{"id":"d63c1155-01e1-48d8-a044-f5d1f2ad44e1","arxiv_id":"1908.01427","paper_version":2,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For all complex parameters, Gevrey solutions of A-hypergeometric systems along coordinate subspaces embed into formal Nilsson solutions; with a cone condition the spaces are equal and have dimension vol(τ).","lead":"This paper proves that every Gevrey solution of an A-hypergeometric system along a coordinate subspace can be rewritten as a formal Nilsson solution in a suitable direction. Under an extra geometric condition the two solution spaces coincide, giving a dimension formula in terms of normalized volume.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Equality theorem rests on unproved extension of [Sai02, Prop 5.4] to non-homogeneous IA; no independent verification is given.","rationale":"The reader's weakest assumption concerned the slice-wise conversion in Theorem 3.3. That step is indeed the technical heart of the containment, but the hypotheses (Aτ pointed, T(τ) refining Γ_{Aτ}) are precisely those required for [DMM12, Theorem 6.4], and Lemma 2.8 supplies the needed uniform degree bound. No gap was found there. The more serious soft spot is in Theorem 3.4, where the proof depends on an explicit assertion that a homogeneous-case proposition extends to non-homogeneous toric ideals, with no proof or reference. This is the only place where the author states validity beyond the published result without support, and a failure there directly falsifies the equality claim. A concrete small-case computation would settle the issue without requiring a full re-proof, and the paper would remain acceptable if the extension holds.","tokens_in":15013,"tokens_out":55627,"duration_ms":526471,"concrete_test":"Check the asserted extension on a non-homogeneous pointed example where pos(Aτ) = pos(A), e.g., A = [[1,0,1],[0,1,1]], τ = {1,2}. Compute a basis of N_w(H_A(β)) for a generic β (e.g., using the algorithm from [DMM12] or the Γ-series in Theorem 2.3) and inspect the log polynomials p_u. If every p_u lies in C[y1,y2], the claim survives; if any p_u involves y3, the equality theorem has a counterexample.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The proof of Theorem 3.4 applies [Sai02, Proposition 5.4] to conclude p_u(y) ∈ C[y_j : j ∈ vert(T_w)] for formal Nilsson solutions of the possibly irregular system H_A(β). The proposition was proved for homogeneous IA; the paper asserts, without proof, that the proof remains valid when IA is not homogeneous. This is load-bearing: if p_u depends on logarithms of variables outside τ, then a basic Nilsson solution φ is not a Gevrey series along Yτ (it would contain logarithms in the normal variables), and the equality in Theorem 3.4 fails. The paper provides no derivation or reference for the non-homogeneous case, and the extension is not a formal consequence of the published statement.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies formal solutions of A-hypergeometric systems H_A(β) and compares two notions: Gevrey series solutions along a coordinate subspace Y_τ and formal Nilsson solutions in a direction w. After presenting preliminaries on regular triangulations, the A-hypergeometric fan, Γ-series, Gevrey solutions, and formal Nilsson solutions, it proves in Section 2 several auxiliary facts about N_w(H_A(β)), including a uniform bound on the degree of logarithmic polynomials (Lemma 2.8). In Section 3, Theorem 3.3 shows, for A_τ pointed of rank d and a weight vector w constructed from a regular triangulation of A_τ, that every Gevrey solution along Y_τ is a formal Nilsson solution of H_A(β) in the direction of w. Theorem 3.4 adds the assumption pos(A)=pos(A_τ) and concludes the spaces are equal, with dimension vol(τ) for β outside the rank-jumping set ε(A); an example shows the condition pos(A)=pos(A_τ) is necessary.","tokens_in":15169,"tokens_out":23214,"duration_ms":222326,"significance":"The main results are a genuine contribution if they hold. Theorem 3.3 removes the generic-parameter assumption from the containment statement in [Fer10], and Theorem 3.4 provides a two-way bridge between Gevrey solutions along coordinate subspaces and formal Nilsson solutions, together with a dimension formula for non-jumping parameters. The proof strategy using slice systems and [DMM12, Theorem 6.4] is coherent, and the auxiliary results in Section 2 (especially the uniform degree bound in Lemma 2.8) are of independent interest. The paper also gives a concrete example demonstrating that the hypothesis pos(A)=pos(A_τ) is necessary. The main weakness is the unproved extension of [Sai02, Proposition 5.4] in Theorem 3.4; if that extension is supplied, the paper should be acceptable.","major_comments":[{"comment":"The proof invokes [Sai02, Proposition 5.4] with the parenthetical assertion that 'its proof is also valid when I_A is not necessarily homogeneous' to conclude p_u(y) ∈ C[y_j : j ∈ vert(T_w)] for basic Nilsson solutions of a possibly irregular system H_A(β). This is load-bearing: it is the step that prevents logarithmic factors in the normal variables, and without it a basic Nilsson solution need not be a formal series along Y_τ, so the equality in Theorem 3.4 would not follow. The published proposition is stated for homogeneous I_A, and the paper neither proves the extension nor gives a reference covering the non-homogeneous case. I request that the author supply a proof of the extension or a precise citation, and state any extra hypotheses (for instance, pointedness or genericity of w) that the proof requires.","section":"Section 3, proof of Theorem 3.4"},{"comment":"The proof of the lower bound dim_C(N_w(H_A(β))) ≥ ∑_{σ∈T_w} vol(σ) for all β consists of the statement that one can 'apply the same procedure as in the proof of [SST00, Theorem 3.5.1]'. This is load-bearing for Corollary 2.2 and hence for the dimension formula in Theorem 3.4. The deformation argument in [SST00] is written for regular systems and a fixed Γ-series basis; in the present setting H_A(β) may be irregular, so one must verify that the limiting series are basic Nilsson solutions in the same direction w and that they remain linearly independent with the claimed cardinality. Please expand this step.","section":"Section 2, Lemma 2.1"}],"minor_comments":[{"comment":"The notation vert(T_w) is used in the proof of Theorem 3.4 but is not defined; a reader must infer that it means the set of column indices appearing as vertices of the triangulation T_w. Please add a definition.","section":"Section 1.2 / Theorem 3.4"},{"comment":"The construction in the paragraph after Lemma 3.2 uses w(τ) both for the weight vector inducing T(τ) and for a newly chosen generic vector in R^τ; the two roles are easy to confuse. Please use distinct notation.","section":"Section 3, before Theorem 3.3"},{"comment":"The statement of Theorem 3.4 does not define the weight vector w; it should say explicitly that w is the vector constructed in Lemma 3.2, rather than leaving this to the preceding paragraph.","section":"Section 3, Theorem 3.4"},{"comment":"In condition ii) of Definition 1.7 the letter C denotes both the support lattice set in the series (1.7) and the strongly convex open cone in C_w; using the same symbol for two objects makes the definition hard to parse. Please choose different letters.","section":"Section 1.6, Definition 1.7"}],"recommendation":"major_revision","confidential_remarks":"Please ask the author to address the [Sai02, Proposition 5.4] issue before acceptance. I do not see a circularity or novelty problem; the self-citations are to the author's own prior work and are used appropriately. The paper fits the journal's scope. My recommendation is major_revision rather than rejection because the defect is local and likely fixable by a proof or a precise reference."