{"id":"b050dc9f-8599-4c7e-a916-b69cffca6923","arxiv_id":"1908.03005","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For zigzag consistent dimers on a torus, the Hochschild cohomology of the Jacobi algebra and the compactly supported Hochschild cohomology of the matrix factorization category are computed in explicit combinatorial terms, with the full BV structure.","lead":"This paper computes the Hochschild cohomology of the Jacobi algebra attached to any zigzag consistent dimer on a torus, including a full description of its Batalin-Vilkovisky structure. It then computes the compactly supported Hochschild cohomology of the matrix factorization category and compares it with the symplectic cohomology of the mirror punctured surface.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 4.5's classification of derivations of J(Q) depends on an imported minimal-path uniqueness result; if that fails, the HH^1 basis and everything built from it collapses.","rationale":"The paper contains extensive detailed arguments and I found no internal contradiction or post hoc data selection: the BV computations, the central-localization machinery, and the spectral-sequence degeneration appear coherent. The open comparison between compactly supported and ordinary Hochschild cohomology is explicitly labeled as unaddressed and does not affect any stated theorem. The single genuinely load-bearing risk is the imported minimal-path uniqueness that underlies Lemma 4.5: the whole additive computation of HH^1, and then HH^2 and HH^3 via degree counts, depends on the assertion that minimal paths are unique up to powers of the potential. If this fails for some zigzag-consistent toric dimer, the basis {partial_i, partial_alpha} is not well-defined and the final theorem would overcount. The proposed concrete test is finite because the matching set is finite, and it checks exactly the place where the proof relies on the imported lemma, without circularly assuming it. Since the paper cites the needed facts from published work rather than proving them, I would not reject the paper, but I would make acceptance conditional on either supplying a proof of the uniqueness statement in this setting or carrying out the computational verification on the paper's explicit dimer examples.","tokens_in":42897,"tokens_out":23080,"duration_ms":267736,"concrete_test":"For the suspended pinch point dimer (Example 2.12), perform a finite linear-algebra check of the disputed ray case: for each adjacent corner matching pair (P_i, P_{i+1}), let gamma be a complex grading supported on P_i cap P_{i+1} and solve the face-sum equations (19) that make chi(f tensor gamma) a derivation of J(Q)[ell^{-1}]. Lemma 4.5 requires that the only solution for every such pair is gamma=0. If a nonzero solution exists, then x_{eta_{i+1}} ell^{-1} gamma is an additional HH^1(J(Q)) class not in the claimed basis, invalidating Theorem 4.6 and the derived HH^2/HH^3 formulas. This check does not invoke [6, Lemma 7.4] or [8, Lemma 3.18], so it isolates whether the load-bearing step itself is sound; repeat it for the dimer in Figure 5 to cover a case with parallel zigzags.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The pivotal step is the derivation classification in Lemma 4.5 and Theorem 4.6. Lemma 4.5 decides whether x_alpha ell^n gamma(a) lies in J(Q) by checking nonnegative degrees in all perfect matchings, and the paper justifies this with the assertion that a class in J(Q)[ell^{-1}] is represented by a minimal path which is unique up to powers of ell, citing [6, Lemma 7.4] and [8, Lemma 3.18] (see Section 2.6 and the start of Section 4.1). The paper does not prove this uniqueness, and the correctness of the set {partial_i, partial_alpha} as a basis depends on it. In particular, in the n=-1, alpha on a ray case, the proof concludes that gamma must be supported on P_i cap P_{i+1} and then invokes Theorem 2.10 to force gamma=0 through the face equations (19). If minimal representatives were not unique in the asserted sense, the degree test could accept or reject a localized derivation incorrectly, so the direct summands in HH^1(J(Q)) would not be a basis. Since HH^2 and HH^3 are computed by dimension counts on the HH^1 and HH^0 degrees (Theorems 4.17 and 4.19), and Theorem 1.3 uses the spectral sequence on these groups, every stated additive result in Theorems 1.2 and 1.3 inherits this risk. The reader's weakest assumption identifies exactly this point, and I agree it is the most load-bearing condition.