{"id":"8dfd1c15-a7f5-4e16-a90d-7f5db53baa52","arxiv_id":"2501.09693","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"On metric graphs with one absorbing point, the path graph maximizes heat content at sufficiently small and sufficiently large times.","lead":"This paper proves that a single straight wire keeps heat longer than any other network of the same length, but only when the time is very short or very long. The proof introduces a new random-walk counting formula for heat, which may help decide the all-times question.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Small-time proof rests on an unpublished monotonicity lemma from [4] that reduces deg(vD) to 1; without it Theorem 2.6(ii) covers only degree-one Dirichlet vertices, and inequality (4.11) has a sign slip that needs correcting.","rationale":"I read the paper as aiming to prove an extremal Faber-Krahn inequality for heat content on metric graphs at asymptotically small and large times, with a new positive random-walk expansion as the main tool for the small-time regime. The large-time proof via Mercer's theorem and the strict eigenvalue Faber-Krahn inequality is straightforward and correct. The random-walk expansion in Theorem 4.1 is plausible and structurally sound: the coefficients α_n increase in n, so (4.1) is indeed a positive combination, and Lemma 4.7 correctly shows that the early-return terms dominate for small t. The load-bearing gap is the reduction to degree-one Dirichlet vertices: the proof of Theorem 2.6(ii) invokes an unproved, unpublished monotonicity result from [4]. Since the subsequent comparison of return-time distributions and the definition of k0 require vD to be a leaf, every graph with deg(vD)>1 — for instance a cycle or a star with a single Dirichlet center — is only covered conditionally on that external lemma. The displayed inequality (4.11) also contains a sign error relative to the preceding line: the tail sum should enter with a minus sign, not a plus sign. This is fixable via Lemma 4.7, so it is not by itself fatal, but it is an internal inconsistency in the written proof. These issues match the reader's weakest assumption and support the CONDITIONAL verdict; they do not justify rejection of the central claim, because the missing lemma is plausible and the sign error is local.","tokens_in":16559,"tokens_out":30544,"duration_ms":335008,"concrete_test":"Implement formula (4.1) for small equilateral graphs with identical edge count N and edge length ℓ — e.g., a cycle with vD of degree 2, a star with vD of degree d, and the path graph of the same N and ℓ — and evaluate Q_t at t much smaller than ℓ^2. Check whether Q_t(G) ≤ Q_t(P) and whether Q_t decreases when deg(vD) increases. A counterexample would refute Theorem 2.6(ii) as stated; success would isolate the missing ingredient to a proof of the [4] monotonicity lemma.","verdict_should_be":"UNCHANGED","load_bearing_attack":"In the proof of Theorem 2.6(ii), after the rational-dependence/equilateral subdivision, the authors assert \"the heat content is decreasing with respect to the degree of vD (cf. [4])\" and therefore suppose deg_G(vD)=1. This is the hinge of the small-time argument. The subsequent comparison of first-return probabilities P_{vD}[τ^G=k] ≤ P_{vD}[τ^P=k] and the choice of k0 as twice the distance to the first vertex of degree at least 3 both use that vD is a leaf; if deg(vD)>1, the initial step and the return-time distributions are different, and the proof does not apply. Since [4] is an unpublished companion paper and the lemma is neither stated nor proved, Theorem 2.6(ii) is presently established only for degree-one Dirichlet vertices. The claim also has a local sign error: the line preceding (4.11) contains a negative tail term -N(ℓ-α_{k0+1}(t)), yet (4.11) writes +N(...); the lower bound as printed is not implied. This error is likely typographical — replacing the sign and inserting the missing factor 1/2 still gives positivity by Lemma 4.7 — but it must be corrected before the proof is fully verifiable.