{"id":"fc96c433-ea02-4ef6-94aa-66aafce056ef","arxiv_id":"2508.07931","paper_version":1,"verdict":"UNVERDICTED","confidence":"LOW","novelty_score":6.0,"correctness_risk":"unknown","formal_verification":"none","parameter_count":0,"one_line_summary":"A weakened form of log-concavity is preserved by generalised heat semigroups, leading to new Schrödinger ground-state and parabolic log-Hessian estimates.","lead":"This paper proves that a weakened form of log-concavity is preserved by a broad class of heat-like semigroups, not just the classical heat flow. The results give new estimates for Schrödinger ground states and parabolic equations, and a new route to functional inequalities in non-log-concave settings.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Unbounded-coefficient HJB regularity and reflection coupling are assumed without evidence; without them, the central propagation and log-Hessian estimates do not follow.","rationale":"The reader's weakest_assumption is essentially the same: the stochastic control representation and HJB regularity in the unbounded-coefficient regime. My concern refines this by pointing to the reflection-coupling step and the two-sided log-Hessian estimates, which are particularly fragile. Since the full text is absent, no further verification is possible; the honest verdict remains UNVERDICTED. The potential significance is real: if the propagation holds, it gives new ground-state log-semiconcavity results beyond log-concave potentials. But the proof hinges on a regime where standard elliptic regularization fails. I recommend keeping UNVERDICTED rather than moving to CONDITIONAL, because the concern is not an identified error in the manuscript but an unverified assumption; the authors may well have handled it. A conditional acceptance would require the omitted proof, which we do not have. Therefore no adjustment to the reader's verdict.","tokens_in":679,"tokens_out":4241,"duration_ms":50985,"concrete_test":"Obtain the full manuscript and verify Proposition/Theorem 3.x (the HJB regularity statement). Then run the following test: choose the Ornstein–Uhlenbeck operator L = Δ - x·∇ on R^d, whose fundamental solution p is explicit. Compute ∇^2 log p(t,x,y) and check whether the claimed time-uniform two-sided bounds hold as t→∞. If they fail, the theorem's hypotheses must exclude this simple unbounded-coefficient case; if the theorem claims to include it, the claim is false. Alternatively, if the paper's definition of 'weak log-concavity' is designed to be testable via a variational formula, verify that the known solution p propagates it in the exact sense stated. This would settle whether the unbounded-coefficient regime is actually covered.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The abstract's central claim is that weak log-concavity propagates along generalised heat semigroups, with consequences including time-uniform two-sided log-Hessian estimates for fundamental solutions of parabolic equations with unbounded coefficients. The proof is announced as relying on a stochastic control representation of the semigroup and a 'second order analysis of reflection coupling along HJB characteristics'. This is the least secure link. Stochastic control representations of the heat semigroup require the Hamiltonian to satisfy coercivity/regularity conditions; reflection coupling requires the controlled diffusions to have bounded coefficients or strong dissipativity so that the coupling time has finite exponential moments. For unbounded coefficients (e.g., a drift b(x) that grows linearly), the value function of the associated HJB equation is only a viscosity solution, not necessarily C^2; its Hessian may be a measure, and the 'log-Hessian estimates' are not meaningful in the classical sense. Moreover, the two-sided uniform-in-time estimates are false in general for parabolic equations with unbounded coefficients unless the coefficients satisfy an anti-dissipativity condition. The abstract gives no indication that such conditions are imposed. If the HJB solutions are not regular enough, the reflection coupling analysis has no rigorous basis, and the propagation of weak log-concavity--if true--would not yield the claimed Hessian bounds. This is a load-bearing concern because the stated novelty and applications hinge on those estimates.