{"id":"e364202b-02d6-47b3-ab5a-95f9863fc317","arxiv_id":"2412.06746","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Positive supersolutions to (-Delta)^s u >= f(u,x) do not exist in exterior domains when f grows like the critical power near 0 for n>2s, or fast enough at infinity for n<=2s.","lead":"This paper proves that certain fractional Laplacian inequalities of the form (-Delta)^s u >= f(u,x) have no positive supersolutions in exterior domains of R^n, under very mild growth conditions on f. The result generalizes classical Liouville theorems to the nonlocal, fractional setting and covers all dimensions, including the previously untreated case n=1 with s>=1/2.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The proof applies Theorem 2.3 to barriers that are unbounded at infinity (e.g. m(r)+epsilon*w and eta_tilde(r)*(Phi_tilde-1)) without the required limiting argument, and the sign estimate (3.19)-(3.21) is also flawed; this gap is load-bearing for Lemma 3.4 and Theorem 3.2.","rationale":"The reader's weakest_assumption correctly identifies the load-bearing gap: the paper repeatedly invokes the comparison principle for barriers that are not in H^s and that are unbounded at infinity, without proving a suitable extension or a limiting argument. This affects Lemma 3.3, Lemma 3.4, and the main Theorem 3.2, and the omitted details are essential because the derived lower bounds are used to show m(r)->+infinity and to obtain the final contradiction. I found an additional concrete defect in the sign estimate (3.19)-(3.21), but the main concern remains the unbounded-barrier comparison. The central theorems may still be true, and likely a truncation-and-limit argument could repair the proof, but as written the argument is incomplete. Therefore the reader's conditional verdict is appropriate and no change is needed.","tokens_in":24523,"tokens_out":12447,"duration_ms":130919,"concrete_test":"Re-derive Lemma 3.4 with truncated barriers w_M=max(-M,-log|x|), apply Theorem 2.3 on bounded domains B_R\\B_r, and pass M->infinity. If the resulting lower bound is uniform in M, the comparison step is valid; if the bound depends on M and no uniform estimate emerges, the proof as written collapses. For the sign issue, recompute the integral in (3.19)-(3.21) directly for representative r and R~ to confirm whether the total is negative for all sufficiently large r.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central issue is the use of the comparison principle (Theorem 2.3) for functions that do not lie in the stated solution classes. In Lemma 3.3, the barrier m(r)+epsilon*Phi with Phi=-|x|^{sigma*} (sigma*>0) is compared with u on R after choosing R~; in Lemma 3.4, the barrier m(r)+epsilon*w with w=-log|x| is used; in Theorem 3.2, the barrier eta_tilde(r)*(Phi_tilde-1) (unbounded above) is compared. Theorem 2.3 is stated only for bounded domains with boundary data in H^s (or for viscosity solutions with L^1(omega) integrability), and the paper gives no approximation or limiting argument to justify comparison when the barrier is unbounded at infinity. The gap is not cosmetic: the lower bound in (3.5)/(3.16) and the bound (3.38) are derived from these comparisons, and they are used to prove m(r)->+infinity and the final contradiction. There is also a concrete sign error in (3.19)-(3.21): the second integral is negative, but the displayed inequalities bound it from the wrong side; the intended dominance of the negative term over the positive one requires a correct lower bound plus a size argument that is not written. Since the paper explicitly leaves the notion of solution vague, one cannot simply invoke a more general comparison principle unless it is stated and proved.