{"id":"29384c2b-3ca1-4911-8c06-1d9034397ebf","arxiv_id":"2502.09461","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For compact metric graphs with Dirichlet conditions, the heat content is shown to equal the volume minus a boundary term plus a weighted sum over Dirichlet-to-Dirichlet paths, for all positive times.","lead":"The authors prove an exact formula for the heat content of a diffusion process on a compact network of wires with absorbing endpoints, expressed through the network volume and closed paths that touch the absorbing set. The formula holds for every time and yields a sharper short-time asymptotic expansion, along with comparison principles across different network topologies.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Path-count estimate (5.5) is false for stars with d>3 Dirichlet leaves, so Theorem 5.1 and Corollary 6.12 are not proved as written; the asymptotic claim is likely fixable with a #VD factor.","rationale":"After checking the algebraic path to Theorem 4.1 (Lemmas 3.6, 3.7, 4.7, 4.9 and the telescoping in (4.33)–(4.35)), I find no gap in the all-time heat-content formula itself. The reader's pinpointed counterexample is correct: directed paths from VD to VD of length 2 on a star with d Dirichlet leaves number d^2, but (5.5) would cap them at 2d. This estimate is used exactly where the small-time remainder is converted into a geometric series, so the proof of Theorem 5.1 is invalid as written; Corollary 6.12 uses the same estimate. The fix is straightforward and does not disturb the main theorem: count starts in VD as well as choices at subsequent vertices, giving #VD dmax^{n-1}. Because the exponential factor e^{-n²ℓmin²/(4t)} decays super-exponentially, the modified geometric-series argument still produces an O(√t e^{-ℓmin²/(4t)}) remainder, with a #VD-dependent constant. A secondary concern is that Proposition 6.5 is stated without proof and rests on an unpublished preprint [14] with overlapping authorship; however, this affects only auxiliary comparison principles, not the central formula, so it does not change the conditional verdict. Overall, the reader's CONDITIONAL verdict is appropriate: the main claim is plausible and the identified flaw is localized and fixable.","tokens_in":43343,"tokens_out":19014,"duration_ms":166382,"concrete_test":"Enumerate all directed paths of combinatorial length 2 in the equilateral star with d=4 Dirichlet leaves (and no other edges). There are 16, so (5.5) fails. Then replace (5.5) by #(Pn∩PVD)≤#VD(dmax)^{n-1} and re-run the derivation of (5.6); verify that the final bound becomes |Lt|≤8#VD√t/√π · e^{-ℓmin²/(4t)}/(1−dmax e^{-ℓmin²/(2t)}), which still implies the O(√t e^{-ℓmin²/(4t)}) asymptotics. If this corrected estimate holds, Theorem 5.1's O-term stands but the literal inequality (5.1) needs a #VD factor.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The proof of Theorem 5.1 bounds the remainder Lt(G;VD) in (5.3) using the claim in (5.5) that #(Pn(G)∩PVD(G))≤2(dmax)^{n-1}. This is false. On an equilateral star with d Dirichlet leaves (Assumption 2.1 holds for d≥1), every ordered pair of leaves (vi,vj) gives a directed path vi→center→vj of combinatorial length 2, so #(P2∩PVD)=d^2, while 2(dmax)^{2-1}=2d. For d>3 the claimed bound is exceeded by a factor d/2. Consequently the estimate (5.6), and hence the stated inequality (5.1) and the O-term (5.2), are not established by the argument in the manuscript. Since Corollary 6.12 invokes (5.5) in (6.14), its proof inherits the same gap. The main all-time formula Theorem 4.1 does not use (5.5), and the asymptotic statement is probably repairable by replacing (5.5) with #(Pn∩PVD)≤#VD(dmax)^{n-1}; the resulting bound then carries an extra factor #VD but preserves the exponential order. Thus the issue is a proof gap in a headline derivative result, not a disproof of the central formula.","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 finite metric graph with Dirichlet conditions imposed on a nonempty set V_D of degree-one vertices. The main result, Theorem 4.1, is an exact all-times formula expressing Q_t(G;V_D) as |G| - (2√t/√π)#V_D plus a scattering-weighted path sum over closed directed/undirected paths that start and