Braided zesting and its applications (Q2042356)

There are several known ways of constructing new (braided) fusion categories out of old ones. This paper is concerned with a recently developed such construction, called \textit{zesting}. As a construction technique it appeared in [\textit{P. Bruillard} et al., Can. Math. Bull. 57, No. 4, 721--734 (2014; Zbl 1342.18013)] with the purpose to categorify certain fusion rule algebra. Then expected modular structure of the zested fusion category obtained in this way is proved in the present paper. More importantly, the general theory of associative zesting for fusion categories (and further for braided, twist and ribbon zestings for categories possessing these additional structures) is developed. Starting with a \(G\)-graded fusion category \(\mathcal{C}=\sum_{g\in G}\mathcal{C}_g\) for a group \(G\), the idea of zesting is to define a new tensor product \[ X_g\stackrel{\lambda}{\otimes} Y_h:=(X_g\otimes Y_h)\otimes\lambda(g,h),\tag{1} \] where \(X_g\in\mathcal{C}_g, Y_h\in\mathcal{C}_h\) are simple objects in their corresponding graded components, and \(\lambda(g,h)\in\mathcal{C}_e\cap\mathcal{C}_{pt}\), that is, it is an invertible object in the trivial component. The zesting of \(\mathcal{C}\) is thus in particular a \(G\)-graded extension of \(\mathcal{C}_e\). For a given choice of \(\lambda:G\times G\to\mathcal{C}_e\cap\mathcal{C}_{pt}\), there are obstructions to \((\mathcal{C}, \stackrel{\lambda}{\otimes}, 1)\) admitting a monoidal category structure. That is, when these obstructions vanish, there are choices of associativity constraints. Fixing one such associative zesting, one can moreover study the existence of braiding (leading to \textit{braided zesting}), which leads to new obstructions and choices. Moreover, for a fixed braided zesting of a ribbon fusion category one may search for a ribbon twist structure leading to \textit{ribbon zesting}. Before describing this process in more details, we observe that in the literature two related constructions can be found. The first one is gauging. If \(\mathcal{C}\) is modular and \(\mathcal{C}_{pt}\cap\mathcal{C}_e\cong\text{Rep}(G)\) is Tannakian with \(G\) abelian, one may set \(\mathcal{B}=(\mathcal{C}_e)_G\) the \(G\)-de-equivariantization of the trivial component, then any modular zesting \(\tilde{\mathcal{C}}\) will be a \(G\)-gauging of \(\mathcal{B}\). In the other related construction, if \(\mathcal{C}_{pt}\cap\mathcal{C}_e\cong\text{Rep}(G)\) as above, one may construct new categories as tensor products over \(G\) by condensing the diagonal algebra in \(\mathcal{C}\boxtimes\text{Rep}(D^\omega G)\). In comparision with the latter two constructions, zesting has several features that these do not, like having explicit fusion rules, producing new categories not necessarily modular or braided, zesting only depends on essentially cohomological choices, and if the resulting category is modular there are explicit formulas for the modular data. We now describe zesting more specifically, without entering into full details. Being more specific, but not entering into full details, Given a faithfully \(G\)-graded fusion category, an \textit{associative \(G\)-zesting} consists of a map \(\lambda: G\times G\to \mathcal{C}_e\cap\mathcal{C}_{pt}\), and for all \(g,h,k\in G\) an isomorphism \(\lambda_{g,h,k}\) satisfying certain equation, called \textit{associative zesting constraint}, which is satisfied for every four elements \(g,h,k,l\in G\), so that both \(\lambda\)'s satisfy normalization conditions. In short, \(\lambda: G\times G\to \mathcal{C}_e\cap\mathcal{C}_{pt}\) and \(\lambda_{g,h,k}\) are specific normalized 2- and 3-cocycle, respectively. In Proposition 3.4 it is proved that under these conditions (1) equips \(\mathcal{C}^\lambda=(\mathcal{C}, \stackrel{\lambda}{\otimes}, 1)\) with a new faithfully \(G\)-graded fusion category structure. In particular, for modular \(\mathbb Z/N\)-graded category \(SU(N)_k\), taking the trivial 2-cocycle \(\lambda\), the associative zesting constraint means that \(\lambda_{g,h,k}\) is a normalized 3-cocycle on \(\mathbb Z/N\), implying that such associative zestings of \(SU(N)_k\) recover those obtained in [\textit{D. Kazhdan} and \textit{H. Wenzl}, ``Reconstructing monoidal categories'', Adv. Soviet Math. 16, 111--136 (1993)] by twisting the associativity morphisms. Dual objects for \(\stackrel{\lambda}{\otimes}\) are defined in Section 3.3 leading to rigid associative zestings. For a braided faithfully \(G\)-graded fusion category \(\mathcal{B}\), where \(G\) is an abelian group, given an associative zesting \(\lambda\) a \textit{braided zesting} is defined by an isomorphism \(t_{g,h}\) associated to \(\lambda\) for each \(g,h\in G\) and a function \(j: G\to\text{Aut}_\otimes(\text{Id}_{\mathcal{B}})\) to the abelian group of tensor natural isomorphisms of the identity functor, so that \((\lambda, j,t)\) satisfy braided zesting conditions (BZ1) and (BZ2) and normalization conditions. Under these conditions, in Proposition 4.4 a braiding for the fusion category \(\mathcal{B}^\lambda\) is given. For a premodular tensor category \(\mathcal{B}\) with a braided zesting \((\mathcal{B}^\lambda, t)\) denotes the corresponding braided fusion category. If one takes trivial 2- and 3-cocycles \(\lambda\), then the braided zestings are simply modifying the braiding by a bicharacter, which is a well-known process. Similarity and equivalence of braided zestings is defined, so that in Proposition 4.7 three claims are proved of the following unifying form. The set of (equivalence classes of) all braided \(G\)-zestings of the form, or similar to, \((\lambda, j,t)\) is the torsor over certain groups. First and second partial obstructions to the existence of a function \(j: G\to\text{Aut}_\otimes(\text{Id}_{\mathcal{B}})\) satisfying (BZ1) are described in Section 4.2. These obstructions automatically vanish when \(G\) is the universal grading (the notion introduced in [\textit{S. Gelaki} and \textit{D. Nikshych}, Adv. Math. 217, No. 3, 1053--1071 (2008; Zbl 1168.18004)]). For a fixed associative zesting \(\lambda\) such that the first and second partial obstructions vanish, in order to solve the equation (BZ1) certain 2-cocycle \(O_1(\lambda):G\times G\to k^\times\) is defined and it is shown in Lemma 4.12 that (BZ1) can be solved if and only if the cohomology class of \(O_1(\lambda)\) vanishes. Assuming that the latter cohomology class vanishes, certain 2-cocycle \(O_2(\lambda)\) is defined. In Theorem 4.15 it is proved that given an associative zesting \(\lambda\), there is to it associated braided zesting if and only if the cohomology class of \(O_1(\lambda)\) vanishes and \(O_2(\lambda)\) has a specific computed form. Starting from a finite abelian group \(G\), a faithfully \(G\)-graded braided tensor category \(\mathcal{B}\) with twist \(\theta\), a braided zesting \((\lambda, j,t)\) and a function \(f:G\to l^\times\), in Proposition 5.1 a twist \(\theta^f\) for \((\mathcal{B}^\lambda, t)\) is characterized, as well as when it is a ribbon twist if so is \(\theta\) for \(\mathcal{B}\). A quadruple \((\lambda, j, t, f )\) where \(f\) satisfies these two characterizations is called a \textit{ribbon zesting}, and \((\mathcal{B}^\lambda, t, f)\) denotes the twist (ribbon) zesting. It is not known if twists or ribbons for braided zesting always exist, though in Remark 5.3 an easy to check requirement for their existence is provided. For a ribbon zesting \((\lambda, j, t, f )\) in Proposition 5.5 a quantum trace of an endomorphism in \((\mathcal{B}^\lambda, t, f)\) is given, while in Theorem 5.7 the \(T\)- and \(S\)-matrix from the modular data of \((\mathcal{B}^\lambda, t, f)\) are computed. In Section 5.2 it is shown that Müger center is not zesting invariant and in Proposition 5.12 sufficient conditions in order for the Müger center of a braided fusion category \(\mathcal{B}\) and of \((\mathcal{B}^\lambda, t)\) to coincide are given. In Section 5.3 the associated braid group representations of a braided fusion category are given, and in Theorem 5.15 the actions of the braid group representations are computed. Section 6 deals with applications to the author's main case of interest, which is for a modular category \(\mathcal{B}\) and \(G=U(\mathcal{B})\) the universal grading. In this case \(\mathcal{B}_e\cap\mathcal{B}_{pt}\) is a symmetric pointed fusion category, that is, it is of the form \(Vec_S\) with \(S\) an abelian group and the braiding is given by the twist or minus twist. Given a \(G\)-graded braided fusion category \(\mathcal{B}\) such that \(Vec_S\subset(\mathcal{B}_e)_{pt}\), in Proposition 6.1 the obstruction to the existence of an associative zesting is computed, so that it vanishes if the order of \(S\) is odd. In Section 6.2.2 for a braided pointed fusion category \(\mathcal{C}(C,\Theta)\) where \(\text{Inv}(\mathcal{B})=C\) (the group of isomorphism classes of invertible objects in \(\mathcal{B}\) under the tensor product) is a cyclic group \(C\) and \(\Theta\) is a ribbon structure, it is characterized in terms of \(\Theta\) when \(\mathcal{C}(C,\Theta)\) is modular, symmetric, Tannakian, and super-Tanankian. In the symmetric case the ribbon is a character and the braiding is the twist or minus twist. Proposition 6.3 gives a parametrization of the equivalence classes of associative zestings of a braided fusion category \(\mathcal{B}\) under certain conditions involving a cyclic group \(C\). Furthermore, Proposition 6.4 characterizes when a pair \((\lambda_a,\lambda_b)\) of associative zestings constructed in the previous Proposition admits a braided zesting with \(j=id\). In this case \(N\) different braided zestings are given explicitly, where \(N\) is the order of \(C\). Being \((\lambda_a,\lambda_b, j=id, t_s)\) a braided zesting constructed in Proposition 6.4, for \(s=1,\dots,N\), and if \(\mathcal{B}\) has a ribbon twist obeying a certain condition, in Proposition 6.5 ribbon zesting and modular data \(T\) and \(S\) are constructed. In Section 6.3 it is explained how modular categories \(SU(N)_k\), which are obtained from quantum groups \(U_Q\mathfrak{sl}_N\) for \(Q=e^{\pi i/(N+k)}\), fit the setting of Section 6.2.2 with \(C=\mathbb Z/N\), then the results of Section 6.1 are applied to \(SU(N)_k\). Concretely, the category \(SU(N)_k\) has a (maximally) pointed subcategory \(\mathcal{P}(N, k)\) with fusion rules like \(\mathbb Z/N\), and \(SU(N)_k\) is (universally) \(\mathbb Z/N\)-graded. The subcategory \(\mathcal{P}(N, k)\) can be identified with the pointed ribbon fusion category \(\mathcal{C}(\mathbb Z/N,\eta)\), where \(\eta\) is a quadratic form given by the twists, which are computed using standard techniques. It is argued that \(\mathcal{P}(N, k)\) is modular if and only if \((k,N)=1\). On the other hand, if \(k=\alpha N\), the following is found. If \(N\) is odd, the categories \(\mathcal{P}(N, \alpha N)\) are Tannakian, but if \(N\) is even, \(\mathcal{P}(N, \alpha N)\) is Tannakian if and only if \(\alpha\) is even, and super-Tannakian otherwise. Moreover, if \(N\) is odd, there are at most \(N^2\) distinct ribbon twist braided zestings, all of which are modular, while if \(N\) is even and \(\alpha\) odd there are at most \(4N^2\) ribbon braided zestings, each of which is modular. It is pointed out that there can be equivalences among braided zestings, which is illustrated by some examples. For \(SU(3)_3\) it is argued that from at most 9 modular categories obtained from \(\mathbb Z/3\)-zesting of \(SU(3)_3\) there are only 3 inequivalent sets of modular data, and presumably only 3 inequivalent modular categories (this is not immediate as modular data is not a complete invariant). Zesting of \(SU(3)_3\) is compared to gauging construction, and consistently 3 gaugings are obtained. For \(SU(4)_4\) eight associative zestings are found, each of them admitting 4 braided zestings, and each of these a unique ribbon structure which is also modular. Thus there are at most 32 distinct modular categories obtained as \(\mathbb Z/4\)-zestings of \(SU(4)_4\). Also \(\mathbb Z/2\)-zestings of \(SU(4)_4\) are considered, in which case for the whole of two associative zestings it is proved that no braided zesting exists. Finally for \(SU(4)_2\) also 8 associative zestings are found, there are 16 braided zestings and 32 ribbon zestings. In Section 7 the authors explain the difference between their braided zestings and those studied in [\textit{A. Davydov} and \textit{D. Nikshych}, ``Braided Picard groups and graded extensions of braided tensor categories'', Preprint, \url{arXiv:2006.08022}]. In the latter reference some particular braided zestings are interpreted as deformations of a braided monoidal 2-functor and its obstruction as an element in an Eilenberg-MacLane cohomology group. This is the obstruction for a symmetric 2-cocycle to admit a braided zesting, while the cohomological obstructions \(O_1(\lambda)\) and \(O_2(\lambda)\) in Theorem 4.15 in the present article are the obstructions that a \textit{fixed associative zesting admits a braided zesting}. Thus, to compute the EM-obstruction one needs first to describe the associative zestings and then check that braided zesting obstructions (BZ1) and (BZ2) vanish. The article finishes with prospects for future applications: a) apply associative zesting to fusion categories that do not admit a braiding and braided zesting with respect to non-universal grading groups and non-cyclic grading groups, b) it is worth investigating if zesting has a meaningful physical interpretation, as it is known that symmetry gauging corresponds to phase transitions of topological phases of matter.