{"id":"2f285ca5-b7aa-42b2-9568-d8b3a26b084b","arxiv_id":"2607.20593","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Kaplansky's second test problem on similarity is solved affirmatively for operators with property (J) in type I_n von Neumann algebras, for all type I_n with n≤3, and for Banach-algebra elements with essentially finite-dimensional commutant.","lead":"The paper proves that, for a large class of operators in matrix-like von Neumann algebras and in Banach algebras, if an operator's double is similar to another operator's double, the two operators are already similar. It settles a special case of Kaplansky's second test problem, a decades-old open question in operator theory.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition 3.21 Case II uses an unjustified implication P_jj = P11 when T_{j,j+1} is not invertible; missing central-projection reduction threatens Theorem 3.27.","rationale":"The central claim is a positive answer to Kaplansky's second test problem under property (J) in type I_n von Neumann algebras. The proof hinges on uniqueness of finite strongly irreducible decompositions (Theorem 3.26), which relies on Proposition 3.21. I examined the radical step flagged by the reader: that step is actually safe, since any subalgebra of block upper triangular matrices has its strictly block upper triangular part as a nilpotent ideal, hence contained in the Jacobson radical. The weaker point is the equality P_jj = P11. The paper's proof of Corollaries 3.22 and 3.23 and Theorem 3.27 uses this to construct Q. If the missing step is merely a spectral localization, the theorem may still be true and the verdict should remain conditional pending the repair; if the equality is actually false in a way that cannot be repaired, the central result would be invalid. The unproved Remark 5.19(1) and the overclaimed LU application are secondary and do not affect the main theorems.","tokens_in":41425,"tokens_out":29585,"duration_ms":257585,"concrete_test":"Verify Proposition 3.21 Case II on A = L∞([0,1]) with T = (x J_2) ⊕ ((1-x)J_2). Compute the commutant of Φ(T), take the idempotent P with P11 = diag(1,0), P22 = [[1,0],[χ_{x=1},0]], P12 = 0. (a) Confirm P ∈ {Φ(T)}' and P22 ≠ P11, so the proof's implication fails. (b) Determine whether an invertible X in the commutant diagonalizes P. (c) If yes, identify the missing argument and check it extends to all of A; if no, Proposition 3.21 is false and Theorem 3.27 collapses.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Proposition 3.21 is the engine behind Theorems 3.26 and 3.27. In Case II (n_1 = ... = n_m = n), after the permutation Φ, the proof obtains P_jj T_{j,j+1} = T_{j,j+1} P_{j+1,j+1} and says 'Since P11 and T_{j,j+1} are diagonal, P_jj = P11.' This is only valid if T_{j,j+1} is invertible. Theorem 3.10 guarantees only R(t^k_{j,j+1}) = I_A, not invertibility. Example: A = L∞([0,1]), T = (x J_2) ⊕ ((1-x)J_2), so D = diag(x,1-x) is not invertible; with P11 = diag(1,0), P22 = [[1,0],[z,0]] and z supported on {x=1}, the intertwining equation holds but P22 ≠ P11. The proof does not supply the standard reduction to a clopen set where all t^k_{j,j+1} are invertible, nor explain how the complementary set is handled, so the diagonal operator Q used later may not lie in the commutant and the radical/diagonalization step is unsupported. This is a genuine gap in a load-bearing lemma.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies Kaplansky's second test problem for similarity: if T⊕T is similar to S⊕S in M2(B), must T be similar to S in B? The authors prove an affirmative answer for operators with property (J) in type In von Neumann algebras (Theorem 3.27), and for elements with essentially finite-dimensional commutant in unital Banach algebras (Theorem 5.13). They also prove a reduction theorem (Theorem 4.3) allowing reduction to Mn(ℓ∞), remove the property (J) assumption for 1 ≤ n ≤ 3 (Proposition 4.7), and discuss applications to weighted shifts and to local unitary equivalence of quantum states.","tokens_in":41800,"tokens_out":30884,"duration_ms":247761,"significance":"If the proofs are completed, this is a substantial contribution to a long-standing problem in operator theory. The Banach-algebra part, culminating in Theorem 