{"id":"7d1ac840-cc10-4391-8b19-6e8f87a6059a","arxiv_id":"1101.5834","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper proves a Thom-Sebastiani theorem and a duality theorem for k[[beta]]-linear dg categories of matrix factorizations, valid without isolated-singularity hypotheses, and derives smoothness, properness, Calabi-Yau, and Knorrer-periodicity consequences.","lead":"Matrix factorizations are the algebraic shadow of a singularity, and this paper proves that combining two of them behaves like adding their potentials, plus a matching duality statement for the functors between them. It gives workers in mirror symmetry and singularity theory a toolkit for computing invariants of these categories without assuming isolated singularities.","discovery_kind":"unification","skeptic_critique":{"model":"selfhosted/deepseek-v4.1-flash","headline":"The reader's f^! concern is well-placed but narrower than stated: 4.1.3 and most of 4.2.2 reduce to classical duality and Lunts, while the least-secure step is the identity-functor identification in 4.2.3(iv), which the author calls 'somewhat opaque' (Remark 4.2.5).","rationale":"The reader's weakest_assumption (the f^! formalism for QC! on derived schemes, §1.4(i)) is a real and acknowledged limitation, and the reader is right that 4.2.2 and 4.2.3 rest on it. But the reader overstates its reach: the headline 4.1.3 reduces to Lunts-type k-linear exterior product plus a two-parameter Cor. 3.2.4 step, and the author's concrete proof of 4.2.2 reduces the perfect-pairing check to classical duality on the smooth M0. The genuinely load-bearing and least-secure step is the identity-functor identification in 4.2.3(iv), which the author himself flags as opaque (Remark 4.2.5) and as convention-dependent between two natural dualities (Remarks 4.2.6-4.2.7). That is where a concrete computation would settle whether the concern lands. The paper is otherwise careful and candid about its gaps (§5 sketch-level, §3.4.4 comparison, Prop. 8.2.4 sign uncertainty), and the circularity burden is low. My read sharpens the location of the concern but does not move the verdict: CONDITIONAL remains appropriate, so UNCHANGED.","tokens_in":74592,"tokens_out":16374,"duration_ms":315994,"concrete_test":"Compute both sides of 4.2.3(iv) in the simplest nontrivial case M = A^1, f = x^2 (crit = {0}, so MF(A^1, x^2) ≃ DSing(k[x]/x^2) is nonzero). Explicitly: (a) compute ω_{Δ,Z} = Δ_* RΓ_0 ω_{A^1} as an object of MF∞(A^2, -x^2+y^2), including its k[[β]]-linear structure; (b) compute the functor it induces on MF(A^1, x^2) via the integral transform of 4.2.3; (c) check it is the identity, not a shift or a β-twist. Repeat with the alternative kernel Δ_* RΓ_0 ω_{M0/M} of Remark 4.2.6 and confirm the two conventions agree on the identity functor. If they match, the concern is resolved for this case; if they differ, the identity-functor identification feeding 8.1.1 and 8.3.4 needs correction.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim has two halves, and the reader's f^! concern lands unevenly on them. For 4.1.3 (Thom-Sebastiani), the proof's chain of identifications reduces the k[[β]]-linear statement to the k-linear exterior-product theorem (Prop. B.3.2, a Lunts-type result) plus a two-parameter version of Cor. 3.2.4 that is asserted rather than proved; f^! is not the workhorse here. For 4.2.2 (duality), the author's concrete proof reduces the perfect-pairing check to classical Grothendieck duality on the smooth M0 plus standard base-change for derived pullback (the Cartesian square M0→M, (M0)^2→(M^2)_0), so the 'no published reference' formalism is less load-bearing than the reader suggests. The genuinely load-bearing and least-secure step is 4.2.3(iv): identifying ω_{Δ,Z} = Δ_* RΓ_Z ω_M with the identity functor. This uses the f^!/f_* base-change machinery (the D1,D2,Δ1,Δ2 diagram), the interaction of RΓ_Z with the S^1/k[[β]]-structure, and the author's own Remark 4.2.5 calls the proof 'somewhat opaque.' Worse, Remark 4.2.6 and 4.2.7 warn that the two natural dualities (ω_{M0} vs. ω_{M0/M}) give different explicit identifications of the trace and identity functors, so the formula in 4.2.3(iv) is convention-dependent and must be checked to give the identity under the specific equivalence chosen, not the other convention's kernel. If this identification is off by a shift or a β-twist, then 4.2.3(iv)-(v) and the trace-dependent applications (smoothness 8.1.1, Calabi-Yau 8.3.4) weaken, though 4.1.3 and 4.2.2 survive.","agreement_with_reader":"partial"},"referee_report":{"model":"selfhosted/deepseek-v4.1-flash","summary":"The paper proves Thom-Sebastiani and duality theorems for the k[[β]]-linear dg-category PreMF(M,f) of coherent complexes on the derived zero fiber of f, and for its 2-periodic version MF(M,f), without isolated-singularity hypotheses. The main results are Theorem 4.1.3 (external tensor product identifies PreMF(M,f) ⊗_{k[[β]]} PreMF(N,g) with PreMF on the zero locus of f⊞g, with support conditions), Theorem 4.2.2 (Grothendieck duality lifts to a k[[β]]-linear anti-equivalence PreMF(M,f)^op ≃ PreMF(M,-f)), and Theorem 4.2.3 (colimit-preserving k[[β]]-linear functors between Ind-completions are identified with PreMF on the zero locus of -f⊞g, with explicit descriptions of the identity functor and evaluation). The proofs use a geometric description of the k[[β]]-linear structure as an action of the derived group B = 0×_{A^1}0, together with a Lunts-type tensor-product theorem and shriek integral transforms for DCoh and QC!. Applications include: a computation of 2-periodic Hochschild invariants (Theorem 8.2.6), smoothness and properness of MF (Theorem 8.1.1), Calabi-Yau structures from volume forms (Theorem 8.3.4), relative Knörrer periodicity for metabolic quadratic bundles (Theorem 9.1.7), and a Clifford-algebra description of PreMF for quadratic bundles (Theorem 9.3.4).","tokens_in":74953,"tokens_out":5874,"duration_ms":114404,"significance":"If correct, the paper settles the expected Thom-Sebastiani and duality statements for matrix factorizations in a generality that goes well beyond isolated singularities, and the applications (smoothness, Calabi-Yau, Knörrer periodicity, Hochschild invariants) are substantial and were open in this generality. The organizing principle — the k[[β]]-linear structure as a derived group action, and the S^1-action / Tate-construction perspective on MF — is novel and unifying, and the paper is careful to reduce k[[β]]-linear statements to k-linear ones where possible. The main theorems 4.1.3 and 4.2.2/4.2.3 have real proofs in §4, and the author is explicit about which parts of the paper are sketch-level. The paper is a significant contribution to the field.","major_comments":[{"comment":"The identification of ω_{Δ,Z} = Δ_* RΓ_Z ω_M with the identity functor is the least-secure step in the paper. The author calls the proof 'somewhat opaque' (Remark 4.2.5) and warns (Remarks 4.2.6-4.2.7) that the two natural dualities (ω_{M0} vs ω_{M0/M}) give different explicit identifications of the trace and identity functors. This identification is load-bearing for Theorem 8.1.1(ii) (smoothness, which reduces to compactness of ω_{Δ,Z}), Theorem 8.2.6 (Hochschild invariants), and Theorem 8.3.4 (Calabi-Yau, whose non-degeneracy check uses [vol_M]^{Tate} being an equivalence). The paper should either verify the identification under the specific convention chosen, or show that the applications are insensitive to the choice of duality (e.g., that a shift or β-twist in ω_{Δ,Z} does not affect the compactness and non-degeneracy conclusions). As written, the reader cannot tell whether the formula in 4.2.3(iv) is correct under the chosen convention or under the other one.","section":"§4.2.3(iv)-(v), Remarks 4.2.5-4.2.7","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The Grothendieck duality / upper-shriek formalism for QC! on derived schemes and nice derived DM stacks is invoked without a published reference; the author states in §1.4(i) that 'a reference for this does not yet exist in the literature' and that it 'will probably appear in a later volume of Jacobi Lurie's DAG series.' Appendix B is the foundation for the tensor-product theorem and the functor-category description, and it is not clear from the text which parts of Appendix B are proved from scratch and which depend on the unpublished f^! formalism. The paper should state precisely which inputs are proved in Appendix B and which are conditional on the forthcoming reference, and ideally give a self-contained proof of the key inputs (or a precise statement of the result used from the forthcoming source). The concrete proof in §4 mitigates this for 4.1.3 and 4.2.2, but not for the parts of Appendix B that feed into 4.2.3.","section":"§1.4(i), Appendix B (Prop. B.3.2, Theorem B.2.4)","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"Section 5 is explicitly sketch-level: the author writes 'we will only sketch these proofs (being cavalier about these compatibilities)' and Theorem 5.2.8 