{"id":"f66b2edf-f245-4211-9186-a6d9f8dae517","arxiv_id":"1908.01731","paper_version":4,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"A selfsimilar Hessian manifold with a potential homothetic vector field is locally a product of radiant Hessian manifolds, each described by a single homogeneous potential function.","lead":"This paper classifies Riemannian manifolds that carry a selfsimilar scaling vector field, and then does the same for Hessian manifolds, which combine such scaling with a flat affine structure. The central result says that when the scaling field is a gradient, the Hessian manifold splits locally into simple radiant pieces, given by explicit coordinate formulas.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Printed normal-form formulas in Prop 3.20 / Thm 1.5(ii) / Thm 3.25 are algebraically wrong for general λ (and even for λ=1 the g_{n+1,n+1} term misses a Q term), so the local structure theorem as stated is false.","rationale":"The reader's conditional verdict is confirmed and sharpened. Among the two flagged issues, the algebraic error in Prop 3.20 is the most load-bearing: it directly falsifies the explicit local normal form in Theorem 1.5(ii) and Theorem 3.25, which is a central advertised result. The integrability of the eigenspace distributions V_i is asserted without proof, but it is a standard consequence of parallel distributions in flat affine geometry and is likely repairable; the coordinate-formula error is concrete and demonstrable. I verified the Hessian by direct differentiation and found missing (a−1) factors and a missing Q term in a_{n+1,n+1}. The error is localized and fixable, so the verdict remains CONDITIONAL rather than REJECT: the broader structural theorem (product decomposition into radiant factors) may survive, but the printed normal forms must be corrected. The reader identified the same algebra concern as 'appears to be algebraically inconsistent'; my check confirms it is actually inconsistent, so agreement is partial: I agree on the algebra issue but do not share the stronger weighting of the integrability gap.","tokens_in":23256,"tokens_out":15385,"duration_ms":128270,"concrete_test":"Take n = 1, λ = 1 (so a = 2), and ψ(u,s) = (u/s)^2 + (u/s)^3, where u = x^1 and s = x^2. Then ρψ = 0 and F = s^2 ψ = u^2 + u^3/s. Directly, g_{22} = ∂²F/∂s² = 2u^3/s^3. The paper's Theorem 3.25 formula gives a_{2,2} = 2ψ − uψ_u = −u^3/s^3, which disagrees with the direct computation. The corrected formula 2ψ − 2S + Q, with S = uψ_u = 2u^2/s^2 + 3u^3/s^3 and Q = u^2ψ_{uu} = 2u^2/s^2 + 6u^3/s^3, yields 2u^3/s^3, matching the direct Hessian. A single symbolic computation of this example therefore confirms that the printed formulas are false.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Proposition 3.20, and therefore Theorems 1.5(ii) and 3.25, is algebraically incorrect. Let s = x^{n+1}, a = 2/λ, and F = s^a ψ, with ψ satisfying ρψ = 0, where ρ = Σ_{i=1}^{n+1} x^i ∂_{x^i}. Writing S = Σ_{i=1}^n x^i ψ_i and Q = Σ_{i,j=1}^n x^i x^j ψ_{ij}, the constraint ρψ = 0 gives sψ_s = −S. A direct differentiation yields ∂_s F = s^{a−1}(aψ − S), hence a_{j,n+1} = ∂_j∂_s F = s^{a−1}((a−1)ψ_j − Σ_i x^i ψ_{ij}) and a_{n+1,n+1} = ∂_s²F = s^{a−2}(a(a−1)ψ − 2(a−1)S + Q). The paper instead prints a_{j,n+1} = s^{a−1}(ψ_j − Σ_i x^i ψ_{ij}) and a_{n+1,n+1} = s^{a−2}(2ψ − S). The proof of Prop 3.20 uses ρ(s^a ψ) = 2s^a ψ instead of a s^a ψ, so the derivation is valid only for a = 2 (λ = 1); moreover even for a = 2 the second displayed formula should be 2ψ − 2S + Q, not 2ψ − S. These entries are the explicit normal form advertised in the abstract and Theorem 1.5, so the central local-structure claim is not established as printed. The error is repairable by replacing the displayed entries, but the current statements are false.