end in V_D. The authors use this formula to derive a short-time asymptotic expansion, a Caccioppoli-type description of the perimeter of a subgraph, several surgery and comparison principles, and a Hadamard-type formula for variations of edge lengths. The proof of the main formula proceeds by integrating a Roth/KPS path-sum representation of the heat kernel and then simplifying the resulting sums through combinatorial lemmas in Section 3.","tokens_in":43620,"tokens_out":9231,"duration_ms":86945,"significance":"The exact heat-content formula is a significant contribution if it is correct: it gives a fully combinatorial description of the heat content at all times, with no fitted parameters, and it yields an exponentially small remainder in the short-time limit, which is stronger than the polynomial remainders typical in the manifold setting. The surgery principles, especially the mirroring formula of Theorem 6.8 and the loop-cut invariance of Proposition 6.4, are likely to be useful for spectral-geometric comparisons of quantum graphs. The paper also contains explicit worked examples, including intervals, stars, lassos, and pumpkin chains, which help to make the path-sum formalism concrete. The central all-time formula of Theorem 4.1 appears derivationally sound; the principal defect is a false path-counting estimate used in the proof of the small-time asymptotics and of Corollary 6.12.","major_comments":[{"comment":"Equation (5.5) is false. It claims #(P_n(G) ∩ P_{V_D}(G)) ≤ 2(d_max)^{n-1} for every n, but the bound does not depend on #V_D. Let G be an equilateral star with d ≥ 4 leaves, all leaves lying in V_D and the centre in V_N; Assumption 2.1 holds. Then every ordered pair of leaves (v_i,v_j) gives a directed path v_i → centre → v_j of combinatorial length 2, so #(P_2(G) ∩ P_{V_D}(G)) = d^2, while the claimed upper bound is 2d. The extra factor 2 is also unnecessary, since P_n(G) already consists of directed paths. Thus the estimate (5.6) is not established by the argument given.","section":"§5, Eq. (5.5)"},{"comment":"Because (5.5) is the only control on the number of paths used to bound the remainder L_t(G;V_D) in (5.3), the small-time bound (5.1) and the O-term (5.2) of Theorem 5.1 are not proved as written. The same false estimate is invoked in (6.14) in the proof of Corollary 6.12, so that result also lacks a valid proof. This is a repairable proof gap rather than a disproof of the asymptotic claim: replacing (5.5) by the valid bound #(P_n(G) ∩ P_{V_D}(G)) ≤ #V_D (d_max)^{n-1} gives an additional factor #V_D in (5.6) and (6.14) but preserves the exponential order, so the stated O(√t e^{-ℓ_min^2/(4t)}) remainder is likely correct. I ask the authors to correct (5.5) and to rerun the estimates in Theorem 5.1 and Corollary 6.12 accordingly.","section":"§5, Eq. (5.6)–(5.8) and §6.4, Eq. (6.14)"}],"minor_comments":[{"comment":"The displayed title contains broken spacing in 'HEA T CONTENT' and 'COMP ACT'; this should be corrected in the final version.","section":"Title"},{"comment":"The statement of Theorem 4.1 in (4.32) is in terms of directed paths with a factor 4√t, while Theorem A and (1.3) use undirected paths with a factor 8√t. This is consistent with the two orientations of each undirected path, but the equivalence should be stated explicitly when the theorem is announced.","section":"Theorem 4.1 and Eq. (1.3)"},{"comment":"In the displayed computation for the lasso graph, the quadruple sum over n,k,ℓ,j and the condition n+k≤m+1 are introduced without explaining how m and the four counters correspond to the path statistics R_{v_D}, R^{(1)}_{v_N}, R^{(2)}_{v_N}, T^{(1)}_{v_N}, T^{(2)}_{v_N}; the example would be much clearer if this dictionary were spelled out.","section":"Example 4.10"},{"comment":"The proof says that imposing an additional Dirichlet condition 'raises the heat content by Proposition 6.2', but Proposition 6.2 states the opposite inequality, Q_t(G;V_D∪{v_0}) ≤ Q_t(G;V_D). The wording should be corrected and the three steps of the surgery argument should be reordered more explicitly.","section":"Corollary 6.3, proof"},{"comment":"The bound (5.15) counts directed topological paths by assigning at most d_max choices at each vertex, but