Singular MASA in Operator Algebras

• Singular MASA is a maximal abelian *-subalgebra in von Neumann or C*-algebras characterized by its rigidity and minimal normalizer structure.
• It exhibits free complementation in free product settings, linking singularity with maximal amenability and the absence of nontrivial symmetries.
• Its construction via iterative methods and precise so-paving properties underpins classification efforts and advances understanding in operator algebras.

A singular MASA (maximal abelian -subalgebra) is a maximal abelian subalgebra within a von Neumann algebra or C-algebra whose normalizer generates only the MASA itself, not a larger subalgebra or the entire ambient algebra. Specifically, for a MASA $A \subset M$, singularity is characterized by the property that the normalizer $$\mathcal N_M(A) = \{u \in \mathcal U(M) \mid uAu^* = A\}$$ satisfies $\mathcal N_M(A)'' = A$, i.e., the only unitaries normalizing $A$ are those contained in $A$ itself. Singular MASAs are distinguished from Cartan (or regular) MASAs, where the normalizer generates the ambient algebra, by their rigidity, lack of symmetries, and essential position in the structure of operator algebras (Boschert et al., 2024, Popa et al., 2014, Pitts, 2021, Popa, 2016).

1. Foundational Concepts and Definitions

Let $(M, \tau)$ be a tracial von Neumann algebra, where $M$ is a von Neumann algebra and $\tau$ is a faithful normal tracial state. A $*$-subalgebra $A \subset M$ is a MASA if $$\mathcal N_M(A) = \{u \in \mathcal U(M) \mid uAu^* = A\}$$0 is abelian and maximal with respect to inclusion: $$\mathcal N_M(A) = \{u \in \mathcal U(M) \mid uAu^* = A\}$$1. The singularity of $$\mathcal N_M(A) = \{u \in \mathcal U(M) \mid uAu^* = A\}$$2 is determined by its normalizer $$\mathcal N_M(A) = \{u \in \mathcal U(M) \mid uAu^* = A\}$$3, which consists of unitaries in $$\mathcal N_M(A) = \{u \in \mathcal U(M) \mid uAu^* = A\}$$4 implementing automorphisms of $$\mathcal N_M(A) = \{u \in \mathcal U(M) \mid uAu^* = A\}$$5 by conjugation. $$\mathcal N_M(A) = \{u \in \mathcal U(M) \mid uAu^* = A\}$$6 is singular if $$\mathcal N_M(A) = \{u \in \mathcal U(M) \mid uAu^* = A\}$$7.

In the C*$$\mathcal N_M(A) = \{u \in \mathcal U(M) \mid uAu^* = A\}$$8algebra context, let $$\mathcal N_M(A) = \{u \in \mathcal U(M) \mid uAu^* = A\}$$9 be a (not necessarily unital) C*$\mathcal N_M(A)'' = A$0algebra and $\mathcal N_M(A)'' = A$1 an abelian C*$\mathcal N_M(A)'' = A$2subalgebra. The inclusion $\mathcal N_M(A)'' = A$3 is singular if the normalizer $\mathcal N_M(A)'' = A$4 equals $\mathcal N_M(A)'' = A$5; in contrast, $\mathcal N_M(A)'' = A$6 is regular if the linear span of normalizers is dense in $\mathcal N_M(A)'' = A$7 (Pitts, 2021). The MASA property ensures that $\mathcal N_M(A)'' = A$8 is maximal abelian in $\mathcal N_M(A)'' = A$9, with singularity reflecting the absence of nontrivial normalizing elements outside of $A$0.

2. Structural Properties and Characterizations

Singular MASAs are structurally rigid and resist extension to regular MASA inclusions by unitization or embedding in larger algebras (Pitts, 2021). They may display divergent behavior with respect to approximate unit properties. One can construct singular MASA inclusions where every approximate unit for $A$1 is an approximate unit for $A$2, as well as cases where this fails. For instance, in $A$3 for $A$4 a non-atomic MASA and $A$5 compact operators, $A$6 is singular and every approximate unit of $A$7 suffices for $A$8. Conversely, there exist singular inclusions, such as $A$9, where the approximate unit property fails.

No regular extension of a MASA in $A$0 exists: one cannot embed $A$1 into a strictly larger algebra where $A$2 remains a MASA but becomes regular (Pitts, 2021).

Cartan MASAs are regular, characterized by having their normalizer generate $A$3 and admitting a faithful conditional expectation to $A$4. Singular MASAs, by contrast, are fixed by the absence of nontrivial automorphisms outside $A$5 itself (Popa et al., 2014, Boschert et al., 2024).