and Lemma 5.5.2 are labelled 'Sketch.' While the main theorems 4.1.3 and 4.2.2/4.2.3 are proved in §4 independently of §5, §5 is the basis for the 'comparison of three viewpoints' (§7) and for the S^1-action perspective that the author says 'formed the basis for the author's thinking on the subject' (§1.5). The paper should either complete these proofs or clearly delimit which results depend on them, so that a reader can tell which statements are established and which are conditional.","section":"§5 (Part 2), Theorems 5.2.7, 5.2.8, Lemmas 5.5.1, 5.5.2","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The comparison to literal matrix factorizations via Orlov's theorem is only stated for flat f; the f=0 case is handled separately and the author acknowledges 'It seems likely that one can remedy these, but the author has not tried to check the details.' This is a gap in the justification of the notation MF(M,f), though not in the main theorems. The paper should either supply the missing comparison or state clearly that the identification with literal matrix factorizations is only established under the flatness hypothesis, and that the notation MF(M,f) is a definition in general.","section":"§3.4.4, Remark 3.4.5","authors_acknowledge":null,"authors_acknowledge_p":null}],"minor_comments":[{"comment":"The abstract uses k[t] with deg t = -2, while the body uses k[[β]] with deg β = -2. The relationship between the two notations should be clarified in the abstract or made consistent with the body.","section":"Abstract","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"There are several typographical errors: 'out study' in §1.2; 'k⊗k[ [bt] ]−' in Lemma 3.1.9 (should be k⊗k[[β]]−); 'once we show identify G' in the proof of Lemma 3.1.9 (should be 'once we show to identify G'); 'the conclusion of the conclusion' in Lemma 5.4.5; 'a a relative form' in Theorem 9.1.7; 'Will will produce' in the proof of Theorem 8.3.4; 'of of' and 'the the' in Remark 8.2.5.","section":"§1.2, Lemma 3.1.9, proof of Lemma 3.1.9, Lemma 5.4.5, Theorem 9.1.7, proof of Theorem 8.3.4, Remark 8.2.5","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The author flags sign uncertainties: Remark 3.1.7 says 'There is a choice of sign appearing here that will probably change at random later,' and Proposition 8.2.4 says '(Warning: I don't know the sign rules well enough to be sure of the sign.)' Given the sign-sensitive HKR and Calabi-Yau computations, these signs should be pinned down, or the dependence of the main results on the sign convention should be stated explicitly.","section":"Remark 3.1.7, Proposition 8.2.4","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"Section 5.4 is titled 'Graded speculation' and is explicitly conditional on a suitable Grothendieck duality over BGm. It should be marked more clearly as speculative (e.g., moved to a remark or an appendix) so that it is not read as a proved result.","section":"§5.4","authors_acknowledge":null,"authors_acknowledge_p":null},{"comment":"The notation 'PreMF' and 'Pre MF' is used inconsistently (e.g., 'Pre MF' in §1.4(ii) and 'PreMF' elsewhere). The paper would benefit from a consistent choice.","section":"Throughout","authors_acknowledge":null,"authors_acknowledge_p":null}],"recommendation":"major_revision","confidential_remarks":"The paper is long (78 pages) and Part 2 (§5-7) is sketch-level, while the main theorems are proved in §4. The journal may wish to consider whether the sketch-level sections are appropriate for the final version, or whether they should be shortened or moved to an appendix with a clear statement of which results are conditional. The identity-functor identification in 4.2.3(iv) is the point most likely to draw scrutiny from a careful reader, and the author's own Remarks 4.2.5-4.2.7 flag it; resolving it (or showing the applications are insensitive to the choice of duality) would substantially strengthen the paper."},"author_rebuttal":null,"desk_editor":{"model":"selfhosted/deepseek-v4.1-flash","letter":"The thing to know: this delivers the Thom-Sebastiani and duality statements people expected for matrix factorizations, without isolated-singularity hypotheses, and it does so by working k[[β]]-linearly on DCoh of the zero fiber rather than in the 2-periodic category. That refinement is the actual contribution — it lets the tensor-product and functor-category theorems be proved in ordinary coherent geometry. The geometric model of the k[[β]]-structure via B = 0×_{A1}0 (Construction 3.1.5, Prop. 3.1.4) is the trick that makes the rest go, and it is a good one. The applications in §8–9 are genuine consequences, not restatements; the metabolic Knörrer periodicity and the Clifford description (9.1.7, 9.3.4) are the parts I'd actually use. Note the classifier says no stated limitations; that's wrong — the author flags gaps in 1.4(i), Remark 3.4.5, Remark 4.2.5, and the sign warning in 8.2.4. He is unusually candid, which counts for something. On the reader's f^! worry: I think it lands less hard than stated. 