{"id":"ee4226d6-8de3-4f35-8b0a-bbbad80c675b","arxiv_id":"2411.15169","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For initial states that are piecewise polynomials on axis-aligned grids, the heat equation solution is a finite sum of products of error functions and Gaussians.","lead":"This paper gives exact formulas for how heat or diffusing matter spreads when the starting distribution is made of polynomial pieces in one, two, or three dimensions. The solutions come out as short lists of Gaussians and error functions, which are easy to compute and useful for checking numerical simulations.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Recurrence (11) is false as stated, so the advertised reduction of χp to Gaussians/error functions is not established; the exactness of formula (7) itself is not affected.","rationale":"Most of the construction is sound: equation (7) follows from repeated integration by parts with derivative jumps, and the tensor-product analogues (21) and (29) are natural extensions whose corner-sign bookkeeping passes a simple rectangle test. The reader correctly identifies the weak point: the sequence (5) defines χp by definite integrals with lower limit 0, but recurrence (11) is derived as if the antiderivative constants were chosen to satisfy it. For p=1, (11) fails at x=0 by √(κt/π). This does not invalidate (7), because the failure terms are lower-limit constants/polynomials whose sums over compact-support jumps should cancel, but that cancellation is never proved. Consequently the advertised finite Gaussian/error-function representation of χp is not established by the printed derivation, and some examples contain numerical factors inconsistent with their t→0 limits. These are correctable issues, so a conditional acceptance with requests for a corrected recurrence and corrected examples is the appropriate verdict.","tokens_in":11057,"tokens_out":17405,"duration_ms":185725,"concrete_test":"Verify Eq. (11) directly from definition (5): compute χ1(0,t)=0 and xχ0(0,t)+2κt χ−1(0,t)=√(κt/π); the inequality settles that the recurrence is false. Then re-derive the §2.5.2 triangular-profile solution using the correct low-order χp expressions and check that at t=0 it returns 1−|x|/a; the printed formula will fail this check unless corrected.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central closed-form claim rests on Eq. (11): p χp(x,t) = x χp−1(x,t) + 2κt χp−2(x,t). With the definition (5), χp+1(x,t)=∫_0^x χp(x′,t) dx′, this recurrence is not an identity. Direct evaluation at p=1, x=0 gives χ1(0,t)=0, while the right-hand side is 2κt χ−1(0,t)=√(κt/π). The recurrence would hold only for antiderivatives with special integration constants; the definite lower-limit-0 integrals acquire lower-limit corrections (constants and polynomial terms) that are not accounted for. Thus the proof that every χp is a finite combination of Gaussians and error functions is invalid as printed. Formula (7) itself is an exact integration-by-parts identity and is independent of this recurrence, and the correction terms should cancel in compact-support sums, but the paper neither states nor proves the required moment cancellations. Consistent with this, example formulas using χp are also wrong: in §2.5.2 the factor 1/2 and the missing 1/a make the t→0 limit 1/2(1−|x|/a) rather than 1−|x|/a for a=1. These are correctable defects, but the advertised closed-form theorem is not established by the derivation given.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper claims an exact solution of the heat equation for initial data that are piecewise polynomials. In one dimension the solution is written as a finite sum of repeated antiderivatives chi_p of the heat kernel, weighted by jumps in derivatives at the segment boundaries (Eq. (7)). Two- and three-dimensional extensions are given for polynomial data on rectangular grids (Eqs. (21) and (29)), together with examples including square, triangular, B-spline, rectangular, bilinear, and trilinear data. The central advertised result is that every chi_p is a finite combination of Gaussians and error functions, so that the exact evolution is closed-form. A section on Schrödinger's equation notes that the same method applies with complex coefficients.","tokens_in":11376,"tokens_out":14305,"duration_ms":137622,"significance":"The core idea of expressing the evolution through derivative jumps is attractive, and the integration-by-parts identity behind Eq. (7) is correct and potentially useful for benchmarking numerical heat-equation solvers. The paper is clearly organized and the intended range of applicability is well explained. However, the proof that the reduced functions are finite combinations of Gaussians and error functions relies on a false recurrence, the alternative formulation rests on a false statement about the heat equation, and the triangular example contains explicit factor