5.13, is elegant and appears to be correct; the radical/Wedderburn argument is transparent and gives a genuinely new general result. The reduction theorem (Theorem 4.3) is also a valuable structural tool. The von Neumann-algebra part rests on a long structural analysis (Proposition 3.21) whose proof is not fully written out, and on a case analysis for n = 3 that contains an unjustified algebraic step. These gaps are local and likely repairable, but they are load-bearing for the main type-In claims.","major_comments":[{"comment":"The proof of this key lemma is incomplete. After Case I and Case II, the text says that Case III follows by a combination of the two, but it only proves the single configuration N=7, n1=3, n2=n3=2, and the concluding 'General Case' paragraph merely reduces to Case III by grouping traces. Since Proposition 3.21 drives Corollaries 3.22 and 3.23 and hence Theorem 3.27, a complete argument for arbitrary block decompositions is needed, not just an illustrative example.","section":"Proposition 3.21, §3.4"},{"comment":"The proof uses a11^{-1} in the construction around the claim '|p22,11| ≥ 1/2 IA up to similarity', while the standing assumption in Case 2 is only R(a11)=IA. In a commutative von Neumann algebra, full range projection does not imply invertibility (for example, multiplication by x on L∞[0,1] has range projection I but no bounded inverse). The argument needs a spectral-localization step (or a different device) before inverting a11. As written, this invalidates the proof of the n=3 case, which is a stated main result.","section":"Proposition 4.7, Case 2.1, §4.2"}],"minor_comments":[{"comment":"The step 'Since P11 and Tj,j+1 are diagonal, Pjj = P11' is terse. It is correct, but only after using that each diagonal entry of Tj,j+1 has range projection IA, so multiplication by that entry is injective on A and forces the off-diagonal entries of Pjj to vanish. Please spell this out; otherwise the claim appears to depend on invertibility of Tj,j+1, which is not available.","section":"Proposition 3.21, Case II, §3.4"},{"comment":"The assertion that a strictly (block) upper-triangular element of the relative commutant lies in the radical is not proved. It is true: the set of such elements is a two-sided ideal and is nilpotent. A one-sentence justification would improve the exposition.","section":"Proposition 3.21, Cases I and II"},{"comment":"The expression 'x = a11^{-1} a22 = (IA - p22,11)^{-1}p12,11' contains an undefined symbol 'a22' and appears garbled. The intended formula should be stated cleanly.","section":"Proposition 4.7, Case 2.1"},{"comment":"The letter Q is used both for the maximal abelian family and for the idempotent X^{-1}(P1⊗0)X, which is confusing. Use different notation for the family and the idempotent.","section":"Theorem 3.27"}],"recommendation":"major_revision","confidential_remarks":"The Banach-algebra theorem and the reduction theorem are strong and likely publishable. The type-In part needs a completed proof of Proposition 3.21 and a fix for the unjustified inversion in Proposition 4.7. I believe both are fixable within the scope of the manuscript, hence major revision rather than rejection."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The paper is a genuine advance. It proves new partial answers to Kaplansky's second test problem: for operators with property (J) in type I_n von Neumann algebras, for all operators in type I_n with n ≤ 3, and for Banach-algebra elements with essentially finite-dimensional commutant. The property (J) framework, the uniqueness theorem for finite strongly irreducible decompositions, and the reduction theorem to M_n(ℓ∞) are real contributions. The Banach-algebra result (Theorem 5.13) is elegant and the proof via Wedderburn and dimension counting is convincing. The self-citations are confined to remarks, so the main results are not riding on them.\n\nThe soft spots are real but not fatal. The abstract's sentence about applying the results to local unitary equivalence is overclaimed: Remark 5.20 merely points at a connection and cites the authors' separate paper, no LU theorem is proved here. Remark 5.19(1) makes an assertion about Jordan operators without proof. The casework in Proposition 4.7 is long and intricate; I did not verify every matrix computation, and the trace/cut-down arguments are compressed. A referee should ask for details there.