Disentangling the Generalized Double Semion model (Q2216770)

There are not very many known constructions of gapped quantum phases of matter that work in arbitrary spacetime dimension. One of the most famous is (a lattice Hamiltonian version of) the construction of Dijkgraaf and Witten, also known as a ``group cohomological'' model: it requires as its input a finite group \(G\) and (a cocycle representative of) a class in \(\mathrm{H}^{n+1}(G; \mathbb{R}/\mathbb{Z})\), where \(n+1\) is the spacetime dimension. Of this famous family, the preeminent member is surely the Toric Code, corresponding to the group \(G = \mathbb{Z}_2\) and the trivial cocycle. The Generalized Double Semion (GDS) model is another family of gapped phases that exist in every dimension, and its microscopic realization is extremely similar to that of the Toric Code. This lovely paper clarifies the relationship between GDS and group cohomological models. There is a fairly general ``ungauging'' procedure that allows one to recognize a model as the result of gauging a symmetry of some other model, and to build that latter model. (In the Toric Code and GDS cases, the recognition step is transparent from the microscopic model: the fluctuating qubits live on codimension-1 cells, and so one simply interprets them as the holonomies of a \(\mathbb{Z}_2\)-bundle. Although the paper doesn't say it, in general to recognize a model as arising from gauging a \(\mathbb{Z}_2\) symmetry, the necessary data is to select a boson-like excitation with \(\mathbb{Z}_2\)-fusion to interpret as the ``charged particle'' in the gauge theory.) This paper calls the resulting ungauged theory the ``dual'' to the original model. It carries an action by \(\mathbb{Z}_2\), and the original model can be reconstructed from the dual together with this action. The Toric Code case is particularly easy: the dual model is trivial(izable), and so is its \(\mathbb{Z}_2\)-action. The first observation of this paper is that the dual of the GDS model is also trivializable, although the \(\mathbb{Z}_2\)-action is not. Trivializing this dual is what the authors mean by ``disentangling'', and I will not elaborate. By definition, a trivializable quantum matter system together with a nontrivial \(G\)-action is called a ``\(G\)-SPT''. The second and main result of this paper is an analysis of the \(\mathbb{Z}_2\)-SPT that arises as the dual to the GDS model. It is not a Dijkgraaf-Witten model, but it is group cohomological in a more general sense. Specifically, a (nice enough) cellulation of an \(n\)-dimensional spatial manifold \(M\) will select some submanifolds that represent the Stiefel-Whitney classes \(w_1,\dots,w_n\) of \(M\). The dual to the GDS model is formed by taking the tensor product of (duals of) Dijkgraaf-Witten models on all of these submanifolds. A high-energy theorist would say that the gauge field has been ``coupled'' to the Stiefel-Whitney classes. The paper ends with discussion of renormalization group flow for the GDS model and of when one can or cannot trivialize the GDS model itself.