{"id":"188af7e8-7e26-41ff-a1b1-cb048860a9be","arxiv_id":"2502.04123","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Continuous K-theory of nuclear modules on Spf(R^hat_I) is isomorphic to lim_n K(R/I^n), via internal projectivity and strongly Mittag-Leffler inverse sequences.","lead":"Efimov proves that the K-theory of Clausen and Scholze's nuclear modules on a formal scheme agrees with the classical continuous K-theory, resolving a conjecture. The proof introduces a general method for computing localizing invariants of inverse limits of dualizable stable categories.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Reader's 5.23 concern is not load-bearing; the key soft spot is Prop. 7.5's identification of Nuc_CS with nuclear objects in an algebraic model.","rationale":"The reader's weakest_assumption targets Prop. 5.23(i) as the input to Thm. 6.1. However, the main theorem for Nuc_CS (Thm. 7.9) is proved via a bounded-product variant that does not need strong Mittag-Leffler; Prop. 5.23 is used for the auxiliary Nuc category and for Cor. 7.11, but not for the K-theory conjecture itself. Moreover, the pro-equivalence (5.13) is adequately justified: the reduction to k=1 by base change x→x^k is legitimate, and the claimed homology computation for k=1 is consistent with the minimal periodic resolution of Z over Z[x]/x^n. Thus the reader's specific concern is unlikely to land. The genuinely load-bearing soft spot is Prop. 7.5, the bridge between the analytic definition of Nuc_CS and the algebraic model used in all later computations. The proof there is compressed, invokes external results [And23, AM24], and the inverse map for basic nuclear objects depends on delicate trace-class witnesses that are not fully spelled out for non-noetherian rings. A failure of Prop. 7.5 would directly undermine Thm. 7.9 and Thm. 0.1, so this is the concern that should be settled. Since the preprint is already CONDITIONAL and the issue is a verification gap rather than a demonstrated error, the verdict remains CONDITIONAL/UNCHANGED.","tokens_in":74393,"tokens_out":26651,"duration_ms":250030,"concrete_test":"Verify Prop. 7.5 in a non-noetherian example, e.g. R = k[[x_1,x_2,...]] with I = (x_1,x_2): check that f_*((⊕_N R^I)^∨) is nuclear and that the constructed map f_*f^*(X)→X is an isomorphism for basic nuclear X. If the argument requires noetherianness or finite generation of R/I over Z, the general statement fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central K-theory conjecture (Thm. 0.1) and its general localizing-invariant form (Thm. 7.9) do not rest on Prop. 5.23(i): Thm. 7.9 is proved by a bounded-product modification of Thm. 6.1 using Prop. 7.12, not on strong Mittag-Leffler. Moreover, Prop. 5.23's reduction to k=1 via x→x^k is valid, since the subsystem indexed by multiples of k is cofinal in n≥k. The actual load-bearing step is Prop. 7.5, which identifies Nuc_CS(R^I) with the nuclear objects of D(ĆSolid R^I). Its proof is sketched: it asserts f_* preserves nuclear objects, citing [AM24, Lemma 2.18], and constructs an inverse to the unit for basic nuclear X using trace-class witnesses. In the non-noetherian case the compact generators and trace-class maps are less explicit, so a gap here would invalidate Thm. 7.9's application to Nuc_CS. This identification is essential and is not independently verified in the preprint.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the category Nuc(R^I) of nuclear modules on an affine formal scheme Spf(R^I), giving three equivalent descriptions: as a dualizable internal Hom, as a rigidification of the I-complete derived category, and as a dualizable inverse limit. The central results are Theorem 6.1, which computes accessible localizing invariants of dualizable inverse limits under a homological-epimorphism hypothesis, and Theorem 7.9, which identifies the localizing invariants of the Clausen-Scholze category Nuc_CS(R^I) with the inverse limit of the same invariants applied to Perf(R_n), where R_n are Koszul DG algebras. In the noetherian case this yields K^cont(Nuc_CS(R^I)) isomorphic to lim_n K(R/I^n), proving a conjecture of Clausen and Scholze. The paper also derives applications to Hochschild homology and to the comparison of the two versions of nuclear modules.","tokens_in":74544,"tokens_out":7195,"duration_ms":74255,"significance":"If the main theorems are correct, the paper resolves a central conjecture in the K-theory of formal schemes and gives a natural categorical home for continuous K-theory: continuous K-theory becomes the localizing invariant of a dualizable category rather than an inverse limit of ad hoc