\n\nThe stress-test concern about Proposition 3.21, Case II, does not hold up. The implication P_jj = P11 from P_jj T_{j,j+1} = T_{j,j+1} P_{j+1,j+1} is justified by the fact that each t^k_{j,j+1} has range projection I_A. If P_jj and P_{j+1,j+1} differed on a nonzero projection p, multiplying by p would force t^k_{j,j+1} p = 0, contradicting the range projection condition. In L∞ the proposed counterexample is an artifact of null sets; in the continuous-function model, equality on a dense set forces equality everywhere. So that particular gap is not real.\n\nThe paper's main weakness is structural rather than a single false step: the proof of Proposition 3.21 asserts that the strictly block-upper-triangular part of the commutant lies in the radical. That is plausible and probably follows from Corollary 3.11 and standard facts about upper-triangular matrix algebras over a semisimple base, but it is not explicitly proved. An editor should send this to a capable referee and ask for that argument to be spelled out, and for the LU sentence to be toned down. But the central theorems appear substantive and likely correct.\n\nRecommendation: send it to peer review. It deserves serious referee time.","headline":"A serious partial answer to Kaplansky's second test problem; the flagged gap in Proposition 3.21 does not survive contact with the paper.","tokens_in":42207,"tokens_out":6893,"would_cite":true,"duration_ms":70185,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["47A45","47A65"],"pacs":[],"model":"deepseek-v4-flash","headline":"Doubling an operator cannot hide a similarity failure: two copies similar forces one copy similar in the settings proved here.","keywords":["Kaplansky's second test problem","similarity","type I_n von Neumann algebra","property (J)","strongly irreducible decomposition","Jacobson radical","essentially finite-dimensional commutant","direct sums"],"falsifier":"Exhibit, for some n≥4 and abelian von Neumann algebra A, an operator T in Mn(A) such that T⊕T is similar to S⊕S in M2(Mn(A)) but T is not similar to S; equivalently, find t≥1 with ρ_n(t)=∞, or produce an idempotent in the relative commutant of a direct sum of strongly irreducible upper-triangular operators whose difference from its diagonal part is not in the radical.","tokens_in":41372,"feed_emoji":"🔁","tokens_out":5084,"duration_ms":56284,"temperature":0.7,"pith_summary":"Kaplansky's second test problem asks whether similarity of doubled elements is enough to conclude similarity of the original elements. This paper proves that the answer is yes for a large class: any operator with property (J) inside a type I_n von Neumann algebra — meaning its relative commutant contains a bounded maximal abelian family of idempotents — and, in the same algebras with n≤3, for every operator without exception. It proves the analogous cancellation for elements of unital Banach algebras whose relative commutant is finite-dimensional modulo the Jacobson radical. The strategy is to decompose operators into strongly irreducible summands, prove that such decompositions are unique up to similarity, and then show that a similarity between doubled operators must match the summands one by one. If the results stand, the long-open test problem is settled affirmatively across these settings and tied to a concrete reduction to matrix-valued functions on ℓ∞.","feed_headline":"Similarity of doubles forces similarity of originals","feed_subtitle":"For operators with property (J), matching T⊕T∼S⊕S forces T∼S; for n≤3, no condition needed.","key_machinery":"Property (J): an operator T in a von Neumann algebra M has property (J) when its relative commutant {T}'∩M contains a bounded maximal abelian family of idempotents. In type I_n algebras this is equivalent to having a finite strongly irreducible decomposition, and the paper's Theorem 3.26 shows such decompositions are unique up to similarity — the key mechanism that turns a similarity between T1⊕T1 and T2⊕T2 into a matching of irreducible summands. The Banach-algebra version replaces idempotent families by the quotient Q(T,B)=(T,B)'/Rad((T,B)'); finite