{"id":"3bcc6e18-3aae-41ba-bccb-350a1f6d757e","arxiv_id":"2501.12101","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Existence via Perron's method and C^{2,α} regularity of flat free boundaries for fully nonlinear one-phase problems with right-hand side and normal-dependent boundary condition.","lead":"Mathematicians proved existence and high-order smoothness for free boundary problems described by very general nonlinear elliptic equations with a forcing term and an interface-speed rule that depends on direction. This generalizes the classical Alt-Caffarelli theory and unifies a wide class of one-phase free boundary models.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Proposition 5.1 uses slope g(0,ν_{x0}) instead of g(x0,ν_{x0}), and the proof's rescaling at x0 needs flatness relative to g(x0,e_d), which is not derived; Theorem 1.5 is therefore not established as written.","rationale":"The reader's weakest-assumption highlighted the external estimate [LZ18, Theorem 1.3] in Proposition 2.7. That concern is reasonable but likely addressable: the limiting operator ~F∞ inherits (H2) from the concavity/convexity assumption, and the oblique vector τ has τ·e_d=1 and is bounded uniformly, so the hypotheses of [LZ18] appear to hold. The more serious issue is internal: Proposition 5.1, which is the direct input to Theorem 1.5, asserts a Taylor expansion with the wrong slope. The free boundary condition at x0 forces the first-order coefficient to be g(x0,ν_{x0}); using g(0,ν_{x0}) is inconsistent with the PDE unless g is constant in x. The proof's rescaling at x0 introduces boundary datum g(x0,e_d), yet the flatness is written with g(0,e_d). The difference is a linear function that cannot be absorbed into the quadratic polynomial p0, and the paper does not show that g(x0,e_d)−g(0,e_d) is small on the free boundary under the flatness assumption. This is not a typo in one line: it affects the initial step of the quadratic improvement iteration at every boundary point, so Theorem 1.5 is not established as written. The main ideas may be salvageable, e.g., by first proving a C^{1,α} estimate that locates the free boundary in a region where g(x0,ν) is close to g(0,ν), but the current manuscript lacks such an argument. For this reason the verdict should move from CONDITIONAL to UNVERDICTED rather than ACCEPT or REJECT: the central claim is plausible but not proven by the given proof.","tokens_in":38355,"tokens_out":31212,"duration_ms":301560,"concrete_test":"Re-run the proof of Proposition 5.1 with the corrected slope g(x0,e_d): from the global flatness (g(0,e_d)x_d ± ε0)_+, derive an initial flatness of the form (g(x0,e_d)x_d + p0(x) − r0^{1+α})_+ ≤ u_{x0,δ} ≤ (g(x0,e_d)x_d + p0(x) + r0^{1+α})_+. If the unavoidable error term [g(0,e_d)−g(x0,e_d)]x_d cannot be bounded by r0^{1+α}/2 uniformly over all x0 ∈ ∂Ω_u ∩ B_{1/2}, then the iteration cannot start. A second check: for F=Δ, f=0, g(x,ν)=1+κx_1 and u near (x_2)_+, determine whether any free boundary point with |g(x0)−1| > Cε0 can satisfy the flatness hypothesis with arbitrarily small ε0; if so, the expansion with slope g(0,ν) contradicts the boundary condition.","verdict_should_be":"UNVERDICTED","load_bearing_attack":"Theorem 1.5 depends on Proposition 5.1, whose statement asserts the Taylor expansion ‖u(x0+x) − g(0,ν_{x0})(x·ν_{x0}) − p_{x0}(x)‖ ≤ C r^{2+α} at each x0 ∈ ∂Ω_u ∩ B_{1/2}. But the free boundary condition at x0 gives |∇u(x0)| = g(x0, ν(x0)), so the correct first-order coefficient is g(x0,ν_{x0}), not g(0,ν_{x0}). The proof reveals the same mismatch: after rescaling u_{x0,δ}(x) = u(x0+δx)/δ, the boundary datum is g_{x0,δ}(0,e_d)=g(x0,e_d), so Proposition 4.1 must be applied to a function flat relative to g(x0,e_d)x_d + p0(x). The proof instead starts from the global flatness (g(0,e_d)x_d ± ε0)_+ and uses p0 ∈ P(e_d, F_{x0,δ}, f_{x0,δ}, g_{x0,δ}) while keeping g(0,e_d)x_d in the flatness. The difference [g(0,e_d) − g(x0,e_d)]x_d is a linear term; the class P in (1.7) contains only quadratic polynomials, so this term cannot be absorbed into p0 unless |g(x0,e_d) − g(0,e_d)| is shown to be O(r0^{1+α}) uniformly for all x0 in the free boundary. The paper provides no such argument. For a point with |x0| large, this difference can be of order 1, invalidating the application of Proposition 4.1. Thus the quadratic improvement iteration that produces the C^{2,α} expansion is not justified for arbitrary free boundary points, and Theorem 1.5 is not proven as written.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the one-phase free boundary problem for fully nonlinear elliptic operators with a right-hand side and a free boundary condition |∇u| = g(x,ν) depending on the normal. It contains three main results: (i) existence of viscosity solutions by Perron's method (Theorem 1.4), together with local Lipschitz regularity and non-degeneracy; (ii) C^{2,α} regularity of flat free boundaries via a quadratic improvement of flatness (Theorem 1.5); and (iii) higher regularity of the free boundary by a hodograph transform (Corollary 1.6). The existence part follows a Perron-type construction with an admissible family of supersolutions and a strict minorant, and the regularity part uses a compactness/linearization argument leading to an oblique-boundary problem for the linearized operator. The hodograph step is standard once C^{2,α} regularity is available.","tokens_in":38670,"tokens_out":25778,"duration_ms":253685,"significance":"If correct, the paper would extend the De Silva-type improvement-of-flatness theory to fully nonlinear one-phase problems with nonzero right-hand side and normal-dependent boundary data, and would in particular give the first C^{2,α} flat-free-boundary theorem in that generality. The Perron existence theorem is a solid contribution in itself, and the paper is largely self-contained: no free parameters are fitted, and the central arguments rely on external regularity theorems rather than circular reasoning. However, the statement and proof of Proposition 5.1 contain a mismatch between the coefficient g(0,ν) and the local value g(x0,ν) that currently breaks the proof of Theorem 1.5. The gap is substantial but appears repairable by reworking the flatness iteration around the local coefficient g(x0,ν).","major_comments":[{"comment":"The first-order coefficient in the claimed Taylor expansion is the wrong one. At a point x0 ∈ ∂Ω_u ∩ B_{1/2} the free boundary condition gives |∇u(x0)| = g(x0, ν_{x0}), so any expansion u(x0+x) = a (x·ν_{x0}) + O(|x|^2) must have a = g(x0, ν_{x0}). Proposition 5.1 instead states, and the proof concludes, a = g(0, ν_{x0}); for g depending on x these differ by an O(1) amount, so the proposition is false as stated unless g is independent of x. The source of the error is visible in the iteration: for the rescaled datum g_n(x,ν) = g_{x0,ρ_n}(x,ν) = g(x0 + ρ_n x, ν), the coefficient in Proposition 4.1 is g_n(0,ν) = g(x0,ν), not g(0,ν). When the proof of Proposition 5.1 asserts the initial flatness (g(0,e_d)x_d − r0^{1+β})_+ ≤ u_{x0,δ} ≤ (g(0,e_d)x_d + r0^{1+β})_+ for u_{x0,δ}(x) = u(x0+δx)/δ, the linear term [g(x0,e_d) − g(0,e_d)]x_d is not absorbable into p0 ∈ P(e_d, F_{x0,δ}, ...), since the class P in (1.7) contains only homogeneous quadratic polynomials and the difference is not small uniformly for |x0| of order 1. Thus the first application of Proposition 4.1 is unjustified, and the iteration proving Theorem 1.5 collapses. The proof needs either a localization argument producing flatness at x0 relative to g(x0,e_d), or a reformulation of the flatness hypothesis, and the