincide for correct statistics and differ for incorrect statistics. In our notation these are the cases ε = − 1 \varepsilon=-1 ε = − 1 and ε = + 1 \varepsilon=+1 ε = + 1 of the single condition θ A 1 = ε id \theta_{A_{1}}=\varepsilon\operatorname{id} θ A 1 = ε id (see § 7 ). Note that the braiding signs agree. The dimensions qdim V − \operatorname{qdim}V_{-} qdim V − in the three cases agree with θ / q \theta/q θ / q computed from our data (see § 7.1 ). NS modules are local and Ramond modules are twisted (non-local), as in Example 6.1.
Odd morphisms. The notion “odd morphism equivalent to even morphism into Π \Pi Π ” is stated in [8 ] for S C \mathcal{SC} SC . For modules we use it as the definition of D s \mathcal{D}^{s} D s (see Remark 4.4). Our Theorems 1.5 (0)–(2) concern possibly non-semisimple C \mathcal{C} C and exact (not necessarily separable) A ‾ \overline{A} A , and give the Müger-center statement for the underlying category of Rep 0 A \operatorname{Rep}^{0}A Rep 0 A .
Creutzig-McRae-Shimizu-Yadav
The comparison concerns the introduction and Sections 1–3 of [9 ] (through Section 3.5), and the two statements Theorem 2.23 and Theorem 2.26 used in § 2.6 . § 4.3 is mentioned below only through the introduction.
The correspondence between left and right modules. [9 ] use left modules ( M , μ M : A ⊗ M → M ) (M,\mu_M:A\otimes M\to M) ( M , μ M : A ⊗ M → M ) , with C A l o c = { μ M c M , A c A , M = μ M } C_A^{\mathrm{loc}}=\{\mu_M c_{M,A}c_{A,M}=\mu_M\} C A loc = { μ M c M , A c A , M = μ M } , and ⊗ A \otimes_A ⊗ A defined through c A , − − 1 c^{-1}_{A,-} c A , − − 1 . Let R R R send a right module ( M , a M ) (M,a_M) ( M , a M ) to the left module ( M , λ M ) (M,\lambda_M) ( M , λ M ) with λ M = a M c A , M \lambda_M=a_Mc_{A,M} λ M = a M c A , M (the left action of § 4.3 ), and be the identity on morphisms. R R R is an isomorphism of categories from right to left A A A -modules, with inverse a M = λ M c A , M − 1 a_M=\lambda_Mc^{-1}_{A,M} a M = λ M c A , M − 1 . Substituting a M = λ M c A , M − 1 a_M=\lambda_Mc^{-1}_{A,M} a M = λ M c A , M − 1 one finds that a M c A , M c M , A = a M a_Mc_{A,M}c_{M,A}=a_M a M c A , M c M , A = a M iff λ M c M , A = λ M c A , M − 1 \lambda_Mc_{M,A}=\lambda_Mc^{-1}_{A,M} λ M c M , A = λ M c A , M − 1 which is equivalent to λ M c M , A c A , M = λ M \lambda_Mc_{M,A}c_{A,M}=\lambda_M λ M c M , A c A , M = λ M , thus R R R restricts to an isomorphism between our category of local right modules and the category of local left modules of [9 ] . For local N N N , λ N = a N c A , N = a N c N , A − 1 \lambda_N=a_Nc_{A,N}=a_Nc^{-1}_{N,A} λ N = a N c A , N = a N c N , A − 1 (§ 4.3 ). Thus, the left action associated with a local module does not depend on whether c c c or c − 1 c^{-1} c − 1 is used to pass between right and left actions, the tensor products ⊗ A \otimes_A ⊗ A defined with c c c and with c − 1 c^{-1} c − 1 coincide on local modules, and we use in this paper the braiding induced by c c c , as in [24 ] .
Results used: for a braided FTC C C C and A A A exact commutative haploid, C A C_A C A is a finite tensor category and C A l o c C_A^{\mathrm{loc}} C A loc a braided FTC, non-degenerate if C C C is [9 ] . Applied to ( B , A ‾ ) (\mathcal{B},\overline{A}) ( B , A ) this is our Theorem 1.5 (0) and the finite-tensor-category part of (1). For a simple commutative algebra in a fusion category, [9 ] gives Corollary 7.3.
