{"id":"a927d903-2272-499b-8817-07742e1feefd","arxiv_id":"1908.02705","paper_version":5,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper constructs the first non-nuclear C*-algebra with both WEP and LLP, as an inductive limit of finite-dimensional subspaces with approximately multiplicative linking maps.","lead":"A mathematician constructed the first example of a non-nuclear C*-algebra that has both the Weak Expectation Property and the Local Lifting Property. The example is built to share its finite-dimensional subspaces with the free group C*-algebra, a setting tied to the open Connes-Kirchberg problem.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The proof of Theorem 7.2 never justifies non-nuclearity: no exactness argument or non-exact choice of (Z_n) appears, and the line 'hence ab∈∪Y_n' only gives membership in the closure.","rationale":"I read the construction carefully and believe the central WEP/LLP mechanism is sound. The reader's flagged concern about C0(B) having the WEP is easily resolved: since C0 is nuclear, C0(B) = C0 ⊗min B has the WEP whenever B does, so that step is not a real obstruction. The bigger issue is that Theorem 7.2's proof asserts non-nuclearity without proving it: the final proof never shows A is non-exact, and the promised use of an initial E1 with exactness constant > 1 is not connected to the formal induction. There is also a small closure gap: the product estimate shows only that ab lies in the closure of ∪Y_n, and the line claiming ab ∈ ∪Y_n is unjustified. Both issues are repairable by taking A to be the norm closure and by choosing (Z_n) to contain non-exact subspaces or to be dense in C0(C), so I do not think the result is false. The conditional verdict remains appropriate, and my read does not change the reader's overall assessment.","tokens_in":25471,"tokens_out":28901,"duration_ms":314249,"concrete_test":"Instantiate Theorem 7.2 with Z_1 = span{u_1, u_2, u_3} ⊂ C0(C), where u_j are free unitary generators, so that Z_1 has exactness constant > 1. Trace Lemma 7.6 to confirm that Y_1 = w_1(Z_1) satisfies d_cb(Y_1, Z_1) ≤ (1+η_1)^2, and check whether this forces A to be non-exact, since exactness constants change by at most this cb-isomorphism factor. If this check succeeds, non-nuclearity is restored by adding a sentence to the proof; if it fails, Theorem 7.2 needs an explicit hypothesis on (Z_n).","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim of Theorem 7.2 is the existence of a non-nuclear C*-algebra with WEP and LLP. The proof establishes, in sequence, that A is closed under products and adjoints, that A has the WEP via Proposition 2.1, and that A has the LLP via local embeddability in C and Proposition 3.7. It never proves that A is non-nuclear or non-exact. The outline in Section 4 promises that 'starting from a f.d. operator space E1 with exactness constant > 1 will ensure that X is not exact', but Lemma 7.6 starts with E1 = Z1 for an arbitrary sequence (Z_n), and the proof of Theorem 7.2 does not impose or use any condition on the exactness constants of the Z_n. The 'Moreover' clause only gives almost isometric copies of the chosen Z_n inside A; without choosing (Z_n) to contain subspaces with unbounded exactness constants, or choosing (Z_n) dense so that A and C are locally equivalent, non-nuclearity does not follow from anything written. A second, smaller gap is that A is defined as the union ∪Y_n, but the product estimate (7.7) gives d(ab, Y_{n+k+1}) ≤ Cδ_{n+k} → 0, which places ab in the closure of ∪Y_n, not necessarily in the union. The sentence 'and hence ab ∈ ∪Y_n' is therefore not justified; the proof should take A to be the closure, after which the WEP and LLP arguments still work. Both gaps are fixable, but the non-nuclearity gap is load-bearing because it is the title property and no argument for it is supplied in the theorem proof.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper constructs a separable C*-algebra A that is claimed to be non-nuclear while having both the Weak Expectation Property (WEP) and the Local Lifting Property (LLP). The construction proceeds inductively: starting from finite-dimensional self-adjoint subspaces Z_n of C0(C*(F_∞)), Lemma 7.1 produces almost multiplicative, almost isometric embeddings T_n : E_n → E_{n+1}; Lemma 7.6 assembles these into a sequence satisfying extension, cb-norm, and approximate multiplicativity conditions; Theorem 7.2 then realizes A inside the ultrapower L = ℓ_∞(C0(C))/c0(C0(C)) as the union of the Y_n = Qθ_n(E_n). The proof claims that A is a C*-subalgebra, has the WEP by