statement of Proposition 5.1 should use g(x0, ν_{x0}).","section":"Section 5, Proposition 5.1"}],"minor_comments":[{"comment":"Please add a short verification that the limiting operator ~F∞ in (4.9) and the oblique vector τ in (4.10) satisfy the hypotheses of [LZ18, Theorem 1.3]. The verification is straightforward — concavity/convexity passes to the uniform limit of the difference quotients in (4.8), and τ·e_d = 1 is preserved — but it should be written out because Proposition 2.7 is a load-bearing external input.","section":"Section 2.6 and Section 4.4"},{"comment":"The displayed formula for ν_n − e_d is garbled: it should first define the unit vector ν_n as the normalization of g_n(0,e_d)e_d + r^{1+α}ν and then expand the difference. The conclusion |ν_n − e_d| ≤ C r^{1+α} is correct, but the displayed computation should be fixed.","section":"Section 4.5, Eq. (4.14)"},{"comment":"The notation [g_{x0,δ}]_{C^{1,β}(B1)} should specify that the norm is taken only in the spatial variable x. Derivatives of g_{x0,δ} with respect to ν are not rescaled by δ and therefore cannot satisfy the stated smallness bound; the later use in the proof of Proposition 4.1 only needs boundedness of the ν-derivatives together with smallness of the x-derivatives.","section":"Section 5, Lemma 5.2"},{"comment":"The conclusion '∂Ω_u ∩ B_1 is a (d−1)-dimensional manifold of class C^{2,α} in B_{1/2}' should read '∂Ω_u ∩ B_{1/2}'.","section":"Section 1.2.2, Theorem 1.5"},{"comment":"The boundary conditions for the auxiliary function v are stated inconsistently: 'v = 0 in B1 \\ D_σ' and 'v = h on ∂B1' overlap on ∂B1 ∩ D_σ. The intended boundary split should be clarified.","section":"Section 3.4, proof of Proposition 3.10"}],"recommendation":"major_revision","confidential_remarks":"The Perron existence part of the paper appears sound and could be published independently. The main problem is Theorem 1.5: the proof of Proposition 5.1 uses g(0,ν) where the local free boundary condition forces g(x0,ν), and this is not a cosmetic typo. I would ask the authors to rework the flatness iteration so that the slope is the local value g(x0,ν), or to state a modified flatness hypothesis for which the proof is valid. The concern that the external estimate [LZ18] might not apply does not seem to land after inspecting the limiting operator; an explicit verification would nevertheless strengthen the paper. I see no evidence of circularity or parameter fitting."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This is a strong paper that genuinely advances the one-phase free boundary theory for fully nonlinear operators with right-hand side and normal-dependent boundary condition. The existence theorem via a carefully chosen admissible family is a real novelty, and the C^{2,α} flatness strategy is the right approach. The technical work in Sections 3–4 is mostly careful, and the reliance on [LZ18] for the linearized problem is appropriate, though the hypotheses should be checked more explicitly.\n\nThe main problem is in Proposition 5.1 and the proof of Theorem 1.5. Proposition 5.1 states the second-order expansion with coefficient g(0,ν_{x0}). At a free boundary point x0, the free boundary condition gives |∇u(x0)| = g(x0,ν_{x0}), so the correct linear coefficient is g(x0,ν_{x0}). The proof rescales around x0 and applies Proposition 4.1 to the rescaled problem, whose boundary data is g_{x0,δ} with g_{x0,δ}(0,e_d) = g(x0,e_d). Proposition 4.1 therefore requires flatness relative to g(x0,e_d). But the proof only derives flatness relative to g(0,e_d) from the global assumption. The difference is a linear term that cannot be absorbed into the quadratic polynomial p0, and for x0 away from the origin it is not small. The paper gives no argument that |g(x0,e_d) − g(0,e_d)| is controlled by the flatness scale. Without that, the iteration cannot start at arbitrary free boundary points.