{"id":"02e63e55-081c-471b-87a9-8a26d068a292","arxiv_id":"1908.07217","paper_version":2,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"Noncommutative Cartan C*-subalgebras are exactly the reduced crossed products by closed, purely outer inverse semigroup actions, and this decomposition is unique up to a canonical refinement.","lead":"The authors characterize Exel's noncommutative Cartan subalgebras in several equivalent ways, including uniqueness of conditional expectations and purely outer inverse semigroup actions. They also prove a uniqueness theorem for the crossed product decomposition, extending Renault's classical theory of commutative Cartan subalgebras to the non-separable case.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified: the Fourier-coefficient step in (8)⇒(3) is load-bearing, but its hypotheses are exactly what (8) provides.","rationale":"The reader's weakest assumption identifies Prop 2.18 as the load-bearing technical step, and I agree it is the most delicate point in the proof of (8)⇒(3). However, the reader frames this as a risk that the decomposition or projections might be unavailable; under the hypothesis (8), the action is closed, so the complementation holds. I examined the construction and found no gap: the Hilbert-module projection is well-defined on the dense union of r(E_t)E_u by the uniqueness of the inner-product formula, the contractivity claim is justified because E_t(x) lies in a slice, and the final annihilation argument in Prop 2.18 is valid. The remaining parts of Theorem 4.3 also check out, including the use of Lemma 4.10 and Prop 3.7 to pass from virtual commutants to uniqueness of conditional expectations. Since the central claim survives scrutiny, no verdict adjustment is needed.","tokens_in":30175,"tokens_out":31837,"duration_ms":319478,"concrete_test":"Re-derive the compatibility of the maps E_t in Prop 2.18 on doubly sliced subspaces: for t,u,v∈S and x∈r(E_t)E_u ∩ r(E_t)E_v, verify directly that the two local orthogonal projections onto E_t agree, using the uniqueness condition x^*E_t(y)=E(x^*y). If this fails for any closed action, E_t is not well-defined and the proof of (8)⇒(3) needs repair.","verdict_should_be":"UNCHANGED","load_bearing_attack":"I examined the proof of (8)⇒(3) in Theorem 4.3. The step using Prop 2.18 is indeed the most delicate: the generalized Fourier coefficients E_t are defined only for closed actions, relying on the complementation of I_{1,t} in s(E_t) guaranteed by Prop 2.17(4), and the contradiction with pure outerness depends on these projections. However, condition (8) explicitly assumes the action is closed and purely outer, so the complementation is available. The construction of E_t appears sound: on a slice, the C*-norm of E_t(x) equals the Hilbert-module norm, so the projection is contractive, and the local decompositions (2.5) are compatible by the uniqueness of the inner-product formula defining E_t. The inference that E_t∘φ vanishes for all t, then φ(I)=0, uses the final criterion in Prop 2.18, which is valid. I found no internal inconsistency or missing hypothesis. The other equivalence directions in Theorem 4.3 are supported by Lemma 4.10, Prop 3.7, and Prop 4.5, which appear sound.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies Exel's noncommutative Cartan subalgebras, i.e. regular C*-inclusions A⊂B with an almost faithful conditional expectation and trivial virtual commutants. The central result, Theorem 4.3, proves the equivalence of eight conditions: uniqueness of conditional expectations on the inclusions I⊂IBI for all ideals I of A, triviality of virtual commutants, a relative commutant condition, slice conditions, and, as the two final conditions, the assertion that any saturated wide grading gives a closed purely outer action with B≅A⋊_r S, and the existence of some closed purely outer inverse semigroup action with B≅A⋊_r S. Theorem 5.6 then shows that the crossed product decomposition is essentially unique: two actions producing the same noncommutative Cartan subalgebra have isomorphic refinements, and the refined action is canonically isomorphic to the tautological action of the slice inverse semigroup Sp(A,B). The paper also connects the theory to aperiodic inclusions and effective dual groupoids, thereby extending Renault's characterisation of commutative Cartan subalgebras to the non-separable case