Proposition 2.1, and has the LLP by local embeddability into C0(C) plus Proposition 3.7. A second part (Section 9) reformulates the construction in the setting of a sequence of C*-algebras for which each C_n is a subquotient of C_{n+1}, and uses this to transfer local properties of the linking maps to the identity map of the resulting algebra.","tokens_in":25860,"tokens_out":8388,"duration_ms":92615,"significance":"If the construction is correct, this is a significant result: it answers an implicit question of Kirchberg and gives new examples of non-nuclear C*-algebras with unique C*-norm on A ⊗ A^{op}. The strategy, combining local lifting with almost multiplicative maps and the LLPs of free group C*-algebras, is original and the estimates in Lemma 7.6 and the product estimate (7.5) are detailed and non-trivial. The paper also gives a general framework (Section 9) that may be useful for transferring other local properties. However, as written, the proof of Theorem 7.2 does not establish the title property (non-nuclearity), and the definition of A as a union rather than a closure creates a gap in the proof that A is a C*-algebra. These issues are load-bearing and require a major revision.","major_comments":[{"comment":"The theorem asserts the existence of a non-nuclear C*-algebra A, but the proof never proves non-nuclearity (or non-exactness). The proof shows that A has the WEP and the LLP and that each Z_n embeds almost isometrically into A; it does not show that C0(C), or any non-exact space, locally embeds into A. Section 4 promises that starting from a finite-dimensional space E1 with exactness constant >1 will ensure non-exactness of X, but Lemma 7.6 starts with E1 = Z1 for an arbitrary sequence (Z_n) and no exactness hypothesis is used in the proof of Theorem 7.2. Since exactness is a local property, if (Z_n) is chosen, for example, to be a d_cb-dense sequence so that A and C are locally equivalent (Remark 7.5), then A is non-exact and hence non-nuclear; but this choice is not stated in the theorem and the argument is not supplied. The non-nuclearity claim therefore needs either a suitable extra hypothesis on (Z_n) together with a proof of local equivalence, or a direct proof that the construction forces non-exactness.","section":"§7, Theorem 7.2"},{"comment":"The proof defines A = ∪Y_n and then uses the product estimate (7.7), which gives d(ab, Y_{n+k+1}) → 0 for a,b ∈ Y_n, to conclude 'hence ab ∈ ∪Y_n'. This inference is invalid: (7.7) only places ab in the closure of ∪Y_n. Since a C*-subalgebra of L must be closed, the construction should define A as the closure of ∪Y_n. The subsequent WEP argument already uses density ('we may obviously assume by density that u(S) ⊂ ∪Y_m'), which is consistent with A being the closure. Defining A as the closure is a fixable modification, but as written the set A is not shown to be closed under products.","section":"§7, after Eq. (7.7)"},{"comment":"The 'Moreover' clause states that for any n and any ε > 0 there is a subspace Z ⊂ A with d_cb(Z_n, Z) < 1+ε for an arbitrary sequence (Z_n). The proof explicitly adds a hypothesis not present in the statement: 'if we arrange the sequence (Z_n) so that each space in it is repeated infinitely many times'. For an arbitrary sequence, a fixed Z_n occurs only once, and the available estimate d_cb(Z_n, Y_n) ≤ (1+η_n)^2 = O(1+ε) is not enough to make the distortion below 1+ε for a prescribed ε when n is fixed. The theorem should either include the repetition hypothesis, or state the approximation property only for a cofinal subsequence of the Z_n.","section":"§7, Theorem 7.2, 'Moreover' clause"}],"minor_comments":[{"comment":"The sentence 'Since B and hence C0(B) has the WEP' is used to produce the extension v of i0u, but no justification is given for why the WEP passes from B to C0(B). This is true (C0(B) = C0 ⊗ B with C0 nuclear), but the implication should be stated or referenced for completeness.","section":"§7, Lemma 7.1"},{"comment":"The proof says 'The proof that A = ∪Y_n is a C*-subalgebra is entirely analogous to that of Theorem 7.2 so we skip it.' Given the closure issue in Theorem 7.2, the analogous proof is not entirely trivial: the same union-versus-closure problem occurs. The theorem and its proof should define A as the closure of ∪Y_n (or state that the closure is taken) and indicate that the remaining arguments are unaffected.","section":"§9, Theorem 9.2"},{"comment":"There are several typos: 'tends pointwise to to 0' in the proof of Proposition 2.1, 'same conclusion hods' in the proof of Theorem 7.2, and 'unitizat ion' in Remark 7.3. These should be corrected.