Papers
Topics
Authors
Recent
2000 character limit reached

Graph Planar Algebra

Updated 12 January 2026
  • Graph planar algebra is a diagrammatic, combinatorial *-algebra constructed from weighted bipartite graphs that encode the standard invariant of finite-index subfactors.
  • It provides an explicit operator-theoretic framework via Hilbert spaces and path-concatenation, ensuring isotopy invariance and adherence to operad gluing axioms.
  • The structure enables embedding of abstract planar algebras, construction of fixed-point subalgebras under automorphism groups, and classification of both classical and exotic subfactors.

A graph planar algebra is a diagrammatic, combinatorial planar *-algebra constructed from a bipartite (or, more generally, weighted and possibly oriented) graph. It is a central object in the theory of subfactor planar algebras as introduced by Jones, encoding the standard invariant of a finite-index subfactor and serving both as a universal receptacle for embeddings of abstract planar algebras and as a powerful tool for constructing and classifying subfactors and module categories. The canonical construction based on a locally finite bipartite graph with Perron–Frobenius weighting facilitates an explicit algebraic and operator-theoretic framework, underlying both deep classification results and the emergence of "exotic" structures across operator algebras, quantum groups, and low-dimensional topology (Burstein, 2010, Jones et al., 2010, Bisch et al., 2024).

1. Construction of the Bipartite Graph Planar Algebra

Let Γ\Gamma be a locally finite bipartite graph with vertex set V=V+VV=V^+\cup V^- and edge set EE. One fixes a weight (spin vector) μ:V(0,)\mu:V\to(0,\infty) subject to the boundedness condition: M<\exists\,M<\infty such that μ(v)/μ(w)<M\mu(v)/\mu(w)<M whenever {v,w}E\{v,w\}\in E. For each n0n\geq 0 and ±\pm, define

Pn,±={A:Loops2n±(Γ)CA bounded},P_{n,\pm} = \{\,A : \operatorname{Loops}^{\pm}_{2n}(\Gamma) \to \mathbb{C} \mid A \text{ bounded}\,\},

where Loops2n±(Γ)\operatorname{Loops}^{\pm}_{2n}(\Gamma) denotes closed walks of length $2n$ based at a vertex in V±V^{\pm}. The total graph planar algebra is P(Γ,μ)=n,±Pn,±P(\Gamma,\mu)=\bigoplus_{n,\pm} P_{n,\pm}, with a grading respected by the planar operad.

Given a shaded planar tangle TT with kk internal disks of types (ni,ϵi)(n_i,\epsilon_i) and external type (n,±)(n,\pm), the multilinear operad action ZTZ_T is defined via a state sum: one sums over all possible labelings σ\sigma of regions by vertices (white by V+V^+, black by VV^-) and contour strings by edges, requiring boundary compatibility. Each internal region (disk DiD_i) yields a closed loop LiL_i; each local maximum or minimum (singularity) contributes a ratio weight μ(v)/μ(w)\mu(v)/\mu(w) of its adjacent regions. The output is

ZT(x1,,xk)(L)=σcompatiblewithLw(σ),w(σ)=(singularitiessμ(vs)μ(ws))i=1kxi(Li).Z_T(x_1,\ldots,x_k)(L) = \sum_{\sigma\,\mathrm{compatible\,with}\,L} w(\sigma), \quad w(\sigma) = \left(\prod_\mathrm{singularities\,s} \frac{\mu(v_s)}{\mu(w_s)}\right) \prod_{i=1}^k x_i(L_i).

One checks that only finitely many states contribute and that the resulting maps satisfy isotopy invariance, *-structure, and the operad gluing axioms, making P(Γ,μ)P(\Gamma,\mu) a planar *-algebra (Burstein, 2010, Jones et al., 2010).

The modulus δ\delta of (Γ,μ)(\Gamma,\mu) is defined by demanding shaded and unshaded circles evaluate identically: e:s(e)=vμ(t(e))2=δμ(v)2,v.\sum_{e:s(e)=v} \mu(t(e))^2 = \delta\,\mu(v)^2,\quad \forall\,v. For finite Γ\Gamma this is the Perron–Frobenius eigenvalue of the adjacency matrix.

2. Algebraic and Operator-Theoretic Realizations

The nn-box spaces Pn,±P_{n,\pm} realize as spaces of bounded operators on Hilbert spaces HnH_n of square-summable length-nn paths starting at vertices of fixed parity. The entire planar algebra acts as a concrete collection of operators commuting with left and right projections corresponding to source and target vertices. The operad multiplication, inclusion, capping, and involution operations are exactly implemented by path-concatenation, contraction, and reversal on H=HnH = \bigoplus H_n (Burstein, 2010).

Graph planar algebras are CC^*-algebras when Γ\Gamma is finite and μ\mu satisfies the eigenvector condition, yielding finite-dimensional semisimple *-algebras, positive definite traces, and spherical structure. In the oriented (or weighted) setting, the construction extends by labeling regions from suitable branching graphs and weighting the state sum by generalized harmonic functions (Koshida, 2023).

3. Automorphism Groups and Fixed-Point Subalgebras

The automorphism group of a bipartite graph planar algebra P(Γ,μ)P(\Gamma,\mu) is classified as follows:

  • Graph-automorphism operators: Any automorphism κ\kappa of Γ\Gamma (preserving V+V^+, VV^-, adjacency, and scaling μ\mu by a constant) yields a unitary UκU_\kappa acting by permuting path-basis vectors.
  • Multiplication operators: A unitary OB(H1)O\in B(H_1) commuting with source/target projections and satisfying revOrev=O\mathrm{rev}\, O\, \mathrm{rev} = O^* may be extended to a grading-preserving unitary UOU_O commuting with concatenation and reversal, acting on loops by twisting by $1$-dimensional representations of π1(Γ)\pi_1(\Gamma).

