{"id":"7d9f1fd6-6c37-4531-8582-e35aa75b69d9","arxiv_id":"1908.09153","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"The logarithmic Schrödinger equation with a deepening potential well with k wells has at least 2^k-1 positive multi-bump solutions for sufficiently large λ.","lead":"Under a potential with several separate wells where the potential vanishes, the paper proves that the logarithmic Schrödinger equation has positive multi-bump solutions for large depth parameter, one solution for every nonempty combination of wells. The result yields at least 2^k-1 positive solutions and shows each cluster converges to a least energy state on its chosen wells as the depth grows.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Central multi-bump construction depends on unproved minimax-level convergence (Cor. 2.4), borrowed from a different nonlinearity; without it Proposition 2.4 cannot force the prescribed bump profile.","rationale":"The reader's weakest assumption identifies the imported properties of F1,F2 and the omitted Lemmas 2.8, Corollary 2.4, Lemmas 3.1 and 3.2. My review narrows this to the most load-bearing item: Corollary 2.4(b), the uniform convergence of the special minimax level bλ,R,Γ → cΓ. This is not a cosmetic omission. Proposition 2.4's deformation argument is built entirely on it: without bλ,R,Γ near cΓ, the deformed path with max below cΓ does not contradict anything, and the conclusion that Φλ,R has a critical point in Aλ_μ ∩ Φ^{cΓ}_{λ,R} ∩ B^λ_{r+1} fails. That critical point is exactly what guarantees the multi-bump profile (each component nontrivial) and the energy bounds used in Section 3 to pass to the limit and identify the least-energy solution on ΩΓ. The lower-bound part of Lemma 2.8 also has a hidden subtlety: the zero point from Lemma 2.7 could in principle have zero restriction on some Ω'_j, and the omitted proof must rule this out. Since the paper cites [1] for these statements, the argument is plausible but not independently verifiable from the text; the correct verdict remains CONDITIONAL, matching the reader. I do not see a fatal flaw, and the central claim is consistent with the literature; the required fix is a written proof of the borrowed minimax-level estimates.","tokens_in":20523,"tokens_out":12002,"duration_ms":124908,"concrete_test":"Re-derive Lemma 2.8 and Corollary 2.4 in full for the logarithmic truncation F1,F2: (1) prove bλ,R,Γ ≥ Σ cλ,j by showing the point from Lemma 2.7 has nonzero restriction on each Ω'_j, using the boundary sign conditions (2.13) and Miranda's theorem; (2) prove bλ,R,Γ ≤ cΓ using γ0 and the disjoint-support decomposition of Φλ,R on Ω'_j; (3) prove cλ,j → cj and obtain bλ,R,Γ → cΓ uniformly for R large. If any step fails for the logarithmic nonlinearity, Proposition 2.4 and hence Theorem 1.1 are incomplete; if all steps go through, the concern is resolved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proof of Theorem 1.1 rests on the special minimax level bλ,R,Γ introduced in Section 2.4 and on Proposition 2.4, which produces a critical point uλ,R with nontrivial bumps on every j∈Γ. Proposition 2.4 needs a strict inequality bλ,R,Γ < cΓ for all large λ and R, and it obtains that by deforming γ0 until the maximax value drops below cΓ; the contradiction is supplied by Corollary 2.4(b), which asserts bλ,R,Γ → cΓ uniformly for R large. However, Lemma 2.8 and Corollary 2.4 are not proved in the paper: the text says 'the proof is the same as that of Proposition 4.2 in [1]' and 'similar to that of Corollary 4.3 in [1]'. The borrowed statements are formulated for a power nonlinearity, and the logarithmic modification changes the estimates. In particular, the lower bound Σ cλ,j ≤ bλ,R,Γ would fail if the zero provided by Lemma 2.7 were allowed to have trivial restriction on some Ω'_j, so the missing argument must rule that out via the boundary sign conditions (2.13). The omitted Lemmas 3.1 and 3.2 are secondary; the minimax-level convergence is the load-bearing step. Without Corollary 2.4, the existence of a critical point in Aλ_μ with the correct multi-bump profile is not established, so Theorem 1.1 is incomplete as written.