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper announces results on propagation of a 'weak' form of log-concavity along generalized heat semigroups, with applications to log-semiconcavity of Schrödinger ground states for non-convex potentials, propagation of functional inequalities, and two-sided log-Hessian estimates for fundamental solutions of parabolic equations with unbounded coefficients, uniform in time. The proofs are said to combine a stochastic control representation with a second-order analysis of reflection coupling along characteristics of quadratic Hamilton-Jacobi-Bellman equations. The abstract also claims novelty in non-log-concave settings and builds on Brascamp-Lieb conditioning/marginalization results.","tokens_in":1018,"tokens_out":2167,"duration_ms":27555,"significance":"If the announced results are correct, they would be a meaningful advance: they extend the classical Prekopa-Leindler preservation of log-concavity to a broader class of semigroups, yield new regularity information on ground states of Schrödinger operators, and provide quantitative log-Hessian bounds for heat kernels with unbounded coefficients. The stochastic-control/HJB approach is appropriate and potentially powerful. The paper promises, in particular, new functional-inequality propagation results that go beyond known log-concave cases. However, the significance can only be fully assessed after checking the regularity and coupling arguments, which are not available from the abstract alone.","major_comments":[{"comment":"The central notion 'weak log-concavity' is never defined in the abstract, so the theorem statements are not checkable. Since this notion is the paper's primary contribution, the manuscript must give a precise definition and clearly state how it relates to classical log-concavity, and why it is preserved by the considered flows. This is load-bearing: without the definition, the propagation claims are only heuristic.","section":"Abstract"},{"comment":"The claimed time-uniform two-sided log-Hessian estimates for fundamental solutions of parabolic equations with unbounded coefficients require structural assumptions that are not stated. In general, HJB value functions only are viscosity solutions and their Hessians may be measure-valued; reflection coupling needs conditions ensuring finite exponential moments of the coupling time (e.g., bounded or strongly dissipative coefficients). The abstract gives no indication that such conditions are imposed. The manuscript should state precise hypotheses on the drift, diffusion, and Hamiltonian, and prove (or cite) the needed C^2 or semiconcavity regularity of the HJB solution.","section":"Abstract"},{"comment":"The proof outline mentions 'second order analysis of reflection coupling along HJB characteristics' but gives no details. Since this is the key technical step, the full paper must provide a self-contained proof or a precise reference. As it stands, the central propagation theorem cannot be verified from the information given. I am not asserting an error, but this is a major missing-support issue in the submitted material.","section":"Abstract"}],"minor_comments":[{"comment":"Minor typographical issues: 'logsemicontinuity' should likely be 'log-semiconcavity' with hyphen; 'eventually study' should be 'finally study' or 'in the last part' to avoid ambiguity. The formatting of 'Pr{\\'e}kopa' should be checked in the final PDF.","section":"Abstract"},{"comment":"The phrase 'To our knowledge, our results are the first of this type' should be supported by a literature review in the introduction, to make the novelty claim verifiable.","section":"Abstract"}],"recommendation":"uncertain","confidential_remarks":"This review is based solely on the abstract; no full text was available. The missing support for the HJB regularity and reflection-coupling steps, along with the undefined notion of weak log-concavity, prevent a reliable verdict. I recommend that the journal obtain the full manuscript before making a decision; the abstract-only limitation is not a reflection on the authors' work, but it precludes a fair assessment."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear [colleague],\n\nThe paper makes a concrete new claim: a weakened log-concavity notion propagates along generalised heat semigroups, and this yields log-semiconcavity of Schrödinger ground states for non-convex potentials, propagation of functional inequalities, and time-uniform two-sided log-Hessian bounds for parabolic fundamental solutions. If the proofs are correct, this is a genuine extension of the Brascamp-Lieb/Prékopa-Leindler program to non-log-concave settings, and the method (stochastic control plus reflection coupling along HJB characteristics) is a promising addition to the toolbox.