and to the noncommutative setting.","tokens_in":30403,"tokens_out":21023,"duration_ms":212299,"significance":"If correct, this is a definitive structural result for Exel's noncommutative Cartan pairs: it reduces the existence and uniqueness of a crossed product decomposition to closedness and pure outerness of an inverse semigroup action, and it upgrades Renault's theorem to the noncommutative and non-separable setting. The generalized Fourier coefficients introduced in Proposition 2.18 are a valuable new tool, and the paper carefully proves the delicate direction (8)⇒(3) of Theorem 4.3 by using closedness to construct the projections E_t and then deriving a contradiction with pure outerness. The non-separable extension, the uniqueness theorem up to refinement, and the applications to ideal detection and simplicity are substantial contributions. I found no internal inconsistency or missing hypothesis in the main argument; the proofs are detailed, and the load-bearing Fourier-coefficient step is justified by exactly the hypotheses supplied by condition (8).","major_comments":[],"minor_comments":[{"comment":"The sentence explaining condition (4) as equivalent to I'∩M(IBI)⊆M(I) 'by Lemma 3.4' is terse and can be misread: one needs Lemma 3.4 to upgrade the inclusion to centrality in M(I), not merely to pass from I' to M(I)'. Please expand this one-sentence justification.","section":"Section 4, remark after Theorem 4.3"},{"comment":"In the surjectivity part of the proof, the step 'J⊆X*E_t implies X·J=E_t·J' is used without explanation. This is plausible by the Rieffel correspondence for Hilbert subbimodules, but it is load-bearing for the uniqueness theorem and deserves a sentence of proof or an exact citation.","section":"Section 5, proof of Theorem 5.6"},{"comment":"There is a typo in 'compact Haudorff object space'; it should be 'Hausdorff unit space'.","section":"Section 3, Example 3.10"},{"comment":"In the statement, the map E_t is written with the symbol '↠' from r(E_t)·(A⋊_r S) to E_t; since E_t is a submodule of the domain and the map is a projection, the notation suggests a surjection onto E_t but could be confused with a quotient map. Using '→' and saying 'the projection onto E_t' would be clearer.","section":"Section 2.4, Proposition 2.18"},{"comment":"In the paragraph proving that (3)–(6) imply (1), the implicit reduction from A-bilinear virtual commutants to I-bilinear virtual commutants for the inclusion I⊂IBI is not spelled out. Since I 2=I, an I-bilinear map is automatically A-bilinear after identifying its implementing multiplier; adding one sentence would remove a possible source of confusion.","section":"Section 4, proof of Theorem 4.3"}],"recommendation":"minor_revision","confidential_remarks":"This is a strong paper. The central theorem is sound as far as I can verify, and the minor comments are local presentation issues. The paper leans on the authors' earlier work [21,22,23] for several technical ingredients, but the main equivalences are proven here and the reliance is normal for this line of research. I support publication after a minor revision that addresses the requested clarifications."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper settles the main open question in Exel's theory: an inclusion A⊂B is a noncommutative Cartan subalgebra exactly when B is a reduced crossed product by a closed and purely outer inverse semigroup action on A. That is the punchline, and it is a real result. Theorem 4.3's multi-way characterization—uniqueness of conditional expectations, trivial virtual commutants, slice conditions, and the outer action condition—is the core, and the proof is direct and detailed. The second main result, Theorem 5.6, gives uniqueness of the crossed product decomposition up to refinement; that subtlety is necessary, as Example 5.8 shows, and the refinement construction via dual groupoids is well handled. The paper also extends Renault's commutative Cartan theory to the non-separable setting, replacing topological principality by effectivity, and connects noncommutative Cartan inclusions with aperiodic inclusions in Section 6. That is genuinely new and useful. The generalized Fourier coefficients in Proposition 2.18 are the main technical tool, and they are constructed carefully. I checked the delicate step in (8)⇒(3): the projection E_t is defined only for closed actions, but