The conditions θ A 2 = i d \theta^2_A=\mathrm{id} θ A 2 = id and θ A = i d \theta_A=\mathrm{id} θ A = id . [26 ] show that D ( A ) = A ∗ D(A)=A^* D ( A ) = A ∗ is local iff θ A 2 = i d \theta^2_A=\mathrm{id} θ A 2 = id [26 ] . This is the condition under which, according to the introduction of [9 ] , C A l o c C_A^{\mathrm{loc}} C A loc is a braided Grothendieck–Verdier category with dualizing object D ( A ) D(A) D ( A ) . For a ribbon structure one needs the stronger condition θ A = i d \theta_A=\mathrm{id} θ A = id together with D ( A ) ∈ ( C A l o c ) ′ D(A)\in(C_A^{\mathrm{loc}})' D ( A ) ∈ ( C A loc ) ′ [26 ] (Proposition 2.8 (iii)). In our setting the relevant twist is θ s \theta^s θ s of B \mathcal{B} B , and the conditions are ( θ A ‾ s ) 2 = i d (\theta^s_{\overline{A}})^2=\mathrm{id} ( θ A s ) 2 = id , i.e., θ A 0 2 = i d \theta^2_{A_0}=\mathrm{id} θ A 0 2 = id and θ A 1 2 = i d \theta^2_{A_1}=\mathrm{id} θ A 1 2 = id (automatic if θ A i = ± 1 \theta_{A_i}=\pm1 θ A i = ± 1 ), resp. θ A ‾ s = i d \theta^s_{\overline{A}}=\mathrm{id} θ A s = id (Proposition 2.6 (c)).
Section 4.3. The introduction of [9 ] states that § 4.3 treats extensions V ⊆ A V\subseteq A V ⊆ A with V V V a Z \mathbb{Z} Z -graded VOA and A A A a Z \mathbb{Z} Z - or 1 2 Z \frac{1}{2}\mathbb{Z} 2 1 Z -graded VOSA containing V V V in its even part, and that Theorem 4.21 is a VOSA generalization of Theorem 1.1 there (a simple Z ≥ 0 \mathbb{Z}_{\geq0} Z ≥ 0 -graded extension of a strongly rational VOA is strongly rational). Corollary 7.3 gives one corresponding categorical ingredient in our formalism.
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 S. Kanade and P. Zusmanovich for valuable 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, and drafting. Revision and writing of final text have been done by the author. The author takes full responsibility for the mathematical content and the final text.
Conflict of interest. The author has no conflicts of interest to declare that are relevant to the content of this article.
References
@page 15/16 starts_mid=0 ends_mid=1
Creutzig-McRae-Shimizu-Yadav
The comparison concerns the introduction and Sections 1–3 of [9 ] (through Section 3.5), and the two statements Theorem 2.23 and Theorem 2.26 used in § 2.6 . § 4.3 is mentioned below only through the introduction.
The correspondence between left and right modules. [9 ] use left modules ( M , μ M : A ⊗ M → M ) (M,\mu_M:A\otimes M\to M) ( M , μ M : A ⊗ M → M ) , with C A l o c = { μ M c M , A c A , M = μ M } C_A^{\mathrm{loc}}=\{\mu_M c_{M,A}c_{A,M}=\mu_M\} C A loc = { μ M c M , A c A , M = μ M } , and ⊗ A \otimes_A ⊗ A defined through c A , − − 1 c^{-1}_{A,-} c A , − − 1 . Let R R R send a right module ( M , a M ) (M,a_M) ( M , a M ) to the left module ( M , λ M ) (M,\lambda_M) ( M , λ M ) with λ M = a M c A , M \lambda_M=a_Mc_{A,M} λ M = a M c A , M (the left action of § 4.3 ), and be the identity on morphisms. R R R is an isomorphism of categories from right to left A A A -modules, with inverse a M = λ M c A , M − 1 a_M=\lambda_Mc^{-1}_{A,M} a M = λ M c A , M − 1 . Substituting a M = λ M c A , M − 1 a_M=\lambda_Mc^{-1}_{A,M} a M = λ M c A , M − 1 one finds that a M c A , M c M , A = a M a_Mc_{A,M}c_{M,A}=a_M a M c A , M c M , A = a M iff λ M c M , A = λ M c A , M − 1 \lambda_Mc_{M,A}=\lambda_Mc^{-1}_{A,M} λ M c M , A = λ M c A , M − 1 which is equivalent to λ M c M , A c A , M = λ M \lambda_Mc_{M,A}c_{A,M}=\lambda_M λ M c M , A c A , M = λ M , thus R R R restricts to an isomorphism between our category of local right modules and the category of local left modules of [9 ] . For local N N N , λ N = a N c A , N = a N c N , A − 1 \lambda_N=a_Nc_{A,N}=a_Nc^{-1}_{N,A} λ N = a N c A , N = a N c N , A − 1 (§ 4.3 ). Thus, the left action associated with a local module does not depend on whether c c c or c − 1 c^{-1} c − 1 is used to pass between right and left actions, the tensor products ⊗ A \otimes_A ⊗ A defined with c c c and with c − 1 c^{-1} c − 1 coincide on local modules, and we use in this paper the braiding induced by c c c , as in [24 ] .
