{"id":"f3d4b3c2-f748-460f-ac59-b5b14331f307","arxiv_id":"1908.06273","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For fixed drift strength and fixed container volume, the ball, together with an inward drift aligned with the expected-lifetime gradient, maximizes the total gradient and Laplacian of the expected exit time.","lead":"This paper shows how a force field should be designed to keep a randomly moving particle inside a container for as long as possible, and proves that the round container is the best shape. It introduces a nonlinear partial differential equation that generalizes the classical torsion function and proves new sharp inequalities for it.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 1 does not establish the reduction to the nonlinear PDE: it freezes the drift at the old solution's gradient and never proves self-consistency or domination by the actual nonlinear solution.","rationale":"The reader correctly identified that the proof needs a more explicit fixed-point reduction in Lemma 1, but the stated weakest assumption was the C^2 regularity and the Zaremba-Hopf-Oleinik boundary behavior. In my reading, the missing reduction is more central: without it, the main theorem's quantifier over all vector fields b does not follow from the domain-level isoperimetric arguments, which are all carried out for the nonlinear PDE. The C^2 regularity concern, by contrast, only affects the boundary regularity threshold and is explicitly acknowledged by the authors. The proposed concrete test is a direct comparison principle for the nonlinear solution against all linear solutions; this is the natural missing step and would settle whether Lemma 1's reduction can be made rigorous. Because the comparison is likely true via standard HJB/viscosity arguments, I do not recommend rejecting the paper; rather, the acceptance should be conditional on supplying that argument. The reader's conditional verdict is therefore unchanged, but the condition should be sharpened from a technical differentiability issue in Lemma 3 to the missing self-consistency step connecting Lemma 1 to the nonlinear PDE.","tokens_in":8451,"tokens_out":14199,"duration_ms":153031,"concrete_test":"Settle the reduction by checking the direct comparison principle: let v solve -Δv - B|∇v| = 1 in a smooth domain and let w solve -Δw + a·∇w = 1 with |a| ≤ B. Test whether v ≥ w by applying viscosity-solution comparison to the Hamilton-Jacobi-Bellman operator -Δv + max_{|a|≤B} a·∇v = 1. If this comparison holds (as the stochastic-control interpretation suggests), then Lemma 1 can be replaced by a two-line argument: the nonlinear solution dominates every linear solution, so the theorem's reduction to the nonlinear PDE is sound. If a counterexample is found, the main theorem's statement over all vector fields is not justified.","verdict_should_be":"CONDITIONAL","load_bearing_attack":"The main theorem quantifies over all vector fields b with fixed L-infinity norm, but the proof's reduction to the nonlinear PDE is not justified as written. Lemma 1 shows only that for a fixed solution w of -Δw + b·∇w = 1, replacing b by -‖b‖∞ ∇w/|∇w| and solving the resulting linear equation for u gives u ≥ w. The new drift is frozen at ∇w/|∇w|, so u solves -Δu - ‖b‖∞ (∇w/|∇w|)·∇u = 1, not -Δu - ‖b‖∞ |∇u| = 1. The self-consistent condition required by the theorem is b = -‖b‖∞ ∇u/|∇u| for the same u. The paper asserts that Lemma 1 reduces the problem to the nonlinear PDE without proving either (i) that iteration of the map w ↦ u converges to a solution of the nonlinear PDE, or (ii) that the actual nonlinear solution dominates every linear solution. All subsequent lemmas (Lemmas 3-6) apply only to solutions of the nonlinear PDE, so the quantifier over all b in the main theorem is not established by the written proof. This is a more load-bearing gap than the C^2 regularity issue: if the reduction fails, the central claim for arbitrary drift fields is unsupported, whereas regularity only narrows the class of domains.