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper studies Riemannian manifolds carrying a homothetic vector field ξ (L_ξ g = 2g), which it calls selfsimilar manifolds, and their Hessian analogues. Theorem 1.1 classifies global selfsimilar manifolds as either generalized cones M × R_{>0} with metric s²g_M + s ds·α + ds², or Euclidean spaces with a vector field ρ + η (ρ radiant, η Killing), and asserts that every selfsimilar manifold is locally isomorphic to a global one. Theorem 1.2 shows that a global selfsimilar manifold with potential ξ is a Riemannian cone or a flat radiant Euclidean space. The main results for Hessian structures are Theorem 1.4 (a selfsimilar Hessian manifold has potential ξ if and only if it is locally a direct product of radiant Hessian manifolds) and Theorem 1.5/3.25, an explicit local normal form g = Hess((x^{n+1})^{2λ^{-1}}ψ) for each radiant factor, with ψ constant along the radiant field and positivity conditions (3.8). Corollary 3.24 claims existence of radiant Hessian manifolds for every λ ≠ 0, 2.","tokens_in":23665,"tokens_out":23608,"duration_ms":195298,"significance":"The intended outcome — a complete local classification of selfsimilar Hessian manifolds with potential homothetic fields, alongside a global classification of selfsimilar manifolds — is a worthwhile contribution to affine differential geometry. The paper's genuine strengths include the global classification argument (Theorem 2.9, where the curvature blow-up under the contracting flow forces flatness), the correctly derived positive-definiteness criterion in Lemma 3.22 and (3.8), and the explicit link to the λ = 2 extensive-Hessian setting of Garcia-Ariza with its thermodynamic motivation. The product splitting of Theorem 1.4 is plausible and its ingredients are standard, though the key integrability step is not proved or cited. However, the explicit normal-form entries in Proposition 3.20, Theorem 1.5(ii), and Theorem 3.25 are algebraically wrong for general λ; since these formulas are exactly the advertised local structure theorem, the central claim needs repair. The errors are localized and correctable, so I judge the paper promising rather than unsalvageable.","major_comments":[{"comment":"The displayed entries of Hess((x^{n+1})^{2λ^{-1}}ψ) in Proposition 3.20, Theorem 1.5(ii), and Theorem 3.25 are incorrect for general λ. Setting s = x^{n+1}, a = 2λ^{-1}, F = s^a ψ, S = Σ_{i=1}^n x^i ψ_i, and Q = Σ_{i,j=1}^n x^i x^j ψ_{ij}, the condition ρψ = 0 gives sψ_s = −S, and direct differentiation yields ∂_s F = s^{a−1}(aψ − S), hence a_{j,n+1} = ∂_j∂_s F = s^{a−1}((a−1)ψ_j − Σ_{i=1}^n x^i ψ_{ij}) and a_{n+1,n+1} = ∂_s²F = s^{a−2}(a(a−1)ψ − 2(a−1)S + Q). The paper prints a_{j,n+1} = s^{a−1}(ψ_j − Σ x^i ψ_{ij}) and a_{n+1,n+1} = s^{a−2}(2ψ − S); the missing factor a−1 changes the formulas for every λ ≠ 1, and even for a = 2 the second entry is wrong (the correct value is 2ψ − 2S + Q). The proof of Proposition 3.20 errs in its first line, where ρ(s^aψ) is evaluated as 2s^aψ; in fact ρ(s^aψ) = a s^aψ. I verified these computations directly; the repair is localized to the three displayed entries, and the inequalities (3.8) are unaffected since they are derived from Lemma 3.22 rather than from the entries.","section":"Prop. 3.20; Thm. 1.5(ii); Thm. 3.25 (Sec. 3.2.1)"},{"comment":"The proof of Theorem 1.4 relies on the assertion, in Section 3.1 immediately after Proposition 3.6, that the affine manifold C is locally isomorphic to a direct product U = ∏ U_i compatible with the decomposition TC = ⊕ V_i into eigen-subbundles of ∇ξ. This is a standard fact — for a flat torsion-free connection the subbundles V_i are parallel, hence integrable, and a parallel splitting of the tangent bundle induces a local product splitting of the affine structure — but the manuscript states it without proof or reference. Because the 'only if' direction of Theorem 1.4 depends on this step, the paper should add a short argument or a citation.","section":"Section 3.1, after Prop. 3.6"},{"comment":"Corollary 3.24 claims the existence of a radiant Hessian manifold for any λ ≠ 0, 2, but the proof covers only λ < 2. With ψ = Σ_{i=1}^n (x^i)^2, the first inequality of (3.8) is (4λ^{-2} − 2λ^{-1})ψ > 0; since ψ > 0 away from 0, this forces 4λ^{-2} − 2λ^{-1} > 0, i.e., λ < 2. The statement needs a different example for λ > 2 (e.g., ψ = exp(Σ_{i=1}^n x^i) − c with c large enough that ψ < 0 on a small convex neighborhood, for which Hess(ψ) > 0) or a restriction of the claimed range.","section":"Corollary 3.24"}],"minor_comments":[{"comment":"The manuscript contains many typos and grammatical errors ('selfsimiar', 'Eucledean', 'wich', 'describ e', 'a positive definite function on M' where a positive function is meant); a careful proofreading pass is needed.","section":"Throughout"},{"comment":"The