errors. The manuscript therefore is not publishable in its present form, although the defects are local and in principle fixable. No code or machine-checked proofs are supplied; the analytic derivation is the content.","major_comments":[{"comment":"The recurrence p chi_p(x,t) = x chi_{p-1}(x,t) + 2 kappa t chi_{p-2}(x,t) is not an identity for the antiderivatives defined with lower limit 0 in Eq. (5). At p=1 and x=0 the left-hand side is chi_1(0,t)=0, while the right-hand side is 2 kappa t chi_{-1}(0,t) = sqrt(kappa t / pi), a nonzero value. The same defect appears in Eq. (10): E_1(z) = int_0^z erf(z') dz' = z erf(z) + pi^{-1/2}(e^{-z^2}-1), whereas (10) gives z erf(z) + 2 pi^{-1/2} e^{-z^2}, omitting the constant -pi^{-1/2}. Consequently the statement in §2.2 that every chi_p is a finite combination of Gaussians and error functions is not established by the derivation as printed; the headline claim of the abstract depends on this reduction.","section":"§2.2, Eq. (11)"},{"comment":"The alternative approach asserts that each chi_p(x,t) satisfies the heat equation and therefore replacing f_p by chi_p gives the evolution. This is false for the lower-limit-0 integrals: for chi_0, d_t chi_0 - kappa d_x^2 chi_0 = -kappa chi(0,t) != 0, and more generally d_t chi_{p+1} - kappa d_x^2 chi_{p+1} = -kappa chi_{p-1}(0,t). The one-dimensional truncated-polynomial formula may still be correct because the boundary constants cancel in the alternating sums, but the paper provides no such cancellation proof, and the product forms used in Eqs. (20) and (28) inherit the same gap. This is load-bearing because the 2D and 3D extensions are derived only through this replacement.","section":"§2.4"},{"comment":"The triangular example contains two concrete errors. The stated representation u(x)=f_1(x-a_0)-2 f_1(x-a_1)+f_1(x-a_2) evaluates to x+a for -a<x<0 and a-x for 0<x<a, i.e., to a u(x), not u(x); a factor 1/a is missing. The printed evolution formula additionally contains an unjustified prefactor 1/2. As a result, the t=0 limit of the plotted evolution is (1/2)(1-|x|/a) for a=1 rather than 1-|x|/a. This error in a central illustrative example needs to be corrected and the figures regenerated.","section":"§2.5.2"}],"minor_comments":[{"comment":"The word 'indeces' in the remark at the end of the section should be 'indices'.","section":"§2.1"},{"comment":"The cross-reference 'Fig.2(a)' in the paragraph after Eq. (24) should be 'Fig.4(a)'.","section":"§3.3.2"},{"comment":"The word 'paralleliped' should be 'parallelepiped'.","section":"§4"},{"comment":"The half-mass radius for the uniform sphere is written as 'a-3/2 ≈ 0.7937 a'; this should be a/2^{1/3} or 2^{-1/3} a.","section":"§4.1.1"},{"comment":"The phrase 'tailored to to a rectangular mesh' contains a duplicated 'to'.","section":"§6"},{"comment":"The notation remark '[The symbols p, q, r are used to denote indeces, never powers.]' is unnecessary and slightly confusing; consider removing it.","section":"§2.2"}],"recommendation":"major_revision","confidential_remarks":"The core identity Eq. (7) is correct and the manuscript is likely salvageable after correcting the recurrence, the heat-equation justification, and the example factors. I recommend major revision and a re-review. The paper fits the journal's scope as a methods-oriented exact-solution note."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nThe bottom line: Eq. (7) is a genuine result and the 2D/3D grid generalization is a natural, useful extension, but the paper's proof that the χ_p reduce to Gaussians and error functions is broken, and several examples are misprinted. I'd send it to a referee, not desk-reject it.\n\nWhat's new: the formula expressing the heat evolution of piecewise polynomial data as a finite sum of jump coefficients times repeated integrals of the heat kernel. That's a correct integration-by-parts identity, and the multidimensional versions (21), (29) are valid for axis-aligned grids. The B-spline example is a nice practical demonstration. For numerical benchmarking this is a handy source of exact solutions.\n\nThe soft spots: Eq. (11) is false as stated. From definition (5), the χ_p are integrals with lower limit 0, and the recurrence that holds for integrals with special lower limits picks up polynomial/constant boundary terms. For example, at p=1, x=0 the left side is 0 but the right side is √(κt/π). So the headline claim that every χ_p is a finite combo of Gaussians and erf is not established by the derivation. That said, the claim itself is true for compact support: the offending constants and polynomials cancel against the moment conditions Σ_i C_i^p a_i^k = 0, but the paper never states or proves this. So this is a fixable gap, not a dead end.\n\nThe examples also need care: the triangular distribution evolution in §2.5.2 is missing a factor 1/a and has an extra 1/2; the half-mass radius for the uniform sphere should be a/2^{1/3}, not a^{-3/2}; and §3.3.1 writes χ where it means χ_0. None of these touch the main formula, but they confirm the manuscript needs a careful revision.