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves nonexistence of positive continuous supersolutions of the fractional semilinear inequality (-Delta)^s u >= f(u,x) in exterior domains of R^n, for s in (0,1). The main results are Theorem 3.2 for n=1 and s in [1/2,1), where nonexistence is driven by the behavior of f as its first argument tends to infinity, and Theorems 4.1 and 4.3 for n>2s, where the decisive hypothesis is the behavior of f near t=0. The proofs combine truncated fundamental-solution barriers with quantitative strong maximum principles and comparison arguments.\n\nThe claimed results are broad and, if fully established, would substantially generalize existing Liouville theorems for fractional elliptic inequalities, including the case n=1 and the borderline s=1/2. The paper gives explicit model nonlinearities and conditions, and the strategy of using local assumptions rather than global ones is attractive. However, several load-bearing comparison and sign estimates are not justified as written, especially in Section 3.","tokens_in":24824,"tokens_out":29195,"duration_ms":328835,"significance":"If the theorems are correct, the paper would be a meaningful contribution: it covers all dimensions n>=1, treats both regimes n<=2s and n>2s, relaxes the hypotheses on f to local conditions, and addresses exterior domains as well as the whole space. The authors correctly identify the different roles of the fundamental solution in the three regimes sigma*<0, sigma*=0 and sigma*>0. The paper also contains useful preparatory material on quantitative strong maximum principles and makes explicit model nonlinearities satisfying the hypotheses. There are no fitted parameters and no circular dependence on the main theorem. The main reservations are technical: the comparison principle is applied to functions outside the stated solution classes, and at least one sign estimate in Lemma 3.4 is incorrect as written. These issues affect the proof of the central claims and must be repaired before the theorems can be accepted.","major_comments":[{"comment":"The comparison principle is repeatedly applied to barriers that are unbounded at infinity and are not contained in the function classes of Theorem 2.3. Examples are mtilde(r)+epsilon*Phi with Phi=-|x|^{sigma*}, mtilde(r)+epsilon*w with w=-log|x|, and eta_tilde(r)*(Phi_tilde-1). Theorem 2.3 is stated for bounded domains with boundary data in H^s, and Remark 2.9 explicitly leaves the notion of solution vague. No truncation, approximation, or limiting argument is supplied to justify comparison for these unbounded functions, even though the comparisons are load-bearing: they produce the inequalities mtilde(r)<=u, (3.38), c(r)Psi_gamma<=u and rho(r)w_gamma<=u, which in turn control m(r) and drive the contradictions. The authors should either state and prove a comparison principle valid for the relevant viscosity classes with the growth appearing in the barriers, or supply a careful limiting argument in each application.","section":"§2, Theorem 2.3; §3, Lemma 3.3, Lemma 3.4, Theorem 3.2 (3.17)-(3.22), (3.35)-(3.38); §4, Lemma 4.5, Theorem 4.1…"},{"comment":"The sign estimate for the truncated logarithmic barrier is not correct as written. The integral over B_r(x)\\B_1(x) in (3.19) is negative, and the displayed chain in (3.20)-(3.21) mixes inequality directions: an upper bound for the whole expression cannot be concluded from a lower bound for the negative integral. In addition, the negative term obtained in (3.21) has size of order epsilon*r*log r/(2\\tilde R)^2, while \\tilde R is allowed to depend on epsilon, r0 and r through condition (3.17). As written, the negativity of (-Delta)^s(mtilde(r)+epsilon w) in B_{\\tilde R}\\Br is not established. This defect directly affects the comparison step (3.22) and hence the s=1/2 branch of Theorem 3.2. A correct estimate must control \\tilde R relative to r and must show that the positive first integral is dominated by the negative contribution.","section":"§3, Lemma 3.4, equations (3.19)-(3.21)"},{"comment":"The lower bounds of the form cmin <= inf_{B_{2r}\\B_r} u are not justified because the proof uses mtilde(r)=inf_{B_r}u and then concludes that 0<mtilde(r1) for some