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper computes the Hochschild cohomology of the Jacobi algebra J(Q) of a zigzag-consistent dimer Q embedded in a torus, in purely combinatorial terms: additive bases for HH^0 through HH^3 are given in Theorem 1.2, the BV algebra structure is described in Section 4.5, and the compactly supported Hochschild cohomology of the matrix factorization category MF(J(Q),ℓ) is computed in Theorem 1.3 and shown to agree additively with the symplectic cohomology of the dual punctured surface. The main technical tools are a central localization theorem for Hochschild cohomology and BV structures (Section 3), a derivation classification for J(Q) (Section 4), and a spectral sequence for the matrix factorization category (Section 5).","tokens_in":1460,"tokens_out":1521,"duration_ms":130835,"significance":"If the main theorems are correct, the paper provides explicit, parameter-free, combinatorially defined bases and BV products for nontrivial noncommutative invariants of dimer models, and it gives concrete evidence for the expected agreement with symplectic cohomology on the mirror side. The manuscript is careful in many places: the localization compatibility with BV structure in Section 3 is written out in detail, the grading arguments in Section 4 are systematic, and the spectral sequence in Section 5 is set up with attention to inverse limits. The central claim is of clear interest to the dimer-model and homological-mirror-symmetry communities, and the results would be a useful reference if the identified gaps are filled.","major_comments":[{"comment":"The classification of derivations of J(Q) is the load-bearing step for Theorems 4.6, 4.17, 4.19, and hence for Theorem 1.2. Its proof relies on two imported facts: that minimal path representatives in J(Q)[ℓ^{-1}] are unique up to powers of ℓ, and that nonnegative perfect-matching degrees characterize membership in J(Q). These are cited to [6, Lemma 7.4] and [8, Lemmas 3.18–3.19] but are not proved or even stated as explicit hypotheses. Because the basis {∂_i, ∂_α} of HH^1 and everything built from it depend on this classification, the paper should either prove these facts or formulate them as explicit assumptions with a precise verification. In the ray case n = -1, the step asserting that a nontrivial γ supported on P_i must equal E_{P_i} also needs proof: the face equations (19) alone show only that γ is supported on P_i, and it is not immediate that the only grading in N supported on P_i is proportional, modulo N^in, to E_{P_i}.","section":"§4.2, Lemma 4.5"},{"comment":"The spectral sequence argument leading to the isomorphisms (31) is too compressed. The text states that the only possible nonzero components of d^2 are d^2_{i,i}, but in a first-quadrant spectral sequence d^2 has bidegree (-1,2); a bidegree-by-bidegree check is needed before concluding that d^2 vanishes. More importantly, after showing d^2 = 0, the paper asserts that the spectral sequence degenerates at the second page. Since the matching-degree argument shows that d^r has degree at least r in all perfect matchings, the vanishing of all higher differentials should be stated uniformly for all r, and the convergence of the inverse-limit exact sequence (29) should be tied explicitly to this degeneration. As written, Theorem 1.3 depends on an unproved degeneration claim.","section":"§5, Proposition 5.2"},{"comment":"The proof of the localization isomorphism for Hochschild cohomology checks the key quasi-isomorphism only for the free bimodule M = A^e and then asserts it for general perfect M. The statement is used for M = A, so the argument should either be written for that case or justified by a finite-cell argument. This is a small but load-bearing gap in the localization theorem on which the later computations rely.","section":"§3.1, Theorem 3.5(2)"}],"minor_comments":[{"comment":"In the ray case, the