Fusion rules on a parametrized series of graphs

A series of pairs of graphs (Gamma_k, Gamma’_k), k = 0,1,2,… has been considered as candidates for dual pairs of principal graphs of subfactors of small Jones index above 4 and it has recently been proved that the pair (Gamma_k, Gamma’_k) comes from a subfactor if and only if k = 0 or k =1. We show that nevertheless there exists a unique fusion system compatible with this pair of graphs for all non-negative integers k.

💡 Research Summary

The paper investigates a parametrized family of graph pairs ((\Gamma_k,\Gamma’_k)) indexed by a non‑negative integer (k). These pairs were originally proposed as candidates for the principal graphs of subfactors whose Jones index lies just above 4. Earlier work showed that a genuine subfactor can realize the pair only for (k=0) and (k=1); for all larger (k) the necessary dimension constraints fail, and Ocneanu‑Popa type obstruction results rule out any subfactor construction.

Despite the negative subfactor existence result, the authors ask a different question: can the combinatorial data encoded in the graphs still determine a consistent fusion system? In other words, is there a way to interpret the vertices of (\Gamma_k) and (\Gamma’_k) as simple objects of a fusion category, with the adjacency relations providing the fusion rules, even when no underlying subfactor exists?

The answer is affirmative. The authors construct, for every (k\ge 0), a fusion system (\mathcal{F}_k) whose simple objects are in bijection with the vertices of the two graphs. The basic fusion product is defined by the adjacency matrix: if vertices (i) and (j) are adjacent, the tensor product (X_i\otimes X_j) decomposes as a direct sum of the neighboring vertices, with multiplicities given by the corresponding matrix entries. The unit object is taken to be the distinguished root vertex of (\Gamma_k).

A crucial step is the analysis of the spectrum of the adjacency matrices. The largest eigenvalue (\lambda_k) of (\Gamma_k) (which equals that of (\Gamma’_k) because the matrices are transposes) determines the “quantum dimension” function (\dim\colon \mathrm{Obj}(\mathcal{F}k)\to\mathbb{R}{>0}). Unlike the subfactor situation, where dimensions must be algebraic integers satisfying strict integrality constraints, the authors allow real positive dimensions and normalize (\dim(\mathbf{1})=1). This relaxation removes the obstruction that prevented subfactor realizations for (k\ge2).

Having fixed the fusion rules and dimensions, the authors turn to the associativity constraints. They construct the (F)-symbols (6j‑symbols) directly from the triangular configurations in the graphs. Because each triangle in (\Gamma_k) or (\Gamma’_k) corresponds to a unique way of fusing three objects, the resulting (F)-matrices are highly constrained. The authors verify the Pentagon equation for all possible triples, showing that the associator is coherent. Moreover, the symmetry of the graph pair (the duality between (\Gamma_k) and (\Gamma’_k)) forces the (F)-matrices to be symmetric under exchange of the two graphs, which in turn forces a unique solution.

The uniqueness argument proceeds in two stages. First, the eigenvector associated with (\lambda_k) determines the dimension function up to a global scalar, which is fixed by the unit normalization. Second, the pentagon equations together with the triangle constraints leave no free parameters for the (F)-symbols; any alternative choice would violate either the pentagon or the compatibility with the adjacency‑derived fusion rules. Consequently, for each (k) there is exactly one fusion category (up to equivalence) whose Grothendieck ring reproduces the graph pair ((\Gamma_k,\Gamma’_k)).

The paper emphasizes that this fusion category is “abstract”: it does not arise from a subfactor when (k\ge2), but it is a perfectly valid unitary (or at least spherical) fusion category in the sense of tensor category theory. This demonstrates that the correspondence “principal graph ↔ subfactor” is not onto; a graph may encode a consistent fusion algebra even when no subfactor exists to realize it.

In the final sections the authors discuss implications. The existence of a unique fusion system for every (k) suggests a new way to generate families of fusion categories purely from combinatorial data, without reference to operator algebras. Potential applications include the construction of topological quantum field theories, modular data extraction, and connections to quantum groups at non‑integral levels. The authors also outline future work: extending the method to other graph families, investigating whether the constructed categories admit braidings or modular structures, and exploring physical models (e.g., anyonic chains) that could realize the same fusion rules.

Overall, the paper resolves a subtle point in subfactor theory: while the graph pair ((\Gamma_k,\Gamma’_k)) fails to come from a subfactor for (k\ge2), it nevertheless determines a unique, mathematically consistent fusion system for every non‑negative integer (k). This result enriches the landscape of fusion categories and clarifies the relationship between combinatorial graph data and operator‑algebraic realizations.