{"id":"d7e1807b-315c-48d7-ba3b-0269c4f421f4","arxiv_id":"2507.16383","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":8.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"For fully nonlinear Loewner-Nirenberg equations with boundary data w=0, the hyperbolic solution is unique when μ_Γ^+>1, while for μ_Γ^+≤1 all solutions form a one-parameter family depending only on x_n.","lead":"This paper classifies every positive solution to a broad family of conformally invariant PDEs in the upper half-space: for one regime the hyperbolic solution is unique, for another there is a whole one-parameter family. It then uses the extra solutions to show that local boundary C^0 estimates fail in the non-unique regime.","discovery_kind":"first_principles","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The symmetry step in Proposition 2.1 rests on an imported comparison principle asserted for non-convex, non-homogeneous pairs without verification; if [36, Thm 3.2] does not cover this setting, the reduction to ODEs and the classification fail.","rationale":"The reader's weakest assumption correctly identifies the comparison principle as the most load-bearing step. I read the ODE analysis and the moving-spheres construction in good faith and found no internal contradiction there; the one-parameter family construction, the blow-up argument for mu > 1, and the asymptotic estimates in Theorem 1.5 all appear coherent. The additional homogeneity gap in Lemma 2.6 is real but affects mainly the claimed relaxed-conditions version of Proposition 2.1; it does not by itself destroy the main classification under the full hypotheses (1.3)–(1.5), (1.8'), because homogeneity is then available. Both issues are repairable: either the comparison principle in [36] is verified to apply, or Proposition 2.1 is restricted to the full assumptions. Hence the original CONDITIONAL verdict is appropriate, and my read does not change it.","tokens_in":31976,"tokens_out":28831,"duration_ms":316869,"concrete_test":"Check the hypotheses of [36, Theorem 3.2] for F(M,p,z) = f(lambda(-zM + (1/2)|p|^2 I)) - 1/2 under (1.3')–(1.5'), (1.6), (1.7). In particular, verify the monotonicity-in-z assumption by computing F(M,p,z_2) - F(M,p,z_1) for z_2 > z_1 on the range of Hessians arising in the doubling/touching argument; an unambiguous sign is required for the comparison principle. If the theorem requires convexity of Gamma or homogeneity of f, supply a proof of Proposition 2.4 under the stated weak assumptions or amend Proposition 2.1 accordingly. Also confirm whether Lemma 2.6 can hold for the non-homogeneous example f(lambda) = (sigma_1(lambda)/3) exp((sigma_1(lambda)-3)/10), where the hyperbolic barrier is not a solution.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Proposition 2.4 is stated under the relaxed assumptions (1.3')–(1.5'), (1.6), (1.7), which drop convexity of Gamma and homogeneity of f, and is justified only by citing [36, Theorem 3.2] via [22, Proposition 2.2]. This comparison principle is load-bearing: Lemma 2.9 uses Proposition 2.3 to compare beta*w with the inverted function w_{nu,R} in D_{nu,R}\\Sigma_s, and Proposition 2.3 is derived from Proposition 2.4. If [36, Theorem 3.2] requires the operator to be proper in the u-variable, or requires Gamma to be convex, then the conclusion w = w(x_n) in Lemma 2.9 is not established and the subsequent ODE reduction in Section 3 has no basis. A separate, concrete gap appears earlier: Lemma 2.6 uses the hyperbolic metric as a lower barrier and thereby assumes f(lambda(1/2 e)) = 1/2, which is guaranteed by homogeneity (1.5) but not by the relaxed (1.5'). For example, f(lambda) = (sigma_1(lambda)/3) exp((sigma_1(lambda)-3)/10) satisfies (1.3')–(1.5'), (1.6), (1.7) on Gamma_1^+ but has f(1/2 e) = (1/2)e^{-3/20} < 1/2. Thus Proposition 2.1 as stated is not proved. The central classification under the full assumptions (1.3)–(1.5), (1.8') is plausibly repairable, but the manuscript should either prove Proposition 2.4 directly or restrict Proposition 2.1 to hypotheses that include homogeneity and convexity.