{"id":"fdb51d07-a9fc-423e-9237-1fff91dddb0f","arxiv_id":"1908.01547","paper_version":2,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"A new fundamental inequality yields W^{1,2} regularity of |Du|^{(p-γ)/2}Du for p-harmonic functions and W^{2,q} regularity for elliptic and parabolic p-Laplace equations with sharp ranges of p.","lead":"This paper proves a new algebraic inequality linking the Laplacian and the infinity-Laplacian, then uses it to establish higher-order Sobolev regularity for solutions of several p-Laplace type equations. A specialist would read it because it settles an open question for the parabolic normalized p-Laplacian in the plane and gives the first sharp second-order regularity results for the parabolic p-Laplace equation.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Central argument is internally consistent; the only load-bearing soft spot is the cited approximation/convergence theory used to pass from ε-regularized solutions to the limit.","rationale":"I checked the algebraic core: Lemma 2.1 follows from Lemma 2.2 by diagonalization, and the Cauchy-Schwarz estimate in Lemma 2.2 is correct. The sign in Lemma 3.1's closing line is a typo (the correct relation is Δ∞u/s = -Δu/(p-2)), but the displayed inequality (3.10) has the correct sign. The elliptic estimates (3.1), (3.2), Lemmas 3.1–3.2 and Corollary 3.3 are internally consistent, and the passage to u in Theorem 1.1 uses standard weak compactness plus a.e. convergence. In the parabolic case, I traced the n=2, p≥6 combination: Lemma 4.6 follows from half of Lemma 4.7 plus Lemma 4.8; the unwanted (u_t)^2, |D^2uDu|^2/s^2, and boundary logarithm terms cancel, leaving the stated estimate. The reverse-Hölder/Gehring step in Theorem 1.3 is legitimate. Thus I found no internal flaw in the main argument. The only load-bearing point is the external approximation theory: uniform gradient bounds and convergence for the ε-regularized parabolic problems with rough boundary data are cited rather than proved. This is exactly the reader's weakest-assumption identification, and it is standard enough that it does not change the ACCEPT verdict.","tokens_in":30804,"tokens_out":46782,"duration_ms":391782,"concrete_test":"Verify directly, or from the cited sources, that solutions of (1.12) and (1.13) satisfy a uniform-in-ε interior C^{1,α} estimate on compact subcylinders of U_T when the parabolic boundary data is only continuous, and that u_ε → u in C^0 with Du_ε bounded uniformly in ε. In particular, check the time slice t=0 used in Lemmas 4.7 and 4.8: if Du_ε is not bounded in a neighborhood of that slice, the ε-boundary terms need re-examination. If the uniform C^{1,α} estimate holds, the transfer (4.2)→(4.1) is justified; if not, Theorem 1.3 lacks a proof.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem 1.3 and Theorem 1.5 both transfer second-order estimates from smooth approximating solutions u_ε to the viscosity/weak solution u by asserting Du_ε ∈ L∞ uniformly and u_ε → u in C^0, then passing to weak limits. For the parabolic normalized p-Laplace approximation (1.12), the paper cites [22] for these properties, but [22] concerns the unregularized homogeneous normalized p-Laplace equation, not the ε-regularized equation with merely continuous parabolic boundary data inherited from a viscosity solution. For the parabolic p-Laplace approximation (1.13), the citations [11,36] likewise concern the degenerate limiting equation rather than the ε-regularized problem with rough boundary data. The interior estimates in Sections 4–5 also integrate over one-sided parabolic cylinders, producing boundary terms at the artificial time slice t=0 (e.g., in Lemmas 4.7 and 4.8); these are controlled using smoothness and uniform L∞ bounds of Du_ε up to that slice. If uniform C^{1,α} estimates for (1.12) or (1.13) fail near such slices, or if u_ε → u holds only in C^0 but not with the gradient convergence needed to identify weak limits of nonlinear expressions like [|Du_ε|^2+ε]^{(p-γ)/4}Du_ε, then the limiting step in Theorem 1.3 would collapse. This is a genuine background assumption, not proved in the paper, but it is standard and likely fillable from the cited literature.