\n\nWho this is for: people who want exact benchmarks for heat solvers, or who teach Green's function methods. It doesn't open a new area. A serious referee should be assigned; the core idea is worth publishing with corrected details.\n\nMy recommendation: send it to peer review with a request for a corrected recurrence lemma (or a direct proof that each χ_p is a finite sum of erf and Gaussian-times-polynomial) and fixed examples.","headline":"Main formula (7) is correct and useful, but the closed-form proof relies on a false recurrence and the examples have fixable errors; worth refereeing after revision.","tokens_in":11848,"tokens_out":10537,"would_cite":false,"duration_ms":130404,"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":"The paper derives exact solutions of the heat equation for piecewise-polynomial initial data, expressed as finite sums of Gaussians and error functions in one, two, and three dimensions.","keywords":["heat equation","exact solution","piecewise polynomial","spline","propagator","error function","Gaussian","Schrödinger equation"],"falsifier":"Compute the first repeated integral of the heat propagator directly from the definition, $\\chi_1(x,t)=\\int_0^x\\chi_0(y,t)\\,dy$ with $\\chi_0(y,t)=\\int_0^y\\chi(z,t)\\,dz$, and evaluate the claimed recurrence $p\\chi_p=x\\chi_{p-1}+2\\kappa t\\chi_{p-2}$ at $p=1$ and $x=0$; the recurrence predicts $\\chi_1(0)=\\sqrt{\\kappa t/\\pi}$, whereas the definition gives $\\chi_1(0)=0$, so the recurrence and the closed-form reduction derived from it fail as stated.","tokens_in":10864,"feed_emoji":"🌡️","tokens_out":7131,"duration_ms":62173,"temperature":0.7,"pith_summary":"This paper claims that the heat equation can be solved exactly, with no numerical stepping, when the initial temperature or density profile is a piecewise polynomial. In one dimension any piecewise polynomial works; in two and three dimensions the pieces must be aligned on a rectangular grid or set of axis-perpendicular planes. The solution at every later time is written as a finite linear combination of Gaussians and error functions, with coefficients given by the jumps in the profile's derivatives at the segment boundaries. The same construction is applied to the force-free Schrödinger equation by allowing the diffusion constant to be imaginary.","feed_headline":"Exact heat evolution reduces to Gaussians and error functions","feed_subtitle":"Piecewise-polynomial initial profiles evolve as a finite sum of Gaussians and error functions in 1D, 2D and 3D.","key_machinery":"The engine of the argument is the sequence of repeated antiderivatives of the heat kernel, defined by $\\chi_{-1}(x,t)=\\chi(x,t)$ and $\\chi_{p+1}(x,t)=\\int_0^x \\chi_p(x',t)\\,dx'$. Integration by parts against a polynomial shifts the derivative from the polynomial to the next $\\chi$, and the process stops when the polynomial derivative vanishes. The boundary terms left behind are exactly the jumps $C_i^p$ of the successive derivatives of the initial profile at the segment endpoints. The paper then invokes the claimed recurrence $p\\,\\chi_p(x,t)=x\\,\\chi_{p-1}(x,t)+2\\kappa t\\,\\chi_{p-2}(x,t)$ to show every $\\chi_p$ is a finite combination of a Gaussian and an error function. The 2D and 3D versions use the same one-dimensional $\\chi_p$ factorized across Cartesian axes, with corner coefficients $C^{p,q}_{i,j}$ and $C^{p,q,r}_{i,j,k}$ built from alternating sums of mixed partial derivatives at grid vertices.","core_discovery":"The central claim is equation (7): if the initial state $\\phi(x,0)$ is a piecewise polynomial of degree at most $n$, then its heat-equation evolution is $\\phi(x,t)=\\sum_{i=1}^N \\sum_{p=0}^n C_i^p \\chi_p(x,t)$, where $C_i^p$ is the jump of the $p$-th derivative of the initial state at the $i$-th boundary and $\\chi_p$ is the $p$-th repeated integral of the heat propagator. Because the $(n+1)$-th derivative of each polynomial is zero, repeated integration by parts terminates after $n$ steps and yields a finite sum. The paper asserts that each $\\chi_p$ is itself a finite combination of the Gaussian $\\exp(-x^2/(4\\kappa t))$ and the error function $\\mathrm{erf}(x/\\sqrt{4\\kappa t})$, so the whole evolution is closed-form. Two- and three-dimensional analogues, equations (21) and (29), express the evolution of grid-aligned piecewise polynomials as products of such one-dimensional terms; spherical symmetry reduces to the one-dimensional problem. The paper also extends the formulas to Schrödinger's equation by taking $\\kappa$ imaginary.","pith_inferences":["The derivation of the closed form for $\\chi_p$ appears to have a boundary-term gap: the stated recurrence conflicts with $\\chi_p(0)=0$ for $p=1$, so a corrected recurrence or an amended antiderivative