4.1.3 reduces to the k-linear exterior-product theorem plus a two-parameter Cor. 3.2.4, and 4.2.2's perfect-pairing check reduces to classical Grothendieck duality on smooth M0 with standard base-change. The genuinely load-bearing and least-secure step is 4.2.3(iv): identifying ω_{Δ,Z} = Δ_* RΓ_Z ω_M with the identity functor. That uses the D1/D2/Δ1/Δ2 base-change diagram, the interaction of RΓ_Z with the k[[β]]-structure, and the author himself calls the proof 'somewhat opaque.' Remarks 4.2.6–4.2.7 make it worse: the two natural dualities (ω_{M0} vs ω_{M0/M}) give different identifications of trace and identity, so the formula is convention-dependent and needs checking against the specific equivalence chosen. If that is off by a shift or β-twist, 4.2.3(iv)–(v) and the trace-dependent applications (8.1.1 smoothness, 8.3.4 Calabi-Yau) weaken; 4.1.3 and 4.2.2 survive. §5 is sketch-level by the author's own admission, and 3.4.4's comparison to literal matrix factorizations is asserted without a natural functor (Remark 3.4.5 admits this). None of that is fatal. This is for people working in matrix factorizations, singularity categories, or the dg-category side of mirror symmetry. It deserves a serious referee, mainly to pin down 4.2.3(iv).","headline":"A real, well-executed k[[β]]-linear Thom-Sebastiani and duality theorem; the weak point is the identity-functor identification in 4.2.3(iv), not the f^! formalism the reader flagged.","tokens_in":75728,"tokens_out":1237,"would_cite":true,"duration_ms":21253,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"pith_extraction":{"msc":["14F05","18G80","13D09","14A20","16E40"],"pacs":[],"model":"selfhosted/deepseek-v4.1-flash","headline":"By keeping a formal degree-(-2) variable before inverting it, the paper reduces tensor products, functor categories, and Hochschild invariants of matrix factorizations to ordinary coherent sheaves on the zero loci of the sum and difference.","keywords":["matrix factorizations","Thom-Sebastiani theorem","Landau-Ginzburg models","Grothendieck duality","derived algebraic geometry","Hochschild homology","Knörrer periodicity","Clifford algebras"],"falsifier":"Take a Landau–Ginzburg pair where $f$ is not flat — for instance $f = xy$ on $mathbb{A}^2$, or a case where $f$ vanishes identically on a component — and compute both sides of the Thom–Sebastiani equivalence independently, checking in particular whether the support condition on $(f boxplus g)^{-1}(0)$ is genuinely needed or whether the equivalence already holds without it; a mismatch would locate the failure.","tokens_in":75807,"feed_emoji":"🧮","tokens_out":4438,"duration_ms":89195,"temperature":0.7,"pith_summary":"The paper's target is the 2-periodic category of matrix factorizations $MF(M,f)$ attached to a map $f: M to mathbb{A}^1$ — the algebraic shadow of a hypersurface singularity, or a Landau–Ginzburg model. The usual formulation is awkward because 2-periodicity throws away information and the standard tools (tensor products, duality, functor categories, Hochschild invariants) are only well understood when the critical locus is an isolated point. The author's move is to keep the bookkeeping: instead of working 2-periodically from the start, he attaches to $(M,f)$ a module structure over $k[[beta]]$ with $deg beta = -2$, calling the result $mathrm{PreMF}(M,f)$; its underlying category is nothing more exotic than coherent complexes on the zero fibre $f^{-1}(0)$, and inverting $beta$ recovers $MF(M,f)$. The main results say this refinement interacts exactly as one would want with the two operations that matter: tensoring two such categories gives the category for the sum potential $f boxplus g$ on the product (Thom–Sebastiani), and colimit-preserving functors between their Ind-completions form a matrix-factorization category for the difference potential $-f boxplus g$ (duality). If that is right, questions about matrix factorizations become questions about coherent sheaves on derived schemes, which is a far more familiar world.","feed_headline":"Tensor products of matrix factorizations become sheaves","feed_subtitle":"Keeping a formal variable before inverting it turns Landau–Ginzburg categories into ordinary coherent sheaves.","key_machinery":"The load-bearing device is a derived group scheme $B = 0 times_{mathbb{A}^1} 0$, the derived self-intersection of the origin in the affine line, equipped simultaneously with a 'loop composition' and a 'pointwise addition' product. The author proves (Proposition 3.1.4) that $QC^!