{"id":"63574c3a-6b43-4cb7-b1c4-7ce931115cb2","arxiv_id":"1908.02970","paper_version":1,"verdict":"REJECT","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"high","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper attempts to prove existence and local uniqueness of multi-peak solutions for the logarithmic Schrödinger equation via Lyapunov-Schmidt reduction, but relies on an invalid Green's function estimate.","lead":"This paper claims to build multi-peak solutions for a logarithmic Schrödinger equation using Lyapunov-Schmidt reduction with a new weighted norm. The key convolution estimate behind the argument appears to be false, which breaks the proof as written.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The key convolution estimate (3.5) is false: the Newtonian potential of a Gaussian decays algebraically, not exponentially, so the *-norm contraction in Proposition 3.2 does not close.","rationale":"The reader identified exactly the same weak point, and my independent check confirms it. The scaling of (3.5) is qualitatively wrong: a Gaussian source is not compactly supported, and its Newtonian potential decays like ε^N/|x-y_j|^{N-2}, not exponentially. This is not a missing constant but a reversal of the decay mechanism, and it enters precisely where the weighted *-norm is used to close the contraction mapping. Since Proposition 3.2 is needed for the reduction step in Theorem 1.1, the submitted proof does not establish the main existence result. No independent verification, machine-checked proof, or reproducible code is present. The local uniqueness section assumes solutions of the type constructed and inherits the same reduction estimates, so it does not rescue the paper. I see no reason to change the reader's REJECT verdict; the appropriate recommendation relative to the reader is UNCHANGED.","tokens_in":24740,"tokens_out":8503,"duration_ms":94618,"concrete_test":"Evaluate (3.5) numerically for N=3, ε=10^-2, y_j=0, x=(1,0,0). Compute I=∫ e^{-|z|^2/(2ε^2)}/(4π|z-x|) dz. Asymptotically I≈√(π/2) ε^3 ≈ 1.25×10^-6, while the bound claimed in (3.5) is of order ε^4 e^{-1/(2ε^2)} ≈ 10^-2180. Since I exceeds the asserted bound by many orders of magnitude, the estimate defining Proposition 3.2 is false. A second check is to redo the contraction argument with the correct algebraic bound ε^N |x-y_j|^{2-N}; the resulting ‖u1‖_* term is not O(R^{-2}) but contains R^{2-N}e^{R^2/2}, which cannot be made small.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing step is (3.5) in Proposition 3.2, where the authors assert that for w(x)=ε^4|x-y_j|^{-2}e^{-|x-y_j|^2/(2ε^2)} one has −Δw ≥ C e^{-|x-y_j|^2/(2ε^2)} and hence ∫ e^{-|z-y_j|^2/(2ε^2)}/|z-x|^{N-2} dz ≤ C w(x). This is incorrect. For |x-y_j| ≫ ε, the Newtonian potential of a Gaussian has the algebraic asymptotic C_N ε^N |x-y_j|^{2-N}, whereas w is exponentially small in |x-y_j|^2/(2ε^2). At |x-y_j|=Rε, the ratio of the true convolution to the weight ∑_j e^{-|x-y_j|^2/(2ε^2)} is of order R^{2-N}e^{R^2/2}, which is unbounded as R→∞. Thus the estimates of u1 in (3.6)–(3.11), and the analogous bounds for u2 and u5, do not follow; Proposition 3.2's conclusion ‖u‖_* ≤ 1/|ln ε|^{1−θ} is unsupported. This is exactly the point where the new *-norm is supposed to close the Lyapunov–Schmidt reduction, so Theorem 1.1 lacks a valid proof. The failure is in the main argument, not a minor typo.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript studies the singularly perturbed logarithmic Schrödinger equation (1.1) in R^N, N≥3, with a bounded positive potential V satisfying (V1) and having k non-degenerate critical points (V2). The authors propose to construct multi-peak positive solutions by Lyapunov-Schmidt reduction, using the unique positive solution U of the limiting equation as building blocks and introducing a weighted sup-norm ||·||_* (defined in (1.8)) to control the remainder R_ε(φ), whose standard H^1 estimate fails because of the logarithmic nonlinearity. The main existence theorem (Theorem 1.1) asserts the existence of k-peak solutions concentrating at the critical points for small ε. A second theorem (Theorem 1.2) asserts local uniqueness of such positive solutions via local Pohozaev identities. The