sentence invoking Theorem 2.10 refers to zigzag cycles of homology -η_i, but the relevant corner matchings are P_i and P_{i+1}; presumably the intended homology class is -η_{i+1}. Please correct the index.","section":"§4.2, Lemma 4.5"},{"comment":"The symbol Z is overloaded: it denotes the center of J(Q), the integer lattice, and the Laurent variable z in the localization. In formulas such as Z·N_C in §4.2 this is confusing; a distinct notation for the center would help.","section":"Throughout"},{"comment":"There are several typographical slips: 'noncommutati ve' in the abstract, 'compactly suppported' in §2.9, and 'Van den Berg' on page 21 for 'Van den Bergh'.","section":"Abstract and §2.9"},{"comment":"The formula in item (4) writes {∂_α, f} = deg_{P_i}(f)x_αℓ^{-1}f, but for a homogeneous f of arbitrary degree the right-hand side should be x_αℓ^{-1} times a scalar depending on the degree of f; please state the degree convention for f explicitly.","section":"§4.5, Proposition 4.21"}],"recommendation":"major_revision","confidential_remarks":"I see no citation-pattern or novelty concerns. The main risk is that the proof is built on imported structural facts about minimal paths in dimer algebras, and the paper's own derivation classification does not fully justify one critical step. If the author can supply the missing arguments indicated in the major comments, the paper would be a solid contribution."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nYou asked about Wong's 'Dimer Models and Hochschild Cohomology.' Short version: this is a real computation with full detail, and it deserves a serious referee. The headline results — explicit additive bases and the full BV structure for HH^*(J(Q)), plus the compactly supported HH of the matrix factorization category — are genuinely new. It is not a repackaging. The paper works through localization, grading arguments, and a spectral sequence carefully, and it does not fake anything I can see. Theorems 1.2, 1.3, 4.6, 4.19, and 5.3 are stated cleanly and proven step-by-step.\n\nThe main caveat, and it is real, sits in Section 4.2. Lemma 4.5 and Theorem 4.6 classify derivations of J(Q) by testing degrees in perfect matchings. The test assumes that classes in the localized algebra are represented by minimal paths that are unique up to powers of the potential, citing [6, Lemma 7.4] and [8, Lemma 3.18]. The paper states these imports honestly, but it does not prove them, and everything in HH^1 — and then HH^2 and HH^3 via degree counts and Van den Bergh duality — inherits that dependency. I do not see an internal error in the surrounding logic, but the proof of Lemma 4.5 would be more convincing if the path-uniqueness were spelled out or if the referee verifies the citations cover exactly the scenario used. This is not a manufactured worry; it is the genuine load-bearing assumption.\n\nOther soft spots are minor. The open question of whether compactly supported HH maps isomorphically to ordinary HH is explicitly flagged as unaddressed, which is fine for a statement about the compactly supported side. There are small typos, e.g., the '1<n' in Theorem 1.2 should likely be '1≤n' based on the internal statement. The paper relies on substantial structural results (Calabi-Yau property, the matrix algebra description in Theorem 2.13, Gulotta's matching theorem), all standard in the field. No sign of circularity or self-citation games.\n\nVerdict: send it to review. I would bring it to a reading group, and if I were in the area I would cite it. The ideal referee knows dimer models well enough to judge whether the imported minimal-path uniqueness is applied in exactly the form needed.","headline":"A serious, detailed computation of Hochschild cohomology and BV structure for toric dimer models; the main theorems are new and mostly well-proven, with one load-bearing imported uniqueness result that a referee should verify.","tokens_in":43740,"tokens_out":2625,"would_cite":true,"duration_ms":29284,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["16E40","16G20","14J33"],"pacs":[],"model":"deepseek-v4-flash","headline":"For a zigzag consistent dimer in a torus, the Hochschild cohomology of its Jacobi algebra and