statement contains 'where a ∈ R' but no parameter a appears in the formula; either delete the phrase or write η = a η₀ for a fixed generator η₀ of so(n).","section":"Theorem 1.1(i)"},{"comment":"The sentence 'Any tangent vector admits a form a ∂/∂ϕ + b ∂/∂ϕ' should read a ∂/∂ϕ + b ∂/∂s.","section":"Example 2.21"},{"comment":"The displayed computation contains typos: the constant factor of g(∂_r, ∂_r) should be a combination of (4λ^{-2} − 2λ^{-1})/(4 − 2λ), and the equation '2^{2λ^{-1}−2} = 0' should read '2λ^{-1} − 2 = 0'; the exponent argument itself is sound.","section":"Proposition 3.28"},{"comment":"The sentence 'Combining Theorem 3.9 and Theorem 3.28, we get Theorem 3.28' is self-referential; the intended statement is presumably Theorem 1.5(i).","section":"End of proof of Theorem 3.28"},{"comment":"References [CT] and [M] do not appear to be cited in the text; either cite them or remove them from the bibliography.","section":"References"},{"comment":"The claim that ∇ξ|_p has the same eigenvalues λ_1,...,λ_k at every point, and the definition of the subbundles V_i, would benefit from a short justification (parallelism of ∇ξ forces the spectrum to be locally constant).","section":"Section 3.1, after Prop. 3.6"}],"recommendation":"major_revision","confidential_remarks":"The algebra in Proposition 3.20 is checkable and the errors are localized; I verified the correct formulas independently, so the repair is feasible within the manuscript's scope. I therefore recommend major revision rather than rejection. The product-splitting step is standard but needs a citation or short proof. The paper is rough in presentation, with many typographical slips that suggest a careful editorial pass. I found no circularity concerns: the derivations are internal, and the reliance on Goldman's radiant-manifold lectures and on Garcia-Ariza's extensive Hessian structures is legitimate and acknowledged."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nBottom line: the paper has a genuinely useful structural idea, but the advertised local normal form is not correct as printed. Theorem 1.5(ii) and Theorem 3.25, via Proposition 3.20, contain algebraic errors. Until those are fixed, the local classification should not be cited as a theorem.\n\nWhat is actually new: Theorem 1.4 — potential homothetic vector field iff locally a direct product of radiant Hessian factors — is a nice organizing statement, and Theorems 1.1 and 1.2 give a clean global picture for selfsimilar manifolds. The connection to Ruppeiner/Weinhold metrics and to Garcia-Ariza's λ=2 extensive Hessian theory is fair and useful. I do not see circularity or self-citation problems; the reliance on Goldman and on Garcia-Ariza is legitimate.\n\nThe splitting argument is sketched rather than proved. The eigenspace distributions V_i are asserted to integrate to a local product without proof. That is a standard fact for parallel distributions on a flat affine manifold, so it is repairable, but the proof should be written out or cited.\n\nWhere it actually goes wrong: Proposition 3.20. Let s=x^{n+1}, a=2/λ, S=Σ_{i≤n} x^i ψ_i, and Q=Σ_{i,j≤n} x^i x^j ψ_{ij}. From ρψ=0 we get sψ_s=-S. Then ∂_s(s^a ψ)=s^{a-1}(aψ-S), not s^{a-1}(2ψ-S). Consequently,\n\na_{j,n+1}=s^{a-1}((a-1)ψ_j-Σ_i x^i ψ_{ij})\n\nand\n\na_{n+1,n+1}=s^{a-2}(a(a-1)ψ-2(a-1)S+Q).\n\nThe printed versions have the wrong coefficient of ψ_j, miss the Q term, and have the wrong coefficient of S. The proof uses ρ(s^aψ)=2s^aψ instead of a s^aψ, so it is only accidentally right for λ=1, and even there the (n+1,n+1) entry is wrong. This is not a cosmetic typo: it invalidates the central local normal form advertised in the abstract and in Theorem 1.5. The good news is that the fix is local and algebraic, not conceptual — replace the displayed entries and re-run the positivity inequalities. Corollary 3.24's existence claim is probably fine, since it only needs the Hessian of a chosen potential.\n\nWho this is for: people working on Hessian geometry, information geometry, thermodynamic geometry, and radiant affine structures. The structural Theorem 1.4 is worth knowing, but I would not rely on the local classification until a corrected version appears.