{"id":"6a85b47b-a751-4b09-b9ad-93c4eafb21cb","arxiv_id":"2501.18234","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Sharp existence and uniqueness theorems are proved for the Liouville equation on R^2 with locally integrable or nonnegative potentials, including a comparison principle and radial uniqueness for positive coupling.","lead":"This mathematics paper proves when the classical Liouville equation on the plane has a solution, and when that solution is unique, under very weak regularity assumptions on the coefficient function. The results tighten old existence thresholds and are applied to conformal geometry, vortex statistics, and anyon models.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 1.2's subsolution construction is internally inconsistent: the defined shift gives normalization 1/|β|, not 1, and logarithmic slope 2 instead of −2β for general β<0, so the Perron comparison fails.","rationale":"The reader's weakest_assumption focused on the uniqueness theorem, specifically condition (1.24) and the missing 4πβ coefficient in ODE (4.5). Those are real issues, but I identify a different, more directly load-bearing gap in the main existence theorem Theorem 1.2: the displayed identity (3.42) asserts both a PDE and a normalization that cannot hold simultaneously for general β<0. This is an internal inconsistency, verifiable by direct substitution, and it breaks the comparison argument that feeds the Perron construction. The flaw is repairable by invoking Theorem 1.1 with the same β rather than the auxiliary β=-1 problem, so the correct disposition remains CONDITIONAL rather than REJECT. Because the reader's stated weakest_assumption was about uniqueness rather than this existence proof gap, my agreement is 'disagree' even though both point to repairable proof gaps.","tokens_in":41662,"tokens_out":20851,"duration_ms":176395,"concrete_test":"Choose β=-2 and a fixed smooth compactly supported radial bump ~W1 with ∫~W1e^{~ψ1}=1 as in the proof. Compute m=∫W1e^{ψ1} for W1=~W1, ψ1=~ψ1-log2; the direct calculation gives m=1/2, contradicting (3.42), and the logarithmic slope of ψ1 is 2 rather than 4. Then rerun the Perron argument using ~ψ1 obtained from Theorem 1.1 for the same β, i.e. -Δ~ψ1=4πβ~W1e^{~ψ1} with ∫~W1e^{~ψ1}=1 and slope -2β; if the rest of the barrier proof then closes, the repair is confirmed.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem 1.2's proof contains an explicit normalization inconsistency that breaks the Perron barrier for general β<0. The proof first produces ~ψ1 satisfying -Δ~ψ1 = -4π~W1e^{~ψ1} with ∫~W1e^{~ψ1}=1. It then sets ψ1(x)=~ψ1(x-x0)-log|β| and W1=~W1, and claims in (3.42) both -Δψ1=4πβW1e^{ψ1} and ∫W1e^{ψ1}=1. The PDE is correct, but the normalization is ∫W1e^{ψ1}=∫~W1e^{~ψ1}/|β|=1/|β|, not 1. Consequently, Lemma 2.2 gives lim_{x→∞} ψ1/log|x| = 2, whereas (3.43) claims the slope is -2β = 2|β|. For |β|>1 these differ, and the constructed ψ2,ε has slope -2β, so the needed inequality ψ1≥ψ2,ε fails at infinity unless β=-1. Thus the cited Perron existence step [N82, Thm. 2.10] is not actually applied to a valid ordered pair of barriers. This is not a missing hypothesis check but an internal inconsistency in a displayed identity; the argument can likely be repaired by applying Theorem 1.1 directly with the same β rather than the auxiliary β=-1 equation, but as written Theorem 1.2 is not proved.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves existence and uniqueness results for the Liouville equation −Δψ = 4πβ V e^ψ on R^2 with the normalization ∫ V e^ψ = 1, for both signs of β. Theorem 1.1 gives radial existence for β > 0 under integrability conditions (1.3) and for β < 0 with nonnegative V; Theorem 1.2 extends the negative-β existence to non-radial V using Perron's method; Theorem 1.3 is a comparison/uniqueness principle for β < 0; Theorem 1.4 proves radial uniqueness for β > 0 under the structural condition (1.24). Applications are given to the Berger–Nirenberg problem, blow-up analysis, statistical mechanics, and the Chern–Simons–Schrödinger model.","tokens_in":41918,"tokens_out":12137,"duration_ms":103832,"significance":"If the proofs are correct, the paper substantially advances the Liouville-equation literature: Theorem 1.1 permits merely locally integrable V, gives explicit