the role of the parameter n and the extra factor d_max^{n+1} are not precisely defined. A short counting lemma analogous to a corrected (5.5) would make this estimate rigorous and easier to verify.","section":"Theorem 5.5, Eq. (5.15)"},{"comment":"In the displayed heat-kernel asymptotics, the notation dist(x,V) is used but the distance to the vertex set V is not defined until later in the same remark; please define it before first use.","section":"Remark 5.2"}],"recommendation":"major_revision","confidential_remarks":"The central all-time formula of Theorem 4.1 appears sound and original, and the algebraic cancellation argument in Section 4 is careful. The false path-counting estimate in (5.5) is a genuine proof gap, but it is local and fixable by replacing it with a bound carrying a factor #V_D; the paper should therefore be sent back for revision rather than rejected. I would also ask the authors to tighten the counting in Theorem 5.5 and to fix the presentation issues listed in the minor comments."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The main formula is new and the derivation is careful, but the proof of the small-time asymptotic has a false counting estimate that needs fixing.\n\nWhat the paper does well: Bifulco and Mugnolo integrate the Roth/KPS path-sum kernel to get an exact all-time heat content formula for compact quantum graphs with Dirichlet vertices. The formula is genuinely new, and the algebraic cancellation in Section 4 checks out. The surgery principles, especially the mirroring result (Theorem 6.8), are substantial and give usable tools. The exposition is modular and mostly clear.\n\nThe soft spot: the proof of Theorem 5.1 relies on the path-count estimate (5.5), which is false. For an equilateral star with d Dirichlet leaves, there are d^2 directed length-2 paths from the Dirichlet set to itself, but (5.5) gives at most 2d. For d>3 the bound is exceeded. The stated O(sqrt(t) exp(-ell_min^2/(4t))) remainder is probably still true, because replacing (5.5) with #V_D * dmax^{n-1} preserves the exponential order with an extra constant factor. But as written, the proof of Theorem 5.1 and of Corollary 6.12, which inherits (5.5) in (6.14), does not establish the estimate. The main all-time formula, Theorem 4.1, does not use (5.5), so the central result stands.\n\nAlso, Proposition 6.5 is stated without proof and deferred to an unpublished preprint with overlapping authorship. That is a citation-practice issue; a referee should ask for a proof or a public version.\n\nThis paper deserves a serious referee. The central formula appears correct and is a real contribution to spectral geometry of quantum graphs. The small-time asymptotic is likely fixable but needs a corrected counting argument. The paper is for researchers working on quantum graphs and heat kernel asymptotics. I would engage with it, but I would want the path-counting issue resolved before relying on Theorems 5.1 and 6.12.","headline":"Solid new heat-content formula for quantum graphs, but the small-time asymptotic proof rests on a false path-count bound; fixable, worth refereeing.","tokens_in":44112,"tokens_out":2730,"would_cite":true,"duration_ms":22855,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["34B45","05C50","35P15","81Q35"],"pacs":[],"model":"deepseek-v4-flash","headline":"The heat content of a compact metric graph with Dirichlet conditions is determined exactly, for all times, by its volume, its number of Dirichlet vertices, and a scattering-weighted sum over closed paths hitting the boundary.","keywords":["heat content","quantum graphs","metric graphs","Dirichlet boundary conditions","small-time asymptotics","heat kernel","scattering coefficients","surgery principles"],"falsifier":"On an equilateral star graph with $d>3$ Dirichlet leaves, compute the heat content exactly from the eigenfunction expansion in the paper and evaluate the right-hand side of the path-sum formula (4.1) with paths enumerated up to large combinatorial length; the two must agree as the truncation grows. To test the asymptotic bound, divide the remainder $Q_t-|G|+2\\sqrt{t}/\\sqrt{\\pi}\\#V_D$ by $\\sqrt{t}\\,e^{-\\ell_{\\min}^2/(4t)}$ and check boundedness as $t\\to0^+$ on the same