r1>r0. Since u is only assumed positive in R\\B_{r0}, the infimum over the full ball B_{r1} may be zero, even if u is positive on every far annulus. This lower bound is load-bearing: it is used in Theorem 3.2 to rule out m(r)->0 in (3.31). The authors need either to assume positivity on all of R, as the theorem statements suggest, or to derive the positive lower bound from positivity in a fixed exterior annulus via a Harnack or barrier argument.","section":"§3, Lemma 3.3, Lemma 3.4 and proof of Theorem 3.2 after (3.30)"},{"comment":"The same interior-positivity issue appears in the lower bound of Lemma 4.5: c(r) is defined using inf_{B_r}u, which may vanish if u is only positive in the exterior domain. Moreover, the choice c(r)=C_gamma/(4C_9) inf_{B_r}u does not by itself guarantee c(r)*Psi_gamma <= u in B_r, because on B_1 the function Psi_gamma equals 1+2C_9/C_gamma and the product of c(r) with this factor may exceed inf_{B_r}u. The lower bound cmin*r^{-n+2s} is used in (4.29)-(4.30) and in the final contradiction of Theorem 4.1, so this needs to be fixed.","section":"§4, Lemma 4.5, equations (4.12)-(4.16) and Theorem 4.1, (4.29)-(4.30)"},{"comment":"The assertion 'Now by (3.38), we have, for all r>r0, eta_tilde(r)=:C0' is not immediate. The function eta_tilde(r) is an infimum over a shrinking exterior domain and is nondecreasing in r; (3.38) alone gives only a global lower bound for u/(Phi_tilde-1), not constancy of the infimum. One can recover constancy by observing that a minimizing sequence for eta_tilde(r) escapes to infinity and therefore also lies eventually in R\\B_s for every fixed s, but this argument is not supplied. Since the final contradiction relies on eta_tilde(r) being independent of r, the claim needs to be stated and proved explicitly.","section":"§3, proof of Theorem 3.2, paragraph containing (3.38)"}],"minor_comments":[{"comment":"There are several typos and small presentation issues: 'the the nonlocal frame-work', 'L ´Hospitals rule', 'Véron' without accent, and 'larger0' with a missing space. These should be corrected.","section":"Throughout"},{"comment":"The constant 1/2 and the factor |B_2\\B_1| are written inconsistently between (4.24) and (4.25); the displayed coefficient 2/\\bar c in (4.25) appears to omit the volume factor. The constants should be written consistently.","section":"§4, equations (4.24)-(4.25)"},{"comment":"The statement says the constants depend on s, u and r0, but the lower-bound constants obtained from mtilde(r1) depend on the particular choice of r1; this should be clarified.","section":"§3, Lemma 3.3 and Lemma 3.4"},{"comment":"In the second part of Lemma 4.5, the argument requires r>3r0 for the fixed annulus B_{3r0}\\B_{2r0} to lie inside B_r, whereas the text only says 'sufficiently large r>2r0'. A minor adjustment of the radius condition is needed.","section":"§4, Lemma 4.5"},{"comment":"The proof of Proposition 2.6 passes to the limit in the weak formulation and uses regularity estimates, but the assumptions under which the approximating sequence exists for general continuous supersolutions are not fully stated. This is connected to the need for a precise solution class.","section":"§2, Proposition 2.6"}],"recommendation":"major_revision","confidential_remarks":"I think the paper addresses an interesting and likely correct set of results, and the overall strategy is coherent. However, the comparison-principle gap and the sign estimate in Lemma 3.4 are genuine obstacles that affect central claims, and the lower-bound proofs have a hidden positivity assumption. These are fixable in principle, but they require substantial rewriting rather than small corrections. I therefore recommend major revision rather than rejection, provided the authors supply the missing approximation/limiting arguments and correct the estimates."