bounds, and covers sharp thresholds, while Theorem 1.3 is a clean comparison principle. The paper is also commendably self-contained, uses standard elliptic/variational machinery without fitted parameters, and draws connections to several active areas. However, two central proofs contain normalization or coefficient inconsistencies that currently invalidate Theorem 1.2 and Theorem 1.4 as stated; these are local and likely repairable, but they are load-bearing for the paper's main claims.","major_comments":[{"comment":"The definition ψ1(x) := ~ψ1(x−x0) − log|β| with W1 := ~W1(x−x0) gives ∫ W1 e^{ψ1} = 1/|β|, not 1 as claimed in (3.42), because e^{−log|β|} = 1/|β|. The PDE −Δψ1 = 4πβ W1 e^{ψ1} is correct, but the normalization identity in (3.42) is false. Consequently Lemma 2.2 yields lim_{x→∞} ψ1/log|x| = 2, whereas (3.43) asserts the slope is −2β = 2|β|. For |β|>1 these differ, and ψ2,ε has slope −2β, so the barrier inequality ψ1 ≥ ψ2,ε can fail at infinity; the application of [N82, Thm. 2.10] is therefore not justified. The argument can likely be repaired by invoking Theorem 1.1 directly with the same β rather than the auxiliary β = −1 equation, but as written Theorem 1.2 is not proved.","section":""},{"comment":"Equation (4.5) is written as ∂_r^2 ψ + (1/r)∂_r ψ + r^n V(r)e^ψ = 0, but the radial reduction of (1.25) is ∂_r^2 ψ + (1/r)∂_r ψ + 4πβ r^n V(r)e^ψ = 0. The missing 4πβ coefficient is not cosmetic: the subsequent definition 2β(s) := ∫_0^∞ r^{n+1}V(r)e^{ψ(r,s)} dr also omits the 2π factor from polar coordinates, so it does not equal the limiting slope for a solution of (1.25). For a solution of (1.25) one has lim_{r→∞}(−r∂_rψ) = 2β while ∫_0^∞ r^{n+1}V e^ψ dr = 1/(2π). The family ψ(·,s) defined by (4.5) therefore does not parametrize the radial solutions of (1.25), and the identities (4.13), (4.14), (4.16), and the use of Corollary 1.2 in (4.6) do not apply to the theorem's equation. Unless the coefficient is restored (or V is rescaled and the normalization ∫|x|^n V e^ψ = 1 is adjusted accordingly), the proof establishes uniqueness only for a different problem.","section":""}],"minor_comments":[{"comment":"The last line of the proof states |x||∇φ(x)| = 4πβ (which is negative for β<0); the intended bound is |x||∇φ(x)| ≤ −2β, as used in (3.4). Please correct the typo.","section":""},{"comment":"Equation (5.17) has an inconsistent normalization: the source term contains e^{−B|x|^2/2} but the integral constraint contains e^{−B|x|^2/4}. From the derivation, the density is |u|^2 = |x|^{2n} e^{ψ_B − B|x|^2/2}, so both occurrences should be e^{−B|x|^2/2} (or the derivation and Corollary 5.8 must be adjusted consistently).","section":""},{"comment":"The hypotheses of [N82, Thm. 2.10] are not stated, and the proof does not verify that the ordered pair (ψ1, ψ2,ε) satisfies them after the normalization issue above is fixed; please either state the theorem or give the verification.","section":""}],"recommendation":"major_revision","confidential_remarks":"The paper is ambitious and contains substantial ideas, especially the positive-β existence theorem and the comparison principle. However, the two identified errors are in the proofs of central theorems: Theorem 1.2's normalization inconsistency and Theorem 1.4's missing coefficient. Both seem repairable, but the revision needs to address them explicitly; otherwise the main claims are not established. I would not recommend rejection, as the underlying methods appear sound and the fixes are likely local."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's what you should know: the radial existence theorem (Thm 1.1) is a genuine step forward—L^1_loc coefficients, sharp beta >= -alpha(V) threshold, explicit bounds, and the counterexamples for beta >= 2 settle a point that was confused in the literature (e.g., [M85] is indeed wrong). The whole-plane comparison principle (Thm 1.3) is useful and appears correct. The author also deserves credit for a variational proof that avoids the usual weighted Sobolev/ODE machinery, and for a wide-ranging applications section that is mostly honest about what it does and doesn't prove.