{"id":"2c8f28ed-20ab-4de3-b720-2286ab3546db","arxiv_id":"2509.03217","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"For the sigma-2 Hessian equation with positive C^1,1 right-hand side, interior Hessian estimates hold in dimension 4, and in n at least 5 under a dynamic semi-convexity condition.","lead":"This paper proves interior Hessian bounds and C^3,alpha regularity for solutions of the sigma-2 Hessian equation with a variable right-hand side in four dimensions. It extends recent results by Shankar-Yuan for the constant right-hand side in dimension 4, and by Qiu for variable right-hand side in dimension 3.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Prop 4.3 applies Harnack to M_{10r}-u, which is not 2-convex (σ_1(D^2(M-u))=-Δu<0); the two-sided oscillation decay and the Alexandrov regularity step are unsupported.","rationale":"The central claim of the paper depends on a compactness argument that requires an a.e. twice-differentiable point y of the viscosity limit u. That point is supplied by Proposition 4.1, whose proof relies entirely on the weighted Hölder estimate (4.1). The proof of (4.1) in turn depends on the two-sided oscillation decay (4.6), and (4.6) is derived by applying Lemma 4.7 to both M_{10r}-u and u-m_{10r}. The application to M_{10r}-u is invalid because M_{10r}-u is not 2-convex: its Hessian is -D^2u, whose σ_1 is -Δu < 0, so its eigenvalue vector is not in Γ_2. Lemma 4.7 (and the underlying Labutin potential estimate) is stated only for nonnegative 2-convex functions, and the Hessian-measure theory used in its proof requires the Γ_2 condition. This is a genuine proof gap, not a mere stylistic issue: the one-sided Harnack inequality that does apply to u-m_{10r} cannot yield the needed two-sided estimate. The paper's algebraic core (the almost Jacobi inequality and the doubling inequality) appears plausible, and the imported Savin-type theorem is a secondary concern; but the Alexandrov regularity step is load-bearing and is not supported as written. The reader's weakest-assumption analysis identifies exactly this point, and I agree with the REJECT verdict: the proof is incomplete as written, though the central claim may be repairable by replacing the two-sided Harnack argument with a measure-based oscillation estimate. Therefore no change to the reader's verdict is needed.","tokens_in":20073,"tokens_out":10077,"duration_ms":87984,"concrete_test":"Independently re-derive (4.6) from the hypotheses that actually hold for v:=M_{10r}-u: v≥0 and σ_2(D^2v)=f, but σ_1(D^2v)=-Δu<0. In particular, check whether Lemma 4.6 (Labutin's potential estimate, [Lab02, Thm 2.1]) remains valid for such v; if not, Lemma 4.7 cannot be invoked. Concretely, take u(x)=|x|^2/2 on B_1⊂R^4 (a smooth admissible solution with Δu=4>0), set v=M-u, and verify that the hypotheses of Lemma 4.6 fail while testing whether the conclusion (4.3) still holds; this isolates the missing Γ_2 condition. If (4.6) cannot be derived without the invalid application, the reader's objection is confirmed.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing gap is in Proposition 4.3, Step 1 (Eq. (4.6)). To prove the two-sided oscillation decay ω_r ≤ θ ω_{10r}+Cr^2, the author applies the Harnack inequality Lemma 4.7 to both M_{10r}-u and u-m_{10r} on B_{10r}. The second function is nonnegative and 2-convex, since subtracting a constant does not change the Hessian; but M_{10r}-u is not 2-convex in the sense defined in Section 2. Indeed, if λ(D^2u)∈Γ_2, then σ_1(D^2(M_{10r}-u)) = -Δu < 0, so the eigenvalues of D^2(M_{10r}-u) lie outside Γ_2 (the cone requires σ_1>0 and σ_2>0). Lemma 4.7 is proved from Labutin's Wolff-potential estimate (Lemma 4.6), which is stated only for nonnegative 2-convex functions; the sign of σ_1 is essential for the Hessian-measure theory of [TW99] used there. Hence the first Harnack application is invalid. The resulting inequality (4.6) is the basis for the weighted Hölder estimate (4.1), which is the key input to Proposition 4.1 (a.e. twice differentiability). Proposition 4.1 supplies the point y used in the compactness argument of Theorems 1.2 and 1.3. Without it, the proof of the main theorems is unsupported. The one-sided Harnack for u-m alone cannot produce two-sided control; adding a large constant to make -u positive does not restore Γ_2-admissibility. This is exactly the type of sign-sensitive step that the paper's own definition of 2-convexity makes visible.