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: the n>2s part of this paper looks solid, and the n=1, s>1/2 result is genuinely new. But the s=1/2 branch has a concrete error, and the comparison principle is applied to unbounded barriers without the needed limiting argument. The paper deserves a serious referee, but it needs a major revision before it's a complete proof.\n\nWhat's new: Theorems 4.1 and 4.3 give exterior-domain Liouville theorems for fractional inequalities under a condition near t=0, with the expected critical exponent n/(n-2s). That is a natural extension of Felmer-Quaas and Chen-Wang-Hajaiej, and the proof follows the standard barrier/comparison pattern. The genuinely new piece is Theorem 3.2 for n=1, s in [1/2,1), where the condition is on f near infinity (power-like for s>1/2, exponential for s=1/2). I haven't seen this in the literature, and it's plausible.\n\nWhere it's soft: Lemma 3.4, which handles s=1/2, has a sign error. In (3.21) the integral is negative (since -log|x-y|<0 for |x-y|>1), and the chain \"≤ then ≥\" is invalid. The bound that makes (-Delta)^s(m+eps w)<0 doesn't follow. This is load-bearing for the s=1/2 case, so that branch is unproven as written. The same lemma, and also Lemma 3.3 and Theorem 3.2, compare u with functions that are unbounded at infinity (m+eps w, m+eps Phi, eta*(Phi-1)) using Theorem 2.3, which is stated only for bounded domains. No approximation or truncation argument is supplied. This may be fixable, but it's not a small omission.\n\nAlso: the paper explicitly leaves the notion of solution vague (Remark 2.9), which is fine in principle, but it means the comparison principle needs to be stated in the generality in which it is used.\n\nThe citation pattern is fine, and the quantitative maximum principle tools are used appropriately. The n>2s case is likely correct and would be publishable on its own.\n\nBottom line: send it to peer review. A good referee should ask for a corrected Lemma 3.4 and a limiting argument for the comparison steps. If those are delivered, this becomes a solid contribution.","headline":"A serious paper with a real gap: the n>2s results are probably right, but the s=1/2 branch of Theorem 3.2 has a sign error and the comparison principle is used without a limiting argument.","tokens_in":25368,"tokens_out":8417,"would_cite":false,"duration_ms":82583,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35R11","35R09","35B53","35D30","35D40"],"pacs":[],"model":"deepseek-v4-flash","headline":"The fractional semilinear inequality admits no positive supersolutions in exterior domains under local conditions on the nonlinearity.","keywords":["Liouville theorem","exterior domain","semilinear inequality","fractional Laplacian","nonexistence","supersolution","fundamental solution","comparison principle"],"falsifier":"Run the comparison step (3.17)--(3.22) explicitly for the barrier $m(r)+\\epsilon(-\\log|x|)$ in the case $s=1/2$: if one can exhibit a continuous positive $u$ with $(-\\Delta)^{1/2}u\\ge 0$ in $\\mathbb{R}\\setminus B_{r_0}$ for which the inequality at infinity fails in the required limiting sense, the proof of the upper bound $m(r)\\le C\\log r$ collapses and the $s=1/2$ branch of Theorem 3.2 is refuted. Equivalently, any explicit continuous positive supersolution of $(-\\Delta)^{1/2}u\\ge g(u)$ in an exterior domain with $g$ satisfying the exponential condition (3.3) would disprove the theorem.","tokens_in":24317,"feed_emoji":"🧮","tokens_out":6737,"duration_ms":64382,"temperature":0.7,"pith_summary":"The paper aims to prove Liouville-type nonexistence theorems for the fractional semilinear inequality $(-\\Delta)^s u \\ge f(u,x)$ in exterior domains $\\mathbb{R}^n \\setminus B_{r_0}$, for all $n\\ge 1$ and $s\\in(0,1)$. The central claim is that no positive continuous supersolution can exist once $f$ obeys only local conditions: when $n>2s$, only the behavior of $f(t,x)$ near $t=0$ matters, and when $n\\le 2s$, only the behavior as $t\\to+\\infty$ matters. This replaces global assumptions on the nonlinearity and covers all dimensions, including the previously untreated