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"First, the substance: this paper removes genericity assumptions from two solution theories of A-hypergeometric systems. Theorem 3.3 shows all Gevrey solutions along a coordinate subspace are formal Nilsson solutions for every β, and Theorem 3.4 gives equality under a cone condition, with dimension vol(τ) outside the rank-jumping set. That is genuinely new and worth having. The Section 2 results, especially the parameter-independent degree bound in Lemma 2.8 and the local-constancy statements in Corollary 2.4, are also useful. The proofs are generally careful and the reduction to slice systems in Theorem 3.3 is well organized. The counterexample in Example 3.6, showing the condition pos(A)=pos(Aτ) is necessary, is a nice touch.\n\nNow the soft spot. The stress-test note lands. In the proof of Theorem 3.4, the author applies [Sai02, Proposition 5.4] to assert that the logarithmic polynomial p_u in a basic Nilsson solution depends only on variables y_j with j∈vert(Tw), and says \"whose proof is also valid when IA is not necessarily homogeneous\"—but gives no proof or reference. That is load-bearing. If p_u involves a logarithm of a variable outside τ, the series is not a Gevrey series along Yτ, and the equality in Theorem 3.4 fails. The published proposition is stated for homogeneous IA. The extension may be true, but it is not a formal consequence of the published statement, and the paper does not show it. A referee should require the author to supply the argument or a citation. This is a finite gap, not a collapse; the rest of the paper stands on its own.\n\nThe citation pattern is fine. Self-citations go to the author's own earlier work and to [DMM12]; they are used as tools, not to assume conclusions. The paper is honest about its limitations.\n\nBottom line: this is a solid paper for specialists in D-modules and hypergeometric systems. It deserves a serious referee, and the referee should press on the Sai02 extension. I would accept it only after that issue is resolved.","headline":"Solid extension of Gevrey-to-Nilsson results to all parameters; one load-bearing assertion about extending [Sai02, Prop 5.4] to non-homogeneous IA needs a proof.","tokens_in":15691,"tokens_out":2729,"would_cite":true,"duration_ms":25971,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["13N10","33C70","14M25","32C38"],"pacs":[],"model":"deepseek-v4-flash","headline":"Every Gevrey solution of an A-hypergeometric system along a coordinate subspace is a formal Nilsson solution, and under a cone condition the two spaces coincide.","keywords":["A-hypergeometric system","Gevrey series","formal Nilsson solution","D-module","regular triangulation","initial ideal","rank-jumping parameter","normalized volume"],"falsifier":"Take an $A$ and $\\tau$ with $\\operatorname{rank}(A_\\tau)=d$ but $A_\\tau$ not pointed; if a Gevrey solution along $Y_\\tau$ were found whose slice coefficients are not formal Nilsson series of the restricted systems, Theorem 3.3 would fail. Conversely, in a case with $\\operatorname{pos}(A_\\tau)\\neq\\operatorname{pos}(A)$, compute both dimensions: the paper's Example 3.6 predicts they differ, so any symbolic computation showing equality for a generic $\\beta$ would indicate the cone condition is not needed.","tokens_in":14839,"feed_emoji":"📐","tokens_out":10263,"duration_ms":93239,"temperature":0.7,"pith_summary":"This paper proves that every Gevrey formal solution of an $A$-hypergeometric system along a coordinate subspace can be rewritten as a formal Nilsson solution: a series of the form $x^v$ times a sum of monomials $x^u$ with polynomial coefficients in logarithms, supported on a cone inside the kernel lattice of $A$. The containment holds for every parameter vector $\\beta$ whenever the restricted matrix $A_\\tau$ is pointed and has full rank, and the direction $w$ is obtained from a regular triangulation of $A_\\tau$ that refines the standard triangulation. If, in addition, the positive cone generated by the columns of $A_\\tau$ equals the positive cone of $A$, the containment becomes an equality: the two solution spaces coincide for all $\\beta$. Outside the rank-jumping set, the common dimension equals the normalized volume of $A_\\tau$, a purely combinatorial quantity independent of $\\beta$. This matters because formal Nilsson solutions are the output of explicit algorithmic constructions, so the result imports