\n\nRecommendation: this deserves peer review, not a desk reject. A serious referee should ask for corrected formulas and a real proof of the product splitting; both are within reach. My own verdict is conditional: positive on the structural theorem, skeptical of the normal form as stated.","headline":"The product-decomposition idea is sound and worth knowing, but the printed normal-form formulas are wrong for general λ (and even for λ=1), so the local classification as stated is false.","tokens_in":24148,"tokens_out":6743,"would_cite":false,"duration_ms":63620,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper proves that a selfsimilar Hessian manifold with a potential homothetic vector field is locally a direct product of radiant Hessian manifolds.","keywords":["selfsimilar manifold","homothetic vector field","Hessian manifold","radiant manifold","Riemannian cone","conical Hessian domain","potential vector field","flat affine connection"],"falsifier":"For a radial-invariant smooth function $\\psi$ on the slice $x^{n+1}=1$, for example $\\psi=(x^1/x^{n+1})^2$ near a point, set $\\lambda=1$, compute the Hessian matrix of $(x^{n+1})^{2\\lambda^{-1}}\\psi$ in the coordinates of Theorem 1.5(ii), and compare the entries $a_{j,n+1}$ and $a_{n+1,n+1}$ with the displayed formulas; any mismatch would settle that the proposed normal form is not the actual Hessian.","tokens_in":23051,"feed_emoji":"📐","tokens_out":14443,"duration_ms":141806,"temperature":0.7,"pith_summary":"The paper sets out to show that a selfsimilar Hessian manifold—a flat affine manifold whose Riemannian metric is locally the Hessian of a function, equipped with a vector field that scales the metric—has a rigid local structure once that scaling field is a gradient. The first main theorem says that the scaling field is a gradient exactly when the manifold is locally a direct product of radiant Hessian manifolds, where a radiant manifold has a vector field $\\rho$ with $\\nabla\\rho=\\mathrm{Id}$. The second main theorem gives explicit flat-coordinate models for each radiant factor: at a zero of the scaling field the metric is Euclidean, and away from zeros it is the Hessian of $(x^{n+1})^{2\\lambda^{-1}}\\psi$, with $\\psi$ constant along the radial direction and two inequalities controlling positive definiteness. If these theorems are right, the local classification of selfsimilar Hessian manifolds with a gradient scaling field is reduced to these coordinate formulas.","feed_headline":"Selfsimilar Hessian manifolds split into radiant products","feed_subtitle":"Potential homothetic fields force local direct products of radiant Hessian cones, with explicit metrics.","key_machinery":"The machinery is the linear operator $\\nabla\\xi$ on the tangent bundle. Because the flow of $\\xi$ preserves the flat affine connection, $\\nabla\\nabla\\xi=0$, so its eigenvalues are constant on connected components. Potentiality of $\\xi$ (equivalently $d\\,\\iota_\\xi g=0$) makes $\\nabla\\xi$ self-adjoint with respect to $g$, so the tangent space splits into mutually orthogonal, parallel eigenspace distributions $V_i$; the paper combines this splitting with the radiant identity $L_\\rho g = g+\\nabla(\\iota_\\rho g)$ (Proposition 3.10) to derive the Hessian potential formula $g=\\operatorname{Hess}\\bigl(g(\\xi,\\xi)/(4-2\\lambda)\\bigr)$ (Proposition 3.11). This identity is what converts the geometric splitting into explicit coordinate formulas for the metric.","core_discovery":"On the paper's own terms, the central discovery is Theorem 1.4: on a selfsimilar Hessian manifold $(C,\\nabla,g,\\xi)$, the homothetic field $\\xi$ is potential, meaning locally $\\xi=\\operatorname{grad} f$, if and only if $(C,\\nabla,g,\\xi)$ is locally isomorphic to a direct product of radiant Hessian manifolds; if $\\xi$ vanishes somewhere, the factor is a single radiant Hessian manifold with $\\xi$ itself as the radiant field. Theorem 1.5 then gives normal forms. At a zero of $\\xi$, flat coordinates make $g=\\sum_i(dx^i)^2$ and $\\xi=\\sum_i x^i\\partial_{x^i}$. At a point where $\\xi$ is nonzero, coordinates can be chosen with $x^{n+1}>0$, $\\xi=\\lambda\\sum_i x^i\\partial_{x^i}$, $\\lambda\\neq 0,2$, and $g=\\operatorname{Hess}\\bigl((x^{n+1})^{2\\lambda^{-1}}\\psi\\bigr)$, where $\\psi$ is constant along the radiant vector field and the positivity of $g$ is exactly the pair of inequalities (3.8).","pith_inferences":["Editorial