quotients. The paper also provides a general mechanism for computing localizing invariants of strongly Mittag-Leffler inverse sequences, with applications beyond the formal-scheme setting, including Hochschild homology computations. The exposition is systematic, the main reductions are explicit, and the K-theoretic identification is not assumed but is derived from categorical constructions. However, the full correctness of the non-noetherian and Nuc_CS statements depends on two technical inputs, Propositions 5.20 and 7.5, whose proofs are either deferred or sketched; the significance is therefore contingent on completing those points.","major_comments":[{"comment":"Proposition 7.5 is the point where the Clausen-Scholze category Nuc_CS(R^I) is identified with the nuclear objects of the algebraic model D(ĆSolid R^I), and this identification is what makes Theorem 7.9 a statement about Nuc_CS rather than about an auxiliary category. The proof is a sketch: it asserts that f_* preserves nuclear objects by checking one object via [AM24, Lemma 2.18], and it constructs an inverse to the unit only for basic nuclear X using trace-class witnesses. No details are given for the coherence of the chosen witnesses, for the independence of the presentations of X, or for the extension to all nuclear objects; in the non-noetherian case the compact generators and trace-class maps are not described explicitly. Since Theorems 0.1 and 7.9 depend on this identification, a complete proof, or a precise reference with the statement, is needed.","section":"§7.2, Proposition 7.5"},{"comment":"Proposition 5.20 is load-bearing for Theorem 5.16: it is used to pass from the abstract ind-approximation of Perf_{I-tors}(R) to the trace-class compatibility that gives the pro-equivalence (5.12) and hence the strongly Mittag-Leffler structure of (D(R_n)). The text explicitly says that a more formal proof will appear in [E]; the explicit proof given here is compressed, especially the Claim that the deformation-algebra filtration quotients are perfect over k after quotienting by Rep_k(C^op ⊗ C, Perf(k)). Because Theorem 0.2 and Corollary 6.3 rely on this, the proof should be completed in this manuscript, or the affected results should be stated as conditional on [E].","section":"§5.3, Proposition 5.20"},{"comment":"Proposition 7.12 is asserted without proof, with only a reference to an unbounded version of the short exact sequence from [KasWin19]. This proposition supplies the bounded-product short exact sequence 0 → bnd∏ Perf(R_n) → bnd∏ D^b(Proj_{ω1}-R_n) → bnd∏ Calk^b_{ω1}(R_n) → 0 that underlies the cofiber sequence in Theorem 7.9(i). The bounded-amplitude version is not automatic from the unbounded one: one has to check that the three bounded subcategories form a short exact sequence. Moreover, the proof of Theorem 7.9 is described only as a modification of the proof of Theorem 6.1, and the bounded-product analogues of the arguments from Section 6 are not stated. Please provide the missing proof or a precise reference.","section":"§7.3, Proposition 7.12 and proof of Theorem 7.9"}],"minor_comments":[{"comment":"There is a typo in the statement: 'for an oobject P' should read 'for an object P'.","section":"§1.5, Proposition 1.30"},{"comment":"The reduction to k=1 by the base change x ↦ x^k is sound, but it should explicitly note that the subsystem indexed by multiples of k is cofinal in n ≥ k; otherwise the step may look like a loss of generality.","section":"§5.2, proof of Proposition 5.23(i)"},{"comment":"Several results are deferred to [E] (Remarks 3.23, 3.24, 6.4, and Section 3.7). A short paragraph at the start of the paper listing which statements are proven here and which are deferred would help the reader assess the scope.","section":"§0 and §3.7"},{"comment":"The identification K^cont(D(R/I^n)) ≅ K(R/I^n) is used without comment; it follows from the setup of [E24], but should be stated explicitly.","section":"§0, Theorem 0.2"},{"comment":"The equivalence between uniform bounded projective amplitude and membership in Stab(∏ Proj_{ω1}-S_j) is called 'essentially a tautology', but it uses the explicit description of Stab of a product; a one-line proof or reference would be helpful.","section":"§7.1, Proposition 7.1"}],"recommendation":"major_revision","confidential_remarks":"This is a substantial paper with a convincing overall strategy and an important target theorem. The main obstacle is that two of the load-bearing technical inputs, Proposition 5.20 and Proposition 7.5, are not fully proved in the manuscript; both are essential for the non-noetherian and Nuc_CS results. I recommend asking the author to complete those proofs or to restate the affected theorems as conditional on [E]. The reader's concern about Proposition 5.23(i) does not appear to be the real weak point: the reduction to k=1 is valid by cofinality. The fit with the journal is appropriate if the missing technical details are supplied."