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the logarithmic Schrödinger equation -Δu + λV(x)u = u log u^2 on R^N, with a nonnegative potential V whose zero set is a bounded set Ω consisting of k disjoint components Ω_1,...,Ω_k. The main result, Theorem 1.1, asserts that for every nonempty Γ ⊆ {1,...,k} and all sufficiently large λ there is a positive solution u_λ whose energy is concentrated on Ω_Γ = ∪_{j∈Γ}Ω_j, and that along λ_n→∞ a subsequence converges strongly in H^1 to a least-energy solution of the limit problem -Δu = u log u^2 on Ω_Γ with zero boundary condition. Corollary 1.1 concludes that there are at least 2^k-1 positive solutions. The proof follows the penalization approach of del Pino and Felmer: the nonlinearity is truncated near zero and modified outside Ω'_Γ, an auxiliary problem is solved on large balls B_R(0), the associated functional is shown to satisfy the (PS) condition, special minimax levels b_{λ,R,Γ} are introduced, and a deformation argument produces critical points with prescribed bumps on each Ω_j; finally, solutions are recovered for the original problem as R→∞ and then as λ→∞.","tokens_in":20897,"tokens_out":15813,"duration_ms":162577,"significance":"If the proof is completed, the result is a meaningful extension of the multi-bump results of Ding and Tanaka from power nonlinearities to the logarithmic nonlinearity, which is technically delicate because the natural energy functional is not well defined or C^1 on the whole space. The paper contains substantial original work: the penalized functional on balls, the (PS)_∞ analysis, the L^∞-bounds via Moser iteration, and the passage from balls to R^N are all carried out in detail. The main issue is completeness: several lemmas that are load-bearing for the central minimax comparison and for the convergence of solutions on expanding balls are stated without proof and referred to earlier papers, and one step in the deformation argument appears to require a stronger choice of the constant r than the one made. These gaps are fixable, but they need to be addressed before the main theorem can be considered established.","major_comments":[{"comment":"The proof of the key minimax comparison is omitted entirely. Lemma 2.8(a) asserts Σ c_{λ,j} ≤ b_{λ,R,Γ} ≤ c_Γ, and Corollary 2.4(b) asserts b_{λ,R,Γ} → c_Γ uniformly for large R; the text says that the proof is 'the same as that of Proposition 4.2 in [1]' and 'similar to that of Corollary 4.3 in [1]'. This is load-bearing: Proposition 2.4 obtains its contradiction precisely from the fact that b_{λ,R,Γ} < c_Γ for large λ,R, while Corollary 2.4(b) forces b_{λ,R,Γ} → c_Γ. Moreover, the cited reference [1] concerns a different, quasilinear-type problem, so the adaptation to the logarithmic penalized functional is not automatic; in particular, the lower bound Σ c_{λ,j} ≤ b_{λ,R,Γ} requires ruling out minimax points whose restriction to some Ω'_j is trivial, using the sign conditions in (2.13). Please provide a complete proof or a precise statement of which estimates transfer and why the logarithmic modification preserves them.","section":"§2.4, Lemma 2.8 and Corollary 2.4"},{"comment":"The deformation argument contains a gap in the treatment of the set B^λ_r. The radius r is only required to satisfy r > max{‖ω_j‖_{H^1_0(Ω_j)}}, but the paths γ_0 contain terms s_j Tω_j with T large, so γ_0([1/T^2,1]^l) need not lie in B^λ_{r/2}. Consequently, the claim in case (2)(ii) that leaving B^λ_r implies a displacement at least r from the initial point u is not justified: if u ∈ B^λ_r but is close to the boundary, a path can leave B^λ_r with arbitrarily small displacement. The proof needs a choice such as r > 2 sup_{t∈[1/T^2,1]^l} ‖γ_0(t)‖, together with a corresponding adjustment of the case analysis, in order for the energy-drop estimate (2.16) to hold. This estimate is what produces the contradiction with Corollary 2.4(b), so the gap directly affects the existence of the critical point in A^λ_{μ,R}.","section":"§2.5, Proposition 2.4"},{"comment":"Lemmas 3.1 and 3.2 are stated without proof, with the comment that they follow by arguments similar to those in Proposition 2.2. These lemmas are needed to pass from solutions u_{λ,R_n} on expanding balls to a solution u_λ on R^N, and in particular to justify the strong H^1 convergence u_{λ,n} → u_λ and the L^1 convergence of F_1(u_{λ,n}) and F'_1(u_{λ,n})u_{λ,n}. For a logarithmic nonlinearity these integrability assertions are not completely routine. Please supply the proofs or give a detailed indication of the modifications of Proposition 2.2 that establish them.","section":"§3, Lemmas 3.1 and 3.2"}],"minor_comments":[{"comment":"The paper states that the required properties of F_1 and F_2 were proved in [21] and [24], but it does not indicate the range of δ for which F_1 is convex and nonnegative. Since the argument later fixes δ small and chooses p ∈ (2,2^*) in (2.2), it would be helpful to state the precise conditions on δ and p that are being used.","section":"§2.1, definitions of F_1 and F_2"},{"comment":"In the paragraph after the definition of A^λ_{μ,R}, the notation 'w = Σ_{j=1}^l w_j ∈ A^λ_{μ,R}' is used before w_j is defined; presumably w_j = ω_j. Please clarify this notation.","section":"§2.5, notation"},{"comment":"In the proof of Proposition 2.3, the consequence drawn from (2.12) is written as '‖u‖_j^2 > τ/(2T)'. Since (2.12) states ‖u‖_j > τ, the square norm is actually > τ^2; the displayed inequality is true only after choosing T large enough so that τ^2 > τ/(2T). Please make that choice explicit.","section":"§2.5, Proposition 2.3"},{"comment":"The functional Φ_λ on E_λ is introduced only after Lemma 3.2, but it is used earlier in the definition of A^λ_μ. Please define Φ_λ before its first use and state its domain and regularity properties clearly.","section":"§3, definition of Φ_λ"}],"recommendation":"major_revision","confidential_remarks":"The central minimax lemmas are delegated to [1], which is by one of the present authors. It would be preferable for the paper to be self-contained in this respect, especially because the omitted statements are not merely technical but are what force the critical point to have the prescribed multi-bump profile. I do not see evidence of circularity or incorrectness in the parts that are proved, and the gaps appear fixable by adding the missing proofs and adjusting the choice of r in Proposition 2.4. The paper should not be accepted in its current form."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This one closes the multi-bump question for the deepening-well logarithmic Schrödinger equation, and the F1/F2 trick is a genuinely useful adaptation of the del Pino–Felmer penalization to a nonlinearity that is not o(t) at zero. The main theorem and the 2^k−1 corollary are plausible and consistent with the surrounding literature. The body is mostly full proofs: the (PS) condition, the boundedness of (PS) sequences, the L∞ estimates, and the convergence analysis are all written out. The authors also correctly identify why [3] could not get multi-bump and where the logarithmic nonlinearity breaks the standard route.\n\nThe soft spot is exactly where the reader's and stress-test notes point: Lemma 2.8 and Corollary 2.4 are stated without proof, with 'the same as Proposition 4.2 in [1]' and 'similar to Corollary 4.3 in [1].' These are not peripheral; Corollary 2.4(b) is what forces the minimax level bλ,R,Γ below cΓ, and that inequality drives Proposition 2.4's construction of a critical point in the annulus Aλ_2μ\\Aλ_μ. Since [1] is a power-nonlinearity paper, the adaptation to the F1/F2 decomposition and the logarithmic equation is not automatic. The authors need to show the estimates actually carry over, or state the modified version and prove it. Lemmas 3.1 and 3.2 are also omitted, but they are less load-bearing; still, they should be included or clearly derived.\n\nOne more thing: Theorem 1.1 says the restriction u|Ωj is a least energy solution of (P∞,Γ) — that is sloppy. The limit should have each component u|Ωj a least energy solution on Ωj (or u itself a least energy solution on the disconnected domain ΩΓ, which means the same thing up to constants). As written it could confuse a reader about the domain.\n\nThe F1/F2 decomposition is imported from [21,24] without verification for the particular constants δ, l, a0. That is acceptable if the cited papers cover it, but a one-line 'these properties hold' with references would be cleaner.\n\nWho this is for: people working on penalization methods for singular or logarithmic Schrödinger equations, and anyone needing multi-bump results for deepening wells. It deserves a serious referee; the result matters and the method is a real extension. My recommendation: send it to review, and the referee should insist that Corollary 2.4 (and ideally the exact adaptation) be proved in the revision. With that added, I would be happy with acceptance.","headline":"Solid extension of the penalization method to the log Schrödinger equation, but the load-bearing minimax estimate is borrowed from a power-nonlinearity paper without proof.","tokens_in":21409,"tokens_out":3775,"would_cite":true,"duration_ms":36911,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35A15","35J10","35B09"],"pacs":[],"model":"deepseek-v4-flash","headline":"A logarithmic Schrödinger equation with a multi-well