\n\nBut the paper as written has a load-bearing error in Theorem 1.2. The proof starts with ~psi1 for beta=-1, then defines psi1(x)=~psi1(x-x0)-log|beta|, W1=~W1, and claims both the PDE and integral W1 e^{psi1} = 1. The PDE is fine, but the integral is integral ~W1 e^{~psi1}/|beta| = 1/|beta|, not 1. Lemma 2.2 then forces lim psi1/log|x| = 2, while (3.43) claims -2beta = 2|beta|. For |beta|>1 the two barriers are no longer ordered, so the Perron step [N82, Thm. 2.10] is not applied to a valid ordered pair. This is an internal inconsistency, not a missing hypothesis check. It's probably repairable—apply Theorem 1.1 directly with the same beta—but as written Theorem 1.2 is not proved.\n\nTheorem 1.4 has a smaller but real problem: the ODE (4.5) drops the 4 pi beta coefficient, so the proof silently rescales V. That can be fixed by rescaling, but it should be stated. Section 5.4 has the same disease: the exponent in (5.17) uses -B|x|^2/2 in the PDE but -B|x|^2/4 in the normalization; as stated Corollary 5.8 is off. The dependence on [N82, Thm. 2.10] without verifying its hypotheses is a minor concern once the barrier is fixed, but it needs to be explicit.\n\nI checked the main variational argument for Theorem 1.1 and it is coherent; the coercivity and monotone convergence steps line up. The criticism of [M85] is backed by a concrete counterexample. The citation pattern is ordinary; the self-citation [ALN24] supplies lemmas, not the target results.\n\nBottom line: this deserves a serious referee. The central theorem is likely correct and important enough that the errors should not sink the paper. The right disposition is major revision, with the author asked to fix the normalization in Thm 1.2 and Thm 1.4, restate the Perron hypotheses, and recompute the quantum mechanics normalization. I'd bring it to the reading group now, but I wouldn't cite it until the revision lands.","headline":"Theorem 1.1 is a real advance, but Theorem 1.2 and 1.4 have load-bearing normalization errors; the paper needs a careful revision before it can be trusted.","tokens_in":42509,"tokens_out":3474,"would_cite":false,"duration_ms":31032,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["47J10","35J20","35Q82","81V27","81V70"],"pacs":[],"model":"deepseek-v4-flash","headline":"Radially symmetric, merely locally integrable potentials still admit weak solutions of the normalized Liouville equation, and a single-radius sign-change condition on $cV(r)+rV'(r)$ forces the radial solution to be unique.","keywords":["Liouville equation","Nirenberg's problem","mean field equation","Chern–Simons–Schrödinger model","radially symmetric solutions","weak solutions","uniqueness","Pohozaev identity"],"falsifier":"Shoot the radial ODE for (1.25) numerically with $\\beta = 1$, $n = 0$, and $V(r) = (1+r^2)^{-3}$, which satisfies (1.23) and (1.24): Theorem 1.4 predicts at most one normalized radial profile, so two different initial heights $\\psi(0) = s$ yielding two profiles with $\\int_{\\mathbb{R}^2} V e^{\\psi} = 1$ would refute the uniqueness claim. As a proof-level check, confirm that the coefficient $4\\pi\\beta$ of (1.25) is absorbed in the ODE (4.5), which as written omits it, or that $V$ is implicitly rescaled before the ODE analysis.","tokens_in":41402,"feed_emoji":"📐","tokens_out":21738,"duration_ms":171017,"temperature":0.7,"pith_summary":"Radially symmetric, merely locally integrable coefficients $V$ still admit weak solutions of the normalized Liouville equation $-\\Delta\\psi = 4\\pi\\beta V e^{\\psi}$ with $\\int_{\\mathbb{R}^2} V e^{\\psi} = 1$, and the paper proves sharp necessary conditions on how fast $V$ must decay. For $\\beta > 0$ it establishes existence under the weighted decay conditions (1.3) provided $\\beta \\geq -\\alpha(V)$, and it shows this threshold is necessary for any radial solution; for $\\beta < 0$, nonnegativity plus the same threshold suffices, and here $V$ need not be radially symmetric. A generalized comparison principle for $\\beta < 0$ yields uniqueness of solutions sharing the same logarithmic growth at infinity. For $\\beta > 0$ the paper proves at most one radially symmetric $C^2$ solution exists when $V$ is nonnegative, radially decreasing, and every combination $cV(r) + rV'(r)$ with $0 < c < n+2$ flips sign exactly once. Why it matters: the criteria are explicit and checkable, exponential decay is shown not to be enough when $\\beta \\geq 2$, and the results translate into thresholds for the