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The non-nuclearity gap is the most serious issue: the title claim is simply not proved for the theorem as stated. The fix is straightforward for an expert (choose (Z_n) dense in the d_cb sense and use exactness as a local property), but it must actually be written out. The closure issue is also easy to fix. The 'Moreover' clause needs a matching hypothesis. Given the author's standing and the detail elsewhere in the paper, I expect these to be repairable within a revision; however, the current version cannot be accepted as the main result is not established."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Pisier's paper is the real thing—first example of a non-nuclear C*-algebra with both WEP and LLP, answering an explicit question from Brown-Ozawa. The construction is original: a sequence of finite-dimensional subspaces with almost multiplicative linking maps, built via the cone algebra trick, and the general framework in Section 9 is a nice bonus. The estimates in Lemma 7.6 and the proof of WEP/LLP are careful and mostly check out.\n\nBut there is a real problem with the advertised headline. Theorem 7.2 states that for any sequence (Z_n) of f.d.s.a. subspaces of C0(C), the constructed A is non-nuclear. The proof never addresses non-nuclearity. It proves A is a C*-algebra with WEP and LLP, and the 'Moreover' clause, but there is no exactness constant argument anywhere. That is not a minor omission: if you take Z_n = 0, the construction gives A = 0, which is nuclear, so the statement as written is false. The intended theorem must include a hypothesis on (Z_n), e.g. that they form a d_cb-dense sequence in C0(C) (then A is locally equivalent to C0(C) and non-exactness follows). The outline in Section 4 suggests the author had this in mind, but it didn't make it into the theorem statement or proof. This has to be fixed.\n\nThere is also a smaller gap: A is defined as the union ∪Y_n, but the product estimate (7.7) only places ab in the closure. The line 'hence ab ∈ ∪Y_n' is not justified; taking A to be the closure fixes it, and the WEP/LLP arguments survive by density. Minor but should be corrected.\n\nOne other thing: Lemma 7.1 uses 'B and hence C0(B) has the WEP' without proof or citation. It's probably standard, but it deserves a reference, since the whole inductive step leans on it.\n\nThe general Theorem 9.2 is sketched—the proof that A is a C*-subalgebra is skipped, and the same closure issue applies. That is a soft spot in the second half, but the framework is clearly valuable.\n\nNet: the core construction is sound and the result is significant. The non-nuclearity gap is fixable but load-bearing; the paper needs a serious revision before publication. I would still send it to a good referee—the ideas are important and the gaps are repairable. This deserves peer review, not desk rejection.","headline":"Important construction with a fixable but load-bearing gap: the proof never actually proves non-nuclearity, and the theorem as stated is false for trivial sequences like Z_n = 0.","tokens_in":26375,"tokens_out":6781,"would_cite":true,"duration_ms":65333,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["46L06","46L07","46L09"],"pacs":[],"model":"deepseek-v4-flash","headline":"A non-nuclear C*-algebra can still have both the weak expectation and local lifting properties.","keywords":["non-nuclear C*-algebra","weak expectation property","local lifting property","Connes-Kirchberg problem","cone algebra","approximately multiplicative maps","operator spaces","inductive limits"],"falsifier":"Exhibit a WEP C*-algebra $B$, a finite-dimensional subspace $S \\subset \\ell^n_1$, and a complete contraction $u : S \\to C_0(B)$ such that for some fixed $\\delta > 0$ every extension $\\tilde{u} : \\ell^n_1 \\to C_0(B)$ with $\\tilde{u}|_S = u$ has $\\|\\tilde{u}\\|_{\\mathrm{cb}} \\geq 1 + \\delta$; this would refute the WEP of the cone algebra and invalidate Lemma 7.1.","tokens_in":25283,"feed_emoji":"🧩","tokens_out":6807,"duration_ms":62789,"temperature":0.7,"pith_summary":"This paper constructs the first example of a separable non-nuclear C*-algebra $A$ that has both the Weak Expectation Property (WEP) and the Local Lifting Property (LLP), two properties that together force a unique C*-norm on the tensor product $A \\otimes A^{\\mathrm{op}}$ and that had not previously been known to coexist outside nuclearity. The construction is designed so that $A$ has exactly the same finite-dimensional operator subspaces as the full C*-algebra $C^*(F_\\infty)$ of the free group on infinitely many generators, the algebra whose possession of the WEP is the open Connes–Kirchberg problem. Because $C^*(F_\\infty)$ is known to have the LLP, the new algebra is a test case: it realizes the WEP at the level of local operator-space structure without settling whether $C^*(F_\\infty)$ itself has the WEP. The paper also proves a general theorem: from a sequence of C*-algebras in which each is a subquotient of the next, with linking maps that “almost allow liftings,” one can form a limit algebra whose identity map inherits the local tensor properties of the inclusions. If the main claim is right, WEP and LLP can coexist with non-nuclearity, and the obstruction to the Connes–Kirchberg problem is not a purely finite-dimensional phenomenon.","feed_headline":"First non-nuclear C*-algebra with both WEP and LLP","feed_subtitle":"It shares its finite-dimensional operator subspaces with C*(F∞), keeping the Connes-Kirchberg problem in play.","key_machinery":"The engine is the cone algebra $C_0(C)$ (the C*-algebra of continuous functions $f : [0,1] \\to C$ with $f(0)=0$, where $C = C^*(F_\\infty)$) together with the notion of an $\\varepsilon$-morphism: a self-adjoint linear map $\\psi : E_0 \\to B_0$ that is almost contractive and almost multiplicative on a finite-dimensional self-adjoint domain. Lemma 6.1 shows that passing to cone algebras makes any quotient map $q$ almost allow liftings: the induced map $q_0 : C_0(C) \\to C_0(B)$ can be locally lifted by $\\varepsilon$-morphisms, via a quasicentral approximate unit of the kernel. Lemma 7.1 is the inductive step: given a finite-dimensional subspace $E$ of $C_0(C)$, it produces a larger subspace $E_1$ and an almost isometric, almost multiplicative inclusion $T : E \\to E_1$ that extends any prescribed $u : S \\to E$ ($S \\subset \\ell^n_1$) up to cb-norm $(1+\\varepsilon)$, using the WEP of $C_0(B)$ to seed the extension $v$. Iterating Lemma 7.6 yields the sequence $(E_n, T_n)$, and the ambient quotient $L = \\ell_\\infty(C_0(C))/c_0(C_0(C))$ turns the almost-multiplicative links into an honest C*-algebra $A$.","core_discovery":"The central claim is Theorem 7.2: there exists a separable non-nuclear (in fact non-exact) C*-algebra $A$ with the WEP and the LLP. The proof builds $A$ inside the quotient algebra $L = \\ell_\\infty(C_0(C))/c_0(C_0(C))$ of the cone algebra of $C = C^*(F_\\infty)$, as an increasing union of finite-dimensional self-adjoint subspaces $Y_n$ that are almost completely isometric to prescribed subspaces $E_n \\subset C_0(C)$. The linking maps $T_n : E_n \\to E_{n+1}$ are chosen to be simultaneously almost isometric and almost multiplicative ($\\varepsilon$-morphisms), so the union becomes a C*-subalgebra while retaining the extension property that characterizes the WEP. Because $A$ has the WEP and locally embeds in $C$, Proposition 3.7 yields the LLP; and since the $E_n$ can be chosen to contain any prescribed dense family of finite-dimensional subspaces of $C_0(C)$, $A$ and $C$ are locally equivalent. The paper further shows the example can be made quasidiagonal and of similarity degree 3, and in Section 9 it abstracts the whole scheme into a general “inductive and projective limit” theorem for sequences of subquotients.","pith_inferences":["The construction suggests that the Connes–Kirchberg problem, if true, cannot be certified by any local property that survives passage to the inductive limit; one would need a global tensor-product argument that distinguishes $C$ from $A$ despite their identical finite-dimensional operator subspace structure.","One could test whether the identity map of $A$ admits the lifting property (LP), not just the LLP; if $A$ fails the LP while having the LLP, it would give a new separation of the two local lifting notions in the separable setting.","The inductive/projective limit picture in Section 9 may transfer to von Neumann algebras or exactness questions: replacing “cone algebra” by another liftable functor would yield algebras whose identities carry properties of the linking maps, possibly producing new counterexamples to permanence of nuclearity or exactness under such limits.","Since the sequence $(E_n)$ can be chosen to contain any prescribed subspaces, one could stress-test the construction by feeding in subspaces known to force large exactness constants, measuring how slowly the almost-multiplicative errors $\\varepsilon_n$ must decay to keep the limit a C*-algebra."],"forward_implications":["$A$ has a unique C*-norm on $A \\otimes A^{\\mathrm{op}}$, because by Kirchberg's