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the heat content Q_t(G;V_D) of a compact metric graph G with nonempty Dirichlet set V_D and connected complement G\\V_D. The main result, Theorem 2.6, asserts a Faber-Krahn-type inequality at large times for every such graph, and at small times under an additional rational-dependence condition on the edge lengths: Q_t(G;V_D) ≤ Q_t(P_{|G|};{0}) for all sufficiently small or large t, with equality only for the path graph of the same total length with the Dirichlet vertex at an endpoint. The large-time proof uses the spectral (Mercer) expansion and the known Faber-Krahn inequality for the first eigenvalue. The small-time proof introduces a new probabilistic representation (Theorem 4.1) expressing the heat content as a positive linear combination of expected return times of a discrete random walk, via the Feynman-Kac formula. The paper is careful to state that the all-times question remains open.","tokens_in":16784,"tokens_out":13036,"duration_ms":124500,"significance":"If the main theorem is fully established, this is a substantial contribution to the spectral geometry of quantum graphs: it provides the first extremal Faber-Krahn-type statement for the heat content, complementing known results for eigenvalues and torsional rigidity. The random-walk expansion of Theorem 4.1 is a genuinely new tool and is likely to be useful beyond this paper. The large-time proof is clean, self-contained, and correctly uses known spectral bounds. The small-time argument is inventive but currently depends on two external or unproved ingredients: an unpublished monotonicity lemma from the companion paper [4] and a displayed inequality that contains a sign error. The paper is honest about its limitations, and the central derivation is not circular; however, the small-time theorem as stated is not yet fully verified from the material included in the manuscript.","major_comments":[{"comment":"The proof reduces to the case deg_G(v_D)=1 by the statement 'since the heat content is decreasing with respect to the degree of v_D (cf. [4])'. This is a load-bearing step: the subsequent comparison of return-time distributions P_{v_D}[τ^G=k] ≤ P_{v_D}[τ^P=k] and the definition of k_0 as twice the distance to the first vertex of degree at least three both use that v_D is a leaf. Since [4] is an unpublished companion paper and the monotonicity lemma is neither stated nor proved in the present manuscript, the proof of Theorem 2.6(ii) is complete only for degree-one Dirichlet vertices. The authors should either include a full statement and proof of the monotonicity result, or give a self-contained argument that circumvents the degree reduction, before the small-time theorem can be accepted as stated.","section":"Section 4, proof of Theorem 2.6(ii)"},{"comment":"The derivation of (4.11) from the preceding inequality is not valid as printed. The line before (4.11) contains a negative tail term -N(ℓ-α_{k0+1}(t)) and a factor 1/2 multiplying the positive term, but (4.11) is written with +N(ℓ-α_{k0+1}(t)) and without the factor 1/2. Replacing the sign and inserting the 1/2 gives the lower bound (1/2)C(P,G)(α_{k0+1}(t)-α_{k0}(t)) - N(ℓ-α_{k0+1}(t)), which is still positive for sufficiently small t by Lemma 4.7. The proof is therefore reconstructable, but the displayed inequality must be corrected.","section":"Section 4, inequality (4.11)"},{"comment":"The equality case is only explicitly argued for the reduced situation deg_G(v_D)=1. If the original graph has deg_G(v_D)>1, the proof first invokes the monotonicity lemma and then shows that absence of vertices of degree ≥3 forces the graph to be a path. The statement that equality implies v_D is an endpoint of the path requires strictness in the monotonicity lemma, which is not stated. The equality claim for arbitrary v_D therefore needs an explicit justification, for instance by stating and proving the strict monotonicity statement or by treating the degree-one case only and separately.","section":"Section 4, equality statement in Theorem 2.6(ii)"}],"minor_comments":[{"comment":"The phrasing 'Theorem 2.6 does not state the existence of some t0 above or below which the heat content of any graph ... will be dominated' is confusing, since the theorem does assert existence of a threshold for each fixed graph. The intended point is that the threshold is not uniform over the class of graphs; the remark should be rewored to say this clearly.","section":"Remark 2.8"},{"comment":"The proof of Lemma 4.7 uses the notation ∫_0^t ∫_0^s dρ_x dρ_y, which is not standard. It would be clearer to write the convolution integral with respect to the distribution functions, e.g., ∫_0^t P[X ≤ t-s] dP[Y ≤ s].","section":"Lemma 4.7"},{"comment":"The notation B_{τ^BM_{v_D}} in (4.4) is introduced without definition; since the sum is over paths with v_n = v_D, the indicator 1{v_n = B_{τ^BM_{v_D}}} should simply be 1{v_n = v_D}.","section":"Section 4, equation (4.4)"},{"comment":"There are several typographical errors, including 'satiesfies' in Definition 2.5, 'correponding' in Remark 2.4, and 'is is' in Section 2.2. These should be corrected in the final version.","section":"General"}],"recommendation":"major_revision","confidential_remarks":"The paper relies at a central point on the unpublished companion paper [4] for the degree-monotonicity lemma. This is a concern for the completeness of the manuscript: the editor may want to verify that [4] is indeed available or in press, and that the lemma is stated in sufficient detail there. The sign error in (4.11) is local and likely fixable, but it must be corrected before the proof is verifiable. The large-time part is solid. Overall, the paper is promising but the small-time theorem as stated is not yet fully established."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear [Colleague],\n\nI've read arXiv:2501.09693 carefully. My take: this is a genuine contribution. The authors prove a Faber-Krahn inequality for the heat content on quantum graphs at large times and, under a rational-edge-length assumption, at small times. The large-time argument via Mercer's theorem is clean and correct. The small-time argument rests on a new random walk expansion (Theorem 4.1) expressing heat content as a positive linear combination of expected return times of discrete simple random walks. That representation is new, and it's a nice technical step—it converts a sum with sign-indefinite Roth terms into one where Lovász's formula E[τ]=2|E|/deg(v) can be applied. The paper is honest about its limitations: it flags the rational-length condition in the theorem, notes the all-time question is open, and even includes a cautious conjecture.\n\nThe soft spots are real but not fatal. First, inequality (4.11) has a sign error: the line before it contains a negative tail -N(ℓ-α_{k0+1}(t)), but (4.11) writes +N(...). There's also a missing factor 1/2 in the first term. Both look typographical; with the correct sign and factor, the lower bound still becomes positive for small t by Lemma 4.7. Needs fixing, but the argument is reconstructable.\n\nSecond, the reduction to degree-one Dirichlet vertices relies on an unpublished monotonicity lemma from [4]. If that lemma holds, the proof works; if not, Theorem 2.6(ii) covers only deg(vD)=1. This is a genuine dependency. A referee will want either a proof in an appendix or assurance that [4] is available. It's likely true, but it's currently an external crutch.\n\nThird, the abstract says the inequality holds at small times without mentioning the rational-length condition. That's an overstatement.\n\nThere are a couple of minor typos (the integration limits in (3.3) are backwards, and (4.11) as noted). None affect the substance.\n\nWho is this for? Spectral geometers working on quantum graphs, and stochastic analysts interested in heat content. It deserves a serious referee. With minor revisions—fix the typos, address the [4] dependency, and refine the abstract—it's publishable.\n\nRecommendation: engage. Send to peer review.","headline":"Solid new Faber-Krahn result for heat content on quantum graphs; small-time proof has a typo and an unpublished dependency, but the core is sound and deserves a serious referee.","tokens_in":17331,"tokens_out":10463,"would_cite":true,"duration_ms":93066,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["34B45","05C81","49Q10"],"pacs":[],"model":"deepseek-v4-flash","headline":"The heat content of a compact metric graph is maximized at small and large times by the interval of equal length, with equality only for the interval.","keywords":["heat content","Faber-Krahn inequality","quantum graphs","random walk expansion","Feynman-Kac formula","shape optimisation","Brownian motion on metric graphs","Mercer theorem"],"falsifier":"Using the exact formula (4.1), evaluate the heat content difference $Q_t(G; v_D)-Q_t(P_{|G|};\\{0\\})$ for an equilateral star with three edges and the Dirichlet vertex at one endpoint as $t\\to0^+$; a negative value at arbitrarily small times would disprove Theorem 2.6(ii), while the proof predicts the difference stays positive.","tokens_in":16324,"feed_emoji":"🔥","tokens_out":11400,"duration_ms":104461,"temperature":0.7,"pith_summary":"The paper asks whether, among quantum graphs (networks of intervals glued at their endpoints) of equal total length, the interval retains the most heat when one boundary point is held at zero temperature. It proves that this Faber–Krahn-type statement