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the Dirichlet problem -Δu + b·∇u = 1 in a bounded C^2 domain Ω, which models the expected exit time of drift-diffusion. For fixed ||b||_∞ and fixed volume, the authors claim that ∫_Ω |∇u| and ∫_Ω -Δu are maximized when Ω is a ball and b = -||b||_∞ ∇u/|∇u|. The proof introduces the nonlinear PDE -Δu - b|∇u| = 1 (with b now a constant equal to the L∞ norm of the drift), defines f(c) as the supremum of ∫|∇u| over domains of volume c, derives a differential inequality for f(c), and shows that the ball saturates it. The argument is extended to L^p norms; the L^p result is credited to Hamel and Russ. The paper is self-contained apart from the gaps noted below, and the ball computation is explicit and parameter-free.","tokens_in":8673,"tokens_out":24773,"duration_ms":220492,"significance":"If the main theorem is fully established, it provides a sharp isoperimetric statement for a natural nonlinear analogue of the torsion function and an explicit optimal drift design for maximizing expected exit times. The level-set induction is elegant, and the explicit ball computation in Lemma 4 is a clear strength. The L^p corollary is already known from Hamel–Russ, so the genuinely new content is the L^1 gradient maximization. The proof is structurally plausible, but the reduction from arbitrary drifts to the nonlinear PDE is currently missing, and the monotonicity issue in Lemma 3 needs a fix; both are load-bearing for the central claim.","major_comments":[{"comment":"Lemma 1 does not establish the stated reduction to the nonlinear PDE. For a solution w of -Δw + b·∇w = 1, the lemma constructs u solving -Δu - ||b||_∞ (∇w/|∇w|)·∇u = 1 and shows u ≥ w; the drift is frozen at ∇w/|∇w|, so u is not a solution of -Δu - ||b||_∞ |∇u| = 1. The manuscript asserts without proof that this 'reduces the problem' to the nonlinear PDE. To justify the quantification over all b in the main theorem, the authors must either prove that the self-consistent solution of the nonlinear PDE dominates every linear solution, or prove convergence of the iteration w ↦ u to a solution of the nonlinear PDE. A direct maximum-principle comparison between the nonlinear solution and w would supply pointwise domination, but it is not included, and pointwise domination alone would still need to be supplemented to control ∫|∇u|. As written, Lemmas 2–6 apply only to nonlinear solutions, so the claim for arbitrary b is unsupported.","section":"3.1, Lemma 1 and the paragraph after it"},{"comment":"The proof of the differential inequality for f uses the step f(|Ω_ε|) ≤ f(c - ε|∂Ω|^2/(f(c)+c) + o(ε)), which requires f to be nondecreasing in c. The text only cites continuity of f at this point; continuity is insufficient to pass from |Ω_ε| ≤ c - εA + o(ε) to the displayed inequality with f evaluated at c - εA. The authors should either prove monotonicity of f (for example by a domain-inclusion comparison for the nonlinear PDE) or replace this step with a limsup argument that avoids monotonicity. This gap affects the derivation of the main estimate for ∫|∇u|.","section":"3.2, proof of Lemma 3"},{"comment":"The manuscript never states an existence theorem for the nonlinear PDE -Δu - b|∇u| = 1 on an arbitrary bounded C^2 domain, and the C^2 regularity up to the boundary used in the volume expansion of Lemma 3 is asserted with a reference ([13]) to fully nonlinear theory that does not directly apply to this semilinear equation. Since Lemma 3's expansion |Ω_ε| = |Ω| - ε∫∂Ω (∂u/∂n)^{-1} dσ + o(ε) relies on this regularity and on ∂u/∂n > 0, the authors should provide a precise existence/regularity statement (e.g., via sub/supersolutions and elliptic regularity) or a specific reference. The issue is fixable by standard methods, but it is currently load-bearing because the theorem ranges over all C^2 domains.","section":"2.1 and 3.2 (existence and regularity)"}],"minor_comments":[{"comment":"The formula '−∆u − b · |∇u| = 1' uses a dot product with a scalar b; once b denotes the L∞ norm, write 'b |∇u|' without the dot.","section":"Abstract"},{"comment":"There are several typos and OCR artifacts in the text, such as 'maximal strenght' and 'nc an'; a careful proofreading pass is needed.","section":"Introduction"},{"comment":"The change of variables ∂/∂c = (1/|∂B_R|) ∂/∂r is written informally; a sentence