those algorithms to the Gevrey setting.","feed_headline":"Gevrey solutions of A-hypergeometric systems are formal Nilsson series","feed_subtitle":"With a cone condition the two spaces coincide, and their dimension is a normalized volume.","key_machinery":"The carrying object is the slice system $H_{A_\\tau}(\\beta-A_\\tau\\alpha)$, obtained by deleting the variables outside $\\tau$ and shifting the parameter by $A_\\tau\\alpha$. The proof writes a Gevrey solution as a sum over $\\alpha\\in\\mathbb{N}^{\\tau}$ of terms $f_\\alpha(x_\\tau)x_\\tau^\\alpha$; each coefficient is a holomorphic solution of the corresponding slice system, and a cited theorem rewrites it as an element of $N_{w(\\tau)}(H_{A_\\tau}(\\beta-A_\\tau\\alpha))$, the space of formal Nilsson series in the restricted variables. The global direction $w$ is built from a regular triangulation $T(\\tau)$ of $A_\\tau$ that refines the standard triangulation $\\Gamma_{A_\\tau}$, extended to a triangulation $T$ of $A$, which ensures $w(\\tau)$ is a perturbation of $(1,\\ldots,1)$. Lemma 2.8 supplies a uniform bound, depending only on $n$, $d$, and $\\operatorname{vol}(A)$, on the degrees of the logarithmic polynomials appearing in any basic Nilsson solution; this bound is what makes it possible to sum the coefficient-wise pieces back into a single formal Nilsson series rather than a formal sum with unbounded log-degree. In the equality direction, a cited proposition shows that when $\\operatorname{pos}(A_\\tau)=\\operatorname{pos}(A)$, the logarithmic polynomials depend only on the $\\tau$-variables, so a Nilsson series in the global direction is visibly a formal series along $Y_\\tau$.","core_discovery":"The central discovery is a bridge between two formal solution theories of the $A$-hypergeometric system $H_A(\\beta)$: Gevrey series along the coordinate variety $Y_\\tau=\\{x_j=0:j\\in\\tau\\}$, and formal Nilsson series in a direction $w$. A Gevrey solution $f=\\sum_{\\alpha\\in\\mathbb{N}^\\tau} f_\\alpha(x_\\tau)x_\\tau^\\alpha$ has coefficients $f_\\alpha$ that are holomorphic solutions of the slice system $H_{A_\\tau}(\\beta-A_\\tau\\alpha)$; known results identify those holomorphic solutions with formal Nilsson solutions of the slice system in a restricted direction $w(\\tau)$, provided $A_\\tau$ is pointed and $w(\\tau)$ is a perturbation of the all-ones vector. The paper shows these slice-wise representations can be reassembled into one formal Nilsson series of $H_A(\\beta)$ in a global direction $w$, proving the containment. When $A$ is pointed and $\\operatorname{pos}(A_\\tau)=\\operatorname{pos}(A)$, the reverse inclusion holds as well, so the two spaces are equal for every $\\beta$; for $\\beta\\notin\\varepsilon(A)$, their dimension is $\\operatorname{vol}(\\tau)$, the normalized volume of $A_\\tau$.","pith_inferences":["One extension not made in the paper: the coefficient-wise rewriting suggests a constructive algorithm—solve each slice system, then reassemble with the uniform log-degree bound—turning the containment theorem into a way to compute a Gevrey solution's Nilsson expansion, not merely to prove it exists.","A second extension: Example 3.6 suggests the failure of equality when $\\operatorname{pos}(A_\\tau)\\neq\\operatorname{pos}(A)$ is controlled by the extra maximal simplices in $T_w$ outside $T(\\tau)$; a natural question is whether the quotient $N_w/(\\text{Gevrey space})$ has a combinatorial basis indexed by those extra simplices.","A third extension: because Lemma 2.8 bounds logarithmic degrees uniformly in $\\beta$, parameter-uniform algorithms could precompute the finite list of possible log-polynomial degrees for a fixed $A$, then sweep $\\beta$; this would make the rank-jumping phenomenon computationally visible as a jump in dimension at $\\varepsilon(A)$."],"forward_implications":["If the containment holds, any algorithm that computes formal Nilsson solutions in a chosen direction can be used to compute the Gevrey solutions along a coordinate subspace, so Gevrey solvability is not a separate computation but a special case of a uniform