Weak Dixmier Property in Operator Algebras

• Weak Dixmier property is an averaging property for unitary orbits in operator algebras that forces elements toward central or relative commutants.
• It employs semigroup formulations of UCP maps and operator-valued weights to extend classical averaging to non-tracial and type III settings.
• The property underpins applications in Popa intertwining, relative solidity, and rigidity of subfactors in advanced von Neumann algebra theory.

The weak Dixmier property is an averaging property for unitary orbits in operator algebras. In its operator-algebraic form, it asserts that conjugation by unitaries generates enough convex averaging to force every element, or at least every element in a specified integrable class, toward a commutant or central subalgebra. For inclusions of von Neumann algebras $N \subset M$, the basic requirement is that the weak$^*$ closed convex hull of $\{uxu^*:u\in \mathcal U(N)\}$ meet $N'\cap M$ for every $x\in M$; when $N=M$, this is the relative counterpart of classical Dixmier averaging to the center. The modern theory includes a complete conditional-expectation case, a type III extension via operator-valued weights, and applications to Popa intertwining, relative solidity, and intermediate subfactor rigidity (Marrakchi, 2019, Isono, 25 Aug 2025).

1. Classical origin and relative formulation

Classically, Dixmier’s theorem states that for a von Neumann algebra $M$, the norm-closed convex hull of the unitary orbit of any element meets the center: $$\overline{\operatorname{co}^{\|\cdot\|}\{uxu^*:u\in \mathcal U(M)\}}\cap Z(M)\neq\varnothing.$$ The relative weak Dixmier property replaces the full unitary group $\mathcal U(M)$ by $\mathcal U(A)$ for a subalgebra $^*$0, replaces the center $^*$1 by the relative commutant $^*$2, and replaces norm closure by $^*$3-weak closure. Thus the relative statement is

$^*$4

In this form, the property says that averaging over unitaries from $^*$5 can push an element of $^*$6 toward something commuting with $^*$7 (Isono, 25 Aug 2025).

For an inclusion $^*$8, the formulation used in the conditional-expectation literature is equivalent: the inclusion has the weak relative Dixmier property if for every $^*$9, the weak$\{uxu^*:u\in \mathcal U(N)\}$0 closed convex hull of $\{uxu^*:u\in \mathcal U(N)\}$1 intersects $\{uxu^*:u\in \mathcal U(N)\}$2. This is a relative version of the classical averaging phenomenon, but it is genuinely subtler than the case $\{uxu^*:u\in \mathcal U(N)\}$3. The property can fail without additional hypotheses, and its behavior reflects deep structural features of the inclusion rather than merely formal convexity (Marrakchi, 2019).

2. Semigroup formulation and averaging maps

A central reformulation replaces explicit convex hulls by compact convex semigroups of UCP maps. For an inclusion $\{uxu^*:u\in \mathcal U(N)\}$4, let $\{uxu^*:u\in \mathcal U(N)\}$5 be the closed convex semigroup generated by $\{uxu^*:u\in \mathcal U(N)\}$6 for $\{uxu^*:u\in \mathcal U(N)\}$7, with closure taken in the topology of pointwise weak$\{uxu^*:u\in \mathcal U(N)\}$8 convergence. Then

$\{uxu^*:u\in \mathcal U(N)\}$9

Moreover, $N'\cap M$0 has the weak relative Dixmier property if and only if $N'\cap M$1 contains a conditional expectation onto $N'\cap M$2 (Marrakchi, 2019).

This semigroup perspective is not only a reformulation; it is the mechanism behind the general existence theorem. A key ingredient is a strengthening of Ellis’ lemma for compact convex semigroups: if $N'\cap M$3 is a compact convex semigroup and $N'\cap M$4 is a minimal idempotent, then

$N'\cap M$5

In the operator-algebraic setting, this yields strong absorption relations among minimal idempotents. If $N'\cap M$6 is a closed convex semigroup and a minimal idempotent $N'\cap M$7 is faithful, then the fixed-point space $N'\cap M$8 is a $N'\cap M$9-subalgebra of $x\in M$0, and $x\in M$1 is a conditional expectation of $x\in M$2 onto $x\in M$3 (Marrakchi, 2019).