paper contains the standard two-step reduction: invertibility of the linearized operator (Proposition 2.1), estimates of the error terms (Lemmas 2.2–2.3 and 3.1), a contraction-mapping fixed point (Proposition 3.3), and a finite-dimensional reduction using Pohozaev identities to locate the peaks (Theorem 4.1). Section 5 then proves local uniqueness by comparing two solutions through their normalized difference.","tokens_in":25101,"tokens_out":13201,"duration_ms":133882,"significance":"If the results were correct, they would provide the first Lyapunov-Schmidt reduction for a logarithmic Schrödinger equation, together with a local uniqueness theorem for multi-peak solutions, a significant advance over purely variational existence results. The authors correctly identify the main technical difficulty (the non-Lipschitz nature of the logarithmic term) and propose an interesting weighted norm to address it. The paper is ambitious, the structure is clear, and the use of local Pohozaev identities for the finite-dimensional reduction is well motivated. However, the central analytical estimate on which the new norm is designed to work is false, so the main theorem is not established.","major_comments":[{"comment":"The convolution estimate (3.5) is false. For w(x)=ε^4 |x-y_j|^{-2} e^{-|x-y_j|^2/(2ε^2)}, the claim that -Δw ≥ C e^{-|x-y_j|^2/(2ε^2)} and hence ∫ e^{-|z-y_j|^2/(2ε^2)} |z-x|^{2-N} dz ≤ C w(x) is incorrect. For |x-y_j| ≫ ε, the Newtonian potential of a Gaussian has the algebraic asymptotic C_N ε^N |x-y_j|^{2-N}, whereas w decays exponentially in |x-y_j|^2/(2ε^2). At |x-y_j| = Rε the ratio of the true convolution to w is of order R^{2-N} e^{R^2/2}, which is unbounded as R→∞. Consequently, the bounds on u1 in (3.6)–(3.11), on u2 in (3.12)–(3.16), on u3 in (3.17)–(3.19), and on u5 in (3.21)–(3.27) do not follow. Proposition 3.2's conclusion ||u||_* ≤ |ln ε|^{-(1-θ)} is therefore unsupported, and the contraction-mapping argument in Proposition 3.3, which relies on Proposition 3.2, collapses. This is a load-bearing error in the proof of Theorem 1.1.","section":"Section 3, Eq. (3.5)"},{"comment":"Formula (2.7) does not define a projection onto E_{ε,y}. For a general u∈H^1, u - Σ_{j,i} ⟨u, ∂U_{ε,y_j}/∂x_i⟩ ∂U_{ε,y_j}/∂x_i is not orthogonal to span{∂U_{ε,y_j}/∂x_i} unless the set {∂U/∂x_i} is orthonormal, which it is not (the Gram matrix has off-diagonal entries of order ε^N). The proof of Proposition 2.1 uses the identity ⟨P_ε L φ, ψ⟩ = ⟨L φ, ψ⟩ for ψ∈E_{ε,y}; this identity does hold with the stated formula for ψ∈E because ⟨∂U_i, ψ⟩ = 0, but P_ε u does not lie in E_{ε,y}, so the reduction equation (3.1) is not an equation in E_{ε,y}. The argument can likely be repaired by using the orthogonal projection with the inverse Gram matrix, but as written this is a gap in the foundational invertibility result on which the whole reduction rests.","section":"Section 2, Eq. (2.7) and Proposition 2.1"},{"comment":"The proof of Theorem 1.2 assumes without proof that every positive solution u_ε^{(i)} concentrated at ξ_1,...,ξ_k has the form u_ε^{(i)} = Σ_{j=1}^k U_{ε,y_{ε,j}^{(i)}} + φ_ε^{(i)} with the estimates from Lemma 5.4 and Proposition 5.3. Proposition 5.3 is proved only for solutions produced by the Lyapunov-Schmidt construction in Theorem 4.1. For the local-uniqueness claim, one must first establish such an expansion for arbitrary positive solutions concentrated at the given points; no argument or citation is provided. Thus Theorem 1.2 is not supported even if the expansion for the constructed solutions were valid.","section":"Section 5, proof of Theorem 1.2"}],"minor_comments":[{"comment":"The title and abstract contain several typographical errors (e.g., 'Schrodinger' for 'Schrödinger', 'Lyap unov' for 'Lyapunov', 'S chrodinger'); these should be corrected.","section":"Title and Abstract"},{"comment":"The notation ⟨R(φ),η⟩_ε is inconsistent with the definition of the H_ε inner product; the estimate in the proof is for ∫ R(φ)η. This should be clarified, as the symbol ⟨·,·⟩_ε usually denotes the H_ε inner product introduced earlier.","section":"Section 2, Lemma 2.3"},{"comment":"The estimate in (3.24) is ambiguous: 'O( ε/R^{N-2} e^{R^2} )' should specify whether the exponential