\n\nThis is a real gap, but I think it is fixable. A compactness argument should show that the global flatness with small ε0 forces g(x0,e_d) to be close to g(0,e_d) uniformly on the free boundary; the heuristic is that a flat solution is close to a half-plane with slope g(0,e_d), so the normal derivative is close to that value, and the free boundary condition then forces g(x0,e_d) close to g(0,e_d). The current manuscript does not contain that argument.\n\nThe Perron existence proof and the higher-regularity bootstrapping look solid. The citation pattern is reasonable; the single self-citation is standard. No fitted parameters. This paper deserves a serious referee, but the referee should prominently flag the Proposition 5.1 issue and request a fix before the regularity theorem can be accepted as stated.","headline":"Substantial paper with a real gap in the pointwise iteration: the slope in the Taylor expansion is evaluated at 0 instead of the free boundary point; fixable but Theorem 1.5 is not proven as written.","tokens_in":39305,"tokens_out":17069,"would_cite":false,"duration_ms":153054,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35R35"],"pacs":[],"model":"deepseek-v4-flash","headline":"Viscosity solutions to fully nonlinear one-phase free boundary problems with a right-hand side and normal-dependent data exist by Perron's method, and flat free boundaries are necessarily $C^{2,\\alpha}$, with higher smoothness following…","keywords":["regularity","free boundary","fully nonlinear","improvement of flatness","one-phase problem","Perron's method","viscosity solutions","oblique boundary condition"],"falsifier":"The decisive test is to inspect the limit operator $\\widetilde{F}_\\infty$ that appears in the compactness step of Proposition 4.7 and ask whether it really satisfies the concavity/convexity and nondegenerate-oblique hypotheses of [LZ18, Theorem 1.3]; exhibiting an admissible sequence $F_n,g_n,u_n$ for which $\\widetilde{F}_\\infty$ violates one of those hypotheses, or producing a flat viscosity solution whose free boundary fails to be $C^{2,\\alpha}$ under (H1)–(H3), would refute Theorem 1.5.","tokens_in":38052,"feed_emoji":"📐","tokens_out":15206,"duration_ms":120946,"temperature":0.7,"pith_summary":"The paper aims to show that the classical one-phase free boundary regularity story survives the move to general fully nonlinear uniformly elliptic operators with a right-hand side and a free boundary condition that depends on the inward normal. It proves existence of viscosity solutions by Perron's method for any continuous nonnegative boundary datum, and it proves that flat free boundaries are automatically $C^{2,\\alpha}$—no variational structure is needed, only a flatness assumption. If the data are smoother, the boundary inherits $C^{k+2,\\beta}$, $C^\\infty$, or analytic regularity. The interest is that flatness alone forces second-order differentiability of the boundary in a setting where the only tools are viscosity methods and an improvement-of-flatness iteration.","feed_headline":"Flat free boundaries are C²,α in fully nonlinear one-phase problems","feed_subtitle":"A quadratic improvement of flatness plus a hodograph transform upgrades boundaries from flat to second-order smooth.","key_machinery":"The central mechanism is the quadratic improvement of flatness, Proposition 4.1, proved by contradiction and compactness. From a solution trapped between $(g(0,e_d)x_d+p(x)-r^{1+\\alpha})_+$ and $(g(0,e_d)x_d+p(x)+r^{1+\\alpha})_+$, the rescaled linearized sequence $(u_n-g_n(0,e_d)x_d-p_n)/r_n^{1+\\alpha}$ satisfies a partial Harnack inequality and converges to a viscosity solution of the linearized problem $\\widetilde{F}(D^2\\widetilde{u})=0$ in $B^+_{1/2}$, $\\nabla\\widetilde{u}\\cdot\\tau=0$ on $B'_{1/2}$. The expansion of this limit, stated as Proposition 2.7 and resting on the external estimate [LZ18, Theorem 1.3], gives a $C^{2,\\alpha_0}$ rate that transfers back to the nonlinear scale; the new normal vector and the new quadratic polynomial are read off from $\\nabla\\widetilde{u}(0)$ and $D^2\\widetilde{u}(0)$. Iterating at every free boundary point produces the $C^{2,\\alpha}$ boundary in Theorem 1.5. For Corollary 1.6, the hodograph transform of [KN77] rewrites the problem as a nonlinear elliptic equation with oblique boundary condition, to which classical regularity results [ADN59, Mor08] apply.","core_discovery":"On its own terms, the paper establishes Theorem 1.5: there exist universal constants $\\varepsilon_0>0$ and $\\alpha_0\\in(0,1)$ such that if $F\\in E(\\lambda,\\Lambda)$, $f\\in C^{0,\\beta}(B_1)$, $g\\in C^{1,\\beta}(B_1,S^{d-1})$ satisfy (H1)–(H3), and a nonnegative continuous viscosity solution $u$ of (1.1) is trapped between $(g(0,\\nu)(x\\cdot\\nu)-\\varepsilon_0)_+$ and $(g(0,\\nu)(x\\cdot\\nu)+\\varepsilon_0)_+$, then $u\\in C^{2,\\alpha}(\\Omega_u\\cap B_{1/2})$ and the free boundary $\\partial\\Omega_u\\cap B_{1/2}$ is a $C^{2,\\alpha}$ manifold. The route is a quadratic improvement of flatness: at each scale the solution lies within error $r^{1+\\alpha}$ of a quadratic polynomial solution of a linearized fully nonlinear oblique problem, and iteration over boundary points yields a second-order Taylor expansion with uniform rate. Corollary 1.6 then applies the hodograph transform to convert the $C^{2,\\alpha}$ free boundary into a $C^{k+2,\\beta}$, $C^\\infty$, or analytic boundary according to the smoothness of $F$, $f$, and $g$. Theorem 1.4 supplies existence of a viscosity solution via Perron's method, with Lipschitz regularity and non-degeneracy of the constructed solution.","pith_inferences":["A natural next target is the analogous two-phase problem with fully nonlinear operators and right-hand side; the same linearized oblique estimate would be the bottleneck, and the paper's partial Harnack argument looks transferable, but that is an extension, not a claim of the paper.","If the hypotheses of the external linearized estimate could be relaxed, the convexity/concavity assumption (H2) in Theorem 1.5 might be weakened; the present proof does not test this.","The dichotomy in the admissible family suggests a template for other non-homogeneous free boundary problems where the right-hand side prevents naive barrier constructions; one could test whether Perron's solutions vary continuously with the strict minorant and the boundary datum."],"forward_implications":["Flat free boundaries are automatically $C^{2,\\alpha}$ even when the operator is fully nonlinear, the right-hand side is nonzero, and the free boundary condition depends on the normal.","The quadratic improvement of flatness yields a uniform second-order Taylor expansion of the solution at every free boundary point, which is exactly the input needed for the hodograph transform.","If $F$, $f$, and $g$ are $C^{k,\\beta}$, $C^{k,\\beta}$, and $C^{k+1,\\beta}$ respectively, the free boundary is $C^{k+2,\\beta}$; for $C^\\infty$ or analytic data the boundary is $C^\\infty$ or analytic.","Perron's method gives a viscosity solution for any continuous nonnegative boundary datum, and the constructed solution is Lipschitz and non-degenerate in compact sets.","The Perron construction deliberately excludes degenerate or collapsed-boundary solutions such as $u(x)=x_d^2/2$ or $u(x)=c|x_d|$, clarifying that the existence result covers a special class of viscosity solutions."],"supporting_citations":[{"why":"Supplies the $C^{2,\\alpha_0}$ estimate for viscosity solutions of