star, where the paper's counting estimate fails.","tokens_in":43114,"feed_emoji":"🌡️","tokens_out":10992,"duration_ms":114476,"temperature":0.7,"pith_summary":"The paper proves an exact, all-time formula for the heat content of a compact metric graph with Dirichlet conditions on a chosen set of degree-one vertices. The formula expresses the heat content as the graph's total length, minus a boundary correction proportional to the number of Dirichlet vertices and to $\\sqrt{t}$, plus a sum over closed paths that touch the Dirichlet set and are weighted by scattering coefficients. The formula turns the heat content into a combinatorial quantity that can be read off from the graph's geometry, and it yields a short-time expansion whose leading coefficients record the volume and the number of Dirichlet vertices. The paper also derives comparison principles under graph surgeries, such as cutting loops, attaching graphs, and mirroring graphs.","feed_headline":"Every quantum graph's heat content is a sum over boundary paths","feed_subtitle":"New formula holds for all times and reads volume and Dirichlet vertices off the short-time expansion.","key_machinery":"The argument rests on a path-sum representation of the heat kernel: the heat kernel $p^{G;V_D}_t(x,y)$ is written as $(4\\pi t)^{-1/2}$ times a sum over directed paths from $x$ to $y$, each weighted by a scattering coefficient $\\alpha(\\vec p)$ and by $e^{-\\ell(\\vec p)^2/(4t)}$. Two combinatorial lemmas, a decomposition lemma and a summation lemma for scattering coefficients of extended paths, together reduce the double integral over $x,y$ to a single sum over closed paths that start and end at the Dirichlet set. The other key object is the function $H(x)=\\pi^{-1/2}e^{-x^2}-x\\,\\mathrm{erfc}(x)$, which arises from integrating the one-dimensional Gaussian kernel and controls the exact remainder in the heat content formula.","core_discovery":"The central claim is that the heat content $Q_t(G;V_D)$ of a compact finite metric graph with Dirichlet conditions on a nonempty set of degree-one vertices equals $$Q_t(G;V_D)=|G|-\\frac{2\\sqrt{t}}{\\sqrt{\\pi}}\\#V_D+8\\sqrt{t}\\sum_{p}\\$\\alpha$(p)\\,H\\!\\left(\\frac{\\ell(p)}{2\\sqrt{t}}\\right)$$ for all $t>0$, where the sum runs over undirected paths that start and end at (possibly different) Dirichlet vertices, $\\alpha(p)\\in[-1,1]$ is a product of scattering coefficients collected at the vertices the path traverses, and $H$ is a monotone, exponentially decaying function. The same formula is equivalent to a sum over directed paths with factor $4\\sqrt{t}$. As a consequence, the small-time expansion begins $$Q_t(G;V_D)=|G|-\\frac{2\\sqrt{t}}{\\sqrt{\\pi}}\\#V_D+O\\!\\left(\\sqrt{t}\\,$e^{{-\\ell_{\\min}}$^2/(4t)}\\right)\\quad(t\\to0^+),$$ with $\\ell_{\\min}$ the minimal edge length. The paper further states a Caccioppoli-type limit in which the heat flowing out of a subregion identifies the number of boundary points, and a set of surgery principles that compare heat contents of different graphs.","pith_inferences":["The exponentially decaying factor in the path sum suggests a practical numerical scheme: truncate the sum at a path-length cutoff and use the $H$ decay to control the error, with convergence expected to be rapid at moderate times.","The Hadamard-type derivative with respect to edge length, combined with the scaling law, could yield monotonicity or isoperimetric-type statements for the heat content under geometric variations, in the spirit of known results for the torsional rigidity.","The failure of the path-counting estimate on stars with many Dirichlet leaves indicates that a corrected proof of the small-time remainder should replace $d_{\\max}$ by a factor involving $\\#V_D$; the asymptotic statement itself may still be true.","Because the derivation uses only a path decomposition of the heat kernel plus the Gaussian kernel on $\\mathbb{R}$, the same integration procedure should extend to Schrödinger operators or to other vertex conditions with adapted scattering coefficients."],"forward_implications":["The heat content at any positive time can be computed from the graph's volume, the number of Dirichlet vertices, and the scattering-weighted set of closed paths hitting the boundary, without solving the heat equation.","As $t\\to0^+$, $Q_t(G;V_D)=|G|-\\frac{2}{\\sqrt{\\pi}}\\#V_D\\sqrt{t}+O\\!