(B)$ with its convolution product is symmetric-monoidally equivalent to $k[[beta]]$-modules, via a Koszul–Tate resolution that identifies the circle action with multiplication by $beta$. The fibre $M_0 = f^{-1}(0)$ carries a right action of $B$ built from the homotopies $f(m) to 0$, and applying $QC^!$ turns this into the $k[[beta]]$-linear structure on $DCoh(M_0)$; an Eckmann–Hilton argument reconciles the two products. The second ingredient is the $QC^! = mathrm{Ind},DCoh$ formalism with its shriek pullback and integral transforms, developed in the appendices, which supplies the tensor-product theorem and the functor-category description used to state and prove the main results.","core_discovery":"The central claim is a pair of theorems. First (Theorem 4.1.3), for Landau–Ginzburg pairs $(M,f)$ and $(N,g)$ and closed subsets $Z_M subset f^{-1}(0)$, $Z_N subset g^{-1}(0)$, external tensor product followed by pushforward along the inclusion $ell: M_0 times N_0 to (f boxplus g)^{-1}(0)$ is a $k[[beta]]$-linear equivalence $mathrm{PreMF}_{Z_M}(M,f) otimes_{k[[beta]]} mathrm{PreMF}_{Z_N}(N,g) xrightarrow{sim} mathrm{PreMF}_{Z_M times Z_N}(M times N, f boxplus g)$, and after inverting $beta$ the analogous statement for the 2-periodic categories, with a version that sums over matching critical values. Second (Theorems 4.2.2 and 4.2.3), Grothendieck duality lifts to a $k[[beta]]$-linear anti-equivalence $mathrm{PreMF}(M,f)^{op} simeq mathrm{PreMF}(M,-f)$, and consequently the dg category of colimit-preserving $k[[beta]]$-linear functors between Ind-completions is equivalent to $mathrm{PreMF}^infty_{Z_M times Z_N}(M times N, -f boxplus g)$, with the identity functor and evaluation written explicitly as the diagonal structure sheaf and its dualizing complex. The author then harvests consequences: the expected 2-periodic Hochschild homology and cohomology computations, with $k[[beta]]$-linear refinements; smoothness of $MF(M,f)$ over $k((beta))$ and properness when $mathrm{crit}(f) cap f^{-1}(0)$ is proper; $m$-Calabi–Yau structures on $MF(M,f)$ from a volume form on an $m$-fold $M$; and a relative Knörrer periodicity that lets metabolic quadratic bundles be discarded, together with a description of matrix factorizations for a non-degenerate quadratic bundle as modules over its Clifford algebra.","pith_inferences":["Because the whole strategy is 'solve the $k[[beta]]$-linear problem, then base-change to $k((beta))$', it suggests looking for $k[[beta]]$-linear lifts of other 2-periodic invariants (negative cyclic, topological Hochschild) rather than computing them 2-periodically from scratch.","The paper notes that $mathrm{PreMF}$ is almost never smooth over $k[[beta]]$; this makes smoothness of $MF$ an artifact of inverting $beta$, and one might expect the same 'smooth only after localization' pattern in other periodic settings.","Remark 4.1.4's description of the total category as a convolution over critical values hints at a concrete computational tool: a convolution algebra on the set of critical values that could be used to assemble global invariants from local ones.","The metabolic Knörrer statement plus the Clifford-algebra description together suggest that the Witt-group invariant of a quadratic bundle is exactly its Clifford algebra class in a 2-periodic Brauer group; that identification could be tested against known Brauer-group computations."],"forward_implications":["Tensor products of matrix factorization categories are computed by coherent complexes on the zero locus of the sum potential, with no isolated-singularity hypothesis and with support conditions tracked explicitly.","Colimit-preserving functors between matrix factorization categories are again a matrix factorization category — for the difference potential — so integral transforms, the identity functor, and evaluation acquire concrete geometric descriptions.","$MF(M,f)$ is smooth over $k((beta))$, and proper when the critical locus meeting $f^{-1}(0)$ is proper, giving a general