The type $x\in M$4 case requires an additional input. For a type $x\in M$5 factor $x\in M$6 with separable predual and faithful normal state $x\in M$7, one has

$x\in M$8

This allows the semigroup machinery to produce faithful minimal idempotents even in the non-semifinite regime, which is the technical point at which the weak relative Dixmier property becomes a type III statement rather than a tracial one (Marrakchi, 2019).

3. Inclusions with conditional expectations

The first general theorem in the modern literature states that if $x\in M$9 admits a faithful normal conditional expectation, then the inclusion has the weak relative Dixmier property. In the form proved in the separable-state reduction, if there exists a faithful normal state $N=M$0 on $N=M$1 such that $N=M$2 is the range of a $N=M$3-preserving conditional expectation $N=M$4, then $N=M$5 has the weak relative Dixmier property. Consequently, every inclusion of von Neumann algebras with a faithful normal conditional expectation has the property, answering a question of Popa (Marrakchi, 2019).

The proof proceeds by successive reductions: separable reduction, center reduction to the case $N=M$6, decomposition to the factor case, and then a distinction between semifinite, type III but not $N=M$7, and type $N=M$8 factors. The semifinite case is already known. In the type III but not $N=M$9 case, one writes

$M$0

for a semifinite von Neumann algebra with expectation $M$1, and combines the known property for $M$2 with amenability of $M$3. In the type $M$4 case, the semigroup argument constructs a faithful minimal idempotent $M$5, and faithfulness upgrades $M$6 to a conditional expectation onto $M$7 (Marrakchi, 2019).

The theorem is nontrivial even at the level of examples. The inclusion $M$8 has the weak relative Dixmier property if and only if $M$9 is AFD. If $$\overline{\operatorname{co}^{\|\cdot\|}\{uxu^*:u\in \mathcal U(M)\}}\cap Z(M)\neq\varnothing.$$0 is a type $$\overline{\operatorname{co}^{\|\cdot\|}\{uxu^*:u\in \mathcal U(M)\}}\cap Z(M)\neq\varnothing.$$1 factor and $$\overline{\operatorname{co}^{\|\cdot\|}\{uxu^*:u\in \mathcal U(M)\}}\cap Z(M)\neq\varnothing.$$2 is its continuous core, then $$\overline{\operatorname{co}^{\|\cdot\|}\{uxu^*:u\in \mathcal U(M)\}}\cap Z(M)\neq\varnothing.$$3 has a trivial bicentralizer if and only if $$\overline{\operatorname{co}^{\|\cdot\|}\{uxu^*:u\in \mathcal U(M)\}}\cap Z(M)\neq\varnothing.$$4 has the weak relative Dixmier property. These examples show that the property detects approximation and centralizer phenomena that are not visible from formal inclusion data alone (Marrakchi, 2019).

An elementary sufficient condition also survives in this general theory: if there exists a faithful normal state $$\overline{\operatorname{co}^{\|\cdot\|}\{uxu^*:u\in \mathcal U(M)\}}\cap Z(M)\neq\varnothing.$$5 on $$\overline{\operatorname{co}^{\|\cdot\|}\{uxu^*:u\in \mathcal U(M)\}}\cap Z(M)\neq\varnothing.$$6 such that

$$\overline{\operatorname{co}^{\|\cdot\|}\{uxu^*:u\in \mathcal U(M)\}}\cap Z(M)\neq\varnothing.$$7

then $$\overline{\operatorname{co}^{\|\cdot\|}\{uxu^*:u\in \mathcal U(M)\}}\cap Z(M)\neq\varnothing.$$8 has the weak relative Dixmier property. This is the familiar tracial or modular-invariant averaging argument, now seen as a special case of the general theorem (Marrakchi, 2019).