condition (8) explicitly provides closedness, so the complementation in Proposition 2.17 is available. The contradiction with pure outerness goes through. The stress-test note's assessment matches my reading. Soft spots are minor. The paper leans heavily on prior results from the same authors' earlier work ([6], [21], [23]), which is normal for this line of research and not circular: the central implications are proven here, and the cited results are themselves published or on arXiv. The exposition is long and technical, but the structure is clear. If I have a caveat, it is that Section 6 shows noncommutative Cartan subalgebras do not always detect ideals, so the theory is less rigid than the commutative case; the paper says this explicitly and does not oversell. Who should read it: anyone working on C*-inclusions, Exel's crossed products, or groupoid C*-algebras. The paper deserves a serious referee and, I expect, acceptance after the usual small corrections. I would bring it to our reading group and would cite it in my own work.","headline":"This paper settles Exel's open question: noncommutative Cartan subalgebras are exactly reduced crossed products by closed, purely outer inverse semigroup actions, with uniqueness up to refinement; it deserves serious refereeing.","tokens_in":703,"tokens_out":1755,"would_cite":true,"duration_ms":31510,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["46L55","20M18","22A22"],"pacs":[],"model":"deepseek-v4-flash","headline":"A regular inclusion with almost faithful conditional expectation is a noncommutative Cartan subalgebra exactly when it is a reduced crossed product by a closed, purely outer inverse semigroup action.","keywords":["noncommutative Cartan subalgebra","inverse semigroup action","Hilbert bimodule","Fell bundle","conditional expectation","reduced crossed product","purely outer action","dual groupoid"],"falsifier":"Take a regular inclusion $A\\subset B$ with an almost faithful conditional expectation and an inverse semigroup grading that is closed but not purely outer, and compute the relative commutant $A'\\cap M(IBI)$ for some ideal $I$. If that relative commutant is ever larger than $Z(M(I))$, the characterization in Theorem 4.3 is false; conversely, if it is always trivial despite the action failing pure outerness, then the generalized Fourier coefficients assumption may be stronger than needed.","tokens_in":29990,"feed_emoji":"🧮","tokens_out":9097,"duration_ms":80302,"temperature":0.7,"pith_summary":"This paper establishes that noncommutative Cartan subalgebras — regular inclusions $A\\subset B$ of C*-algebras with an almost faithful conditional expectation and only trivial virtual commutants — are exactly the reduced crossed products $A\\rtimes_r S$ by inverse semigroup actions that are closed and purely outer. The characterization is proved through eight equivalent conditions, weaving together uniqueness of conditional expectations over ideals, triviality of relative commutants, and slice-theoretic forms of outerness. On top of this, the crossed product decomposition is unique up to a canonical refinement, so the inclusion determines the dynamical data that generated it. This matters because it extends the classical commutative theory of Cartan subalgebras to noncommutative and non-separable algebras, and it converts structural questions about such inclusions into questions about inverse semigroup dynamics.","feed_headline":"Noncommutative Cartan subalgebras are purely outer crossed products","feed_subtitle":"One theorem pins down when an inclusion is a reduced crossed product by a closed, purely outer inverse semigroup action.","key_machinery":"The carrying object is an inverse semigroup action on a C*-algebra by Hilbert bimodules — equivalently a saturated Fell bundle over the inverse semigroup — whose reduced crossed product is the ambient algebra. Two properties of such actions are decisive: closedness, meaning the canonical weak conditional expectation on the crossed product is $A$-valued (equivalently, the unit space of the dual groupoid is closed), and pure outerness, meaning no non-zero slice built from the annihilator of $I_{1,t}$ is isomorphic to an ideal as a Hilbert bimodule. The technical engine for the hard implication is a family of generalized Fourier coefficients $\\mathcal{E}_t$, defined only for closed actions, which project $r(E_t)\\cdot(A\\rtimes_r S)$ onto the slice $E_t$; these turn the hypothetical presence of a virtual commutant into a concrete Hilbert sub-bimodule that violates pure outerness. For uniqueness, the refinement of an action replaces $S$ by the inverse semigroup of bisections of the dual groupoid, keeping the crossed product and dual groupoid unchanged.","core_discovery":"The central claim is Theorem 4.3. For a regular C*-subalgebra $A\\subset B$ with an almost faithful conditional expectation $E:B\\to A$, the following are equivalent: uniqueness of conditional expectations $IBI\\to I$ for every ideal $I$ of $A$; faithfulness plus uniqueness of faithful conditional expectations; triviality of all virtual commutants — the definition of being a noncommutative Cartan subalgebra; the relative commutant identity $A'\\cap M(IBI)=Z(M(I))$ for every ideal; two slice-theoretic purity conditions; and, decisively, the existence of a closed and purely outer inverse semigroup action on $A$ with an $A$-preserving isomorphism $A\\rtimes_r S\\cong B$. Moreover, if these hold, every saturated, wide grading of $B$ with unit fibre $A$ is automatically closed and purely outer, and the grading yields the same reduced crossed product. The proof of the direction from an action to trivial virtual commutants uses generalized Fourier coefficients for closed actions to show that any virtual commutant would force a non-purely-outer sub-bimodule; Theorem 5.6 then shows the action is unique up to refinement, and the refined action is canonically the tautological action of the slice inverse semigroup.","pith_inferences":["The generalized Fourier coefficients suggest a practical computational criterion: to test whether a virtual commutant exists in a concrete reduced crossed product, it suffices to check whether $\\mathcal{E}_t(\\varphi(I))=0$ for all $t$; this could be applied to examples where the dual groupoid is non-Hausdorff or the action is only known to be almost closed.","Because refinement depends only on the dual groupoid's bisections, an isomorphism of Cartan pairs should force an isomorphism of dual groupoids even in settings beyond those explicitly treated, for instance for inclusions with non-unital $A$ or with almost faithful but not faithful expectations; the paper's non-separable techniques seem to leave room for this.","The failure of closedness for non-Hausdorff groupoid algebras suggests that a 'weak Cartan' theory targeting multiplier or injective-hull conditional expectations may be the right framework for non-Hausdorff dual groupoids; testing the paper's conditions on such examples would show whether the complementation hypothesis can be relaxed.","One testable consequence of the aperiodic comparison: for Cartan inclusions where $A$ is prime or has an essential Type I ideal, ideal-detection and support properties follow automatically; it would be worth checking whether these properties persist for the broader class of closed, purely outer actions when the dual groupoid is effective."],"forward_implications":["If $A\\subset B$ is a noncommutative Cartan subalgebra, then every saturated, wide inverse semigroup grading of $B$ with unit fibre $A$ is closed and purely outer, and the canonical map $A\\rtimes S\\to B$ descends to an isomorphism $A\\rtimes_r S\\cong B$.","Conversely, any closed and purely outer inverse semigroup action produces a noncommutative Cartan inclusion, so the class of examples is exactly the reduced crossed products of such actions.","Two inverse semigroup actions that present the same Cartan pair have isomorphic refinements and isomorphic dual groupoids; the refined action is canonically the tautological slice action, making $\\mathrm{Sp}(A,B)$ an intrinsic invariant of the inclusion.","When the primitive ideal space of $A$ is Hausdorff, being Cartan is equivalent to having a unique conditional expectation $E:B\\to A$, and $B$ is then the reduced section algebra of a Fell bundle over a Hausdorff, etale, locally compact groupoid, unique up to isomorphism.","In the commutative case the characterization recovers the classical groupoid model without separability: maximal Abelian subalgebras with faithful conditional expectation correspond to twists over effective, Hausdorff, etale groupoids."],"supporting_citations":[{"why":"defines noncommutative Cartan subalgebras and proves every such inclusion in a separable algebra is