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper claims a priori interior Hessian estimates for the sigma-2 Hessian equation sigma_2(D^2u)=f(x,u,Du) with positive C^{1,1} right-hand side in dimension 4, and in all higher dimensions under the dynamic semi-convexity condition (1.2). It also claims interior C^{3,alpha} regularity for viscosity solutions in dimension 4. The strategy follows Shankar-Yuan's doubling method adapted to variable right-hand sides: the author proves an almost Jacobi inequality (Prop 2.4), a doubling inequality (Prop 3.1), an Alexandrov-type a.e. twice-differentiability theorem for viscosity solutions (Prop 4.1), and then combines these with a generalized Savin small-perturbation theorem (Theorem 5.1) in a compactness argument to prove Theorems 1.2 and 1.3.","tokens_in":20410,"tokens_out":6050,"duration_ms":54861,"significance":"If correct, the main theorem would be a substantial generalization of the Shankar-Yuan Hessian estimates to variable right-hand sides and would establish interior regularity in dimension 4. The algebraic core, especially Proposition 2.4 and Proposition 3.1, is carefully derived; the treatment of the remainder terms from differentiating f appears technically sound, and the paper makes an honest effort to supply full details of the doubling argument. However, the proof of the Alexandrov regularity theorem contains a sign-sensitive gap in the application of the Harnack inequality, and that gap is load-bearing for the compactness argument in the main theorems.","major_comments":[{"comment":"The oscillation decay estimate omega_r <= theta omega_{10r} + C r^2 is obtained by applying Lemma 4.7 to both M_{10r}-u and u-m_{10r}. The function M_{10r}-u is nonnegative but it is not 2-convex in the admissible sense defined in Section 2: its Laplacian equals -Delta u < 0, so the eigenvalues of D^2(M_{10r}-u) are not in Gamma_2. Lemma 4.7 derives from Lemma 4.6, Labutin's Wolff-potential estimate, which is stated only for nonnegative 2-convex functions, and the sign of sigma_1 is essential in that theory. Therefore the first Harnack application is invalid and the two-sided estimate (4.6) is unsupported. Since (4.6) is the basis for the weighted Holder estimate (4.1), Proposition 4.1 (a.e. twice differentiability), and the compactness step in the proofs of Theorems 1.2 and 1.3, the main results are not established by the given argument. A one-sided Harnack estimate for u-m alone cannot produce two-sided control, and adding a large constant to make -u positive does not restore Gamma_2-admissibility.","section":"Section 4, Proposition 4.3, Step 1 (Eq. (4.6))"}],"minor_comments":[{"comment":"The statement writes 'sigma_2(D^2)' where the argument u is missing; it should read 'sigma_2(D^2u)=f'.","section":"Section 4, Lemma 4.7"},{"comment":"The paper uses '2-convex' in the admissible Gamma_2 sense in Section 2 and switches to the Hessian-measure context of Trudinger-Wang in Section 4 without explicitly reconciling the two conventions; this is especially confusing because the validity of Lemma 4.6 depends on the cone condition.","section":"Section 4, Definition 4.5 and Lemma 4.6"},{"comment":"In the verification that the operator G satisfies hypothesis H2 of Theorem 5.1, the ellipticity constants should be controlled in terms of D^2Q and the uniform bounds on f; the paragraph merely asserts this, and a reader needs an explicit check that the constants depend only on the stated quantities.","section":"Section 6, Step 2"}],"recommendation":"reject","confidential_remarks":"The paper depends on the author's unpublished preprint [Fan25] for the generalized Savin theorem. Even setting that aside, the gap in Proposition 4.3 is a fundamental sign incompatibility with the admissibility cone, and the proof of Proposition 4.1 cannot be repaired by a local correction in the manuscript. The main theorems are therefore unsupported."