is true at the two extremes of the time range: for all sufficiently large times, and for all sufficiently small times provided the edge lengths are rationally dependent (multiples of a common unit). In both regimes the interval is the unique maximizer, meaning any other graph of the same length loses strictly more heat. The result matters because heat content has no variational characterization, so it cannot be compared by the usual Rayleigh-quotient tools; the authors introduce a positive random-walk expansion that reduces the comparison to return times of a discrete random walk.","feed_headline":"Path graphs maximize heat content at small and large times","feed_subtitle":"At very short and very long times, the interval beats every other graph of equal volume.","key_machinery":"The key identity is Theorem 4.1, a random-walk expansion of the heat content on an equilateral metric graph. It expresses $Q_t(G; v_D)$ as $\\frac{\\deg(v_D)}{2}$ times a positive linear combination of terms built from the first-return time $\\tau_{v_D}$ of the symmetric discrete random walk on the combinatorial graph, with coefficients $\\alpha_n(t)$ that depend only on the edge length and the time, not on the topology. The argument works because two facts meet: Proposition 3.7 gives the explicit value $\\mathbb{E}_{v_D}[\\tau_{v_D}] = 2\\#E/\\deg(v_D)$, and Lemma 4.7 shows that as $t\\to0^+$ the short-path terms dominate the long-path terms through the hierarchy $(\\ell-\\alpha_k(t))/(\\alpha_k(t)-\\alpha_{k-1}(t))\\to0$. The resulting comparison formula (Corollary 4.9) separates the combinatorial return-time probabilities from the universal coefficients, which is exactly what makes the path graph win at small times.","core_discovery":"The paper's central result, Theorem 2.6, is an extremal Faber–Krahn inequality for the heat content $Q_t(G; V_D)$ on a compact metric graph: among graphs of fixed total length $|G|$ with a nonempty Dirichlet set and connected complement, the path graph $P_{|G|}$ with the Dirichlet condition at an endpoint maximizes the heat content for all sufficiently large $t$, and for all sufficiently small $t$ when the edge lengths are rationally dependent. Equality at those times forces $G$ to be the path graph itself. The large-time half follows from Mercer's theorem: the heat content is a sum of $e^{-t\\lambda_k}|\\langle \\varphi_k,1\\rangle|^2$, and the known fact that the interval minimizes the first Dirichlet eigenvalue makes the first term dominate. The small-time half is proved by a new positive expansion of $Q_t$ as a linear combination of expected return times of the symmetric discrete random walk.","pith_inferences":["A natural next step, not taken in the paper, is to test whether the rational-dependence condition in the small-time theorem can be dropped by subdividing edges in a controlled limit, since the expansion is insensitive to topology apart from return-time probabilities.","One could isolate the unpublished monotonicity assumption by computing the heat content of two equilateral graphs that differ only by adding a Neumann pendant edge at the Dirichlet vertex; if the heat content rose with degree, the small-time theorem would need rephrasing for degree-one Dirichlet vertices.","The same positive expansion may apply to other heat-type functionals, for example the time-integrated heat content (torsional rigidity), where a Faber–Krahn inequality is already known by different methods.","A numerical search over pitchfork-shaped graphs—a long edge with the Dirichlet endpoint and two short stubs at the other end—would be the sharpest probe of whether the inequality holds at intermediate times, as the paper itself identifies these as the most delicate candidates."],"forward_implications":["For large times, every compact metric graph of fixed total length carries no more heat than the interval of the same length, and equality identifies the interval.","For small times, the same statement holds within the class of graphs with rationally dependent edge lengths, again with the interval as unique maximizer.","The random-walk expansion gives a comparison formula in which two graphs can be ordered by their return-time probabilities alone, without revisiting the heat equation.","The large-time bound is a transfer of the classical Faber–Krahn theorem for the first eigenvalue to the heat content through the Mercer expansion.","Between the two extremes, the question of whether the interval maximizes heat content at every time remains