factor is in the numerator or denominator, and the power of R in the denominator should be checked against the computation.","section":"Section 3, Eq. (3.24)"},{"comment":"The expansion (5.14) for U_{ε,y^{(1)}} - U_{ε,y^{(2)}} is schematic; a rigorous statement would use the mean value theorem with explicit bounds on the second derivatives of U. As written, the factor (y^{(1)}-y^{(2)})/ε is correctly identified as o(1) by (5.2), but the pointwise validity of the displayed formula in the support of the gradient should be justified.","section":"Section 5, Eq. (5.14)"}],"recommendation":"reject","confidential_remarks":"The manuscript's main theorem depends on a false convolution estimate (3.5), which is not a minor fix because the exponential weight in the *-norm cannot dominate the algebraic decay of the Newtonian potential. In addition, the local uniqueness theorem assumes an expansion for arbitrary solutions that is not proved. Substantial revision would be needed to make the argument sound."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Here's the short version: Theorems 1.1 and 1.2 are not proved. The load-bearing step is the convolution estimate (3.5) in Proposition 3.2, and it is wrong. The Newtonian potential of a Gaussian does not decay exponentially; for |x−y_j| ≫ ε it behaves like C_N ε^N / |x−y_j|^{N−2}, while the barrier w(x)= ε^4 |x−y_j|^{-2} e^{−|x−y_j|^2/(2ε^2)} is exponentially small. So the ratio LHS/w is unbounded. In fact the pointwise inequality −Δw ≥ C e^{−|x−y_j|^2/(2ε^2)} is itself false at large distances because −Δw turns negative. Either way, the estimates on u1 and u5 in Proposition 3.2 do not follow, the contraction in the weighted *-norm doesn't close, and the existence theorem collapses. Since Theorem 1.2 assumes those solutions, it goes down with it.\n\nWhat is genuinely new: this is indeed the first Lyapunov–Schmidt reduction attempt for the logarithmic Schrödinger equation. The weighted norm ||·||_* is a reasonable idea—the log nonlinearity does create a singularity when the base concentration sum is small, and some weighted sup norm is natural. The use of local Pohozaev identities in Section 5 to avoid a global variational argument is standard in the modern literature but sensibly applied. The invertibility proof for P_ε L_ε and the estimates on l_ε and R_ε in Section 2 look standard and plausible.\n\nWhere the soft spots are, in proportion: the one fatal flaw is the estimate above. It is not a typo; the claimed inequality is used in exactly the place the new norm is supposed to work. Away from that, I have concerns about the hand-waving in Lemma 5.2 and the maximum-principle argument at the end of Theorem 1.2, but those are secondary. The citation pattern is fair; the paper cites the relevant variational, penalization, and local-Pohozaev literature.\n\nWho this is for: an expert in singularly perturbed elliptic PDEs or logarithmic Schrödinger equations. It shows the shape of a potential reduction proof, but as submitted it is not a reliable source.\n\nRecommendation: I would not desk-reject—the claim of a first reduction is important and the error is specific enough to be checkable. Send it to a referee with strong harmonic analysis background. My own recommendation to the editor would be major revision at best; the convolution estimate is central and needs a genuinely different argument, not a tweak.","headline":"The first reduction-based attack on logarithmic Schrödinger multi-peak existence has a false convolution estimate at its core, so the main theorems are unproven; still worth a referee's time.","tokens_in":25551,"tokens_out":5454,"would_cite":false,"duration_ms":53552,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35B25","35J10","35J60"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper proves that for any finite set of non-degenerate critical points of the potential, the logarithmic Schrödinger equation admits positive k-peak solutions for sufficiently small ε, and any two such solutions concentrating at the…","keywords":["logarithmic Schrödinger equation","multi-peak solutions","Lyapunov–Schmidt reduction","weighted norm","local Pohozaev identities","local uniqueness","singular perturbation"],"falsifier":"Compute