Results used: for a braided FTC C C C and A A A exact commutative haploid, C A C_A C A is a finite tensor category and C A l o c C_A^{\mathrm{loc}} C A loc a braided FTC, non-degenerate if C C C is [9 ] . Applied to ( B , A ‾ ) (\mathcal{B},\overline{A}) ( B , A ) this is our Theorem 1.5 (0) and the finite-tensor-category part of (1). For a simple commutative algebra in a fusion category, [9 ] gives Corollary 7.3.
The conditions θ A 2 = i d \theta^2_A=\mathrm{id} θ A 2 = id and θ A = i d \theta_A=\mathrm{id} θ A = id . [26 ] show that D ( A ) = A ∗ D(A)=A^* D ( A ) = A ∗ is local iff θ A 2 = i d \theta^2_A=\mathrm{id} θ A 2 = id [26 ] . This is the condition under which, according to the introduction of [9 ] , C A l o c C_A^{\mathrm{loc}} C A loc is a braided Grothendieck–Verdier category with dualizing object D ( A ) D(A) D ( A ) . For a ribbon structure one needs the stronger condition θ A = i d \theta_A=\mathrm{id} θ A = id together with D ( A ) ∈ ( C A l o c ) ′ D(A)\in(C_A^{\mathrm{loc}})' D ( A ) ∈ ( C A loc ) ′ [26 ] (Proposition 2.8 (iii)). In our setting the relevant twist is θ s \theta^s θ s of B \mathcal{B} B , and the conditions are ( θ A ‾ s ) 2 = i d (\theta^s_{\overline{A}})^2=\mathrm{id} ( θ A s ) 2 = id , i.e., θ A 0 2 = i d \theta^2_{A_0}=\mathrm{id} θ A 0 2 = id and θ A 1 2 = i d \theta^2_{A_1}=\mathrm{id} θ A 1 2 = id (automatic if θ A i = ± 1 \theta_{A_i}=\pm1 θ A i = ± 1 ), resp. θ A ‾ s = i d \theta^s_{\overline{A}}=\mathrm{id} θ A s = id (Proposition 2.6 (c)).
Section 4.3. The introduction of [9 ] states that § 4.3 treats extensions V ⊆ A V\subseteq A V ⊆ A with V V V a Z \mathbb{Z} Z -graded VOA and A A A a Z \mathbb{Z} Z - or 1 2 Z \frac{1}{2}\mathbb{Z} 2 1 Z -graded VOSA containing V V V in its even part, and that Theorem 4.21 is a VOSA generalization of Theorem 1.1 there (a simple Z ≥ 0 \mathbb{Z}_{\geq0} Z ≥ 0 -graded extension of a strongly rational VOA is strongly rational). Corollary 7.3 gives one corresponding categorical ingredient in our formalism.
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 S. Kanade and P. Zusmanovich for valuable 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, and drafting. Revision and writing of final text have been done by the author. The author takes full responsibility for the mathematical content and the final text.
Conflict of interest. The author has no conflicts of interest to declare that are relevant to the content of this article.
References
Institute of Mathematics, Czech Academy of Sciences, Žitná 25, Prague, Czech Republic
Email address: zuevsky@yahoo.com