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves a pointwise algebraic inequality (Lemma 2.1, Eq. (1.2)) controlling the structure of ΔvΔ∞v in terms of |D^2v| and Dv, and then uses this inequality as the central tool to establish second-order Sobolev regularity for p-harmonic functions (Theorem 1.1, Corollary 1.2), for viscosity solutions of the parabolic normalized p-Laplace equation (Theorem 1.3), and for weak/viscosity solutions of the parabolic p-Laplace equation (Theorem 1.5). The results include improved ranges of γ and p, a complete answer for n=2 to an open question of Høeg and Lindqvist, and a sharpness example for the range p<3 in Theorem 1.5. The proofs pass through ε-regularized smooth solutions and then take limits using known convergence and uniform gradient bounds.","tokens_in":31113,"tokens_out":13868,"duration_ms":135161,"significance":"If the claims hold, this is a significant contribution to the regularity theory of degenerate elliptic and parabolic equations. The fundamental inequality (1.2) is new, is proved from scratch via an elementary spectral decomposition, and is likely to be a useful tool beyond the applications considered here. The paper also gives explicit quantitative estimates and a clean sharpness calculation for the parabolic p-Laplace equation. The central analytic arguments are self-contained apart from standard approximation facts, and the main results improve previously known ranges in the elliptic case and are the first higher-integrability results in the parabolic normalized case for n=2.","major_comments":[{"comment":"The passage from the regularized solutions u_ε to the limit solution u relies on the assertions that u_ε ∈ C^∞(U_T) ∩ C^0(overline{U_T}), Du_ε ∈ L^∞(U_T) uniformly in ε, and u_ε → u in C^0(U_T). In Section 4 this is attributed to [22], and in Section 5 to [11,36]. However, the cited papers concern the unregularized degenerate equations, not the ε-regularized equations (1.12) and (1.13) with boundary data inherited from a viscosity solution. These approximation properties are load-bearing: without them the weak-limit identification of D^2u_ε and u_{ε,t} in Theorems 1.3 and 1.5 collapses. The authors should either provide precise statements or proofs from the standard theory of uniformly parabolic quasilinear equations, or give correct references that cover the regularized problems uniformly in ε. This gap is likely fillable but must be addressed in the manuscript.","section":"Sections 4 and 5, Eqs. (1.12) and (1.13)"}],"minor_comments":[{"comment":"The range p ∈ (1,2) ∪ (2, 3+2/(n−2)) is stated without comment for n=2, where the expression 3+2/(n−2) is undefined; it should be interpreted as the whole interval (1,∞), and this convention should be stated explicitly.","section":"Theorem 1.3 and Corollary 1.2"},{"comment":"The statement 'u_t, D^2u ∈ L^q_loc(Ω)' should read 'u_t, D^2u ∈ L^q_loc(Ω_T)', since the estimates are on space-time cylinders Q_r ⊂ Ω_T.","section":"Theorem 1.3, Eq. (1.7)"},{"comment":"The notation Q_r and Q(0,r) is used inconsistently: the definition Q_r(z,s) = (s−r^2,s) × B(z,r) gives a cylinder ending at time s, but the proof of Lemma 4.1 appears to translate time so that the bottom of the cylinder is t=0. Please clarify the time normalization and the appearance of integrals over B_{2r} at t=0 in Lemmas 4.7 and 4.8.","section":"Section 4, Lemma 4.1 and proof of Theorem 1.3"},{"comment":"The boundary terms involving ln[|Du_ε(x,0)|^2+ε] are controlled only after invoking uniform boundedness of Du_ε and the prefactor ε; this should be stated explicitly, since ln[|Du_ε|^2+ε] is not bounded below uniformly as ε→0.","section":"Lemma 4.6 and Lemma 4.7"},{"comment":"In the abstract, the formula for γn,p appears as min{p + (n−1)/n, 3+(p−1)/(n−1)}, while Theorem 1.1 states min{p + n/(n−1), 3+(p−1)/(n−1)}. The theorem version is the correct one; the abstract should be corrected.","section":"Abstract and Section 1.1"}],"recommendation":"major_revision","confidential_remarks":"This is a strong paper with a new structural inequality and substantial results. The main concern is the approximation step in Sections 4 and 5, where the cited regularity and convergence results for the unregularized equations do not directly cover the ε-regularized problems. If the authors supply precise references or short proofs for the needed uniform properties of u_ε, the paper should be acceptable. The remaining issues are presentation-level."