the integral $I(x)=\\int_{\\mathbb{R}^N}|z-x|^{2-N}e^{-|z-y|^2/(2\\varepsilon^2)}\\,dz$ for $|x-y|\\gg\\varepsilon$. For $N\\ge 3$ it behaves like $(2\\pi\\varepsilon^2)^{N/2}|x-y|^{2-N}$, an algebraic decay, not the exponential barrier $\\varepsilon^4|x-y|^{-2}e^{-|x-y|^2/(2\\varepsilon^2)}$ used in estimate (3.5). The contradiction would show that the bounds on $u_1$ and $u_2$ in Proposition 3.2 do not follow from the stated argument.","tokens_in":24559,"feed_emoji":"⚛️","tokens_out":10825,"duration_ms":114691,"temperature":0.7,"pith_summary":"The paper proves that the singularly perturbed logarithmic Schrödinger equation $-\\varepsilon^2\\Delta u + V(x)u = u \\log u^2$ has positive multi-peak solutions concentrated at any prescribed non-degenerate critical points of $V$, for $\\varepsilon$ small. This is the first Lyapunov–Schmidt reduction for a logarithmic Schrödinger equation; the difficulty is that the logarithmic nonlinearity makes the variational functional fail to be $C^1$, and the usual remainder estimate fails in $H^1$. The authors introduce a weighted sup norm, the star norm, that measures a correction $\\phi$ against the sum of the Gaussian-shaped bumps used as building blocks, and run the contraction argument in that norm. They then use local Pohozaev identities to select the peak locations and to obtain local uniqueness. The result is a general reduction scheme for logarithmic nonlinearities and a uniqueness theorem for multi-peak bound states.","feed_headline":"New weighted norm yields multi-peak logarithmic Schrödinger solutions","feed_subtitle":"A log-singular equation that resisted reduction now falls to Lyapunov–Schmidt plus local Pohozaev identities.","key_machinery":"The load-bearing object is the weighted norm $\\|\\phi\\|_*=\\sup_x (\\sum_{j=1}^k e^{-|x-y_j|^2/(2\\varepsilon^2)})^{-1}|\\phi(x)|$. It replaces the ordinary $H^1$ norm as the metric for the contraction step, making the nonlinear remainder $R_\\varepsilon(\\phi)$ contractive, $\\|R_\\varepsilon(\\phi)\\|_*=O(\\|\\phi\\|_*^2)$, which is false for the $H^1$ norm. Around this sit four standard pieces: the explicit ground state $U(x)=e^{(\\omega+N-|x|^2)/2}$ and the non-degeneracy of its linearization, whose kernel is spanned by $\\partial U/\\partial x_j$; the projection $P_\\varepsilon$ onto the orthogonal complement of that kernel; the invertibility of $P_\\varepsilon L_\\varepsilon$ with a uniform bound; and a family of local Pohozaev identities that convert the vanishing of the finite-dimensional Lagrange multipliers into the equations $\\nabla V(y_j)=O(|\\log\\varepsilon|^{-(1-\\theta)})$, solvable because $V$ has non-degenerate critical points. The same identities, combined with a blow-up analysis and the maximum principle, drive the uniqueness proof.","core_discovery":"The central claim is Theorem 1.1: for $N\\ge 3$ and a potential $V$ satisfying (V1) and (V2), equation (1.1) has, for sufficiently small $\\varepsilon$, a positive $k$-peak solution whose peaks tend to the non-degenerate critical points $\\xi_1,\\dots,\\xi_k$, and Theorem 1.2: any two such positive solutions concentrating at the same points coincide for small $\\varepsilon$. The construction writes the solution as $\\sum_{j=1}^k U_{\\varepsilon,y_j}+\\phi$, where each $U_{\\varepsilon,y_j}$ is the explicit Gaussian-type solution of the rescaled limiting equation, and solves for $\\phi$ in a subspace orthogonal to the approximate kernel. The paper also proves sharp concentration information: $|y_{\\varepsilon,j}-\\xi_j|=o(\\varepsilon)$ and $\\|\\phi_\\varepsilon\\|_\\varepsilon=O(\\varepsilon^{N/2+2})$.","pith_inferences":["The same weighted-norm reduction should extend to logarithmic equations with fractional Laplacians or logarithmic Schrödinger–Newton systems, provided the convolution kernel estimate is re-derived for the new Green's function.","One could test the predicted $o(\\varepsilon)$ peak-location rate numerically in one dimension by continuing the two solutions in $\\varepsilon$; agreement would support the local-uniqueness conclusion.","The Pohozaev treatment of the finite-dimensional problem suggests the existence result may persist for potentials whose critical points are saddles of any