\\left(\\sqrt{t}\\,e^{-\\ell_{\\min}^2/(4t)}\\right)$, so the first two coefficients of the short-time expansion are purely geometric.","For a closed connected subset $H$ of $G\\setminus V_D$ whose boundary avoids vertices of degree at least three, the rescaled heat flowing from $H$ into its complement tends to $\\#\\partial H$ as $t\\to0^+$.","Cutting a loop at its midpoint leaves the heat content unchanged at all times; mirroring a graph $m$ times multiplies the heat content by $m$; attaching a pendant graph increases the heat content; and lengthening an edge increases the heat content for small times.","Scaling all edge lengths by $s>0$ gives $Q_t(sG;V_D)=s\\,Q_{t/s^2}(G;V_D)$ for all $t>0$."],"supporting_citations":[{"why":"Supplies the path-sum expansion of the heat kernel on a metric graph, the starting point for the integrated formula.","marker":"[48]"},{"why":"Extends the path-sum formula to Dirichlet conditions at distinguished vertices, giving the heat kernel representation used in Proposition 4.2.","marker":"[28]"},{"why":"Establishes the classical small-time heat content asymptotics on domains whose graph analogue is proved here.","marker":"[6]"},{"why":"Provides the manifold heat content asymptotics that motivate the short-time expansion and the Caccioppoli-type limit.","marker":"[8]"}],"fun_headline_variants":["Heat content of quantum graphs: exact sum over boundary paths","All-time heat content formula for compact quantum graphs","Quantum graph heat content: boundary paths tell all","Exact heat content for all times on quantum graphs"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof of the small-time asymptotics relies on the estimate that the number of directed paths of combinatorial length $n$ that start and end at the Dirichlet set is at most $2(d_{\\max})^{n-1}$, where $d_{\\max}$ is the maximal vertex degree; this estimate is false for an equilateral star with more than three Dirichlet leaves, so the remainder bound as written does not follow for such graphs.","fun_headline_variants_meta":{"raw":{"variants":["Heat content of quantum graphs: exact sum over boundary paths","All-time heat content formula for compact quantum graphs","Quantum graph heat content: boundary paths tell all","Exact heat content for all times on quantum graphs"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000461,"raw_usage":{"total_tokens":2310,"prompt_tokens":954,"completion_tokens":1356,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":570,"completion_tokens_details":{"reasoning_tokens":1294}},"tokens_in":570,"tokens_out":1356,"duration_ms":8496,"temperature":1.0,"reasoning_tokens":1294,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T21:26:11.562860+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"On an equilateral star graph with $d>3$ Dirichlet leaves, compute the heat content exactly from the eigenfunction expansion in the paper and evaluate the right-hand side of the path-sum formula (4.1) with paths enumerated up to large combinatorial length; the two must agree as the truncation grows. To test the asymptotic bound, divide the remainder $Q_t-|G|+2\\sqrt{t}/\\sqrt{\\pi}\\#V_D$ by $\\sqrt{t}\\,e^{-\\ell_{\\min}^2/(4t)}$ and check boundedness as $t\\to0^+$ on the same star, where the paper's counting estimate fails.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the path-sum expansion of the heat kernel on a metric graph, the starting point for the integrated formula."},{"cited_title":"Kostrykin, J","cited_arxiv_id":null,"evidence_quote":"Extends the path-sum formula to Dirichlet conditions at distinguished vertices, giving the heat kernel representation used in Proposition 4.2."},{"cited_title":"van den Berg and E.B","cited_arxiv_id":null,"evidence_quote":"Establishes the classical small-time heat content asymptotics on domains whose graph analogue is proved here."},{"cited_title":"van den Berg and P.B","cited_arxiv_id":null,"evidence_quote":"Provides the manifold heat content asymptotics that motivate the short-time expansion and the Caccioppoli-type limit."}],"review_version":1}