3. Singular MASAs in Free Product and Group Factor Settings

Let $A$6 be tracial abelian von Neumann algebras and $A$7 their free product. Any MASA of the form $A$8, where $A$9, $(M, \tau)$0, and $(M, \tau)$1, is freely complemented (FC) in $(M, \tau)$2. If each $(M, \tau)$3 is purely non-separable, then any purely non-separable singular MASA in $(M, \tau)$4 is FC. This result gives an explicit classification of singular MASAs in terms of their free-complemented structure.

In the context of free group factors $(M, \tau)$5, known maximal amenable MASAs—such as the generator MASAs $(M, \tau)$6, cyclic MASAs $(M, \tau)$7 for $(M, \tau)$8, the radial MASA $(M, \tau)$9, and semicircular families $M$0—admit Haar unitaries $M$1 that are $M$2-free from $M$3. This settles Popa’s weak FC conjecture in all known cases: for amenable subalgebras $M$4, there exists a Haar unitary $M$5 in $M$6 that is free from $M$7.

The equivalence (in the nonseparable setting) of singularity, maximal amenability, and free complementation provides a dichotomy: non-separable singular MASAs correspond precisely to freely complemented (hence maximal amenable) subalgebras, arising by re-assembly of the generator MASAs (Boschert et al., 2024).

4. Construction and Existence of Singular MASAs

An iterative procedure enables the construction of singular MASAs with prescribed properties in separable $M$8-factors (Popa, 2016). One builds $M$9 as an increasing union of finite-dimensional abelian subalgebras $\tau$0, refining at each stage to ensure maximal abelianity and to avoid bimodule intertwining with a countable family of subalgebras $\tau$1. This process secures the singularity of $\tau$2 by ensuring $\tau$3 for each $\tau$4.

An s-MASA is a MASA $\tau$5 such that the algebra $\tau$6 is itself maximal abelian in $\tau$7. Local s-thin approximation characterizes when factors admit s-MASAs: for every finite partition $\tau$8, finite set $\tau$9, and $*$0, there is a finer partition $*$1 and $*$2 making every $*$3 approximable in $*$4 within $*$5 by sums of $*$6. Any such factor has uncountably many non-intertwinable singular s-MASAs (Popa, 2016).

5. Paving Properties and Singular MASAs

The so-paving property reflects the capability to approximate elements in the strong operator topology via partitioning by projections in the MASA. For a singular MASA $*$7 in a $*$8-factor $*$9, Popa and Vaes demonstrated that so-paving holds with optimal bounds: for any self-adjoint $A \subset M$0, $A \subset M$1, one can find projections $A \subset M$2 giving a partition of unity with

$A \subset M$3

for $A \subset M$4. This $A \subset M$5 dependence is sharp (Popa et al., 2014). So-paving for singular MASAs is equivalent to norm paving after passing to the ultrapower inclusion $A \subset M$6, provided a normal conditional expectation exists.

Kadison–Singer norm-paving over diagonal MASAs, Cartan MASAs in amenable $A \subset M$7-factors, and singular MASAs are unified under the so-paving framework. Popa–Vaes conjecture that any MASA $A \subset M$8 with normal conditional expectation has so-paving with uniform bound $A \subset M$9.

6. Rigidity, Maximal Amenability, and Classification Implications

Singular MASAs in the non-separable free product setting exhibit strong rigidity: they arise as re-assemblies of generator MASAs via free complementation. Every (purely non-separable) singular MASA is freely complemented and hence maximally amenable (Boschert et al., 2024). The entire class of such MASAs can be parametrized by partitions of unity in the abelian core $$\mathcal N_M(A) = \{u \in \mathcal U(M) \mid uAu^* = A\}$$00. This structure also gives rise to injective homomorphisms from automorphism groups of the “sans-core” into the outer automorphism group of the ambient algebra, reflecting significant symmetry constraints.

A salient implication is that maximal amenability for subalgebras is closely tied to their being freely complemented. The dichotomy “singular ⇔ freely complemented” in the non-separable setting suggests a near-complete classification of singular MASAs as those constructed by free re-assembly from standard Cartan pieces and validated by free product techniques (Boschert et al., 2024).

7. Open Problems and Future Directions

Outstanding questions pertain to the extension of so-paving to MASAs lacking conditional expectations and the determination of universal constants for the paving bound $$\mathcal N_M(A) = \{u \in \mathcal U(M) \mid uAu^* = A\}$$01 in all cases. The growth rate of required projections in multi-paving scenarios remains unresolved, especially for MASAs whose normalizer is neither trivial (singular) nor maximal (Cartan) (Popa et al., 2014). A plausible implication is that as understanding of singular MASAs deepens through free product and so-paving frameworks, broader classification and rigidity results will emerge, potentially enabling full resolution of Popa’s conjectures regarding amenable subalgebras and normalizer structure in von Neumann algebras.