construction.","Under the cone equality $\\operatorname{pos}(A_\\tau)=\\operatorname{pos}(A)$, bases for $N_w(H_A(\\beta))$ are also bases of the Gevrey solution space, giving explicit dimension and basis formulas for all $\\beta$.","Outside the rank-jumping set $\\varepsilon(A)$, the dimension of the common space is $\\operatorname{vol}(\\tau)$, so the Gevrey solution space has a stable, purely combinatorial dimension independent of $\\beta$.","For non-rank-jumping parameters, the formal Nilsson space depends only on the cone $C_w$ of the $A$-hypergeometric fan, not on the particular weight vector; hence the Gevrey space too is constant on such cones.","Corollary 2.6 gives a lower bound, sharp for generic $\\beta$, on the dimension of convergent series inside $N_w$, measured by simplices contained in facets of the standard triangulation; this relates formal solution spaces back to holomorphic solutions."],"supporting_citations":[{"why":"Identifies holomorphic solutions of the pointed system with formal Nilsson solutions for perturbations of the all-ones vector; this is the coefficient-wise bridge in Theorem 3.3.","marker":"[DMM12, Theorem 6.4]"},{"why":"Shows each coefficient $f_\\alpha(x_\\tau)$ of a Gevrey solution is a holomorphic solution of the slice system $H_{A_\\tau}(\\beta-A_\\tau\\alpha)$.","marker":"[Fer10, Lemma 6.11]"},{"why":"Supplies the initial-form argument: initial forms of solutions are solutions of the initial ideal, used to locate exponents and prove support cones.","marker":"[SST00, Theorem 2.5.5]"},{"why":"Ensures log-polynomials of formal Nilsson solutions depend only on the $\\tau$ variables, a step in showing Nilsson series are series along $Y_\\tau$.","marker":"[Sai02, Proposition 5.4]"},{"why":"Gives the derivative formula for $x^v p(\\log x)$, used both in the uniform degree bound and in the support-cone argument.","marker":"[Sai02, Lemma 5.3]"},{"why":"Defines the rank-jumping set $\\varepsilon(A)$, used in stating the dimension result for $\\beta$ outside this set.","marker":"[MMW05]"},{"why":"Proves constancy of the rank of the initial ideal for $\\beta$ outside $\\varepsilon(A)$, used in Corollary 2.2 for the dimension equality.","marker":"[SW08, Theorem 4.28]"}],"fun_headline_variants":["A-hypergeometric Gevrey solutions are formal Nilsson series","When cones align, Gevrey equals Nilsson for A-hypergeometric","A-hypergeometric: Gevrey solutions are always Nilsson"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof assumes the restricted matrix $A_\\tau$ is pointed and full-rank, and that the weight vector $w(\\tau)$ is a perturbation of the all-ones vector, so that a cited theorem converts holomorphic solutions of every slice system into formal Nilsson solutions; if pointedness or the perturbation condition fails, the coefficient-wise bridge collapses.","fun_headline_variants_meta":{"raw":{"variants":["A-hypergeometric Gevrey solutions are formal Nilsson series","When cones align, Gevrey equals Nilsson for A-hypergeometric","A-hypergeometric: Gevrey solutions are always Nilsson"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001965,"raw_usage":{"total_tokens":7632,"prompt_tokens":849,"completion_tokens":6783,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":465,"completion_tokens_details":{"reasoning_tokens":6719}},"tokens_in":465,"tokens_out":6783,"duration_ms":48870,"temperature":1.0,"reasoning_tokens":6719,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:13:16.349052+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take an $A$ and $\\tau$ with $\\operatorname{rank}(A_\\tau)=d$ but $A_\\tau$ not pointed; if a Gevrey solution along $Y_\\tau$ were found whose slice coefficients are not formal Nilsson series of the restricted systems, Theorem 3.3 would fail. Conversely, in a case with $\\operatorname{pos}(A_\\tau)\\neq\\operatorname{pos}(A)$, compute both dimensions: the paper's Example 3.6 predicts they differ, so any symbolic computation showing equality for a generic $\\beta$ would indicate the cone condition is not needed.","supporting_citations":[],"review_version":1}