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper classifies positive viscosity solutions w of the fully nonlinear Loewner–Nirenberg problem f(λ(−A_w)) = 1/2, λ(−A_w) ∈ Γ in R^n_+, with w = 0 on ∂R^n_+. Under the standard hypotheses (1.3)–(1.8) the authors prove that the hyperbolic solution w^(0)(x)=x_n is the unique solution when μ_Γ^+ > 1, and that when μ_Γ^+ ≤ 1 the solution set is a one-parameter family {w^(a)(x_n)}_{a≥0} with w^(a)(1)=1+a, each solution depending only on x_n. The proof has three stages: C^0 bounds via barriers, C^{2,α} and C^1 regularity up to the boundary via Savin's small-perturbation theorem, and a moving-spheres argument with spheres centred in the lower half-space. The resulting one-dimensional ODE is analysed in Section 3, including existence, uniqueness, asymptotics and a first integral. As applications, the paper gives counterexamples to local boundary C^0 and gradient estimates when μ_Γ^+ ≤ 1 and a counterexample to the comparison principle on unbounded domains.","tokens_in":32330,"tokens_out":7491,"duration_ms":83925,"significance":"If the main classification is correct, it is a substantial contribution: it gives a complete Liouville theorem for a large class of conformally invariant fully nonlinear equations on the half-space, identifies the sharp parameter μ_Γ^+ as the criterion for uniqueness, and shows a surprising failure of local boundary estimates. The moving-spheres construction with centres below the boundary is a genuine novelty, and the ODE analysis is detailed and largely self-contained. The paper also honestly discusses limitations and connections to prior work. However, several load-bearing statements are made under relaxed assumptions (1.3′)–(1.5′) that the proofs do not actually support; the central classification under the full hypotheses (1.3)–(1.8) is plausible and likely repairable, but the manuscript in its current form overclaims its generality.","major_comments":[{"comment":"The regularity argument near ∂R^n_+ relies on the identity G(∇²û, ∇û, û, x) = x_n²(f(λ(−A_u)) − (1/2)e^{2u}) and on the normalization G(0,0,0,x) = 0. Both require f(tλ) = tf(λ), i.e. homogeneity. Under the stated relaxed assumptions (1.3′)–(1.5′), (1.6), (1.7), the rescaled operator G does not have this form, and the uniform ellipticity at (0,0,0,x) is not justified. Thus Lemma 2.8, and with it the C^1 up-to-boundary information used in Lemma 2.10, is not proved in the stated generality. This is a separate gap from the barrier issue, since even if one fixes the lower and upper C^0 bounds, the Savin-type perturbation argument in Step 1 of Lemma 2.8 requires the rescaling identity that homogeneity provides.","section":"§2.4, definition of G and Lemma 2.8"}],"minor_comments":[{"comment":"The sentence 'This follows from the fact that G is uniformly elliptic at (0,0,0,x) for x bounded away from 0 and infinity' should be expanded: one should explicitly note that (1/2)e ∈ Γ and that ∂f/∂λ_i > 0 at λ = e/2 by (1.6), since this is the actual hypothesis justifying the uniform ellipticity.","section":"§2.4, near equation (2.6)"},{"comment":"There are a few minor typographical issues: in the line after (2.6), 'for e.g. r > 2δ^{-1}' appears to mean r > 2/δ; in the proof of Lemma 3.5 the interval notation in 'Dom(ψ) ∩ (0, ∞) = (0, ∞)' is redundant; and the displayed equation for G in §2.4 has a mismatch between the matrix argument and the preceding definition of  that would benefit from a brief explanatory sentence.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The central classification under the standard hypotheses (1.3)–(1.5), (1.8′) is likely correct and valuable, but the paper currently states Theorem 1.3 and Proposition 2.1 under weaker hypotheses that the proofs do not support. The main fix is to restrict the statements or add the missing arguments; I do not see a reason to reject the paper outright. I would also ask the authors to double-check that the comparison principle in [36] indeed applies verbatim to the viscosity setting with non-convex Γ and non-homogeneous f, since that is the single most consequential external input in the proof."