its BV structure are determined by the dimer's fan and perfect matchings.","keywords":["dimer models","Jacobi algebras","Hochschild cohomology","Batalin-Vilkovisky structure","matrix factorizations","toric geometry","noncommutative crepant resolutions","mirror symmetry"],"falsifier":"Find a zigzag consistent dimer in a torus with two minimal paths between the same vertices that have the same homotopy class and the same degrees in every corner matching but are not equal in $J(Q)$ up to a power of $\\ell$; the claimed bases for $HH^1$, $HH^2$, and $HH^3$ would then contain spurious classes. Concretely, checking the suspended pinch point by a direct bar-complex computation of $HH^1$ would either confirm or contradict the dimension formula.","tokens_in":42690,"feed_emoji":"🧮","tokens_out":17161,"duration_ms":159718,"temperature":0.7,"pith_summary":"This paper claims that for a zigzag consistent dimer drawn on a torus, the Hochschild cohomology of the associated Jacobi algebra—the algebra built from quiver paths modulo the dimer's face relations—is completely described by dimer combinatorics: the antizigzag fan, the corner matchings, and the powers of zigzag cycles. It gives explicit additive bases for $HH^0$ through $HH^3$ and shows that the Batalin-Vilkovisky operator, the Gerstenhaber bracket, and the cup products are determined by the same data. From this it computes the compactly supported Hochschild cohomology of the matrix-factorization category of the dimer's Landau-Ginzburg model and identifies it additively with the cohomology of the punctured mirror surface together with even and odd polynomial copies attached to each zigzag cycle. If correct, the computation turns deformation-theoretic invariants of noncommutative toric resolutions and their mirrors into finite combinatorial data.","feed_headline":"Dimer combinatorics fixes Hochschild cohomology and its BV structure","feed_subtitle":"The dimer's fan and matchings yield explicit bases and a full Batalin-Vilkovisky structure for the Jacobi algebra.","key_machinery":"The load-bearing mechanism is comparison with central localization. For a Calabi-Yau algebra $A$, the paper proves that Hochschild cohomology commutes with localization away from a central element: $HH^*(\\hat A)$ is the localization of $HH^*(A)$, and the BV operators are related by $\\hat{\\Delta}(f \\otimes s^{-1}) = \\Delta(f) \\otimes s^{-1} - \\{s, f\\} \\otimes s^{-2}$. For a dimer, $J(Q)[\\ell^{-1}]$ is Morita equivalent to the group algebra of the torus fundamental group, i.e. Laurent polynomials in three variables, whose Hochschild cohomology is the algebra of polyvector fields with a divergence operator as BV differential. The hard lifting step—deciding which classes on the localization come from $J(Q)$—is carried out with perfect-matching gradings (integral gradings from selecting one arrow in each face) and the antizigzag fan, whose cones $\\sigma_i$ and semigroup algebras $S_i$ index the non-local summands; minimal-path uniqueness then makes the basis well defined.","core_discovery":"The central discovery is that, under zigzag consistency, the entire Hochschild cohomology of the Jacobi algebra $J(Q)$ of a toric dimer is additively built from the center $Z$, the lattice $N^{out}$ of one-parameter subgroups of outer automorphisms, and powers of zigzag cycles. In the paper's notation, $HH^0(J(Q)) \\cong Z$, $HH^1(J(Q)) \\cong Z \\otimes N^{out}_{\\mathbb{C}} \\oplus \\bigoplus_i S_i$, while $HH^2(J(Q))$ and $HH^3(J(Q))$ are direct sums of the corresponding wedge powers of $N^{out}_{\\mathbb{C}}$, the semigroup algebras $S_i = \\mathbb{C}[\\mathrm{Int}\\,\\sigma_i \\cap \\mathbb{Z}^2]$, powers of zigzag cycles, and, in the top degree, vertex generators. The paper further shows that the BV operator is unique up to scaling, is computed by a divergence operator on the localization $J(Q)[\\ell^{-1}]$, and satisfies explicit formulas on all basis elements; together with the cup products and Gerstenhaber bracket, this determines the full BV algebra structure.","pith_inferences":["The formulas suggest that $HH^*(J(Q))$ depends only on the antizigzag fan together with the corner-matching data, so dimers with the same fan and