Commuting Squares of Tracial von Neumann Algebras

• Commuting squares are structured diagrams of finite-dimensional C*-subalgebras in tracial von Neumann algebras that satisfy conditional expectation compatibility.
• Their iterative construction via the Jones basic method generates hyperfinite II₁ subfactors, linking analytic properties to fusion categories and planar algebras.
• They provide practical insights through analytic obstructions like the noncommutative Minkowski inequality and have applications in tensor network states for topological quantum matter.

A commuting square of tracial von Neumann algebras is a foundational structure in subfactor theory, encoding compatibility between two pairs of nested algebras. In its finite-dimensional form, a commuting square consists of four C*-subalgebras embedded in a tracial finite von Neumann algebra, with all inclusions satisfying a conditional-expectation commutation property. When iterated via the Jones basic construction, such commuting squares yield hyperfinite type II₁ subfactors of finite index and finite depth, situating them at the intersection of functional analysis, tensor category theory, and topological quantum matter. Their combinatorial and categorical characterizations provide insight into fusion categories, subfactor planar algebras, and recent analytic obstructions for realizing fusion rings as unitary tensor categories.

1. Definitions and Basic Structure

A finite-dimensional commuting square in a tracial von Neumann algebra is the diagram

$$\begin{array}{ccc} B_{00} & \subset & B_{01} \ \cap & & \cap \ B_{10} & \subset & B_{11} \end{array}$$

with each $B_{ij}$ a finite-dimensional C*-subalgebra of $B_{11}$, equipped with a faithful, normalized trace $\operatorname{tr}(1)=1$ (Kawahigashi, 2021). For each inclusion $X\subset Y$, there is a unique trace-preserving conditional expectation $E_X:Y\to X$ such that $\operatorname{tr}(x y)=\operatorname{tr}(x E_X(y))$ for $x\in X$, $y\in Y$.

The square is a commuting square if any (and thus all) of the following equivalent conditions hold:

• $E_{B_{01}}(x) = E_{B_{00}}(x)$ for all $x\in B_{10}$,
• $E_{B_{01}}(B_{10})=B_{00}$,
• $$E_{B_{00}}=E_{B_{01}}\circ E_{B_{10}}=E_{B_{10}}\circ E_{B_{01}}$$ on $B_{11}$, and others reflecting conditional expectation compatibility.

Symmetry (or non-degeneracy) requires that the linear span of products $B_{01} B_{10}$ is $B_{11}$, equivalently that all four Bratteli diagrams associated to the inclusions in the square are connected (Kawahigashi, 2021, Bakshi et al., 2019).

2. Iteration and Subfactor Construction

Given a symmetric commuting square, one can iterate the Jones basic construction,

$$B_{01} \subset B_{02} \ \cap\qquad\,\,\,\, \cap \ B_{11} \subset B_{12}$$

where $B_{0,2} = \langle B_{01}, e_{B_{00}}\rangle$ on $L^2(B_{01})$, $B_{1,2} = \langle B_{11}, e_{B_{10}}\rangle$ on $L^2(B_{11})$, and so on (Kawahigashi, 2021). This generates a doubly-indexed grid $\{B_{k,l}\}_{k,l}$ of finite-dimensional von Neumann algebras equipped with canonical Jones projections, forming an infinite "lattice" of commuting squares.

By restricting to a minimal projection $p\in B_{00}$ and renormalizing the trace, one obtains algebras $A_{k,l}=p B_{k,l} p$ whose completions along rows and columns yield ascending towers whose inductive limits are hyperfinite II₁ factors $A_{k,\infty}, A_{\infty,l}$.

These limits produce, for each $k$, inclusions $A_{k,0} \subset A_{k+1,0}$ and $A_{0,l} \subset A_{0,l+1}$—each a II₁ subfactor of finite index. The entire bi-tower encodes fine-grained structure about the subfactor's standard invariant, depth, and index.

3. Characterization via Fusion Categories and Sato’s Construction

Any finite-dimensional symmetric commuting square with inductive limit a finite-index, finite-depth hyperfinite II₁ subfactor is in bijection with a pair of Morita-equivalent unitary fusion categories (Kawahigashi, 2021). Given a II₁ subfactor $N \subset M$ of finite index:

• The $N$-$N$ bimodule category generated by $L^2(M)$ under bimodule tensor product forms a unitary fusion category $\mathcal{C}_N$ ("even part").
• The $M$-$M$ bimodules generate $\mathcal{C}_M$; $\mathcal{C}_N$ and $\mathcal{C}_M$ are Morita equivalent.
• The global index is the global (Perron–Frobenius) dimension $$\operatorname{Dim}(\mathcal{C}_N)=\sum_{X\in\operatorname{Irr}(\mathcal{C}_N)}(\operatorname{dim}_{PF} X)^2$$.

Sato’s theorem states that for pairs of subfactors $A\subset B$, $C\subset D$ with Morita-equivalent fusion categories, and a suitable $P$–$A$ bimodule $X$ implementing the equivalence,

$$A_{k,l} = \operatorname{End}_{A-A}\left( D^{\otimes_C k/2} \otimes_C X \otimes_A B^{\otimes_A l/2} \right)$$

yields a symmetric grid of commuting squares whose row and column limits are $A\subset B$ and $C\subset D$. Every finite-dimensional commuting square of this form arises—up to isomorphism—from such a categorical datum.