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Two things you should know. First, this is a strong paper: it completes the classification of positive viscosity solutions to the fully nonlinear Loewner-Nirenberg equation on the half-space, and the answer has a clean phase transition at mu_Gamma^+ = 1. When mu_Gamma^+ > 1 the hyperbolic solution is unique; when mu_Gamma^+ <= 1 there is a one-parameter family w^(a)(x_n), and those extra solutions kill local boundary C^0 and gradient estimates. Second, the proof is mostly sound, but the version of Theorem 1.3 that claims symmetry under the relaxed assumptions (1.3')-(1.5') is not proved as written.\n\nThe genuinely new part is the use of moving spheres centered in the lower half-space, which bypasses the Hopf lemma failure at the boundary. The C^0 barriers, the Savin-based regularity, and the ODE analysis are careful and credible. The one-parameter family comes from the ODE, not from fitting the conclusion, so there is no circularity. The counterexample to comparison on unbounded domains in Appendix B is a nice bonus.\n\nSoft spots, in proportion. The relaxed-assumption theorem overreaches. Lemma 2.5 and Lemma 2.9 both use homogeneity: the barrier needs f(1/2 e) = 1/2, and the beta-scaling argument needs f(beta^2 lambda) = beta^2 f(lambda) to make beta w a supersolution. Under (1.3')-(1.5') alone neither is available, so the symmetry conclusion w = w(x_n) is not established in that generality. This does not touch the main classification under the full assumptions (1.3)-(1.5), which include homogeneity. The fix is to restrict Proposition 2.1 and the first part of Theorem 1.3 to the standard assumptions, or supply a genuine relaxed proof.\n\nThe other soft spot is the imported comparison principle. Propositions 2.3-2.4 are load-bearing and are passed off to [36, Thm 3.2] and [22, Prop 2.2]. The authors should verify that the non-convex, non-homogeneous pairs allowed by (1.3')-(1.5') actually satisfy the hypotheses of that theorem. If not, the moving-spheres step still works under the standard assumptions, so the main results survive.\n\nMinor: the indexing in the parameter family is off between the statement of Theorem 1.1 (w^(a)(1) = 1+a) and the proof (where the parameter is w(1) = a). Cosmetic, but confusing. The uniform ellipticity assertion for G in Section 2.4 is sketched quickly, though it looks repairable.\n\nWho should read this: anyone working on fully nonlinear Yamabe/Loewner-Nirenberg problems, boundary regularity, or Liouville theorems for conformally invariant equations. It deserves a serious referee; with the relaxed-assumption claim trimmed and the comparison-principle citation checked, it should be publishable in a strong journal.","headline":"A serious classification paper with a real phase transition; the main theorems hold up, but the relaxed-assumption symmetry claim and the imported comparison principle need referee attention.","tokens_in":32889,"tokens_out":8869,"would_cite":true,"duration_ms":81471,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35J60","35J70","35B53","53C21"],"pacs":[],"model":"deepseek-v4-flash","headline":"All half-space solutions to the fully nonlinear Loewner-Nirenberg problem are classified, and local boundary estimates fail when a cone parameter is at most one.","keywords":["Liouville theorem","Loewner-Nirenberg problem","fully nonlinear elliptic equations","Schouten tensor","method of moving spheres","viscosity solutions","boundary estimates","negative curvature conformal metrics"],"falsifier":"Integrate the first-order system $w'=\\sqrt{2K(bw)}$ (equivalently the ODE (3.9)) for a pair $(f,\\Gamma)$ with $\\mu_\\Gamma^+\\leq 1$: the paper predicts that for every $b>0$ the solution with $w(0)=0$ satisfies $w'(0)=1$ and $w(1)<\\infty$. Finding any $b>0$ whose solution has $w'(0)\\neq 