justifying it via the coarea formula or the implicit function theorem would improve clarity.","section":"3.2, Lemma 4 proof"},{"comment":"Reference [13] is not the standard source for C^{2,α} regularity of semilinear elliptic equations with Lipschitz gradient dependence; a reference such as Gilbarg–Trudinger would be more appropriate.","section":"References"},{"comment":"The paragraph on existing results is compressed; a brief statement of the Hamel–Russ rearrangement result and its hypotheses would help the reader see exactly why the L^p corollary follows.","section":"2.2"}],"recommendation":"major_revision","confidential_remarks":"The missing reduction in §3.1 is the main obstacle; if the authors can supply the comparison argument and fix the monotonicity issue in Lemma 3, the result is likely correct. The novelty is concentrated in the L^1 gradient estimates, since the L^p corollary is already known; the presentation would be strengthened by clearly separating the new result from the known one. The paper fits the journal's scope."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a real result, but the written proof has a hole in the first step and a hand-wave in the middle. The main theorem—ball maximizes ∫|∇u| and ∫−Δu for solutions of −Δu + b·∇u = 1 with ‖b‖∞ fixed—is new and, as far as I can tell, true. The level-set induction (Lemmas 3–5) is a genuinely clever way to get an isoperimetric inequality for the nonlinear PDE −Δu − b|∇u| = 1, and the explicit ball computation in Lemma 4 checks out. The Lp result is credited to Hamel–Russ, which is honest.\n\nNow the soft spots. The stress on Lemma 1 is correct: that lemma shows that if w solves the linear equation with drift b, then solving again with the drift frozen at −M∇w/|∇w| gives a pointwise larger solution. It does not show that the self-consistent nonlinear solution dominates every linear solution. The paper says 'this reduces to the nonlinear PDE' without proving that. There is a simple fix: if u solves −Δu − M|∇u| = 1, then for any b with |b| ≤ M, −Δu + b·∇u = 1 + (M|∇u| + b·∇u) ≥ 1, so u is a supersolution and u ≥ w for every w. Then the coarea formula gives ∫|∇w| ≤ ∫|∇u| and hence ∫−Δw ≤ ∫−Δu. But that argument isn't in the paper. As written, the quantifier over all vector fields b is not justified.\n\nThe second issue is Lemma 3. The function f(c) is a supremum over domains, and the proof differentiates it without saying which derivative. A rigorous version would use Dini derivatives or a variational argument. For a C2 domain the expansion of |Ωε| is fine, but 'rearranging and letting ε→0' is not a proof of differentiability. This is a technical gap, not a fatal one.\n\nMinor: the abstract says 'for fixed volume, ‖∇u‖L1 and ‖Δu‖L1 are maximized if Ω is the ball' without mentioning that the drift must be the aligned one. The theorem statement is precise; the abstract isn't.\n\nBottom line: the central idea is good, the result is likely correct, and the proof strategy will be reused. With a revision that (i) proves the pointwise comparison lemma and (ii) states the derivative argument in terms of Dini derivatives, this would be a solid paper. I'd send it to a referee, not desk-reject.","headline":"A new and likely true isoperimetric theorem for a nonlinear elliptic PDE, but the written proof has a fixable gap in the reduction from arbitrary drifts to the nonlinear equation and a hand-waved differentiation step.","tokens_in":9251,"tokens_out":10795,"would_cite":true,"duration_ms":102617,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35B51","49K20","60J60"],"pacs":[],"model":"deepseek-v4-flash","headline":"The ball, with drift pointing radially inward, is the optimal trap for Brownian motion with a drift of fixed strength.","keywords":["drift diffusion","exit time","isoperimetric inequality","torsion function","Brownian motion","expected lifetime","nonlinear elliptic PDE","level-set method"],"falsifier":"Compute, for a fixed volume and fixed maximal drift strength, the solution of $-\\Delta u - b|\\nabla u| = 1$ on a non-ball domain such as a long thin rectangle or an ellipse with the same area as a reference ball, and compare $\\int_\\Omega |\\nabla