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Read this one. It is a genuinely strong regularity paper. The central object is the pointwise inequality (1.2), a dimension-dependent bound on |D^2v Dv|^2 - Δv Δ_∞v - 1/2[|D^2v|^2 - (Δv)^2]|Dv|^2. The proof via eigenvalues and Cauchy-Schwarz is short, self-contained, and correct. Everything else hangs off it, and the paper is honest about what is new and what is reproof. Theorem 1.3 answers the Høeg-Lindqvist question for n=2 with higher integrability q>2; Theorem 1.5 gives W^{2,2} for the parabolic p-Laplacian with sharp range p<3, with a clean sharpness example. Those are real advances. The proofs are long and technical, but the structure is clear: Lemma 2.1, pointwise estimates for ε-regularized solutions, uniform L^2 bounds on D^2u_ε and u_ε,t, then Gehring/compactness to pass to the limit.\n\nThe soft spot, as flagged, is the approximation step. The paper cites [22] and [11,36] for convergence and uniform gradient bounds of the regularized problems, but those references treat the limiting degenerate equations, not the ε-regularized ones with rough boundary data. This is a genuine gap, though a standard one; I expect it can be filled by combining the cited interior estimates with standard boundary arguments. Similarly, in Section 5 the uniform L∞ bound on Du_ε is asserted without a precise citation. The boundary terms at t=0 in Lemmas 4.7 and 4.8 are controlled only if Du_ε is bounded uniformly up to the initial slice; the paper does not discuss this. These are not fatal, but a referee should ask for precise statements.\n\nCorollary 1.2 partly reproves known Cordes-condition results, and the paper says so. That is fine.\n\nOverall: the main theorem is new, the proof is mostly self-contained, and the cited background is standard. I would send it to a serious referee and would cite it if I worked in this area.","headline":"A solid, self-contained regularity paper built on a new algebraic inequality; the main theorems are new and the proofs are mostly careful, with one standard-but-unproven approximation step that a referee should ask about.","tokens_in":31705,"tokens_out":2374,"would_cite":true,"duration_ms":25251,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35B65","35J92","35K92","35D40"],"pacs":[],"model":"deepseek-v4-flash","headline":"A dimension-weighted Hessian inequality yields second-order Sobolev regularity for p-Laplacian equations and fully settles the planar parabolic normalized case.","keywords":["p-Laplacian","normalized p-Laplace equation","infinity-Laplacian","second-order regularity","Sobolev regularity","viscosity solutions","parabolic equations","Hessian estimates"],"falsifier":"Take an explicit planar solution of the parabolic normalized $p$-Laplace equation, for instance a radial or self-similar profile, and compute whether $\\int_{Q_r}(|D^2u|^q+|u_t|^q)\\,dx\\,dt$ is finite for some $q>2$; a single example where the integral diverges for every $q>2$ would refute Theorem 1.3. Similarly, a $p$-harmonic function for which $|Du|^{(p-\\gamma)/2}Du$ fails to lie in $W^{1,2}_{\\rm loc}$ for some $\\gamma<\\gamma_{n,p}$ would refute Theorem 1.1.","tokens_in":30603,"feed_emoji":"📐","tokens_out":19182,"duration_ms":166363,"temperature":0.7,"pith_summary":"This paper establishes a pointwise structural inequality for smooth functions in $n$ dimensions: the algebraic interaction between the Hessian, the Laplacian $\\Delta v$, and the infinity-Laplacian $\\Delta_\\infty v=D^2vDv\\cdot Dv$ is controlled by a dimension-dependent error term, and the control is an exact identity in the plane. From this single inequality it derives second-order Sobolev regularity for solutions of the $p$-Laplace equation, the parabolic normalized $p$-Laplace equation, and the degenerate parabolic $p$-Laplace equation. In two dimensions it completely answers an open question by showing that viscosity solutions of the parabolic normalized $p$-Laplace equation have spatial Hessian and time derivative in $L^q_{\\rm loc}$ for some $q>2$ for every $p\\ne2$. Because the argument avoids convexity or monotonicity of the $p$-Laplacian, the same inequality governs all three equation classes and yields explicit estimates that remain uniform as $p$ approaches $2$.","feed_headline":"One inequality unlocks second-order regularity for p-Laplacian","feed_subtitle":"In two dimensions it fully answers an open problem on the parabolic normalized p-Laplace equation.","key_machinery":"The load-bearing object is the dimension-weighted pointwise inequality (2.1), which bounds the algebraic expression $|D^2vDv|^2-\\Delta v\\,\\Delta_\\infty v-\\tfrac12(|D^2v|^2-(\\Delta v)^2)|Dv|^2$ by the nonnegative quantity $|D^2v|^2|Dv|^2-|D^2vDv|^2$ times $(n-2)/2$. Its proof diagonalizes the Hessian at a point, reducing the inequality to an $n$-term eigenvalue statement whose direction is the unit vector $Dv/|Dv|$, and then invokes an elementary vector inequality. The argument applies this inequality to smooth regularized solutions of the elliptic $p$-Laplace equation and of the two parabolic equations under study; it yields uniform-in-$\\varepsilon$ estimates on $|D^2u_\\varepsilon|^2$, $|u_{\\varepsilon,t}|^2$, and weighted products of Hessian and gradient. Those uniform estimates are what pass to the limiting solution through compactness and a higher-integrability step.","core_discovery":"The paper's central claim is that the planar identity\n$$\n|$D^{2}$vDv|^2-\\$\\Delta$ v\\,\\Delta_\\infty v=\\tfrac12(|$D^{2}$v|^2-(\\$\\Delta$ v)^2)|Dv|^2\n$$\nadmits a higher-dimensional replacement: for every smooth $v$,\n$$\n\\Bigl||$D^{2}$vDv|^2-\\$\\Delta$ v\\,\\Delta_\\infty v-\\tfrac12(|$D^{2}$v|^2-(\\$\\Delta$ v)^2)|Dv|^2\\Bigr|\n\\le \\frac{n-2}{2}\\bigl(|$D^{2}$v|^2|Dv|^2-|$D^{2}$vDv|^2\\bigr),\n$$\nwhere $\\Delta_\\infty v=D^2vDv\\cdot Dv$. Using this inequality on regularized solutions, the paper proves that $p$-harmonic functions satisfy $|Du|^{(p-\\gamma)/2}Du\\in W^{1,2}_{\\rm loc}$ for $\\gamma<\\min\\{p+\\frac{n}{n-1},\\,3+\\frac{p-1}{n-1}\\}$; that viscosity solutions of the parabolic normalized $p$-Laplace equation have $D^2u,u_t\\in L^q_{\\rm loc}$ for some $q>2$ in the exponent range $(1,2)\\cup(2,3+\\frac{2}{n-2})$, which for $n=2$ means every $p\\ne2$; and that for the degenerate parabolic $p$-Laplace equation, $D^2u\\in L^2_{\\rm loc}$ and $u_t\\in L^2_{\\rm loc}$ for $1<p<3$, with the upper endpoint sharp.","pith_inferences":["Because the fundamental inequality is purely algebraic and needs no convexity or monotonicity of the $p$-Laplacian, it may transfer to other equations whose operators interpolate between Laplacian and infinity-Laplacian, such as game-theoretic or image-processing parabolic models; the paper does not explore these applications.","One way to probe sharpness is to compute the weighted gradient quantity for explicit power-type $p$-harmonic functions and check whether $W^{1,2}$ integrability fails as $\\gamma$ approaches the paper's upper bound; the paper does not carry out this endpoint test.","The authors conjecture that for $n\\ge3$ and $p\\ge3+\\frac{2}{n-2}$, the expression $|D^2u|^2-(\\Delta u)^2$ can change sign for some $p$-harmonic function; if verified, the exponent ranges in the second-order results would be sharp, and the method's limitation would be intrinsic rather than technical."],"forward_implications":["The exponent range for the weighted gradient quantity in Theorem 1.1 improves the earlier bound $\\gamma\\le2$ for every $p\\ne2$ and every dimension $n\\ge2$.","For the parabolic normalized $p$-Laplace equation in the plane, the open question on second-order regularity is settled for all $p\\ne2$, with the stronger conclusion that the integrability exponent is $q>2$ rather than merely $2$.","For $n\\ge3$, viscosity solutions of the parabolic normalized equation have $D^2u$ and $u_t$ in $L^q_{\\rm loc}$ for some $q>2$ throughout $p\\in(1,2)\\cup(2,3+\\frac{2}{n-2})$, a wider range than the coefficient-degeneracy approach could reach.","For the degenerate parabolic $p$-Laplace equation, the range $p\\in(1,3)$ is sharp for spatial $W^{2,2}$-regularity: an explicit solution has $|D^2w|$ comparable to $|x_1|^{(2-p)/(p-1)}$, which is in $L^2_{\\rm loc}$ exactly when $p<3$.","All estimates