1$, or a finite-time blow-up, would falsify Theorem 1.1; alternatively, exhibiting two bounded-domain viscosity solutions with boundary order $w_21$ the hyperbolic solution $w^{(0)}=x_n$ is the unique positive viscosity solution, while if $\\mu_\\Gamma^+\\leq 1$ the solutions are exactly the increasing one-parameter family $\\{w^{(a)}(x_n)\\}_{a\\geq 0}$, with $w^{(0)}$ the minimal member. In the non-unique regime every solution depends only on the distance to the boundary, each $w^{(a)}$ gives a metric locally complete near the boundary, and only $w^{(0)}$ is globally complete. It follows that local boundary $C^0$ and gradient estimates fail when $\\mu_\\Gamma^+\\leq 1$, which is the regime $\\Gamma=\\Gamma_k^+$ with $k\\geq n/2$ for the Gårding cones.","feed_headline":"One cone parameter decides: unique solution or a whole family","feed_subtitle":"Below the threshold, the solution set is a one-parameter family and local boundary estimates fail.","key_machinery":"The proof rests on a variant of the method of moving spheres in which the spheres are centred in the lower half-space rather than on the hyperplane $\\{x_n=0\\}$, so the comparison domain is bounded and neither a Hopf lemma along the boundary nor a 'touching at infinity' argument is needed. Three ingredients make this work: the comparison principle imported as Propositions 2.3 and 2.4; the barriers giving $w\\geq x_n$ globally and $w\\leq x_n+C x_n^2$ near the boundary; and the small-perturbation theorem of [45] giving $C^{2,\\alpha}$ regularity up to the boundary, following the scheme of [37]. After symmetry reduces the problem to an ODE in $t=x_n$, the key object is the function $\\varphi$ defined by $f(\\varphi(s),1,\\dots,1)=1/(2s)$, which converts the PDE into $ww''=(w')^2(1-\\varphi((w')^2/2))/2$; the first integral $G(s)=((s-\\tfrac12)/s)^{1/n}e^{B(s)}$ and its inverse $K=G^{-1}$ control the solution, and the asymptotic $K(x)=x^{1+\\mu_\\Gamma^++o(1)}$ decides whether solutions blow up in finite time ($\\mu_\\Gamma^+>1$) or exist globally ($\\mu_\\Gamma^+\\leq 1$).","core_discovery":"The central discovery is that all continuous viscosity solutions of (1.2) are one-dimensional: under the relaxed assumptions (1.3′)–(1.5′), (1.6) and (1.7) any such solution satisfies $w=w(x_n)$, and under the standard assumptions together with the weak inequality (1.8′) it must coincide either with $w^{(0)}$ or with one of the $w^{(a)}$ from Theorem 1.1. The family has $w^{(a)}(1)=1+a$, $(w^{(a)})'\\geq 1$, $(w^{(a)})''\\geq 0$, and $(w^{(a)})'(0)=1$; the metrics $g_{w^{(a)}}$ are locally complete near the boundary but incomplete on the whole half-space for $a>0$. The same result implies uniqueness of the solution $u=x_n^{-(n-2)/2}$ to the semilinear Yamabe-type equation (1.9), and Theorem 1.5 shows that $\\inf_{t\\in[\\varepsilon,E]}w^{(a)}(t)$ and $\\inf_{t\\in[\\varepsilon,E]}(w^{(a)})'(t)$ diverge as $a\\to\\infty$, giving counterexamples to local boundary $C^0$ and gradient estimates.","pith_inferences":["The dichotomy at $\\mu_\\Gamma^+=1$ may be a general phenomenon for conformally invariant Dirichlet problems modelled on the hyperbolic metric: the same parameter should govern compactness of solution families and the validity of boundary estimates for other fully nonlinear curvature equations.","Because the moving-spheres argument runs from centres below the boundary, it sidesteps the failure of the Hopf lemma; this variant may transfer to other degenerate-elliptic boundary problems where solutions touch at the boundary with equal normal derivatives.","The failure of local boundary $C^0$ estimates for $\\mu_\\Gamma^+\\leq 1$ suggests that any existence theorem for the fully nonlinear Loewner-Nirenberg problem on compact manifolds in that range must use global geometric data rather than boundary-local barriers."],"forward_implications":["For $\\mu_\\Gamma^+>1$, the hyperbolic metric $x_n^{-2}|dx|^2$ is the unique locally complete conformal metric of the prescribed curvature type on the half-space; in the