smooth-and-proper statement beyond the isolated case.","A volume form on $M$ produces an $m$-Calabi–Yau structure on $MF(M,f)$, so Calabi–Yau structures on Landau–Ginzburg models come from honest differential forms on the total space.","Metabolic quadratic bundles can be eliminated: tensoring with the structure sheaf of a Lagrangian subbundle gives an equivalence from $mathrm{Perf}(X)((beta))$ to $MF(Q,q)$, and non-degenerate quadratic bundles give an equivalence with Clifford-algebra modules.","The 2-periodic Hochschild homology and cohomology of matrix factorizations are identified with Tate constructions on the Hochschild invariants of $DCoh(M)$, with $k[[beta]]$-linear refinements computed by HKR and local cohomology."],"supporting_citations":[{"why":"Supplies the tensor-product theorem and functor-category description for quasi-coherent complexes on perfect stacks, which the paper's Ind-coherent analogues are modelled on.","marker":"[BZFN]"},{"why":"Supplies the coherent/Ind-coherent versions of the tensor-product and integral-transform theorems over a field, which Appendix B extends to derived schemes and support conditions.","marker":"[L1]"},{"why":"Supplies the foundations of dg-categories and their Morita localization used throughout, including the identification of Ind-completions with dg-modules.","marker":"[T1]"},{"why":"Supplies the ∞-categorical machinery (presentable categories, limits and colimits in $Pr^L$ and $Pr^R$, Barr–Beck) on which the proofs of the main theorems rest.","marker":"[L6]"},{"why":"Supplies Orlov's theorem identifying the category of singularities of a hypersurface with matrix factorizations, which justifies calling $mathrm{PreMF} otimes_{k[[beta]]} k((beta))$ the matrix factorization category.","marker":"[O2]"},{"why":"Supplies the statement that Drinfeld–Verdier sequences are preserved by tensor product, used to compare $mathrm{Perf}$, $DCoh$, and $DSing$ after base change.","marker":"[D]"},{"why":"Supplies Kuznetsov's homological projective duality and the algebra $B$ of a projective quadric, which the paper identifies with its Clifford algebra construction.","marker":"[K]"},{"why":"Supplies the vanishing of the Brauer class in the metabolic case, which the paper's Knörrer periodicity result re-proves in the 2-periodic setting.","marker":"[KO]"}],"fun_headline_variants":["Tensor products of matrix factorizations match sheaves on zero loci","Thom-Sebastiani theorem for matrix factorizations via k[[beta]]","Duality swaps a potential with its negative for matrix factorizations","Matrix factorizations get smoothness, properness, and Calabi-Yau structures","Relative Knörrer periodicity discards metabolic quadratic bundles"],"cache_read_input_tokens":0,"weakest_assumption_plain":"The paper depends on a Grothendieck duality and upper-shriek formalism for Ind-coherent complexes on derived schemes and derived Deligne–Mumford stacks that the author says has no written reference yet, and several supporting results (Theorem 5.2.7, Lemma 5.5.2, Proposition 4.1.6) are presented only as sketches.","fun_headline_variants_meta":{"raw":{"variants":["Tensor products of matrix factorizations match sheaves on zero loci","Thom-Sebastiani theorem for matrix factorizations via k[[beta]]","Duality swaps a potential with its negative for matrix factorizations","Matrix factorizations get smoothness, properness, and Calabi-Yau structures","Relative Knörrer periodicity discards metabolic quadratic bundles"]},"model":"selfhosted/deepseek-v4.1-flash","effort":"low","attempts":1,"cost_usd":0.0,"raw_usage":{"total_tokens":1922,"prompt_tokens":1192,"completion_tokens":730,"prompt_tokens_details":null,"completion_tokens_details":{"reasoning_tokens":651}},"tokens_in":1192,"tokens_out":730,"duration_ms":11673,"image_pages":[],"temperature":1.0,"reasoning_budget":50,"reasoning_tokens":651,"transport_provider":"selfhosted","cache_read_input_tokens":0,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-11-10T18:08:24.825801+00:00","model_set":{"reader":"selfhosted/deepseek-v4.1-flash"},"falsifier":"Take a Landau–Ginzburg pair where $f$ is not flat — for instance $f = xy$ on $mathbb{A}^2$, or a case where $f$ vanishes identically on a component — and compute both sides of the Thom–Sebastiani equivalence independently, checking in particular whether the support condition on $(f boxplus g)^{-1}(0)$ is genuinely needed or whether the equivalence already holds without it; a mismatch would locate the failure.","supporting_citations":[],"review_version":1}