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things to know. First, the paper genuinely extends the recent n=4 sigma-2 Hessian estimate to variable C^{1,1} right-hand sides, which is a natural next step after Shankar-Yuan and Qiu. Second, the proof as written has a real gap in Proposition 4.3: the Harnack inequality from Lemma 4.7 is applied to M_{10r}-u, and that function is not 2-convex in the paper's own sense. Its Laplacian is -\\Delta u < 0, so \\sigma_1 < 0 and it sits outside \\Gamma_2. That invalidates the two-sided oscillation decay (4.6), the weighted Holder estimate (4.1), and hence the Alexandrov regularity theorem that supplies the twice-differentiable point in the compactness argument. The one-sided Harnack for u-m cannot give two-sided control. This is not a cosmetic issue; it is load-bearing.\n\nWhat the paper does well: the derivation of the almost Jacobi inequality in Proposition 2.4 is careful, and the algebra with the remainder terms from f checks out. The doubling inequality in Proposition 3.1 is a clean adaptation of the Guan-Qiu test function. The n >= 5 dynamic semiconvex case is also new, and the paper is honest about what is imported from other works.\n\nSoft spots, in proportion: the Harnack gap is the main one and it is severe. There is also a dependency on the author's own preprint [Fan25] for the generalized Savin theorem; that is an external stated result rather than a circular restatement, but it does mean the main theorem rests on an unverified companion paper. That is worth flagging to whoever referees this.\n\nBottom line: I think the central claim is probably true and repairable, but the proof as written does not support it. The paper deserves a serious referee and a major-revision decision rather than a desk rejection.","headline":"A real generalization of the n=4 sigma-2 Hessian estimate to variable right-hand sides, but the Alexandrov regularity step applies Harnack to a function that is not 2-convex, leaving the main theorem unsupported as written.","tokens_in":21059,"tokens_out":1945,"would_cite":false,"duration_ms":19042,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35B45","35B65","35J60","35J96"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims a priori interior Hessian estimates for the sigma-2 equation with a positive C^{1,1} variable right-hand side in dimension 4, and for n >= 5 under a dynamic semi-convexity condition.","keywords":["sigma-2 Hessian equation","interior Hessian estimates","Jacobi inequality","doubling method","Alexandrov regularity","viscosity solutions","C^{1,1} right-hand side","dimension four"],"falsifier":"Check the application of Lemma 4.7 in Proposition 4.3, Step 1: for $M_{10r}=\\sup_{B_{10r}}u$, the function $M_{10r}-u$ has laplacian $-\\Delta u<0$, so it is not a nonnegative 2-convex solution in the admissible class required by the Harnack inequality. The two-sided estimate $\\omega_r\\le\\theta\\omega_{10r}+Cr^2$ therefore does not follow from the cited lemma; finding a valid replacement for this step, or a counterexample to the estimate, would settle whether Theorem 1.2 is proved.","tokens_in":19805,"feed_emoji":"📐","tokens_out":11691,"duration_ms":96193,"temperature":0.7,"pith_summary":"The paper aims to prove that smooth solutions to the $\\sigma$-2 Hessian equation $\\sigma_2(D^2u)=f(x,u,Du)$, with $f$ positive and $C^{1,1}$ and with $\\Delta u>0$, satisfy an interior Hessian bound: $|D^2u(0)|$ is controlled by a constant depending only on $\\|f\\|_{C^{1,1}}$, $\\|1/f\\|_{L^\\infty}$, and $\\|u\\|_{C^1}$. If this is true, the estimate is exactly what converts weak viscosity solutions into classical ones: Theorem 1.4 states that any continuous viscosity solution in $B_1\\subset\\mathbb{R}^4$ with $\\Delta u>0$ in the viscosity sense belongs to $C^{3,\\alpha}_{\\mathrm{loc}}$ for every $\\alpha<1$. The proof extends the doubling method developed for $\\sigma_2=1$ in dimension 4 to variable right-hand sides, combining an almost Jacobi inequality