line case $n=1$, $s\\ge 1/2$, with the whole-space case included as a special case. If the theorems are correct, the decisive object is the fundamental solution of the fractional Laplacian, whose exponent $\\sigma_*=-n+2s$ changes sign exactly at the critical split $2s=n$.","feed_headline":"No positive supersolutions for the fractional semilinear inequality","feed_subtitle":"Only local behavior of the nonlinearity matters: near zero when n>2s, near infinity when n<=2s, in every dimension.","key_machinery":"The engine is the fundamental solution $\\Phi(x)=|x|^{\\sigma_*}$, $-\\log|x|$, or $-|x|^{\\sigma_*}$ according as $\\sigma_*=-n+2s<0$, $=0$, or $>0$, together with carefully cut versions of it supported on annuli. These cut fundamental solutions are used as barriers that are subsolutions for $(-\\Delta)^s$ on the exterior region and small on the inner ball; the quantitative strong maximum principle (Lemmas 2.4 through 2.8) then converts pointwise lower bounds on annuli into growth information on $m(r)=\\inf_{B_{2r}\\setminus B_r}u$, and the comparison principle (Theorem 2.3) is applied to force a contradiction between the lower and upper bounds on $m(r)$ and the behavior of $f$.","core_discovery":"In the authors' own formulation, the main discovery is Theorem 3.2 for $2s\\ge n$ (that is, $n=1$, $s\\in[1/2,1)$) and Theorems 4.1 and 4.3 for $2s<n$, $n\\ge 1$: under hypotheses (f1')--(f3') or (f1)--(f2)/(f2')--(f4'), the inequality $(-\\Delta)^s u\\ge f(u,x)$ has no continuous positive solution in any exterior domain. For the model nonlinearity $f(t,x)=|x|^{-\\gamma}g(t)$ with $\\gamma<2s$, the nonexistence conditions become $\\liminf_{t\\to+\\infty} e^{bt}g(t)>0$ for every $b>0$ when $n=1$, $s=1/2$; $\\liminf_{t\\to+\\infty} t^{-\\tilde\\alpha_*}g(t)>0$, with $\\tilde\\alpha_*=1+(2s-\\gamma)/(-\\sigma_*)$, when $n=1$, $s>1/2$; and $\\liminf_{t\\to 0} t^{-\\tilde\\alpha_*}g(t)>0$ when $n>2s$. The theorems are stated so that the conclusion holds for classical, weak, and viscosity notions of supersolution.","pith_inferences":["If the dichotomy is sharp, analogous nonexistence thresholds should hold for fractional Hardy operators and for systems of fractional inequalities, with the same local conditions near zero or infinity replacing global growth assumptions.","A natural test is whether the exponent $\\tilde\\alpha_*$ is optimal: for $f(t)=t^p$ it predicts the threshold $p\\ge 1/(1-2s)$ on the line $n=1$, $s>1/2$, a concrete number that separate computations could probe.","For $s=1/2$, the condition that $g$ not decay faster than any exponential suggests the borderline may be controlled by a Laplace-transform type condition on $1/g$, which would require new estimates beyond the paper."],"forward_implications":["The critical split $2s=n$ is forced by the fundamental solution: for $n>2s$ the fundamental solution decays at infinity and the nonlinearity is tested near $t=0$, while for $n\\le 2s$ it grows or is logarithmic at infinity and the nonlinearity is tested near $t=+\\infty$.","For power-type nonlinearities the results recover and extend the known critical exponent $p=n/(n-2s)$ in dimensions $n>2s$, and they provide a new nonexistence statement on the line $n=1$, $s\\ge 1/2$.","Because the only ingredients are comparison and maximum principles, the conclusion applies to classical, weak, and viscosity supersolutions and to inequalities interpreted in both divergence and nondivergence forms.","Taking the deleted ball arbitrarily small shows that the whole-space Liouville theorem is a special case of the exterior-domain result."],"supporting_citations":[{"why":"Supplies the exterior-domain strategy of proving nonexistence by comparing infima over annuli and the local analogues of the quantitative maximum principle.","marker":"[3]"},{"why":"Provides the local quantitative maximum principle that the fractional version in Lemma 2.4 extends.","marker":"[9]"},{"why":"Introduces cutting the fundamental solution of the