come with constants that do not blow up as $p\\to2$, so the results connect continuously to the classical theory at $p=2$."],"supporting_citations":[{"why":"Supplies the planar identity (1.1) that inequality (2.1) generalizes to higher dimensions.","marker":"[24]"},{"why":"Raises the open question on second-order regularity for the parabolic normalized $p$-Laplace equation that is completely answered when $n=2$.","marker":"[17]"},{"why":"Established the earlier $\\gamma\\le2$ range for the weighted gradient quantity that Theorem 1.1 improves.","marker":"[4]"},{"why":"Give the uniform $L^\\infty$ gradient bounds and $C^{0,\\alpha}$ convergence of the elliptic regularization used to pass to the limit.","marker":"[33, 26, 10]"},{"why":"Provides the uniform gradient bound and $C^0$ convergence for the regularized normalized parabolic equation.","marker":"[22]"},{"why":"Provide the convergence and regularity facts for the regularized degenerate parabolic $p$-Laplace equation.","marker":"[11, 36]"},{"why":"Documents the coefficient-degeneracy condition that earlier gave restricted-range second-order estimates.","marker":"[29]"},{"why":"Supplies the higher-integrability lemma that upgrades the $L^2$ estimate to $L^q$ for $q>2$.","marker":"[15, 16]"},{"why":"Supplies the parabolic embedding estimate used to turn the $L^2$ bound into higher integrability.","marker":"[25]"}],"fun_headline_variants":["New inequality proves p-Laplacian second-order regularity","Second-order regularity for p-Laplacian via inequality","Open question on p-Laplacian resolved in plane","Sharp p-range for parabolic p-Laplacian regularity","Fundamental inequality yields p-harmonic regularity"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof works first on smooth approximate solutions; if those approximations did not have uniformly bounded gradients and did not converge to the true solution, the second-order estimates would not carry over to the limit.","fun_headline_variants_meta":{"raw":{"variants":["New inequality proves p-Laplacian second-order regularity","Second-order regularity for p-Laplacian via inequality","Open question on p-Laplacian resolved in plane","Sharp p-range for parabolic p-Laplacian regularity","Fundamental inequality yields p-harmonic regularity"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000369,"raw_usage":{"total_tokens":2202,"prompt_tokens":1391,"completion_tokens":811,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":1007,"completion_tokens_details":{"reasoning_tokens":737}},"tokens_in":1007,"tokens_out":811,"duration_ms":7987,"temperature":1.0,"reasoning_tokens":737,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:10:11.596447+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take an explicit planar solution of the parabolic normalized $p$-Laplace equation, for instance a radial or self-similar profile, and compute whether $\\int_{Q_r}(|D^2u|^q+|u_t|^q)\\,dx\\,dt$ is finite for some $q>2$; a single example where the integral diverges for every $q>2$ would refute Theorem 1.3. Similarly, a $p$-harmonic function for which $|Du|^{(p-\\gamma)/2}Du$ fails to lie in $W^{1,2}_{\\rm loc}$ for some $\\gamma<\\gamma_{n,p}$ would refute Theorem 1.1.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the planar identity (1.1) that inequality (2.1) generalizes to higher dimensions."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Raises the open question on second-order regularity for the parabolic normalized $p$-Laplace equation that is completely answered when $n=2$."},{"cited_title":"Bojarski and T","cited_arxiv_id":null,"evidence_quote":"Established the earlier $\\gamma\\le2$ range for the weighted gradient quantity that Theorem 1.1 improves."},{"cited_title":"Jin and L","cited_arxiv_id":null,"evidence_quote":"Provides the uniform gradient bound and $C^0$ convergence for the regularized normalized parabolic equation."},{"cited_title":"Maugeri, D","cited_arxiv_id":null,"evidence_quote":"Documents the coefficient-degeneracy condition that earlier gave restricted-range second-order estimates."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the parabolic embedding estimate used to turn the $L^2$ bound into higher integrability."}],"review_version":1}