tensor-product criterion WEP plus LLP implies $A \\otimes_{\\min} B = A \\otimes_{\\max} B$ for every $B$, in particular for $B = A^{\\mathrm{op}}$.","$A$ and $C = C^*(F_\\infty)$ are locally equivalent: every finite-dimensional operator subspace of $C$ embeds almost completely isometrically into $A$, and conversely by the LLP of $A$.","Choosing the prescribed subspaces densely, $A$ witnesses that the Connes–Kirchberg problem is not decided by any finite-dimensional obstruction: the WEP of $C^*(F_\\infty)$ would follow from the WEP of $A$ only if the WEP were a purely local property, which it is not known to be.","The example can be modified to be unital, quasidiagonal, and of similarity degree 3, so the non-nuclearity is compatible with strong finite-dimensional approximation properties.","The general Theorem 9.2 supplies, for any suitable local property of an injective *-homomorphism $i : C \\to B$ (with $C$ having the LLP and the quotient map $q$ almost allowing liftings), a C*-algebra $A$ for which the identity map has that property and $A$ is locally equivalent to $C$."],"supporting_citations":[{"why":"Supplies the WEP/LLP tensor-product characterizations via nuclear pairs and the lemma that any separable C*-algebra embeds in a separable WEP algebra.","marker":"[10]"},{"why":"Gives Theorem 1.4 that the pair $(B(\\ell_2), C^*(F_\\infty))$ is nuclear and that $C^*(F_\\infty)$ has the LLP.","marker":"[11]"},{"why":"Provides Lemma 13.4.4, the cone-algebra lifting trick underlying Lemma 6.1.","marker":"[5]"},{"why":"Introduces asymptotic morphisms and deformations, the idea behind the $\\varepsilon$-morphism strategy.","marker":"[6]"},{"why":"Disproves WEP implies LLP and supplies the $d_{SC}$ formula connecting local embeddability to tensor norms.","marker":"[9]"},{"why":"Gives a simpler proof of Kirchberg's theorem and the extension results used in Propositions 2.1 and 3.7.","marker":"[17]"},{"why":"Introduced the Weak Expectation Property and the notion of nuclearity it extends.","marker":"[12]"},{"why":"Provides the dec-norm techniques used to prove that local embeddability plus WEP implies LLP in Proposition 3.7.","marker":"[8]"}],"fun_headline_variants":["Non-nuclear C*-algebra with WEP and LLP: a first example","First C*-algebra achieving WEP and LLP despite non-nuclearity","C*-algebra with WEP and LLP, not nuclear, quasidiagonal","New construction: C*-algebra with WEP and LLP, non-nuclear"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The construction assumes, without proof in Lemma 7.1, that the cone algebra $C_0(B)$ of a WEP C*-algebra $B$ again has the WEP, so that the map $i_0u$ admits the almost-isometric extension $v$ that seeds the inductive step; it also relies on Kirchberg's lemma that any separable C*-algebra embeds in a separable WEP algebra.","fun_headline_variants_meta":{"raw":{"variants":["Non-nuclear C*-algebra with WEP and LLP: a first example","First C*-algebra achieving WEP and LLP despite non-nuclearity","C*-algebra with WEP and LLP, not nuclear, quasidiagonal","New construction: C*-algebra with WEP and LLP, non-nuclear"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000794,"raw_usage":{"total_tokens":3578,"prompt_tokens":1108,"completion_tokens":2470,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":724,"completion_tokens_details":{"reasoning_tokens":2388}},"tokens_in":724,"tokens_out":2470,"duration_ms":18291,"temperature":1.0,"reasoning_tokens":2388,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:37:41.304533+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Exhibit a WEP C*-algebra $B$, a finite-dimensional subspace $S \\subset \\ell^n_1$, and a complete contraction $u : S \\to C_0(B)$ such that for some fixed $\\delta > 0$ every extension $\\tilde{u} : \\ell^n_1 \\to C_0(B)$ with $\\tilde{u}|_S = u$ has $\\|\\tilde{u}\\|_{\\mathrm{cb}} \\geq 1 + \\delta$; this would refute the WEP of the cone algebra and invalidate Lemma 7.1.","supporting_citations":[{"cited_title":"Connes and N","cited_arxiv_id":null,"evidence_quote":"Introduces asymptotic morphisms and deformations, the idea behind the $\\varepsilon$-morphism strategy."},{"cited_title":"Junge and G","cited_arxiv_id":null,"evidence_quote":"Disproves WEP implies LLP and supplies the $d_{SC}$ formula connecting local embeddability to tensor norms."},{"cited_title":"Operator algebras and their connection with Topology and Ergodic The ory","cited_arxiv_id":null,"evidence_quote":"Provides the dec-norm techniques used to prove that local embeddability plus WEP implies LLP in Proposition 3.7."}],"review_version":1}