Nirenberg, mean-field, and Chern–Simons–Schrödinger problems.","feed_headline":"Locally integrable potentials now solve the Liouville equation","feed_subtitle":"Locally integrable coefficients were out of reach; decay thresholds and a one-radius sign change settle both questions.","key_machinery":"The argument runs on three devices. First, a Moser-type energy functional $E_n[\\phi] := \\frac{1}{2}\\int_{D(0,n)}|\\nabla\\phi|^2 - 4\\pi\\beta\\log\\left(\\int_{D(0,n)} V e^{\\phi}\\right)$ over radially symmetric $\\phi \\in H^1_0(D(0,n))$ with $\\int V e^{\\phi} > 0$: the decay conditions make it coercive and bounded above uniformly in $n$, and the minimizers $\\varphi_n$, after subtracting the normalization $\\log\\int V e^{\\varphi_n}$, converge to a global weak solution. Second, for the uniqueness theorem, the radial flow $\\psi(r,s)$ solving $\\partial_r^2\\psi + r^{-1}\\partial_r\\psi + r^n V(r)e^{\\psi} = 0$ with $\\psi(0,s)=s$, $\\partial_r\\psi(0,s)=0$, its sensitivity $\\varphi(r,s) = \\partial_s\\psi$, and the P-function $P(\\psi) := r\\partial_r\\psi(\\tfrac{1}{2}r\\partial_r\\psi + \\beta) + r^{n+2}Ve^{\\psi}$: condition (1.24) forces $P(\\psi) \\geq 0$, and the cases where $\\varphi$ has no zero, two zeros, or one crossing each end in contradiction, so $\\beta(s)$ is strictly monotone and two radial solutions cannot coexist. Third, the comparison principle of Theorem 1.3 is proved by a maximum principle applied at a minimizing point of the ratio $(\\psi_1-\\psi_2)/\\log(|x|-1)$, which turns the monotonicity of $F$ into an algebraic contradiction.","core_discovery":"The central claim is that solvability of (1.1) is governed by the weighted-decay index $\\alpha(V) := \\sup\\{\\alpha : \\int_{\\mathbb{R}^2\\setminus D(0,1)} |V(x)|\\,|x|^{2\\alpha}\\,dx < \\infty\\}$ together with the local behavior of $V$ at the origin. For $\\beta > 0$, existence holds when $V^+$ satisfies the weighted integrability conditions (1.3) near zero and infinity, $V^-$ is integrable against the weight $|x|^{-2\\beta}$, and $\\beta \\geq -\\alpha(V)$; the paper proves the condition on $V^-$ is necessary, so a radial solution forces $\\beta \\geq -\\alpha(V)$. For $\\beta < 0$, nonnegative locally integrable $V$ suffices under the same threshold, with no radial symmetry assumed. Uniqueness: for $\\beta < 0$, a comparison principle extends the classical maximum principle to the full plane using the asymptotic ratio $(\\psi_1-\\psi_2)/\\log|x|$, and forces two solutions with identical logarithmic growth to coincide. For $\\beta > 0$ with $V$ nonnegative, radially decreasing and satisfying $r^{n+2+\\delta}V(r) \\leq C$, at most one radially symmetric $C^2$ solution exists whenever every $cV(r) + rV'(r)$ with $0<c<n+2$ changes sign from nonnegative to nonpositive at a single radius; the zero-counting argument on $\\varphi = \\partial_s\\psi$ is what this condition makes possible.","pith_inferences":["Because the paper proves $\\beta \\geq -\\alpha(V)$ is necessary for radial solutions, $\\alpha(V)$ acts like a dimension of the measure $V\\,dx$ at infinity; the natural conjecture, which Question 1.1 leaves open, is that the same index gates nonradial existence for $\\beta \\geq 2$.","Condition (1.24) is verifiable from the graph of $rV'(r)/V(r)$ alone; a plausible strengthening is that it forces every solution to be radially symmetric for $n>0$ under the asymptotic $\\psi/\\log|x| \\to -2\\beta$, which the paper explicitly says it could not prove.","For the Chern–Simons–Schrödinger model the paper analyzes only monomials $|z|^{2n}e^{-B|z|^2/2}$; Question 1.3's perturbation problem $f_\\lambda = z^n + \\lambda g$ would test whether the unique radial family survives when the potential breaks symmetry, and Corollary 5.8 gives the exact parameter window in which to look.","The explicit, data-continuous bounds behind (1.4) hint that the construction is quantitative enough to yield blow-up rates for rough potentials, which would extend Corollary 5.4's compactness beyond Hölder coefficients."],"forward_implications":["Corollary 1.1 gives explicit, sharp growth windows for radial power-type potentials: $l_0 < 2\\beta-2$ controls $V^-$ at infinity, $l_1 > \\beta-2$ controls $V^+$ at the origin, and $l_2 < \\beta-2$ controls $V^+$ at infinity, with counterexamples at the boundary