\n\nWhat is clearly new from the abstract: the notion of weak log-concavity itself, the first propagation results in non-log-concave settings, and the control-theoretic route to Hessian estimates. The paper also studies conditioning and marginalisation following Brascamp-Lieb, which is a natural and useful extension. I can't verify the proofs, because only the abstract is available, but the claims are specific enough that a referee can test them.\n\nThe main soft spot is the two-sided log-Hessian estimate for unbounded coefficients. The stress-test note is on target: stochastic control representations typically assume coercivity of the Hamiltonian, and reflection coupling needs strong enough regularity of the controlled diffusion. With only a viscosity solution to the HJB equation, the 'Hessian' may not be a classical object, and uniform-in-time bounds can fail without dissipativity or anti-dissipativity conditions. The abstract does not state what conditions are imposed. This could be an abstract omission, but it is exactly where a referee should dig first. Similarly, the definition of weak log-concavity is absent from the abstract, so I cannot judge how much weaker it is or whether the results are truly of Prekopa-Leindler type.\n\nA minor point: 'first of this type' is a strong claim, and the abstract gives no comparison to the existing literature beyond Brascamp-Lieb and Prekopa-Leindler. That's normal for an abstract, but it means the novelty assessment should wait for the full introduction.\n\nOverall, I'd say this is a serious piece of work that deserves a rigorous referee. The core idea is clear and the consequences are concrete. My uncertainty is about the technical hypotheses behind the HJB/coupling argument, not about the honesty or coherence of the exposition.\n\nMy recommendation: send it to peer review, explicitly asking the referee to check (a) the second-order regularity of the HJB solution in the unbounded-coefficient case, (b) the conditions on the coefficients for the reflection-coupling argument, and (c) the exact definition and non-vacuity of weak log-concavity. If those hold up, it's a publishable contribution to mathematical analysis and probability.\n\nBest,\n\n[You]","headline":"Plausible and potentially important advance, but the unbounded-coefficient HJB regularity step is the key thing a referee must check; abstract-only.","tokens_in":1431,"tokens_out":3354,"would_cite":false,"duration_ms":35272,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"Generalised heat semigroups preserve a weak form of log-concavity, yielding log-semiconcavity for Schrödinger ground states and two-sided Hessian bounds for parabolic equations.","keywords":["weak log-concavity","heat semigroup","Hamilton-Jacobi-Bellman equation","reflection coupling","Schrödinger ground state","log-Hessian estimates","functional inequalities","stochastic control"],"falsifier":"Pick a smooth, non-convex potential with unbounded Hessian, such as $V(x) = \\frac14 |x|^4 - \\frac12 |x|^2$, and run the corresponding generalised heat flow on a Gaussian initial condition. If the eigenvalues of $\\nabla^2 \\log u(t,x)$ are unbounded above or below as $t$ grows for some choice of coefficients, the claimed time-uniform log-Hessian bounds fail.","tokens_in":661,"feed_emoji":"🔥","tokens_out":13503,"duration_ms":140098,"temperature":0.7,"pith_summary":"Classical heat flow preserves log-concavity (a positive function whose logarithm is concave), but this property is lost for more general semigroups. The paper isolates a slightly weaker condition—log-concavity up to a controlled quadratic error—and proves that it propagates along generalised heat semigroups, including semigroups with unbounded coefficients. The proof treats the semigroup as the value function of a stochastic control problem and uses a reflection coupling to bound the Hessian of the logarithm of the solution. From that, the authors derive log-semiconcavity of Schrödinger ground states for non-convex potentials, propagation of functional inequalities along the flow, and time-uniform two-sided log-Hessian estimates for parabolic fundamental solutions.","feed_headline":"Weak log-concavity survives generalised heat flows","feed_subtitle":"A slightly weaker concavity condition survives generalised heat flow, yielding new Schrödinger and parabolic estimates.","key_machinery":"The key machinery is the stochastic control interpretation of the generalised heat semigroup, which expresses the solution as the value function of an optimal control problem. The associated dynamic programming equation is a quadratic Hamilton-Jacobi-Bellman (HJB) equation, and the paper studies it via a reflection coupling of the controlled diffusions—a coupling that mirrors one trajectory in the other until they coalesce. A second-order (Hessian-level) analysis of this HJB solution along the coupled characteristics yields the quantitative bounds on $\\nabla^2 \\log u$ that carry the propagation result.","core_discovery":"The paper's central claim is that a slightly weakened notion of log-concavity—one that allows the Hessian of the logarithm to deviate from zero by a controlled amount—is propagated by generalised heat semigroups. The proof uses a stochastic control representation of the semigroup and a reflection-coupling argument along the characteristics of the corresponding Hamilton-Jacobi-Bellman equation, and it applies in the unbounded-coefficient regime. From this propagation result the paper derives log-semiconcavity of Schrödinger ground states for non-convex potentials, preservation of functional inequalities along the flow, and two-sided, time-uniform log-Hessian estimates for parabolic fundamenta","pith_inferences":["If the uniform log-Hessian bounds extend to the invariant measure, they could yield quantitative exponential convergence for diffusion semigroups with non-convex drifts, a regime where classical convexity arguments stop.","The reflection-coupling technique may transfer to nonlocal or degenerate generators, suggesting weak log-concavity is the natural propagation condition for a wider class of Markov semigroups.","Tracking the constant in the weak-log-concavity bound as a function of the potential's Hessian would give a quantitative trade-off between the concavity lost by the potential and the smoothness gained from the heat part, relevant for optimal transport and concentration inequalities."],"forward_implications":["Functional inequalities that are known for weakly log-concave functions will be preserved along the generalised heat flow, extending the classical log-concave regime.","Schrödinger ground states corresponding to non-convex potentials satisfy a log-semiconcavity bound, so their superlevel sets and concentration properties can be compared with those of log-concave densities up to a Gaussian factor.","The two-sided log-Hessian estimates hold uniformly in time for parabolic fundamental solutions with unbounded coefficients, giving a priori regularity that can be used in long-time convergence analysis.","Weak log-concavity is preserved under conditioning and marginalisation in the settings studied, generalising the classical stability of log-concavity to non-log-concave settings."],"supporting_citations":[],"fun_headline_variants":["Weak log-concavity persists in generalised heat flows","Weaker concavity survives heat flow, new Schrödinger bounds","HJB method preserves weak log-concavity, yields estimates","New log-semiconcavity for non-convex Schrödinger potentials","Time-uniform log-Hessian bounds via reflection coupling"],"cache_read_input_tokens":2816,"weakest_assumption_plain":"The main proof assumes that the Hamilton-Jacobi-Bellman equation associated with the generalised heat semigroup admits a sufficiently regular (second-order) solution in the unbounded-coefficient case, so that the Hessian computations along the stochastic trajectories are legitimate.","fun_headline_variants_meta":{"raw":{"variants":["Weak log-concavity persists in generalised heat flows","Weaker concavity survives heat flow, new Schrödinger bounds","HJB method preserves weak log-concavity, yields estimates","New log-semiconcavity for non-convex Schrödinger potentials","Time-uniform log-Hessian bounds via reflection coupling"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000787,"raw_usage":{"total_tokens":3317,"prompt_tokens":760,"completion_tokens":2557,"prompt_tokens_details":{"cached_tokens":256},"prompt_cache_hit_tokens":256,"prompt_cache_miss_tokens":504,"completion_tokens_details":{"reasoning_tokens":2469}},"tokens_in":504,"tokens_out":2557,"duration_ms":20810,"temperature":1.0,"reasoning_tokens":2469,"cache_read_input_tokens":256,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-05T21:44:29.767579+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Pick a smooth, non-convex potential with unbounded Hessian, such as $V(x) = \\frac14 |x|^4 - \\frac12 |x|^2$, and run the corresponding generalised heat flow on a Gaussian initial condition. If the eigenvalues of $\\nabla^2 \\log u(t,x)$ are unbounded above or below as $t$ grows for some choice of coefficients, the claimed time-uniform log-Hessian bounds fail.","supporting_citations":[],"review_version":1}