Sato originally assumed $X$ irreducible and $A_{0,0}=\mathbb{C}$; the full characterization only requires $A_{0,0}$ a finite sum of matrix blocks and $X$ such that its generated bimodule categories are the even parts.

4. Planar Algebras and Smoothness

Non-degenerate smooth commuting squares of II₁ factors with all inclusions of equal finite index $\delta^2$ are in one-to-one correspondence with inclusions of subfactor planar algebras of modulus $\delta>1$ (Bakshi et al., 2019). The Guionnet–Jones–Shlyakhtenko (GJS) construction associates to a planar algebra $P$ a tower $M_0(P)\subset M_1(P) \subset \cdots$ of II₁-factors, with the tower’s standard invariant given by $P$.

Given an inclusion $Q\subset P$ of *-planar algebras, the associated GJS towers yield a commuting square,

$$M_0(Q)\subset M_1(Q) \ \cap\qquad\ \cap \ M_0(P)\subset M_1(P)$$

which is non-degenerate and smooth. The smoothness (i.e., the embedding $N'\cap K_k\subset L'\cap M_k$ at each step) ensures the inclusion $P_k\subset Q_k$ of planar algebra vector spaces, and closure under planar tangles is ensured by the commuting-square property.

Necessary and sufficient conditions for this correspondence include: extremality, equal index in all inclusions, non-degeneracy (density of span $KL$ in $M$), and smoothness (commutant inclusion at each stage).

5. Analytic Obstructions: Noncommutative Minkowski Inequality

Recent results establish a noncommutative analogue of Minkowski’s integral inequality for commuting squares of tracial von Neumann algebras (Lim, 20 Jan 2026). Let $(M,\operatorname{tr})$ be a tracial von Neumann algebra, and consider the commuting square

$L \subset M \ \cap\qquad\,\,\cap \ N \subset K$

with trace-preserving conditional expectations $E_L$, $E_K$, $E_N$, fulfilling $E_L\circ E_K = E_N$.

For $x\in M_+, 1\leq p<\infty$, under mild commutation hypotheses, the NC–Minkowski inequality states: $$E_L\big(E_K(x)^p\big)^{1/p} \leq E_K\big(E_L(x^p)^{1/p}\big) \subseteq N.$$ This noncommutative convexity constraint generalizes the classical Fubini–Minkowski inequality, and equality cases characterize joint commutativity and proportionality in $M$.

Combinatorial Specialization. For quadruples of multi-matrix inclusion graphs (encoding the corners $A_{00}, A_{01}, A_{10}, A_{11}$), this yields necessary conditions expressed as arrays of inequalities on the associated inclusion matrices and traces. For fusion rings, an analogous family of inequalities provides an obstruction to unitary categorification: if a based ring fails them, no unitary spherical fusion category realization of the ring exists.

6. Applications to Tensor Categories and Topological Order

The commuting-square data admit a 4-tensor formulation in the context of matrix product operators (MPOs) and tensor network states. A bi-unitary connection $W: V_{ij}^{ab}$, with indices $(i,j)$ for "vertical" graphs and $(a,b)$ for "horizontal" graphs, encodes a commuting square if it satisfies two bi-unitarity conditions: $$\sum_{j,b} W_{ij}^{ab} \overline{W_{i'j}^{a'b}} = \delta_{i,i'}\delta_{a,a'},\quad \sum_{i,a} W_{ij}^{ab} \overline{W_{ij'}^{ab'}} = \delta_{j,j'}\delta_{b,b'}.$$ These tensors define local weights for 2D tensor-network states or MPO projectors, and the main result asserts that exactly those bi-unitary 4-tensors whose induced fusion graphs yield Morita-equivalent fusion categories correspond to commuting-square subfactor constructions of finite index and depth—thus giving rise to genuine 2D topological order (Kawahigashi, 2021).

7. Examples, Classification, and Recent Developments

Specific constructions include Temperley–Lieb algebras (with even–odd planar subalgebra inclusions), group–subgroup subfactors (crossed-product factors), and trivial subalgebra situations. The iteration of analytic, combinatorial, and categorical criteria has led to classification advances for amenable subfactors of finite depth.

The noncommutative Minkowski inequality provides a novel, highly effective numerical test for realizability of commuting squares via fusion data: Lim and collaborators identified numerous fusion rings excluded by this test but not by previous obstructions. For example, $\approx 20.36\%$ of low-rank fusion rings in the Vercleyen–Slingerland database fail this criterion, with about $1.9\%$ escaping all earlier analytic or cyclotomic constraints (Lim, 20 Jan 2026).

The relationship between symmetric commuting squares, Morita equivalence of fusion categories, and bi-unitary connections underpins modern approaches in topological phases of matter, subfactor theory, and tensor categorical categorification.

---

Table: Characterizations and Correspondences

Description	Algebraic/Categorical Data	Reference
Symmetric commuting square (finite dim, II₁ limit)	Pair of Morita-equivalent unitary fusion categories	(Kawahigashi, 2021)
Non-degenerate smooth commuting square (all II₁)	Inclusion of subfactor planar algebras (same modulus)	(Bakshi et al., 2019)
Analytic/combinatorial obstruction	Failure of NC-minkowski or derived inequalities	(Lim, 20 Jan 2026)