1. The question
Can a complicated operator algebra contain a simpler subalgebra that imposes exactly the same commutation constraints on everything around it? The answer depends on what counts as “around it.” In modular theory, the surrounding algebra includes not only the original operators but also unitaries recording a canonical time evolution.
Family 290 claims that an expected inclusion of von Neumann algebras contains an amenable expected subalgebra preserving the commutant in this enlarged continuous core. A companion manuscript develops a bounded-recovery principle: a signal detected on narrow spectral bands of Hilbert-space vectors can also be detected by uniformly bounded algebra elements.
The papers connect these two statements to bicentralizers. They also give alternative arguments for the absolute bicentralizer theorem, explicitly crediting an earlier September 2026 announcement by Houdayer and Marrakchi. The relative inclusion theorem and the analytical recovery tool are the family’s distinct contributions.
2. Background and essential definitions
A von Neumann algebra is an adjoint-closed algebra of bounded operators closed in the weak operator topology. For a subalgebra , its relative commutant in is
An inclusion is expected if there is a faithful normal conditional expectation . This is a positive, identity-preserving projection compatible with multiplication by elements of ; it acts like averaging away the information outside . Faithfulness detects positive elements, and normality is the relevant continuity for monotone operator limits.
A faithful normal state assigns normalized positive expectations to operators. It determines a modular automorphism group , a canonical real-parameter evolution. Its fixed algebra
is the state’s centralizer. The continuous core adjoins unitaries implementing this evolution. Changing the faithful state gives a canonical core identification fixing , so the commutant assertion below is independent of that choice.
An amenable von Neumann algebra here means an injective one, a particularly tractable class possessing strong approximation and extension properties. This does not mean that the larger algebra is amenable.
The bicentralizer consists of operators commuting asymptotically with every uniformly bounded sequence that asymptotically centralizes the state. State centralization is measured in the norm of the predual, the space of normal linear functionals. A scalar bicentralizer means that only multiples of the identity pass all these tests.
3. Development history and previous partial results
Connes introduced the bicentralizer problem in the classification of injective factors. Haagerup proved triviality for amenable type factors, completing the uniqueness theorem for the injective factor of that type. Type III factors lack the finite trace used in theory; the subscript one describes the case with full modular scaling behavior.
Popa’s work on maximal abelian subalgebras and irreducible inclusions supplied a related strategy: find a tractable subalgebra retaining the required commutation constraints. Haagerup linked bicentralizer triviality to states with irreducible centralizers. “Irreducible centralizer” means its relative commutant is scalar; it does not mean that the centralizer itself is scalar.
Masuda introduced a relative bicentralizer in a finite-index setting. Ando, Haagerup, Houdayer and Marrakchi developed a canonical relative bicentralizer flow. Marrakchi’s later work formulated the general relative conjecture and related it to commutants inside the continuous core and to averaging of unitary orbits.
A preceding theorem of Marrakchi supplies an expected amenable preserving . M1 seeks equality inside , which is stronger. One cannot simply replace the upper algebra by its core in that theorem: the original expectation onto does not automatically extend to an expectation from the core.
On September 10, 2026, Houdayer and Marrakchi announced absolute bicentralizer triviality for every type factor admitting a faithful normal state, without a separability restriction in the statement. Both manuscripts acknowledge that result. Their absolute arguments are alternatives for the separable-predual case, rather than claims of first resolving the absolute problem.
4. What the new manuscripts claim
M1: preserve the full core commutant
Suppose is a unital inclusion with separable preduals and a faithful normal conditional expectation . M1 states that there is a unital amenable , itself the range of a faithful normal conditional expectation from , such that
This allows arbitrary centers and types. The smaller algebra imposes exactly the same commutation constraints as , even on the added modular operators.
M1 also obtains an expected maximal abelian subalgebra satisfying
The symbol means the von Neumann algebra generated by the two indicated subalgebras. The expectations specify the compatible embeddings of the cores. A further conclusion that is maximal abelian in uses the additional core condition stated in M1; it is not asserted for every inclusion solely from the displayed formula.
M2: recover spectral signals with bounded operators
Assume a faithful normal state has scalar centralizer, . On the standard Hilbert space put , , and , where is the modular operator.
Fix an arbitrary bounded Hilbert-space operator and a real spectral point . Suppose unit vectors have -spectral support in , with , and
Here is the fixed translation-invariant mean obtained from symmetric interval averages as specified by M2, rather than an assumed ordinary time limit.