semilinear case this recovers the uniqueness of $u=x_n^{-(n-2)/2}$ for (1.9).","For $\\mu_\\Gamma^+\\leq 1$, every positive viscosity solution is one of the $w^{(a)}$; the hyperbolic solution is the minimal solution and is the only one whose metric is complete on all of $\\mathbb{R}_+^n$.","For the Gårding cones $\\Gamma_k^+$, the threshold $\\mu_\\Gamma^+\\leq 1$ is exactly $k\\geq n/2$, so non-uniqueness begins halfway up the cone hierarchy.","Local boundary $C^0$ and gradient estimates fail when $\\mu_\\Gamma^+\\leq 1$: $\\inf_{t\\in[\\varepsilon,E]}w^{(a)}(t)$ and $\\inf_{t\\in[\\varepsilon,E]}(w^{(a)})'(t)$ blow up as $a\\to\\infty$.","When $\\mu_\\Gamma^+\\leq 1$, the comparison principle fails on unbounded domains and no lower-semicontinuous admissible supersolution can blow up along a boundary patch (Propositions B.1 and B.3)."],"supporting_citations":[{"why":"supplies the comparison principle imported as Propositions 2.3 and 2.4, on which the moving-spheres comparison and the ODE uniqueness depend.","marker":"[36]"},{"why":"provides the small-perturbation theorem used to obtain $C^{2,\\alpha}$ regularity near the boundary.","marker":"[45]"},{"why":"gives the boundary-regularity scheme used in Lemma 2.8.","marker":"[37]"},{"why":"introduces the parameter $\\mu_\\Gamma^+$ and the Harnack theory that sets the threshold.","marker":"[35]"},{"why":"gives the bounded-domain comparison principle version used as Proposition 2.4 in the fully nonlinear Loewner-Nirenberg context.","marker":"[22]"},{"why":"characterises conformally invariant operators, underpinning the conformal-invariance step in the moving-spheres transformation.","marker":"[27]"},{"why":"establishes existence for $\\mu_\\Gamma^+>1$ via local boundary estimates whose failure in the $\\mu_\\Gamma^+\\leq 1$ regime is the application here.","marker":"[15]"},{"why":"gives Liouville and counterexample results for the degenerate equation at $\\mu_\\Gamma^+=1$, marking the boundary between the two regimes.","marker":"[12]"}],"fun_headline_variants":["Cone parameter decides: unique solution or one-parameter family","One cone number flips uniqueness to a family, breaks boundary estimates","Below cone threshold, solutions form a family and estimates fail","Threshold cone parameter: single solution or family, no boundary estimates","Cone parameter >1: unique; else family and no local estimates"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument depends on a comparison principle for viscosity solutions on bounded domains, assumed to hold even when the defining function is neither convex nor homogeneous; if that principle fails, the conclusion that every solution depends only on distance to the boundary collapses.","fun_headline_variants_meta":{"raw":{"variants":["Cone parameter decides: unique solution or one-parameter family","One cone number flips uniqueness to a family, breaks boundary estimates","Below cone threshold, solutions form a family and estimates fail","Threshold cone parameter: single solution or family, no boundary estimates","Cone parameter >1: unique; else family and no local estimates"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00147,"raw_usage":{"total_tokens":6099,"prompt_tokens":1319,"completion_tokens":4780,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":935,"completion_tokens_details":{"reasoning_tokens":4702}},"tokens_in":935,"tokens_out":4780,"duration_ms":36663,"temperature":1.0,"reasoning_tokens":4702,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-06T15:12:06.966087+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Integrate the first-order system $w'=\\sqrt{2K(bw)}$ (equivalently the ODE (3.9)) for a pair $(f,\\Gamma)$ with $\\mu_\\Gamma^+\\leq 1$: the paper predicts that for every $b>0$ the solution with $w(0)=0$ satisfies $w'(0)=1$ and $w(1)<\\infty$. Finding any $b>0$ whose solution has $w'(0)\\neq 1$, or a finite-time blow-up, would falsify Theorem 1.1; alternatively, exhibiting two bounded-domain viscosity solutions with boundary order $w_2