matching polygon should have isomorphic Hochschild cohomology; the paper does not explicitly state this invariance.","If the natural map from compactly supported to ordinary Hochschild cohomology of the matrix-factorization category is an isomorphism, Theorem 1.3 would upgrade to a full computation of $HH^*(MF(J(Q),\\ell))$, matching the A-side wrapped category on the cochain level rather than only additively.","A testable extension is to replace the torus by a higher-genus surface: the localization comparison would express $HH^*(J(Q))$ in terms of the string-topology BV structure on the free loop space of the surface, provided an analog of the minimal-path-uniqueness lemma holds.","The multiplicative discrepancy in the five-punctured-sphere example predicts that the filtration on symplectic cohomology recording parallel zigzags is the B-side shadow of the BV algebra's noncommutativity; this could be checked by comparing the filtered ring structure on the A-side."],"forward_implications":["The additive formulas give explicit bases for $HH^0$ through $HH^3$: the ranks and generator degrees are read directly from the antizigzag fan, the corner matchings, and the number of parallel zigzag cycles.","The complete BV algebra structure means the Gerstenhaber bracket, the BV operator, and the cup product on $HH^*(J(Q))$ are computable for any toric dimer, so deformation-theoretic invariants of the associated noncommutative crepant resolution are explicit.","All first-order Calabi-Yau deformations of $J(Q)$ are deformations of the superpotential, classified up to gauge equivalence by boundary cycles, minimal interior cycles, and powers of antizigzag cycles.","For the matrix-factorization category, the compactly supported Hochschild cohomology is additively $H^*(\\Sigma^\\vee \\setminus Q_0^\\vee, \\mathbb{C})$ plus even and odd polynomial generators $Z_j^n$ attached to each zigzag; this reproduces the symplectic cohomology of the punctured mirror surface as a BV complex.","In the four-punctured-sphere example the computed BV algebra is isomorphic to the symplectic cohomology algebra, while in the five-punctured-sphere example the additive and BV-operator structures agree but the ring structures differ, pinpointing where the B-side remembers more than the punctures do."],"supporting_citations":[{"why":"Establishes the path-uniqueness lemma (minimal paths differ by powers of the potential) and links zigzag consistency to cancellation, on which the lifting argument rests.","marker":"[6]"},{"why":"Supplies the existence and uniqueness of minimal paths in toric dimers and the center description used in the proofs of HH^0 and HH^3.","marker":"[8]"},{"why":"Provides the antizigzag-fan description of the center, the perfect-matching and N-out machinery, and the noncommutative crepant resolution context.","marker":"[10]"},{"why":"Proves that zigzag consistency makes the Jacobi algebra Calabi-Yau of dimension 3, which supplies the BV structure and the duality between homology and cohomology.","marker":"[14]"},{"why":"Fixes the corner matchings and identifies zigzag homology classes with outward normal vectors, giving the boundary combinatorics of the matching polygon.","marker":"[23]"},{"why":"Identifies the localized Jacobi algebra with matrices over the surface group algebra, yielding the Laurent-polynomial model in the torus case.","marker":"[7]"},{"why":"Provides the curved-algebra spectral sequence and the vanishing of ordinary Hochschild cohomology for nonzero curvature, used in the Section 5 computation.","marker":"[13]"},{"why":"Gives the isomorphism between compactly supported Hochschild cohomology of the curved algebra and of its matrix-factorization category.","marker":"[29]"},{"why":"Supplies the mirror dual dimer, the bijection between zigzag cycles and dual vertices, and the mirror-symmetry equivalence motivating the matrix-factorization computation.","marker":"[9]"},{"why":"Provides the duality identifying Hochschild cohomology with Hochschild homology via capping with a volume, used to transfer homology computations to cohomology.","marker":"[34]"}],"fun_headline_variants":["Dimer combinatorics yields full BV structure","Hochschild cohomology computed