dimensionality of this quotient forces a finite Wedderburn block decomposition, enabling the same matching argument.","core_discovery":"On its own terms, the paper establishes two cancellation theorems. Theorem 3.27: if T1 has property (J) in a type I_n von Neumann algebra M and T1⊕T1 is similar to T2⊕T2 in M2(M), then T1 is similar to T2 in M. Theorem 5.13: in any unital Banach algebra B, if T1 has essentially finite-dimensional commutant and T1^(n) is similar to T2^(n) in Mn(B), then T1 is similar to T2 in B. The proof hinges on showing that every finite strongly irreducible decomposition of an operator with property (J) is unique up to similarity: any two bounded maximal abelian families of idempotents in the relative commutant are conjugate by an invertible element of that commutant. For 1≤n≤3, the paper shows property (","pith_inferences":["If the same diagonalization mechanism works without property (J) in higher dimensions, Question 1.1 would be resolved; the natural test case is n=4 in Mn(ℓ∞), where the similarity function ρ_4 could be studied numerically or constructively.","The proof's reliance on the unproved containment of the strictly upper-triangular commutant in the radical suggests a precise stress test: check that containment in a concrete 3+2 block example, since Proposition 3.21's conclusion depends on it.","The Banach-algebra theorem may extend to elements whose commutant is finite-dimensional modulo some other two-sided ideal, provided the quotient remains semisimple enough for a Wedderburn-type decomposition to apply."],"forward_implications":["In every type I_n von Neumann algebra, operators with property (J) pass Kaplansky's second test problem: T1⊕T1∼T2⊕T2 implies T1∼T2.","For type I_1, I_2, and I_3 algebras the test problem has an unconditional affirmative answer, since property (J) holds automatically.","The reduction theorem makes the full type I_n question equivalent to a numerical condition on the similarity function ρ_n, so it suffices to analyze Mn(ℓ∞).","In unital Banach algebras, elements with finite-dimensional commutant modulo the radical also pass the test; in particular this covers strictly cyclic quasinilpotent weighted shifts.","As the paper notes, these results connect to local unitary equivalence of quantum states, where a related cancellation question has been posed."],"fun_headline_variants":["Doubles similar? Then originals similar, with property (J)","Kaplansky's second test: solved for property (J) operators","Similarity of 2x2 blocks implies similarity in type I_n","When T⊕T ~ S⊕S forces T ~ S","New cancellation: finite commutant yields similarity"],"cache_read_input_tokens":2304,"weakest_assumption_plain":"In Proposition 3.21, the assertion that an idempotent in the commutant of a direct sum of strongly irreducible upper-triangular operators is similar to the diagonal formed from its diagonal entries uses the claim that the strictly upper-triangular part of that commutant lies in its radical; this inclusion is asserted but not proved, and the later theorems inherit it.","fun_headline_variants_meta":{"raw":{"variants":["Doubles similar? Then originals similar, with property (J)","Kaplansky's second test: solved for property (J) operators","Similarity of 2x2 blocks implies similarity in type I_n","When T⊕T ~ S⊕S forces T ~ S","New cancellation: finite commutant yields similarity"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000198,"raw_usage":{"total_tokens":1224,"prompt_tokens":784,"completion_tokens":440,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":528,"completion_tokens_details":{"reasoning_tokens":353}},"tokens_in":528,"tokens_out":440,"duration_ms":4032,"temperature":1.0,"reasoning_tokens":353,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-01T10:11:47.771398+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Exhibit, for some n≥4 and abelian von Neumann algebra A, an operator T in Mn(A) such that T⊕T is similar to S⊕S in M2(Mn(A)) but T is not similar to S; equivalently, find t≥1 with ρ_n(t)=∞, or produce an idempotent in the relative commutant of a direct sum of strongly irreducible upper-triangular operators whose difference from its diagonal part is not in the radical.","supporting_citations":[],"review_version":1}