extension: a global version of the product theorem would need to control the affine holonomy that can permute the factors when the eigenspace distributions fail to integrate globally; the local theorem leaves this monodromy unaddressed.","Editorial extension: the inequalities (3.8) can be read as a radial-invariant strong-convexity condition; one could test numerically whether the admissible functions $\\psi$ form a convex cone for fixed $\\lambda$, and identify the boundary as a degenerate-Hessian limit.","Editorial extension: because the normal form contains an arbitrary radial-invariant function $\\psi$, it gives an explicit construction kit for Hessian metrics of prescribed homothetic weight $\\lambda$, which may feed examples in affine sphere theory or Monge–Ampère equations."],"forward_implications":["Locally, the classification of selfsimilar Hessian manifolds with a potential vector field reduces to the radiant case; no other local model is needed.","At a zero of the potential homothetic field, the structure is Euclidean: $g=\\sum(dx^i)^2$ and $\\xi$ is the radial vector field.","Away from zeros, every radiant factor is a conical Hessian domain, with metric $g=\\operatorname{Hess}\\bigl((x^{n+1})^{2\\lambda^{-1}}\\psi\\bigr)$ and positivity explicitly controlled by inequalities (3.8).","Because condition (1.3) cannot hold for $\\lambda=2$, the positive-definite normal form excludes the degenerate extensive-Hessian case, which lives at $\\lambda=2$.","Radiant Hessian manifolds exist for every $\\lambda\\neq 0,2$ (Corollary 3.24), so the classification is nonempty and the parameter $\\lambda$ is genuinely free."],"supporting_citations":[{"why":"Supplies the radiant-Hessian identity $L_\\rho g = g+\\nabla(\\iota_\\rho g)$ used to derive the Hessian potential formula, and sets the $\\lambda=2$ context.","marker":"[G-A]"},{"why":"Supplies the radiant-manifold facts that $\\nabla\\rho=\\mathrm{Id}$ makes the flow affine and that $\\nabla\\nabla\\xi=0$, the input for the eigenspace decomposition.","marker":"[Go]"},{"why":"Supplies the definition of Hessian manifolds and the Hessian-metric formalism in which all normal forms are expressed.","marker":"[Sh]"},{"why":"Supplies the affine-atlas formulation of flat affine manifolds used to write radiant coordinates and the local product structure.","marker":"[FGH]"}],"fun_headline_variants":["Selfsimilar Hessian manifolds: radiant product decomposition","Potential homothetic fields split selfsimilar Hessian manifolds","Radiant Hessian products from selfsimilar manifolds","Selfsimilar Hessian manifolds are radiant products locally","Homothetic potential forces radiant Hessian splitting"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing unproved step is the assertion, made right after Proposition 3.6, that the directions in which the scaling field acts with different rates separate into independent coordinate factors of a local product; the coordinate formulas in Theorem 1.5(ii) also rest on Hessian computations that the text does not display.","fun_headline_variants_meta":{"raw":{"variants":["Selfsimilar Hessian manifolds: radiant product decomposition","Potential homothetic fields split selfsimilar Hessian manifolds","Radiant Hessian products from selfsimilar manifolds","Selfsimilar Hessian manifolds are radiant products locally","Homothetic potential forces radiant Hessian splitting"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000605,"raw_usage":{"total_tokens":2800,"prompt_tokens":901,"completion_tokens":1899,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":517,"completion_tokens_details":{"reasoning_tokens":1821}},"tokens_in":517,"tokens_out":1899,"duration_ms":12811,"temperature":1.0,"reasoning_tokens":1821,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:07:01.633904+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For a radial-invariant smooth function $\\psi$ on the slice $x^{n+1}=1$, for example $\\psi=(x^1/x^{n+1})^2$ near a point, set $\\lambda=1$, compute the Hessian matrix of $(x^{n+1})^{2\\lambda^{-1}}\\psi$ in the coordinates of Theorem 1.5(ii), and compare the entries $a_{j,n+1}$ and $a_{n+1,n+1}$ with the displayed formulas; any mismatch would settle that the proposed normal form is not the actual Hessian.","supporting_citations":[],"review_version":1}