via dimer fan","Zigzag consistency fixes Hochschild and BV","Dimer matchings dictate cohomology and BV"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole computation rests on the imported lemma that, in the localized algebra, a path is determined by its homotopy class and its degrees in the perfect matchings, so minimal paths are unique up to powers of the potential; if two non-equivalent minimal paths shared those data, the derivation basis and everything built from it would fail.","fun_headline_variants_meta":{"raw":{"variants":["Dimer combinatorics yields full BV structure","Hochschild cohomology computed via dimer fan","Zigzag consistency fixes Hochschild and BV","Dimer matchings dictate cohomology and BV"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000223,"raw_usage":{"total_tokens":1447,"prompt_tokens":922,"completion_tokens":525,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":538,"completion_tokens_details":{"reasoning_tokens":465}},"tokens_in":538,"tokens_out":525,"duration_ms":5467,"temperature":1.0,"reasoning_tokens":465,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:28:09.753839+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Find a zigzag consistent dimer in a torus with two minimal paths between the same vertices that have the same homotopy class and the same degrees in every corner matching but are not equal in $J(Q)$ up to a power of $\\ell$; the claimed bases for $HH^1$, $HH^2$, and $HH^3$ would then contain spurious classes. Concretely, checking the suspended pinch point by a direct bar-complex computation of $HH^1$ would either confirm or contradict the dimension formula.","supporting_citations":[{"cited_title":"Consistency conditions for dimer models","cited_arxiv_id":null,"evidence_quote":"Establishes the path-uniqueness lemma (minimal paths differ by powers of the potential) and links zigzag consistency to cancellation, on which the lifting argument rests."},{"cited_title":"A dimer ABC","cited_arxiv_id":null,"evidence_quote":"Supplies the existence and uniqueness of minimal paths in toric dimers and the center description used in the proofs of HH^0 and HH^3."},{"cited_title":"Dimer models and Calabi-Yau algebra s","cited_arxiv_id":null,"evidence_quote":"Provides the antizigzag-fan description of the center, the perfect-matching and N-out machinery, and the noncommutative crepant resolution context."},{"cited_title":"Consistency conditions for brane tilings","cited_arxiv_id":null,"evidence_quote":"Proves that zigzag consistency makes the Jacobi algebra Calabi-Yau of dimension 3, which supplies the BV structure and the duality between homology and cohomology."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Fixes the corner matchings and identifies zigzag homology classes with outward normal vectors, giving the boundary combinatorics of the matching polygon."},{"cited_title":"Calabi-Yau algebras and weighted quiver polyhedra","cited_arxiv_id":null,"evidence_quote":"Identifies the localized Jacobi algebra with matrices over the surface group algebra, yielding the Laurent-polynomial model in the torus case."},{"cited_title":"Curved A∞ algebras and Landau-Ginzburg models","cited_arxiv_id":null,"evidence_quote":"Provides the curved-algebra spectral sequence and the vanishing of ordinary Hochschild cohomology for nonzero curvature, used in the Section 5 computation."},{"cited_title":"Hochsch ild (co)homology of the second kind I","cited_arxiv_id":null,"evidence_quote":"Gives the isomorphism between compactly supported Hochschild cohomology of the curved algebra and of its matrix-factorization category."},{"cited_title":"Noncommutative mirror symmetry for punc tured surfaces","cited_arxiv_id":null,"evidence_quote":"Supplies the mirror dual dimer, the bijection between zigzag cycles and dual vertices, and the mirror-symmetry equivalence motivating the matrix-factorization computation."},{"cited_title":"A relation between Hochschild hom ology and cohomology for Gorenstein rings","cited_arxiv_id":null,"evidence_quote":"Provides the duality identifying Hochschild cohomology with Hochschild homology via capping with a volume, used to transfer homology computations to cohomology."}],"review_version":1}