potential admits at least 2^k−1 positive multi-bump solutions for sufficiently deep wells.","keywords":["logarithmic Schrödinger equation","multi-bump solutions","deepening potential well","variational methods","mountain pass","penalization method","Palais-Smale condition","least energy solution"],"falsifier":"Compute explicitly the functions F1 and F2 for the given δ, l, a0 and check whether F1 is convex and nonnegative for all s, and whether F2′(s)/s is nondecreasing and F2′(s)≥0 for s>0 as claimed. If any of these properties fails for the specific choices, the auxiliary problem (Mλ,R) does not have the mountain-pass geometry and the main theorem is not established. Alternatively, a numerical simulation for a concrete potential V with two wells could test the claim of existence of multi-bump solutions for large λ, checking whether positive solutions indeed concentrate on each chosen well.","tokens_in":20272,"feed_emoji":"🧮","tokens_out":1469,"duration_ms":16589,"temperature":0.7,"pith_summary":"The paper proves that a logarithmic Schrödinger equation with a deepening potential well (a nonnegative potential V whose zero set has k disjoint bounded components) has at least 2^k−1 positive solutions for large λ. These solutions concentrate on any chosen nonempty subset of the wells, with each bump resembling a least-energy solution on the corresponding component. The proof works despite the energy functional not being well-defined on unbounded domains, by first solving an auxiliary penalized problem on large balls and then passing to the limit.","feed_headline":"A log Schrödinger equation with k wells has at least 2^k−1 positive solutions","feed_subtitle":"For deep enough wells, each subset of wells supports a bump, giving exponentially many multi-bump states.","key_machinery":"The auxiliary problem (Mλ,R) on a ball BR(0), obtained by splitting the nonlinearity into F1 and F2: F2(s) − F1(s) = ½s² log s², with F1 convex, even, nonnegative and F2 modified outside Ω′Γ so that its derivative grows at most linearly. The key mechanism is the comparison of minimax levels: the mountain-pass type level bλ,R,Γ for the auxiliary functional lies between ∑cλ,j and cΓ, and converges to cΓ as λ→∞. This level control forces the solutions to have nonzero components on each chosen well and to vanish outside Γ, leading to multi-bump structure.","core_discovery":"The central claim is that for any nonempty subset Γ of {1,…,k}, there exists λ* such that for all λ ≥ λ* problem (Pλ) has a positive solution uλ. As λ tends to infinity along a sequence, a subsequence converges strongly in $H^{1}$ to a function u that vanishes outside ΩΓ and is a least-energy solution of the limit problem −Δu = u log u² on ΩΓ with Dirichlet boundary conditions. The immediate corollary is that for large λ the equation has at least 2^k−1 positive solutions, one for each nonempty subset of the wells.","pith_inferences":["A likely testable extension is to replace the logarithmic nonlinearity by a family of subcritical nonlinearities that converge to the logarithmic law, checking whether the number of bumps persists under the limit.","The exponential growth in the number of solutions suggests that similar multi-bump counting may hold for other nonlocal or fractional logarithmic Schrödinger equations, provided a suitable penalization can be constructed.","The paper's technique of proving boundedness of Palais-Smale sequences via the logarithmic Sobolev-type inequality (Lemma 2.2) might be applicable to other problems where the energy functional is not coercive in the usual sense.","The uniformity of the level convergence bλ,R,Γ → cΓ as λ→∞ could be sharpened to explicit rates, which would give quantitative information about the size of the bumps."],"forward_implications":["For any prescribed subset of the k wells, there is a positive solution whose mass is concentrated in those wells, with negligible mass elsewhere.","The number of positive solutions grows exponentially with the number of wells: at least 2^k−1.","The solutions are obtained without requiring the energy functional to be finite on all of H^1(R^N), overcoming the lack of well-definedness by an approximation on balls.","The method extends the penalization approach of del Pino and Felmer to logarithmic nonlinearities, which do not satisfy the usual polynomial asymptotics near zero.","The limit profiles are least-energy solutions of the Dirichlet problem on the union of the selected wells, establishing a precise concentration