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a strong paper that deserves a serious referee. It proves the Clausen-Scholze conjecture (Thm 0.1), sets up a reusable theorem (Thm 6.1/Cor 6.2) computing localizing invariants of dualizable inverse limits under a homological epimorphism condition, and introduces a genuinely new categorical object, Nuc(R^I), with three equivalent descriptions. The internal projectivity of proper ω1-compact dualizable categories (Thm 3.6) is the conceptual engine; the Raynaud-Gruson analogy is illuminating, not just window dressing.\n\nThe paper is honest about its own reliance on hard technical inputs: Proposition 2.2 (permuting limits and colimits) is pushed to Appendix A, and Proposition 5.20 is explicitly deferred to the author's forthcoming paper [E]. That is a real but proportionate concern. The reader's worry about Prop 5.23(i) being load-bearing does not hold up: Theorem 7.9 is proved via the bounded-product modification (Prop 7.12), and the reduction to k=1 in (5.13) is valid by cofinality of the subsystem.\n\nThe soft spot that actually matters is Prop 7.5: the identification of Nuc_CS(R^I) with the nuclear objects of D(ĆSolid R^I). The proof is a sketch: it asserts f_* preserves nuclear objects via [AM24, Lemma 2.18] and constructs an inverse to the unit for basic nuclear X using trace-class witnesses. That is the essential bridge from the Clausen-Scholze solid-module category to the algebraic model that feeds Theorem 7.9. In the noetherian case the compact generators are explicit and the argument is plausible; the non-noetherian case is where I would want to see the details. This is the main reason I'd call the verdict conditional rather than accepted.\n\nOverall: the central argument has structure and the pieces fit. The paper deserves a serious referee, and I'd take it to reading group. My recommendation: send to an expert in stable ∞-categories/K-theory, and ask specifically for a documented proof of Prop 7.5 and an indication of the status of Prop 5.20. With those, I'd expect this to become standard reference.","headline":"A serious, original paper that proves the Clausen-Scholze conjecture on K-theory of nuclear modules; the main new machinery is real, but one essential identification (Prop 7.5) is sketched and should be filled in before publication.","tokens_in":75145,"tokens_out":1855,"would_cite":true,"duration_ms":21569,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["18F25","19D55","14B20"],"pacs":[],"model":"deepseek-v4-flash","headline":"Continuous K-theory of affine formal schemes is computed by a natural category of nuclear modules.","keywords":["continuous K-theory","nuclear modules","formal schemes","localizing invariants","dualizable categories","Mittag-Leffler inverse sequences","inverse limits","Koszul complexes"],"falsifier":"Check the homology of the cone of the natural map $\\mathbb Z \\otimes^{\\mathbf L}_{\\mathbb Z[x]} \\mathbb Z \\to \\mathbb Z \\otimes^{\\mathbf L}_{\\mathbb Z[x]/x^{n+1}} \\mathbb Z$ in $D(\\mathbb Z \\otimes^{\\mathbf L}_{\\mathbb Z[x]} \\mathbb Z)$ for large $n$; Proposition 5.23 asserts that this map is zero in degrees $\\ge 2$ through the explicit vanishing argument around equation (5.14). A nonzero homology class in any such degree would break the pro-equivalence (5.13), so the sequence $(D(R_n))$ would not be strongly Mittag-Leffler and Theorem 6.1 would not apply.","tokens_in":74099,"feed_emoji":"🧮","tokens_out":15693,"duration_ms":143707,"temperature":0.7,"pith_summary":"Classical continuous K-theory of an affine formal scheme is defined as an inverse limit $\\varprojlim_n K(R/I^n)$, a definition that has long felt ad hoc because no single natural category seemed to have that spectrum as its K-theory. This paper proves that such a category does exist, resolving a conjecture recorded in [CS20]: the dualizable category $\\operatorname{Nuc}_{\\mathrm{CS}}(R^{\\wedge}_I)$ of nuclear solid modules has continuous K-theory isomorphic to $\\varprojlim_n K(R/I^n)$ in the noetherian case. The proof introduces a closely related category $\\operatorname{Nuc}(R^{\\wedge}_I)$ with three equivalent universal descriptions and shows that the two nuclear categories have the same finitary localizing invariants. The payoff is that continuous K-theory becomes the localizing