open."],"supporting_citations":[{"why":"Supplies the Feynman–Kac representation of the heat semigroup as survival probability before hitting the Dirichlet vertex, which is the starting point of the stochastic proof.","marker":"[1]"},{"why":"Provides the heat-content path formula on combinatorial graphs and the monotonicity in Dirichlet-vertex degree used to reduce to the degree-one case.","marker":"[4]"},{"why":"Gives the extremal result that the interval minimizes the first Dirichlet eigenvalue, which is the spectral input for the large-time comparison.","marker":"[13]"},{"why":"Supplies the one-dimensional hitting-time and reflection-principle formulas that define the coefficients $\\alpha_n(t)$.","marker":"[15]"},{"why":"Establishes the Brownian-motion-on-metric-graph framework from which the discrete random walk of successively hit vertices is extracted.","marker":"[18]"},{"why":"Provides the expected-return-time formula $\\mathbb{E}_{v_D}[\\tau_{v_D}] = 2\\#E/\\deg(v_D)$ that evaluates the infinite sums in the random-walk expansion.","marker":"[21]"},{"why":"Proves the Faber–Krahn inequality for the first eigenvalue of a metric graph, which is the comparison used in the large-time Mercer argument.","marker":"[25]"}],"fun_headline_variants":["Intervals maximize heat content at short and long times","Random walk expansion proves heat-content extremal bound","Heat-content Faber-Krahn holds at time extremes on graphs","Path graphs beat all graphs in heat content for extreme t","Extremal heat content: intervals trump other graphs at t→0,∞"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The small-time proof relies on an unpublished monotonicity lemma from [4]—that heat content decreases when the Dirichlet vertex's degree increases—and on the explicit hypothesis that all edge lengths are rationally dependent.","fun_headline_variants_meta":{"raw":{"variants":["Intervals maximize heat content at short and long times","Random walk expansion proves heat-content extremal bound","Heat-content Faber-Krahn holds at time extremes on graphs","Path graphs beat all graphs in heat content for extreme t","Extremal heat content: intervals trump other graphs at t→0,∞"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00094,"raw_usage":{"total_tokens":4018,"prompt_tokens":943,"completion_tokens":3075,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":559,"completion_tokens_details":{"reasoning_tokens":3005}},"tokens_in":559,"tokens_out":3075,"duration_ms":23034,"temperature":1.0,"reasoning_tokens":3005,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T19:48:08.645482+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Using the exact formula (4.1), evaluate the heat content difference $Q_t(G; v_D)-Q_t(P_{|G|};\\{0\\})$ for an equilateral star with three edges and the Dirichlet vertex at one endpoint as $t\\to0^+$; a negative value at arbitrarily small times would disprove Theorem 2.6(ii), while the proof predicts the difference stays positive.","supporting_citations":[{"cited_title":"Becker, F","cited_arxiv_id":null,"evidence_quote":"Supplies the Feynman–Kac representation of the heat semigroup as survival probability before hitting the Dirichlet vertex, which is the starting point of the stochastic proof."},{"cited_title":"Bifulco and D","cited_arxiv_id":null,"evidence_quote":"Provides the heat-content path formula on combinatorial graphs and the monotonicity in Dirichlet-vertex degree used to reduce to the degree-one case."},{"cited_title":"Friedlander","cited_arxiv_id":null,"evidence_quote":"Gives the extremal result that the interval minimizes the first Dirichlet eigenvalue, which is the spectral input for the large-time comparison."},{"cited_title":"Karatzas and S","cited_arxiv_id":null,"evidence_quote":"Supplies the one-dimensional hitting-time and reflection-principle formulas that define the coefficients $\\alpha_n(t)$."},{"cited_title":"Kostrykin, J","cited_arxiv_id":null,"evidence_quote":"Establishes the Brownian-motion-on-metric-graph framework from which the discrete random walk of successively hit vertices is extracted."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the expected-return-time formula $\\mathbb{E}_{v_D}[\\tau_{v_D}] = 2\\#E/\\deg(v_D)$ that evaluates the infinite sums in the random-walk expansion."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Proves the Faber–Krahn inequality for the first eigenvalue of a metric graph, which is the comparison used in the large-time Mercer argument."}],"review_version":1}