u|\\, dx$ or $\\int_\\Omega -\\Delta u\\, dx$ with the ball's value; if either integral exceeds the ball's value, the theorem is false. A numerical or explicit counterexample would settle it.","tokens_in":8193,"feed_emoji":"🎯","tokens_out":11216,"duration_ms":92363,"temperature":0.7,"pith_summary":"This paper asks how to keep a Brownian particle inside a bounded container for as long as possible when an external flow of prescribed maximum strength can push it around. The answer it proposes is sharp: among all containers of a given volume and all flows of a given maximum speed, the expected survival time is largest when the container is a ball and the flow always points straight toward the center of the ball. In any container, the best flow is not chosen in advance but is nonlinearly tied to the survival-time function itself, pointing along its gradient; this reduces the problem to a single nonlinear PDE. The paper proves that for this PDE the ball maximizes the total variation of the survival-time function and the integral of its Laplacian, and it recovers the known fact that the ball maximizes all $L^p$ norms of the survival time.","feed_headline":"A ball with inward-pointing drift traps Brownian motion longest","feed_subtitle":"Sharp isoperimetric result: the ball maximizes survival-time gradients among equal-volume domains with fixed drift strength.","key_machinery":"The central object is the nonlinear PDE $-\\Delta u - b|\\nabla u| = 1$ with $b = \\|b\\|_{L^\\infty}$ constant, called a nonlinear analogue of the torsion function (the torsion function solves $-\\Delta u = 1$). Its defining property for the proof is invariance under adding constants: if $u$ solves it on $\\Omega$, then $u - \\varepsilon$ solves the same PDE on the superlevel set $\\Omega_\\varepsilon = \\{u \\geq \\varepsilon\\}$. This lets the authors induct over level sets, trading the shape optimization for a differential inequality $f'(c) \\leq b(f(c)+c)/(c_d c^{(d-1)/d})$ for $f(c)$, the supremum of $\\int_\\Omega |\\nabla u|\\, dx$ over domains of volume $c$. Two classical ingredients make the induction valid: the Zaremba-Hopf-Oleinik boundary point lemma, which ensures the inward normal derivative of $u$ does not vanish, and $C^2$ regularity of solutions, which makes the volume expansion $|\\Omega_\\varepsilon| = |\\Omega| - \\varepsilon \\int_{\\partial\\Omega} (\\partial u/\\partial n)^{-1} d\\sigma + o(\\varepsilon)$ exact enough. The ball is the extremal case because its radial solution satisfies the same ODE with equality at every step.","core_discovery":"On its own terms, the paper establishes the following theorem: for bounded $C^2$ domains with fixed volume and vector fields with fixed $L^\\infty$ norm, the solution $u$ of $-\\Delta u + b\\cdot\\nabla u = 1$ with $u=0$ on the boundary maximizes both $\\int_\\Omega |\\nabla u|\\, dx$ and $\\int_\\Omega -\\Delta u\\, dx$ when $\\Omega$ is the ball and $b = -\\|b\\|_{L^\\infty} \\nabla u/|\\nabla u|$. The optimal flow condition is derived from a maximum-principle comparison: replacing any candidate $b$ by the feedback field aligned with $\\nabla u$ can only increase $u$ pointwise. This reduces the shape problem to the nonlinear PDE $-\\Delta u - b|\\nabla u| = 1$, whose invariance under adding constants lets the proof peel off level sets and obtain a differential inequality for the maximal gradient integral; the ball saturates the inequality via an explicit radial ODE. A corollary, already known from a general rearrangement principle, is that the $L^p$ norms of $u$ are also maximized by the ball.","pith_inferences":["The authors leave the optimal boundary-regularity threshold open; a natural test is whether the volume expansion and differential inequality persist on $C^1$ or Lipschitz domains, where corners might change trapping efficiency.","Because the proof uses only level-set invariance and the boundary-point lemma, it may extend to other functionals of $u$ that are additive over level sets, such as moments of the exit time or the work done by the drift.","A concrete numerical probe of the isoperimetric claim would be to solve the nonlinear PDE on ellipses of the same area as a ball and