invariant of a canonical dualizable category, so the same machinery computes Hochschild homology and other invariants of formal schemes as inverse limits.","feed_headline":"Continuous K-theory equals K-theory of nuclear modules","feed_subtitle":"The inverse limit used to define continuous K-theory is shown to be the K-theory of a natural dualizable category.","key_machinery":"The central object is the dualizable inverse limit, an inverse limit taken in the category of dualizable stable categories rather than in the larger category of presentable categories. A sequence of dualizable categories is called strongly Mittag-Leffler when the pro-system of composite functors $F_{mn}F^R_{mn}$ is essentially constant and the limiting functors are strongly continuous with left adjoints. For such sequences the paper proves that the map from $\\varprojlim_n C^\\kappa_n$ to $\\varprojlim_n \\operatorname{Calk}^{\\mathrm{cont}}_\\kappa(C_n)$ is a homological epimorphism (Theorem 5.4), where $\\operatorname{Calk}^{\\mathrm{cont}}_\\kappa(C_n)$ is the Calkin category, a quotient that encodes the difference between the original category and its $\\kappa$-compact approximation. Theorem 6.1 then expresses any accessible localizing invariant of the dualizable inverse limit as an inverse limit of invariants of the terms. The category $\\operatorname{Nuc}(R^{\\wedge}_I)$ is exhibited as exactly such a dualizable inverse limit of $\\operatorname{Perf}(R_n)$, and the internal projectivity of the proper, $\\omega_1$-compact category $D_{I\\text{-}\\mathrm{tors}}(R)$ (Theorem 3.6) is what makes the internal-Hom description preserve short exact sequences. Dualizable here means the category has a dual and evaluation/coevaluation functors, so it behaves like a finite-dimensional object in the tensor category of stable categories.","core_discovery":"The paper's central claim is that for a commutative ring $R$ and a finitely generated ideal $I=(a_1,\\ldots,a_m)$, the continuous K-theory of the affine formal scheme $\\operatorname{Spf}(R^{\\wedge}_I)$ is the continuous K-theory of the dualizable category $\\operatorname{Nuc}_{\\mathrm{CS}}(R^{\\wedge}_I)$ of nuclear solid modules. Theorem 7.9 states this in full generality: for any accessible localizing invariant $\\Phi$ and the Koszul DG algebras $R_n=\\operatorname{Kos}(R;a_1^n,\\ldots,a_m^n)$, there is an isomorphism $\\Phi^{\\mathrm{cont}}(\\operatorname{Nuc}_{\\mathrm{CS}}(R^{\\wedge}_I)) \\simeq \\varprojlim_n \\Phi(\\operatorname{Perf}(R_n))$; in the noetherian case this becomes $\\varprojlim_n \\Phi(\\operatorname{Perf}(R/I^n))$, and for $\\Phi=K$ it is the isomorphism conjectured in [CS20]. To reach it, the paper constructs its own category $\\operatorname{Nuc}(R^{\\wedge}_I)$ and proves three equivalent definitions: as the dualizable internal Hom $\\operatorname{Hom}^{\\mathrm{dual}}_R(D_{I\\text{-}\\mathrm{tors}}(R),D(R))$, as the rigidification of the derived $I$-complete category, and as the dualizable inverse limit $\\varprojlim_n^{\\mathrm{dual}} D(R_n)$. It then proves that $\\operatorname{Nuc}_{\\mathrm{CS}}(R^{\\wedge}_I)$ and $\\operatorname{Nuc}(R^{\\wedge}_I)$ have the same finitary localizing invariants, which yields the same formulas for all such invariants and gives $\\operatorname{HH}(\\operatorname{Nuc}(\\mathbb Z_p)/\\mathbb Z) \\simeq \\operatorname{HH}(\\operatorname{Nuc}_{\\mathrm{CS}}(\\mathbb Z_p)/\\mathbb Z) \\simeq \\mathbb Z_p$.","pith_inferences":["Editorial: The dualizable-inverse-limit formulation suggests a sheaf-theoretic extension: defining relative invariants of arbitrary schemes by internal Hom over the base, as the paper's Remark 3.24 sketches for future work, would turn continuous K-theory into a Zariski sheaf with quotient functors for open immersions.","Editorial: The strongly Mittag-Leffler condition is likely a general principle for categorical limits: any localizing invariant that commutes with countable products should convert a dualizable inverse limit into an inverse limit once the pro-system of endofunctors is essentially constant. Testing this on I-adic completions of noncommutative DG algebras would show whether the mechanism is specific","Editorial: Since the paper notes, following [AM24], that the nuclear category is unchanged by the choice of analytic structure on $R^{\\wedge}_I$, the localizing-invariant formulas should be independent of that choice; computing Hochschild homology under different analytic structures would make this explicit."],"forward_implications":["Continuous K-theory of affine formal schemes is no longer an ad hoc inverse limit: it is $K^{\\mathrm{cont}}$ of a natural