index, a regime where minimization arguments fail."],"forward_implications":["For any finite set of non-degenerate critical points of $V$, equation (1.1) admits a positive solution with a peak near each point for all sufficiently small $\\varepsilon$.","In the single-peak case with a strict global minimum point, the ground state of the equation is unique.","The Lyapunov–Schmidt reduction is now a viable tool for logarithmic nonlinearities, not just polynomial or saturable ones, once the correction is measured in the weighted norm.","The $o(\\varepsilon)$ location of the peaks shows the constructed solutions are not merely abstract limits but concentrate with sharp precision."],"supporting_citations":[{"why":"Supplies the unique positive solution $U$ of the limiting equation and the non-degeneracy of its linearization; this is the building block for the approximate solutions.","marker":"[14]"},{"why":"Origin of the local Pohozaev identity technique used to choose the peak points $y_j$ and to handle non-degenerate critical points of $V$.","marker":"[22]"},{"why":"Gives the blow-up and maximum-principle framework that the uniqueness proof adapts to prove Theorem 1.2.","marker":"[8]"},{"why":"The classical Lyapunov–Schmidt reduction whose failure for logarithmic nonlinearities motivates the new weighted norm.","marker":"[5]"}],"fun_headline_variants":["New norm enables first Lyapunov-Schmidt reduction for log Schrödinger","Multi-peak solutions for logarithmic Schrödinger via reduction and Pohozaev","Weighted norm overcomes log singular term in Schrödinger multi-peak proof","First reduction construction for singularly perturbed logarithmic Schrödinger","Local Pohozaev identities prove uniqueness of multi-peak log solutions"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof rests on the assumption that the influence of each Gaussian-shaped bump on far-away points dies off exponentially fast; if that convolution estimate fails, the weighted contraction argument that produces the correction term does not close.","fun_headline_variants_meta":{"raw":{"variants":["New norm enables first Lyapunov-Schmidt reduction for log Schrödinger","Multi-peak solutions for logarithmic Schrödinger via reduction and Pohozaev","Weighted norm overcomes log singular term in Schrödinger multi-peak proof","First reduction construction for singularly perturbed logarithmic Schrödinger","Local Pohozaev identities prove uniqueness of multi-peak log solutions"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000246,"raw_usage":{"total_tokens":1514,"prompt_tokens":894,"completion_tokens":620,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":510,"completion_tokens_details":{"reasoning_tokens":525}},"tokens_in":510,"tokens_out":620,"duration_ms":6275,"temperature":1.0,"reasoning_tokens":525,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:31:46.463580+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the integral $I(x)=\\int_{\\mathbb{R}^N}|z-x|^{2-N}e^{-|z-y|^2/(2\\varepsilon^2)}\\,dz$ for $|x-y|\\gg\\varepsilon$. For $N\\ge 3$ it behaves like $(2\\pi\\varepsilon^2)^{N/2}|x-y|^{2-N}$, an algebraic decay, not the exponential barrier $\\varepsilon^4|x-y|^{-2}e^{-|x-y|^2/(2\\varepsilon^2)}$ used in estimate (3.5). The contradiction would show that the bounds on $u_1$ and $u_2$ in Proposition 3.2 do not follow from the stated argument.","supporting_citations":[{"cited_title":"d’A venia, E","cited_arxiv_id":null,"evidence_quote":"Supplies the unique positive solution $U$ of the limiting equation and the non-degeneracy of its linearization; this is the building block for the approximate solutions."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Origin of the local Pohozaev identity technique used to choose the peak points $y_j$ and to handle non-degenerate critical points of $V$."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the blow-up and maximum-principle framework that the uniqueness proof adapts to prove Theorem 1.2."},{"cited_title":"Bahri, Critical Points at Infinity in Some Variational Problems","cited_arxiv_id":null,"evidence_quote":"The classical Lyapunov–Schmidt reduction whose failure for logarithmic nonlinearities motivates the new weighted norm."}],"review_version":1}