a reduced inverse semigroup crossed product; this is the starting point being sharpened.","marker":"[14]"},{"why":"gives the commutative Cartan characterization via twisted etale groupoids that the paper extends to the noncommutative, non-separable setting.","marker":"[26]"},{"why":"introduces pure outerness and aperiodicity for Hilbert bimodules and Fell bundles, supplying the dynamical property used in the characterization.","marker":"[21]"},{"why":"provides the dual groupoid, essential crossed product, and aperiodic inclusion technology used throughout, including ideal-detection results.","marker":"[23]"},{"why":"constructs the reduced crossed product and the canonical weak conditional expectation for Fell bundles over inverse semigroups, on which the closedness condition rests.","marker":"[6]"},{"why":"equates inverse semigroup actions by Hilbert bimodules with saturated Fell bundles and gives the duality with groupoid actions, used for the refinement and Fell bundle results.","marker":"[7]"},{"why":"studies uniqueness of conditional expectations for discrete crossed products, the comparison point for conditions (1)-(2) in the main theorem.","marker":"[27]"}],"fun_headline_variants":["Noncommutative Cartan subalgebras are purely outer crossed products","Purely outer inverse semigroup actions characterize Cartan subalgebras","Unique crossed product decomposition for noncommutative Cartan subalgebras","Cartan subalgebras = reduced crossed products by purely outer actions","Characterizing noncommutative Cartan via purely outer actions"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument that a closed, purely outer crossed product has only trivial virtual commutants depends on the generalized Fourier coefficients $\\mathcal{E}_t$, whose construction requires each ideal $I_{1,t}$ to be complemented in $s(E_t)$; if that complementation fails, the projection machinery and the contradiction establishing (8)$\\Rightarrow$(3) break down.","fun_headline_variants_meta":{"raw":{"variants":["Noncommutative Cartan subalgebras are purely outer crossed products","Purely outer inverse semigroup actions characterize Cartan subalgebras","Unique crossed product decomposition for noncommutative Cartan subalgebras","Cartan subalgebras = reduced crossed products by purely outer actions","Characterizing noncommutative Cartan via purely outer actions"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000489,"raw_usage":{"total_tokens":2378,"prompt_tokens":887,"completion_tokens":1491,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":503,"completion_tokens_details":{"reasoning_tokens":1399}},"tokens_in":503,"tokens_out":1491,"duration_ms":10471,"temperature":1.0,"reasoning_tokens":1399,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:22:28.735877+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a regular inclusion $A\\subset B$ with an almost faithful conditional expectation and an inverse semigroup grading that is closed but not purely outer, and compute the relative commutant $A'\\cap M(IBI)$ for some ideal $I$. If that relative commutant is ever larger than $Z(M(I))$, the characterization in Theorem 4.3 is false; conversely, if it is always trivial despite the action failing pure outerness, then the generalized Fourier coefficients assumption may be stronger than needed.","supporting_citations":[{"cited_title":"Math.17 (2011), 331–382, available athttp://nyjm.albany.edu/j/2011/17-17.html","cited_arxiv_id":null,"evidence_quote":"defines noncommutative Cartan subalgebras and proves every such inclusion in a separable algebra is a reduced inverse semigroup crossed product; this is the starting point being sharpened."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"gives the commutative Cartan characterization via twisted etale groupoids that the paper extends to the noncommutative, non-separable setting."},{"cited_title":"Essential crossed products for inverse semigroup actions: simplicity and pure infiniteness","cited_arxiv_id":"1906.06202","evidence_quote":"provides the dual groupoid, essential crossed product, and aperiodic inclusion technology used throughout, including ideal-detection results."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"studies uniqueness of conditional expectations for discrete crossed products, the comparison point for conditions (1)-(2) in the main theorem."}],"review_version":1}