M2 claims a subsequence , elements , and constants , with
whose vectors have support in . Thus the signal survives on algebra elements with one uniform operator-norm bound.
M2 applies this to a spectral-intertwining rigidity theorem and derives for type factors with separable predual.
5. Why the advance matters
Importance: 5/5. If correct, M1 resolves the general relative bicentralizer conjecture in its separable-predual expected-inclusion setting. The core formulation retains information invisible to ordinary commutants and makes the theorem useful across different operator-algebra types.
The bounded-recovery result addresses a recurrent analytical obstacle. Hilbert-space approximation can produce vectors represented by algebra elements whose operator norms grow without bound. Such approximations cannot be inserted into hypotheses about uniformly bounded sequences. M2 supplies a way across that gap while retaining increasingly narrow spectral localization.
The significance is not that the absolute bicentralizer conjecture was previously untouched. The earlier Houdayer–Marrakchi announcement is part of the baseline. The family adds a relative core statement and a general recovery mechanism with its own precise assumptions.
6. The main idea
Turn commutation into comparison of multiplication maps
For the type part of M1, let be the relative bicentralizer and the vector representing the state. On suitable tensor-product vectors, consider the two orders of multiplication:
The manuscript proves these maps are isometries and relates them through the generators of modular evolution and bicentralizer flow. This comparison packages the failure of ordinary commutation into an explicit spectral phase.
A tensor argument controls words of arbitrary length and intervening flow shifts uniformly. Randomized repeated tests eliminate a possible discrepancy between the proposed spectral comparison and the actual multiplication maps. Passing to the core then identifies the fixed algebra of the relative flow with the desired core commutant.
Existing characterizations convert that fixed-point result into the expected amenable subalgebra. Other factor types and central decompositions extend the conclusion beyond the initial type case.
Average, truncate and restore the band
M2 starts with spectral vectors that need not come from uniformly bounded algebra elements. It first approximates them by individual bounded representatives; their bounds may depend on the vector.
Modular orbit averages and independent phases provide second- and fourth-moment control. Fourth moments limit how much mass can hide in very large operator values, allowing one fixed polar-truncation threshold to produce a uniformly bounded sequence without losing the signal.
Truncation can spoil spectral localization. A final Fourier filter restores support near the original spectral point. Its norm estimate is independent of the shrinking interval width, which is why the uniform bound survives this last step. The output band may be four times wider, but still shrinks to the same point.
Extract rigidity from joint spectra
For a state-preserving action satisfying the manuscript’s asymptotic intertwining condition, its implementing unitaries commute with modular evolution. Their generators have a joint spectral support in the plane. Bounded recovery lets the sequence-level intertwining hypothesis constrain this support.
Applying the spectral comparison in both directions forces a restrictive symmetric phase relation. The scalar-centralizer assumption rules out the unwanted modular eigenvalues, leaving the action trivial. Standard bicentralizer reductions then yield the absolute application.
7. Scope, limitations, and open questions
M1 requires separable preduals and a faithful normal conditional expectation for the original inclusion. It does not claim the same conclusion for arbitrary nonexpected inclusions. The core equality is independent of the state used to construct the core, but not independent of these structural hypotheses.
M2 requires a scalar centralizer, not merely an irreducible one. Its input is a positive averaged limsup for a fixed and fixed spectral point; its conclusion passes to a subsequence and allows constants depending on the data. It is not a norm-bounded representation of every spectral vector, nor a universal quantitative recovery rate.
The expected maximal abelian subalgebra obtained for a factor need not preserve the particular scalar-centralizer state used in the recovery argument. Keeping those two existence statements separate avoids an extra conclusion absent from the manuscripts. Extensions beyond the stated hypotheses and sharper recovery bounds are distinct further questions.
8. Sources and further reading
- Houdayer and Marrakchi, The classification of flows on II1 factors and Connes’ bicentralizer problem, the September 10 absolute theorem acknowledged by both sources.
- Marrakchi, Kadison’s problem for type III subfactors and the bicentralizer conjecture, for the relative-core characterization.
- Marrakchi, Ergodicity of the bicentralizer flow and Kadison’s problem, for the preceding expected-subalgebra result used by M1.
Source manuscripts
- M1. OpenAI, Expected amenable subalgebras preserving core commutants (September 23, 2026).
- M2. OpenAI, Bounded recovery for modular spectral averages (September 23, 2026).
Sources consulted on 25 October 2026 at repository revision 1581d4c. Download the manuscript citations.