fractional Laplacian, the barrier construction the paper adapts to exterior domains.","marker":"[17]"},{"why":"Supplies the weak formulation and integration-by-parts formula used to compare solutions against cut-off fundamental solutions.","marker":"[10]"},{"why":"Provides the comparison principle and the weak-viscosity equivalence that justify the maximum-principle steps for different solution notions.","marker":"[32]"},{"why":"Supplies the fractional Hopf lemma and boundary regularity estimates used in the quantitative maximum principle.","marker":"[29]"}],"fun_headline_variants":["No positive supersolutions for fractional elliptic inequalities","Semilinear fractional inequality rules out positive supersolutions","Positive supersolutions absent for fractional elliptic inequalities","Fractional inequalities: no positive supersolutions in exterior domains"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the comparison principle, stated for bounded domains, can be applied to barriers that are unbounded at infinity, such as $m(r)+\\epsilon(-\\log|x|)$ and $\\eta(\\tilde\\Phi-1)$; the required limiting argument at infinity is not written out, and if those comparisons fail for merely continuous positive supersolutions the $s=1/2$ branch of Theorem 3.2 would not follow.","fun_headline_variants_meta":{"raw":{"variants":["No positive supersolutions for fractional elliptic inequalities","Semilinear fractional inequality rules out positive supersolutions","Positive supersolutions absent for fractional elliptic inequalities","Fractional inequalities: no positive supersolutions in exterior domains"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000287,"raw_usage":{"total_tokens":1655,"prompt_tokens":886,"completion_tokens":769,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":502,"completion_tokens_details":{"reasoning_tokens":710}},"tokens_in":502,"tokens_out":769,"duration_ms":7847,"temperature":1.0,"reasoning_tokens":710,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T19:22:56.564689+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the comparison step (3.17)--(3.22) explicitly for the barrier $m(r)+\\epsilon(-\\log|x|)$ in the case $s=1/2$: if one can exhibit a continuous positive $u$ with $(-\\Delta)^{1/2}u\\ge 0$ in $\\mathbb{R}\\setminus B_{r_0}$ for which the inequality at infinity fails in the required limiting sense, the proof of the upper bound $m(r)\\le C\\log r$ collapses and the $s=1/2$ branch of Theorem 3.2 is refuted. Equivalently, any explicit continuous positive supersolution of $(-\\Delta)^{1/2}u\\ge g(u)$ in an exterior domain with $g$ satisfying the exponential condition (3.3) would disprove the theorem.","supporting_citations":[{"cited_title":"Armstrong, S","cited_arxiv_id":null,"evidence_quote":"Supplies the exterior-domain strategy of proving nonexistence by comparing infima over annuli and the local analogues of the quantitative maximum principle."},{"cited_title":"Brezis, X","cited_arxiv_id":null,"evidence_quote":"Provides the local quantitative maximum principle that the fractional version in Lemma 2.4 extends."},{"cited_title":"Felmer, A","cited_arxiv_id":null,"evidence_quote":"Introduces cutting the fundamental solution of the fractional Laplacian, the barrier construction the paper adapts to exterior domains."},{"cited_title":"Bucur, E","cited_arxiv_id":null,"evidence_quote":"Supplies the weak formulation and integration-by-parts formula used to compare solutions against cut-off fundamental solutions."},{"cited_title":"Servadei, E","cited_arxiv_id":null,"evidence_quote":"Provides the comparison principle and the weak-viscosity equivalence that justify the maximum-principle steps for different solution notions."},{"cited_title":"Ros-Oton, Nonlocal elliptic equations in bounded domains: a survey, Publ","cited_arxiv_id":null,"evidence_quote":"Supplies the fractional Hopf lemma and boundary regularity estimates used in the quantitative maximum principle."}],"review_version":1}