the fully nonlinear oblique problem (Proposition 2.7), the external input that makes the quadratic improvement of flatness work.","marker":"[LZ18]"},{"why":"Introduces the improvement-of-flatness strategy with a right-hand side that Proposition 4.1 adapts to the fully nonlinear normal-dependent setting.","marker":"[DeS11]"},{"why":"Provides the quadratic improvement of flatness for the Laplacian with right-hand side whose contradiction-and-compactness structure is followed and modified.","marker":"[DFS19]"},{"why":"Gives the hodograph transform used in Corollary 1.6 to upgrade $C^{2,\\alpha}$ regularity of the free boundary to higher smoothness.","marker":"[KN77]"},{"why":"Defines the classical one-phase problem that the present problem generalizes.","marker":"[AC81]"},{"why":"Initiates Perron's method for free boundary problems, the basis of the existence proof in Theorem 1.4.","marker":"[Caf88]"},{"why":"Extends Perron's method to fully nonlinear homogeneous concave operators, a precedent extended here to non-homogeneous, gradient- and $x$-dependent operators.","marker":"[Wan03]"},{"why":"Handles Perron's method for two-phase problems with distributed sources, contributing the right-hand-side framework used in the existence argument.","marker":"[DFS15b]"},{"why":"Provides the classical boundary regularity estimates for the oblique-boundary elliptic problem arising after the hodograph transform in Corollary 1.6.","marker":"[ADN59]"},{"why":"Supplies the uniformly elliptic viscosity framework and Harnack estimates used throughout the paper.","marker":"[CC95]"}],"fun_headline_variants":["Quadratic flatness gives C²,α free boundaries in one-phase problems","Perron existence and flatness improve free boundaries to C²,α","Hodograph transform upgrades flat C²,α free boundaries to smooth","Existence and C²,α regularity for fully nonlinear one-phase problems","From flat to C²,α: a quadratic improvement in one-phase free boundaries"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument rests on an external $C^{2,\\alpha}$ estimate for the linearized oblique problem (Proposition 2.7, citing [LZ18, Theorem 1.3]) that requires the operator to be concave or convex in the Hessian and the oblique direction to stay nondegenerate; the paper asserts these hypotheses pass to the limiting operator through (H2) and compactness but does not verify them in detail.","fun_headline_variants_meta":{"raw":{"variants":["Quadratic flatness gives C²,α free boundaries in one-phase problems","Perron existence and flatness improve free boundaries to C²,α","Hodograph transform upgrades flat C²,α free boundaries to smooth","Existence and C²,α regularity for fully nonlinear one-phase problems","From flat to C²,α: a quadratic improvement in one-phase free boundaries"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000613,"raw_usage":{"total_tokens":2838,"prompt_tokens":920,"completion_tokens":1918,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":536,"completion_tokens_details":{"reasoning_tokens":1819}},"tokens_in":536,"tokens_out":1918,"duration_ms":15100,"temperature":1.0,"reasoning_tokens":1819,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T17:32:57.227247+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"The decisive test is to inspect the limit operator $\\widetilde{F}_\\infty$ that appears in the compactness step of Proposition 4.7 and ask whether it really satisfies the concavity/convexity and nondegenerate-oblique hypotheses of [LZ18, Theorem 1.3]; exhibiting an admissible sequence $F_n,g_n,u_n$ for which $\\widetilde{F}_\\infty$ violates one of those hypotheses, or producing a flat viscosity solution whose free boundary fails to be $C^{2,\\alpha}$ under (H1)–(H3), would refute Theorem 1.5.","supporting_citations":[],"review_version":1}