check that $\\int_\\Omega |\\nabla u|\\, dx$ stays below the ball's value, which would also show how the optimum is approached as the domain becomes round."],"forward_implications":["In any fixed domain, replacing the drift by $b = -\\|b\\|_{L^\\infty} \\nabla u/|\\nabla u|$ increases the expected lifetime pointwise, so optimal traps can always be taken to push in the direction of increasing lifetime.","Among equal-volume domains, the ball maximizes both $\\int_\\Omega |\\nabla u|\\, dx$ and $\\int_\\Omega -\\Delta u\\, dx$, giving sharp bounds on the total variation of expected lifetime.","The same level-set argument shows that the ball maximizes $\\|u\\|_{L^p}$ for every positive integer $p$, with an elementary derivation for the $L^\\infty$ case.","On the ball the optimal configuration is explicit: a radial solution of a one-dimensional ODE gives the exact maximal values as functions of volume and drift strength."],"supporting_citations":[{"why":"supplies the Zaremba-Hopf-Oleinik boundary point lemma used to show the inward normal derivative of the solution does not vanish.","marker":"[20]"},{"why":"provides a further treatment of the same boundary point lemma for $C^2$ domains.","marker":"[25]"},{"why":"textbook maximum-principle and boundary point lemma background that underlies Lemma 2.","marker":"[28]"},{"why":"gives the $C^2$ regularity of solutions needed to justify the asymptotic volume expansion of level sets.","marker":"[13]"}],"fun_headline_variants":["Ball wins for Brownian trap with optimal inward drift","Optimal drift makes ball the best Brownian trap","Nonlinear torsion: ball maximizes drift-diffusion lifetime","Ball is isoperimetric maximizer for expected lifetime under drift","Steepest-descent drift: ball is the champion trap shape"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that on every candidate domain the survival-time solution is smooth up to the boundary and meets it with nonzero inward derivative, so the level-set layers have the precise volume the proof needs; this is proved for $C^2$ domains, and the true minimal regularity is open.","fun_headline_variants_meta":{"raw":{"variants":["Ball wins for Brownian trap with optimal inward drift","Optimal drift makes ball the best Brownian trap","Nonlinear torsion: ball maximizes drift-diffusion lifetime","Ball is isoperimetric maximizer for expected lifetime under drift","Steepest-descent drift: ball is the champion trap shape"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000763,"raw_usage":{"total_tokens":3430,"prompt_tokens":1034,"completion_tokens":2396,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":650,"completion_tokens_details":{"reasoning_tokens":2314}},"tokens_in":650,"tokens_out":2396,"duration_ms":17745,"temperature":1.0,"reasoning_tokens":2314,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T12:53:36.094421+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute, for a fixed volume and fixed maximal drift strength, the solution of $-\\Delta u - b|\\nabla u| = 1$ on a non-ball domain such as a long thin rectangle or an ellipse with the same area as a reference ball, and compare $\\int_\\Omega |\\nabla u|\\, dx$ or $\\int_\\Omega -\\Delta u\\, dx$ with the ball's value; if either integral exceeds the ball's value, the theorem is false. A numerical or explicit counterexample would settle it.","supporting_citations":[{"cited_title":"Kuran, On positive superharmonic functions in α -admissible domains, J","cited_arxiv_id":null,"evidence_quote":"supplies the Zaremba-Hopf-Oleinik boundary point lemma used to show the inward normal derivative of the solution does not vanish."},{"cited_title":"Nazarov, A centennial of the Zaremba-Hopf-Oleinik l emma","cited_arxiv_id":null,"evidence_quote":"provides a further treatment of the same boundary point lemma for $C^2$ domains."},{"cited_title":"Pucci, J","cited_arxiv_id":null,"evidence_quote":"textbook maximum-principle and boundary point lemma background that underlies Lemma 2."},{"cited_title":"Caffarelli and X","cited_arxiv_id":null,"evidence_quote":"gives the $C^2$ regularity of solutions needed to justify the asymptotic volume expansion of level sets."}],"review_version":1}