for $\\log\\Delta u$, a doubling inequality, almost-everywhere twice differentiability for viscosity solutions, and a generalized small perturbation theorem in a compactness argument. This would generalize Qiu's three-dimensional estimate with variable right-hand side and Shankar-Yuan's four-dimensional estimate for $\\sigma_2=1$.","feed_headline":"Bounded Hessian for sigma-2 equations in dimension 4","feed_subtitle":"A data-only constant controls D^2u(0), turning weak viscosity solutions into C^{3,α} functions.","key_machinery":"The machinery that carries the proof is the almost Jacobi inequality for $b=\\log\\Delta u$ with respect to the linearized operator $\\Delta_F=F^{ij}\\partial_{ij}$, where $F^{ij}=\\Delta u\\,\\delta^{ij}-u^{ij}$. In dimension four, Proposition 2.4 gives $$\\Delta_F b\\ge \\varepsilon|\\nabla_F b|^2 - C\\$Gamma^{2}$(1+\\$\\Delta$ u)+\\sum_i f_{p_i}b_i,\\qquad \\varepsilon=\\frac{2}{9}\\left(\\frac12+\\frac{\\lambda_{\\min}}{\\$\\Delta$ u}\\right),$$ with $\\Gamma=\\|u\\|_{C^1}+1$. The coefficient $\\varepsilon$ may degenerate when $\\lambda_{\\min}/\\Delta u$ nears $-1/2$, and the proof compensates by using the largeness of $\\lambda_{\\min}^2$ in those cases. This inequality feeds the maximum-principle test function of Guan-Qiu to produce the doubling inequality $\\sup_{B_2}\\Delta u\\le C\\exp(C\\|u\\|_{C^1(B_3)}^6)\\sup_{B_1}\\Delta u$. The remaining machinery is the weighted Holder estimate obtained from Labutin's potential estimate, which yields the Alexandrov-type theorem that viscosity solutions are twice differentiable almost everywhere, and the generalized Savin small-perturbation theorem that provides $C^{2,\\alpha}$ control at small scales; compactness plus the doubling inequality converts that small-scale control into the global Hessian bound.","core_discovery":"The central claim is Theorem 1.2: every smooth solution of $\\sigma_2(D^2u)=f(x,u,Du)$ in $B_1\\subset\\mathbb{R}^4$ with $\\Delta u>0$ satisfies $|D^2u(0)|\\le C$, where $C$ depends only on $\\|f\\|_{C^{1,1}}$, $\\|1/f\\|_{L^\\infty}$, and $\\|u\\|_{C^1(B_1)}$. In dimensions $n\\ge 5$, the same bound, with $C$ also depending on $n$, is claimed under the dynamic semi-convexity condition $\\lambda_{\\min}(D^2u)\\ge -c(n)\\Delta u$ with $c(n)=(\\sqrt{3n^2+1}-n+1)/(2n)$ (Theorem 1.3). From these a priori estimates the paper derives interior regularity: viscosity solutions in dimension four with $\\Delta u>0$ are $C^{3,\\alpha}_{\\mathrm{loc}}$ for every $\\alpha<1$ (Theorem 1.4). The proof is built from an almost Jacobi inequality for $\\log\\Delta u$, a doubling inequality, Alexandrov-type almost-everywhere twice differentiability, and a generalized Savin small-perturbation theorem, joined in a compactness argument.","pith_inferences":["A natural next step the paper does not spell out is to attempt dimension $n=5$ without the dynamic semi-convexity assumption; the degenerating coefficient $\\varepsilon$ isolates exactly the ratio $\\lambda_{\\min}/\\Delta u$ where the current case analysis is needed.","Because the doubling inequality has exponent $\\|u\\|_{C^1(B_3)}^6$, a sharper test-function argument might lower this power and yield estimates depending on weaker norms, which would matter for boundary regularity.","The Harnack-based oscillation decay in Proposition 4.3 is the step where the paper switches from smooth solutions to viscosity solutions; if that step can be replaced by a direct potential-theoretic argument, the Alexandrov regularity part would apply to larger classes of $k$-convex functions.","The explicit constant $c(n)$ in (1.2) arises from the roots of a quadratic in $y=F_{ii}/\\Delta u$; optimizing that quadratic is a concrete way to sharpen or generalize the admissible class in higher dimensions."],"forward_implications":["In dimension 4, a smooth solution with $\\Delta u>0$ to $\\sigma_2(D^2u)=f(x,u,Du)$ satisfies $|D^2u(0)|\\le C$, with $C$ depending only on $\\|f\\|_{C^{1,1}}$, $\\|1/f\\|_{L^\\infty}$, and $\\|u\\|_{C^1(B_1)}$ (Theorem 1.2).","Consequently, viscosity solutions of the same equation in $B_1\\subset\\mathbb{R}^4$ with positive $C^{1,1}$ $f$ and $\\Delta u>0$ belong to $C^{3,\\alpha}_{\\mathrm{loc}}(B_1)$ for every $\\alpha\\in(0,1)$ (Theorem 1.4).","In dimensions $n\\ge 5$, the same Hessian bound holds for solutions satisfying the dynamic semi-convexity condition $\\lambda_{\\min}(D^2u)\\ge -c(n)\\Delta u$ (Theorem 1.3).","The doubling inequality $\\sup_{B_2}\\Delta u\\le C\\exp(C\\|u\\|_{C^1(B_3)}^6)\\sup_{B_1}\\Delta u$ means Hessian control on one ball propagates to larger balls, which is what turns small-scale perturbation control into a global interior estimate.","The estimate is a priori and universal in the data, so it is preserved under smooth approximation; this is the route from smooth solutions to viscosity solutions via the standard approximating and Evans-Krylov arguments."],"supporting_citations":[{"why":"Supplies the doubling method for $\\sigma_2=1$ in dimension four, including the almost Jacobi inequality and Alexandrov regularity framework that the paper adapts.","marker":"[SY25]"},{"why":"Provides the three-dimensional Hessian estimate for $\\sigma_2=f$ whose Jacobi-inequality and doubling strategy the paper generalizes to dimension four.","marker":"[Qiu24]"},{"why":"Gives potential estimates used to obtain Holder estimates for viscosity solutions, replacing the gradient estimates in the Alexandrov regularity proof.","marker":"[Lab02]"},{"why":"Supplies the small perturbation theorem that underlies the epsilon-regularity argument at small scales.","marker":"[Sav07]"},{"why":"Provides, together with [Fan25], the generalized Savin small perturbation theorem for non-homogeneous fully nonlinear equations.","marker":"[LZ24]"},{"why":"The author's companion preprint proving the generalized small perturbation theorem used in the compactness argument.","marker":"[Fan25]"},{"why":"Provides the Alexandrov-type theorem for $k$-convex functions that is the template for the almost-everywhere twice differentiability proof.","marker":"[CT05]"},{"why":"Supplies the test function and Jacobi inequality scheme used to derive the doubling inequality.","marker":"[GQ19]"},{"why":"Provides the Hessian measure and Holder estimate machinery for $k$-convex functions used in the weighted Holder estimates.","marker":"[TW99]"}],"fun_headline_variants":["Interior Hessian bound for sigma-2 in dimension 4","Hessian control for sigma-2 equations in 4D","Sigma-2: bounded Hessian in 4D without extra convexity","4D sigma-2: Hessian estimates from weak to C^3"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that a two-sided Harnack-based oscillation decay is valid for the admissible solutions under study, i.e., that both the solution shifted up to its running maximum and shifted down to its running minimum lie in the class to which the Harnack inequality applies; if either fails, the Alexandrov regularity step and hence the main theorem are unsupported.","fun_headline_variants_meta":{"raw":{"variants":["Interior Hessian bound for sigma-2 in dimension 4","Hessian control for sigma-2 equations in 4D","Sigma-2: bounded Hessian in 4D without extra convexity","4D sigma-2: Hessian estimates from weak to C^3"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000298,"raw_usage":{"total_tokens":1699,"prompt_tokens":891,"completion_tokens":808,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":507,"completion_tokens_details":{"reasoning_tokens":729}},"tokens_in":507,"tokens_out":808,"duration_ms":7186,"temperature":1.0,"reasoning_tokens":729,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-15T16:37:18.418005+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Check the application of Lemma 4.7 in Proposition 4.3, Step 1: for $M_{10r}=\\sup_{B_{10r}}u$, the function $M_{10r}-u$ has laplacian $-\\Delta u<0$, so it is not a nonnegative 2-convex solution in the admissible class required by the Harnack inequality. The two-sided estimate $\\omega_r\\le\\theta\\omega_{10r}+Cr^2$ therefore does not follow from the cited lemma; finding a valid replacement for this step, or a counterexample to the estimate, would settle whether Theorem 1.2 is proved.","supporting_citations":[],"review_version":1}