dualizable category of nuclear modules.","Every accessible localizing invariant of the nuclear category is the inverse limit of the same invariant on the Koszul truncations $R_n$, so Hochschild homology, topological Hochschild homology, and similar invariants are computable by the same formula.","The two versions of nuclear modules, the original one and the new $\\operatorname{Nuc}(R^{\\wedge}_I)$, carry identical finitary localizing invariants, so the more tractable three-definition category can be used for computations.","In the noetherian case the Koszul DG algebras are pro-equivalent to the ordinary quotients $R/I^n$, which recovers the classical continuous K-theory $\\varprojlim_n K(R/I^n)$.","The Hochschild homology computation $\\operatorname{HH}(\\operatorname{Nuc}(\\mathbb Z_p)/\\mathbb Z) \\simeq \\mathbb Z_p$ shows that nuclear categories can have finite Hochschild homology even when the cotangent complex of the base algebra is very large."],"supporting_citations":[{"why":"Defines the original category $\\operatorname{Nuc}_{\\mathrm{CS}}(R^{\\wedge}_I)$ of nuclear solid modules and records the conjecture, proved here, that its continuous K-theory is $\\varprojlim_n K(R/I^n)$.","marker":"[CS20]"},{"why":"Supplies the continuous K-theory functor $K^{\\mathrm{cont}}$ on dualizable categories and the Calkin-category construction on which the paper's computations rely.","marker":"[E24]"},{"why":"Provides the finitary localizing invariants and commutation with countable products used to compare the two nuclear categories and to extend the K-theory result to arbitrary invariants.","marker":"[Cor23]"},{"why":"Supplies the framework of rigid monoidal categories and dualizable modules over them in which the three equivalent definitions of $\\operatorname{Nuc}(R^{\\wedge}_I)$ are formulated.","marker":"[GaiRoz17]"},{"why":"Provides the rigidity criterion in terms of trace-class colimits and the rigidification functor used for the second definition of $\\operatorname{Nuc}(R^{\\wedge}_I)$.","marker":"[Ram24b]"},{"why":"Establishes the identification of $\\operatorname{Nuc}_{\\mathrm{CS}}(R^{\\wedge}_I)$ with nuclear objects in a solid derived category, a step used in the proof of Theorem 7.9.","marker":"[And23]"},{"why":"Supplies the general theory of nuclear solid modules and the invariance of the nuclear category under changes of analytic structure used in Section 7.","marker":"[AM24]"}],"fun_headline_variants":["Continuous K-theory = K-theory of nuclear modules","Nuclear modules realize continuous K-theory","Localizing invariants commute with inverse limits","Inverse limits yield continuous K-theory via nuclear modules","A dualizable category matches continuous K-theory"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the inverse system $D(\\operatorname{Kos}(R;a_1^n,\\ldots,a_m^n))$ is stable enough (strongly Mittag-Leffler), which in the model case $R=\\mathbb Z[x]$, $I=(x)$ is exactly the pro-equivalence $\\mathbb Z[x]/x^k \\otimes^{\\mathbf L}_{\\mathbb Z[x]} \\mathbb Z[x]/x^k \\simeq \\varprojlim_{n\\ge k} \\mathbb Z[x]/x^k \\otimes^{\\mathbf L}_{\\mathbb Z[x]/x^n} \\mathbb Z[x]/x^k$; if that pro-equivalence fails for some $k$, the main limit formula loses its proof.","fun_headline_variants_meta":{"raw":{"variants":["Continuous K-theory = K-theory of nuclear modules","Nuclear modules realize continuous K-theory","Localizing invariants commute with inverse limits","Inverse limits yield continuous K-theory via nuclear modules","A dualizable category matches continuous K-theory"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001196,"raw_usage":{"total_tokens":5087,"prompt_tokens":1256,"completion_tokens":3831,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":872,"completion_tokens_details":{"reasoning_tokens":3759}},"tokens_in":872,"tokens_out":3831,"duration_ms":26195,"temperature":1.0,"reasoning_tokens":3759,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-08T23:26:02.861993+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Check the homology of the cone of the natural map $\\mathbb Z \\otimes^{\\mathbf L}_{\\mathbb Z[x]} \\mathbb Z \\to \\mathbb Z \\otimes^{\\mathbf L}_{\\mathbb Z[x]/x^{n+1}} \\mathbb Z$ in $D(\\mathbb Z \\otimes^{\\mathbf L}_{\\mathbb Z[x]} \\mathbb Z)$ for large $n$; Proposition 5.23 asserts that this map is zero in degrees $\\ge 2$ through the explicit vanishing argument around equation (5.14). A nonzero homology class in any such degree would break the pro-equivalence (5.13), so the sequence $(D(R_n))$ would not be strongly Mittag-Leffler and Theorem 6.1 would not apply.","supporting_citations":[],"review_version":1}