values.","Corollary 1.2 shows the criterion $n > \\beta-2$ is necessary and sufficient for $V(r) = r^n V_0(r)$ with nonnegative radially decreasing $V_0$; in particular $V = e^{-r^2}$ has a solution if and only if $\\beta < 2$, so exponential decay does not rescue $\\beta \\geq 2$.","For $\\beta < 0$, Theorems 1.2 and 1.3 provide existence for general nonnegative locally integrable $V$, a monotone family $\\psi_\\beta + \\log|\\beta|$, and a comparison principle that gives uniqueness for solutions with the same logarithmic growth.","Theorem 1.4 establishes radial uniqueness for $\\beta > 0$ under the one-radius sign-change condition (1.24), which for $n=0$ becomes full uniqueness since all solutions are then radial.","The applications yield concrete thresholds: Berger–Nirenberg solvability (Corollaries 5.1–5.2), blow-up compactness (Corollaries 5.3–5.4), mean-field solvability iff $n > (-\\beta-8\\pi)/4\\pi$ (Corollary 5.5), and Chern–Simons–Schrödinger existence iff $2n > \\beta-2$ (Corollary 5.8), with the $B\\to0$ and $B\\to\\infty$ concentration behavior."],"supporting_citations":[{"why":"introduces the $\\psi_1 = \\psi + \\psi_0$ shift with $f = -\\Delta\\psi_0$ that makes the $\\beta>0$ variational problem translation-invariant; the existence proof says it applies this approach.","marker":"[M85]"},{"why":"supplies the radial ODE flow and the P-function zero-counting scheme that the proof of Theorem 1.4 explicitly follows.","marker":"[Lin00]"},{"why":"proves all solutions are radial when $n=0$, which converts Theorem 1.4's radial uniqueness into genuine uniqueness in Remark 1.13.","marker":"[CK94]"},{"why":"provides the asymptotic expansion and Pohozaev-type argument behind the necessity direction of Corollary 1.2.","marker":"[CL93]"},{"why":"gives the Perron-method existence result, Thm. 2.10, on which Theorem 1.2 is built.","marker":"[N82]"},{"why":"supplies the Brezis–Merle uniform estimates and the $4\\pi$ dichotomy used in Corollary 5.6's blow-up analysis.","marker":"[BM91]"},{"why":"Moser's sharp exponential inequality, together with [T67], gives the coercivity estimate (2.7) needed for the $\\beta<0$ minimizers in Lemma 2.1.","marker":"[M71]"},{"why":"provides the regularized Newtonian-potential estimates (Lemmas 3.6, 3.7, 3.15) used to read logarithmic asymptotics of solutions in Lemma 2.2 and Corollary 1.2.","marker":"[ALN24]"}],"fun_headline_variants":["Weighted decay index settles Liouville solvability","Sign change condition unlocks uniqueness for Liouville","New threshold for Liouville existence and uniqueness","Liouville equation solved for locally integrable potentials"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The uniqueness theorem rests entirely on condition (1.24): for every constant $c \\in (0, n+2)$, the combination $cV(r) + rV'(r)$ must switch from nonnegative to nonpositive at some radius $r_c$, and if that single-sign-change structure fails, the zero-counting contradiction in the proof collapses.","fun_headline_variants_meta":{"raw":{"variants":["Weighted decay index settles Liouville solvability","Sign change condition unlocks uniqueness for Liouville","New threshold for Liouville existence and uniqueness","Liouville equation solved for locally integrable potentials"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000188,"raw_usage":{"total_tokens":1305,"prompt_tokens":891,"completion_tokens":414,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":507,"completion_tokens_details":{"reasoning_tokens":359}},"tokens_in":507,"tokens_out":414,"duration_ms":4442,"temperature":1.0,"reasoning_tokens":359,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T00:16:08.331071+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Shoot the radial ODE for (1.25) numerically with $\\beta = 1$, $n = 0$, and $V(r) = (1+r^2)^{-3}$, which satisfies (1.23) and (1.24): Theorem 1.4 predicts at most one normalized radial profile, so two different initial heights $\\psi(0) = s$ yielding two profiles with $\\int_{\\mathbb{R}^2} V e^{\\psi} = 1$ would refute the uniqueness claim. As a proof-level check, confirm that the coefficient $4\\pi\\beta$ of (1.25) is absorbed in the ODE (4.5), which as written omits it, or that $V$ is implicitly rescaled before the ODE analysis.","supporting_citations":[],"review_version":1}