behavior."],"supporting_citations":[{"why":"Provides the decomposition F1,F2 of the logarithmic nonlinearity and its properties, which are fundamental to the auxiliary problem.","marker":"[21]"},{"why":"Also establishes the properties of F1,F2, used to justify the convexity and growth conditions.","marker":"[24]"},{"why":"The penalization method of del Pino and Felmer is adapted here to handle the logarithmic nonlinearity and prove multi-bump existence.","marker":"[18]"},{"why":"Prior work by the authors using an auxiliary problem for logarithmic Schrödinger equations; the boundedness of (PS) sequences is inspired by it.","marker":"[7]"},{"why":"Provides the minimax level comparison and the deformation arguments used in Section 2.4 and Proposition 2.4.","marker":"[1]"},{"why":"The original multi-bump result for polynomial nonlinearities with deepening potential well, which this paper extends to the logarithmic case.","marker":"[20]"},{"why":"The logarithmic Sobolev inequality used to prove boundedness of Palais-Smale sequences (Lemma 2.2 and Corollary 2.1).","marker":"[17]"}],"fun_headline_variants":["Deep wells give 2^k−1 positive solutions","Log Schrödinger equation yields 2^k−1 bump solutions","Multi-bump states in log Schrödinger with k wells","Each well subset yields a positive bump solution","At least 2^k−1 positive solutions for log Schrödinger"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The paper assumes without proof that the decomposition of the logarithmic nonlinearity into F1 and F2 satisfies all the listed properties (convexity, growth, monotonicity) for the particular constants δ, l, a0 chosen; if those properties fail, the auxiliary problem loses its geometry and the whole construction collapses.","fun_headline_variants_meta":{"raw":{"variants":["Deep wells give 2^k−1 positive solutions","Log Schrödinger equation yields 2^k−1 bump solutions","Multi-bump states in log Schrödinger with k wells","Each well subset yields a positive bump solution","At least 2^k−1 positive solutions for log Schrödinger"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000667,"raw_usage":{"total_tokens":2997,"prompt_tokens":854,"completion_tokens":2143,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":470,"completion_tokens_details":{"reasoning_tokens":2063}},"tokens_in":470,"tokens_out":2143,"duration_ms":13281,"temperature":1.0,"reasoning_tokens":2063,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T11:20:59.511464+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute explicitly the functions F1 and F2 for the given δ, l, a0 and check whether F1 is convex and nonnegative for all s, and whether F2′(s)/s is nondecreasing and F2′(s)≥0 for s>0 as claimed. If any of these properties fails for the specific choices, the auxiliary problem (Mλ,R) does not have the mountain-pass geometry and the main theorem is not established. Alternatively, a numerical simulation for a concrete potential V with two wells could test the claim of existence of multi-bump solutions for large λ, checking whether positive solutions indeed concentrate on each chosen well.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the decomposition F1,F2 of the logarithmic nonlinearity and its properties, which are fundamental to the auxiliary problem."},{"cited_title":"Squassina, A","cited_arxiv_id":null,"evidence_quote":"Also establishes the properties of F1,F2, used to justify the convexity and growth conditions."},{"cited_title":"del Pino, P.L","cited_arxiv_id":null,"evidence_quote":"The penalization method of del Pino and Felmer is adapted here to handle the logarithmic nonlinearity and prove multi-bump existence."},{"cited_title":"Alves, C","cited_arxiv_id":null,"evidence_quote":"Prior work by the authors using an auxiliary problem for logarithmic Schrödinger equations; the boundedness of (PS) sequences is inspired by it."},{"cited_title":"Alves, Existence of multi-bump solutions for a class of quasilinea r problems, Adv","cited_arxiv_id":null,"evidence_quote":"Provides the minimax level comparison and the deformation arguments used in Section 2.4 and Proposition 2.4."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"The original multi-bump result for polynomial nonlinearities with deepening potential well, which this paper extends to the logarithmic case."},{"cited_title":"del Pino, J","cited_arxiv_id":null,"evidence_quote":"The logarithmic Sobolev inequality used to prove boundedness of Palais-Smale sequences (Lemma 2.2 and Corollary 2.1)."}],"review_version":1}