The group of all planar *-automorphisms Aut(P)\operatorname{Aut}(P) factors uniquely as EAE\rtimes A, where EE is the group of multiplication operators (isomorphic to the dual of H1(Γ)H^1(\Gamma)), and AA is the group of automorphisms of Γ\Gamma (Burstein, 2010).

For any subgroup GAut(P)G\leq\operatorname{Aut}(P), the fixed points Pn,±G={xPn,±g(x)=x gG}P^G_{n,\pm} = \{x\in P_{n,\pm}\mid g(x)=x \ \forall\, g\in G\} form a planar *-subalgebra. If GG acts transitively on V+V^+ and VV^-, PGP^G is a subfactor planar algebra with the same modulus as PP, realized as the standard invariant of a finite-index extremal II1_1 subfactor. The index of the corresponding subfactor is δ2\delta^2 (Burstein, 2010).

4. Structural Theorems and Embedding Results

The bipartite graph planar algebra P(Γ,μ)P(\Gamma,\mu) arises as the relative commutant planar algebra for the basic construction tower of a strongly Markov inclusion of finite-dimensional algebras with Bratteli diagram Γ\Gamma. There is a canonical *-isomorphism between the tower's relative commutant planar algebra and P(Γ,μ)P(\Gamma,\mu) (Jones et al., 2010).

Every finite-depth subfactor planar algebra QQ_{\bullet} of modulus dd embeds as a planar *-subalgebra of the graph planar algebra P(Γ)P(\Gamma) of its principal graph Γ\Gamma. This embedding identifies the Temperley–Lieb generators and respects all operad tangles (Jones et al., 2010, Morrison, 2013, Bisch et al., 2024).

The same mechanism, extended to commuting square subfactors and more general weighted or oriented graphs (e.g., Young graphs with harmonic functions), yields powerful obstructions for standard invariants and organizes the construction of subfactors with prescribed combinatorial and analytic properties (Bisch et al., 2024, Koshida, 2023).

5. Applications and Classification Frameworks

The fixed-point planar subalgebra construction under automorphism subgroups provides uniform realizations for a range of classical and exotic subfactors:

  • Group-subgroup (star) subfactors: Γ\Gamma is a star graph, GSnG\leq S_n acts on leaves; PGP^G recovers the standard invariant of MGMM^G\subset M, index nn.
  • Diagonal subfactors: Cayley-type bipartite graphs encode automorphism-generated inclusions.
  • Bisch–Haagerup and Wassermann subfactors: Γ\Gamma reflects group-theoretic and module-coset structures, with GG acting by graph or multiplication operators (Burstein, 2010, Brothier, 2016).
  • Infinite-depth and non-amenable cases: Tree graphs and suitable automorphism actions yield subfactor planar algebras with Haagerup property and complete metric approximation property (CMAP) (Brothier, 2015).

In the broader context, every symmetric enveloping inclusion of a subfactor planar algebra arising from a Hecke pair (G,K)(G,K) (where KK is a compact open almost normal subgroup of GG) can be described as a crossed product, providing analytic control over amenability, approximation, and representation-theoretic properties (Brothier, 2016, Brothier, 2015).

6. Computational and Category-Theoretic Connections

Graph planar algebra embeddings are used to facilitate explicit computation and the construction of new diagrammatic relations in the study of module categories over quantum groups and fusion categories. By embedding the planar algebra of a generating object into the graph planar algebra of its fusion graph, one translates skein-theoretic constraints into concrete finite-dimensional linear algebra, enabling the discovery of new quantum subgroups, such as the type G2G_2 quantum subgroups via computational solution of relations among diagrams in the ambient graph planar algebra (Hill, 8 Jan 2026).

Moreover, the construction extends via oriented branching graphs, harmonic functions, and combinatorial data to categorification contexts such as the Khovanov Heisenberg category, with explicit state sums on tangles recovering the relevant algebraic and representation-theory invariants (Koshida, 2023).

7. Impact and Open Directions

The graph planar algebra acts as a universal combinatorial host for subfactor planar algebras, unifying their construction, providing embedding theorems, and enabling new analytic and combinatorial approaches to classification. Key impacts include:

  • Transparent derivation of known and "exotic" subfactors via group action fixed points.
  • Uniform strategy for enforcing finiteness and sphericality via symmetry.
  • Concrete analytic inheritance of approximation properties (amenability, Haagerup, CMAP) from acting groups to fixed-point subfactors (Brothier, 2015).
  • Obstructions for principal graphs of subfactors via embedding criteria, guiding the classification of standard invariants (Morrison, 2013).

Open problems include the detailed classification of which automorphism group actions lead to non-isomorphic fixed-point planar subalgebras, the construction of amenable or non-amenable subfactors from infinite graphs, and the systematic generation of continuous families of examples via π1(Γ)\pi_1(\Gamma) character twisting. Further computational developments—leveraging the diagrammatic presentation and the translation to module category theory—continue to expand the reach of graph planar algebra techniques (Burstein, 2010, Hill, 8 Jan 2026).

Whiteboard

Topic to Video (Beta)

Follow Topic

Get notified by email when new papers are published related to Graph Planar Algebra.