4. Operator-valued weights and the weak relative Dixmier property for integrable positive elements

A substantial extension replaces conditional expectations by faithful normal semifinite operator-valued weights. Let $$\overline{\operatorname{co}^{\|\cdot\|}\{uxu^*:u\in \mathcal U(M)\}}\cap Z(M)\neq\varnothing.$$9 be an inclusion of von Neumann algebras equipped with a faithful normal semifinite operator-valued weight

$\mathcal U(M)$0

Then for every positive element $\mathcal U(M)$1 such that $\mathcal U(M)$2, the weak Dixmier property relative to $\mathcal U(M)$3 holds: $\mathcal U(M)$4 Equivalently, if

$\mathcal U(M)$5

then

$\mathcal U(M)$6

This is the weak relative Dixmier property for integrable positive elements (Isono, 25 Aug 2025).

The result strictly extends the conditional-expectation theorem. Marrakchi proved the weak relative Dixmier property for inclusions with a faithful normal conditional expectation $\mathcal U(M)$7. The operator-valued-weight version works when $\mathcal U(M)$8 may be infinite and no conditional expectation need exist. When $\mathcal U(M)$9 happens to be a conditional expectation, the theorem recovers Marrakchi’s setting as a special case (Isono, 25 Aug 2025).

The semigroup packaging is sharpened accordingly. Writing $\mathcal U(A)$0 for the point-$\mathcal U(A)$1-weak closure of the convex semigroup generated by $\mathcal U(A)$2, $\mathcal U(A)$3, the paper proves

$\mathcal U(A)$4

and hence there exists $\mathcal U(A)$5 such that

$\mathcal U(A)$6

Thus the conclusion is not merely pointwise existence of an averaged commutant element for each $\mathcal U(A)$7, but the existence of a UCP averaging map from the Dixmier semigroup sending all integrable elements into the relative commutant (Isono, 25 Aug 2025).

The theorem also yields a useful equivalence. The following are equivalent:

1. $\mathcal U(A)$8 for all nonzero positive $\mathcal U(A)$9.
2. For every $^*$00 with $^*$01, $^*$02.
3. For every $^*$03 with $^*$04,

$^*$05

The same zero-averaging statements also hold for all $^*$06 by linearity. Technically, the proof splits $^*$07 into semifinite and type III parts via a central projection $^*$08: the semifinite part is handled by standard Hilbert space averaging, while the type III part uses minimal idempotents in the weak Dixmier semigroup together with operator-valued weight techniques. This separation of semifinite and type III regimes is one of the main technical contributions of the operator-valued-weight theory (Isono, 25 Aug 2025).

5. Type III applications: intertwining, relative solidity, and intermediate subfactors

The operator-valued-weight version is designed for type III applications. One consequence is a tracial-free reformulation of Popa’s intertwining-by-bimodules criterion. In the non-tracial setting, intertwining can be characterized by the nonexistence of certain nets of unitaries or, equivalently, by the nonexistence of suitable UCP maps in the weak Dixmier semigroup. A representative condition is that there is no net $^*$09 such that

$^*$10

strongly for all $^*$11. The new averaging map $^*$12 in the basic construction supplies the replacement for the tracial analytic criterion (Isono, 25 Aug 2025).

A second application is an extension of Ozawa’s relative solidity theorem to the type III setting. For a crossed product

$^*$13

with $^*$14 amenable and $^*$15 bi-exact, the theorem proves that for any $^*$16 with expectation,

$^*$17

This removes finiteness assumptions on the target algebra and places relative solidity within the type III crossed-product framework (Isono, 25 Aug 2025).

A third application is a Galois-type correspondence for intermediate subfactors of crossed products by totally disconnected groups. Under a proper-outerness/minimality assumption relative to a compact open subgroup $^*$18, every intermediate subfactor between $^*$19 and $^*$20 is of the form

$^*$21

for a closed subgroup $^*$22. The result applies even in type $^*$23 situations, where traces or invariant states may be absent, and it resolves a question of Boutonnet and Brothier on the structure of intermediate subfactors (Isono, 25 Aug 2025).

These applications indicate the role of the weak relative Dixmier property in current von Neumann algebra theory: it functions as an averaging principle robust enough to survive the loss of traces, and therefore as a structural substitute for tracial averaging in rigidity and decomposition problems.

6. Relation to $^*$24-algebraic Dixmier averaging and local variants

In unital $^*$25-algebras, the standard object is the Dixmier set

$^*$26

with norm closure. The Dixmier property requires

$^*$27

A unital $^*$28-algebra has the Dixmier property if and only if it is weakly central, every simple quotient $^*$29 has at most one tracial state, and every extreme tracial state of $^*$30 factors through some simple quotient $^*$31. The same circle of ideas also contains the singleton Dixmier property and the stronger uniform Dixmier property, which imposes uniform bounds on averaging complexity across the algebra (Archbold et al., 2016).

This $^*$32-algebraic theory is not the same as the weak relative Dixmier property for inclusions of von Neumann algebras, but the themes are parallel: unitary averaging, convex hulls, central or commutant targets, and trace-theoretic obstructions. In particular, the full Dixmier property in $^*$33-algebras is governed by weak centrality and tracial structure, while the weak relative Dixmier property for inclusions is governed by expectations, operator-valued weights, and commutant-valued averaging (Archbold et al., 2016).

A further development studies element-wise variants. A Dixmier element in a unital $^*$34-algebra is one for which $^*$35, and the approximate version asks for

$^*$36

A parallel theory replaces unitary mixing operators by unital completely positive elementary operators $^*$37, leading to the Magajna set $^*$38 and the condition $^*$39. For selfadjoint $^*$40, these approximate and exact local conditions are characterized by numerical-range data $^*$41 on Glimm quotients and, in the Dixmier case, by additional tracial constraints on quotients admitting tracial states (Archbold et al., 2021).

Several classes of elements are always locally averageable. Self-commutators $^*$42 and quasinilpotent products $^*$43, $^*$44 with $^*$45 quasinilpotent are always averageable to $^*$46, hence belong to both the Dixmier-type and Magajna-type local sets. Every quasinilpotent element lies in the local Dixmier set. At the same time, the local theory shows that commutator structure alone is insufficient: the paper gives a commutator which does not lie in the Magajna set (Archbold et al., 2021).

7. Terminological extensions and adjacent uses

Outside the operator-algebraic unitary-orbit setting, “Dixmier” terminology is used for several distinct rigidity and trace phenomena. In fully symmetric operator ideals, the relevant statement is not a unitary-orbit averaging property but a weak$^*$47-density theorem: if a fully symmetric operator ideal $^*$48 admits a trace, then Dixmier traces exist on $^*$49, and the set of Dixmier traces is weak$^*$50-dense in the set of all fully symmetric traces on $^*$51. Here the order structure is Hardy–Littlewood submajorization rather than unitary conjugacy, and the target objects are traces rather than central elements or relative commutants (Sukochev et al., 2014).

In Poisson algebra theory, a paper titled “Relative Dixmier property for Poisson algebras” defines a different notion entirely. For families of Poisson algebras, $^*$52 is $^*$53-Dixmier if every injective Poisson morphism

$^*$54

has image $^*$55 inside $^*$56. The paper explicitly notes that it does not define a separate notion literally called “Weak Dixmier Property”; the operative notion is the relative rigidity statement above, which is strictly stronger than ordinary Dixmier-ness and is used for cancellation and Hopf-coaction rigidity (Huang et al., 26 Jan 2026).

A similar caveat appears in the affinoid-envelope setting. The paper on affinoid Dixmier modules does not introduce a formal property called “Weak Dixmier Property,” but proves that for sufficiently high deformations $^*$57, weakly rational ideals become exactly Dixmier annihilators: $^*$58 This is a deformed Dixmier–Moeglin statement about primitive ideal theory, not an averaging property for unitary orbits (Jones, 2021).

These adjacent usages show that “Dixmier” terminology has broadened well beyond its original convex-averaging context. The operator-algebraic weak Dixmier property remains the specific assertion about averaging unitary conjugacy classes into a center or relative commutant, but the same name now also appears in trace density, embedding rigidity, and primitive-ideal classification.