definition would be needed to fully justify the 'finite combination of Gaussians and error functions' claim; the main jump-coefficient formula (7) does not depend on this recurrence.","The same integration-by-parts machinery should carry over to other parabolic equations whose propagators are Gaussians, such as the Ornstein-Uhlenbeck process, with the Gaussian replaced by the appropriate kernel and the error functions by their generalized counterparts.","One testable extension is to allow non-axis-aligned polyhedral boundaries in 2D and 3D: the factorized product structure will break, but the boundary-jump description may still hold with orientation-dependent weights.","If the recurrence is repaired, the method could be packaged as a drop-in exact-evolution routine for spline-based initial data, competing with finite-difference and spectral solvers on accuracy rather than speed."],"forward_implications":["Any spline or piecewise-polynomial initial profile can be evolved exactly, so the result is immediately useful for converting fitted data curves into exact heat evolutions.","Because the expressions are finite sums, they do not accumulate iteration error and can serve as reference solutions for benchmarking numerical heat-equation solvers.","Bilinear and trilinear interpolations of rectangular-grid data evolve exactly, giving a practical route from discrete observational data to an exact time-dependent field.","With imaginary diffusion constant, the same jump coefficients give exact free-particle Schrödinger evolution of piecewise-polynomial wavefunctions in terms of Fresnel integrals.","The exact solutions show explicitly how an arbitrary bounded, finite-support initial profile relaxes toward the Gaussian with the same mass and median width, quantifying the approach to the asymptotic regime."],"supporting_citations":[{"why":"Supplies the heat propagator (fundamental solution) that is the starting point $\\chi_{-1}$ for the repeated-integral sequence.","marker":"[1]"},{"why":"Supplies the error-function identities and the repeated-integral recurrence similar to the one used to reduce $\\chi_p$ to Gaussians and error functions.","marker":"[3]"},{"why":"Provides the prior piecewise-polynomial wave-function treatment and the coefficient formulas for 2D and 3D grids that the paper extends to the heat equation.","marker":"[6]"}],"fun_headline_variants":["Heat equation solved exactly for spline initial profiles","Piecewise polynomials evolve as Gaussian and error function sums","Closed-form heat evolution for grid-aligned splines","Exact heat flow from polynomial and spline data","Heat equation's exact answer: finite Gaussians and error functions"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof that the repeated integrals of the propagator reduce to a finite combination of Gaussians and error functions relies on recurrence (11), but that recurrence is not actually derived from the definition of those integrals; it already fails for $p=1$ at $x=0$, so the closed-form claim is not established by the derivation given in the paper.","fun_headline_variants_meta":{"raw":{"variants":["Heat equation solved exactly for spline initial profiles","Piecewise polynomials evolve as Gaussian and error function sums","Closed-form heat evolution for grid-aligned splines","Exact heat flow from polynomial and spline data","Heat equation's exact answer: finite Gaussians and error functions"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000665,"raw_usage":{"total_tokens":3014,"prompt_tokens":903,"completion_tokens":2111,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":519,"completion_tokens_details":{"reasoning_tokens":2034}},"tokens_in":519,"tokens_out":2111,"duration_ms":14835,"temperature":1.0,"reasoning_tokens":2034,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T20:58:30.386143+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Compute the first repeated integral of the heat propagator directly from the definition, $\\chi_1(x,t)=\\int_0^x\\chi_0(y,t)\\,dy$ with $\\chi_0(y,t)=\\int_0^y\\chi(z,t)\\,dz$, and evaluate the claimed recurrence $p\\chi_p=x\\chi_{p-1}+2\\kappa t\\chi_{p-2}$ at $p=1$ and $x=0$; the recurrence predicts $\\chi_1(0)=\\sqrt{\\kappa t/\\pi}$, whereas the definition gives $\\chi_1(0)=0$, so the recurrence and the closed-form reduction derived from it fail as stated.","supporting_citations":[{"cited_title":"Encyclopedia of Mathematics, encyclopediaofmath.org","cited_arxiv_id":null,"evidence_quote":"Supplies the heat propagator (fundamental solution) that is the starting point $\\chi_{-1}$ for the repeated-integral sequence."},{"cited_title":"and Stegun I.A","cited_arxiv_id":null,"evidence_quote":"Supplies the error-function identities and the repeated-integral recurrence similar to the one used to reduce $\\chi_p$ to Gaussians and error functions."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the prior piecewise-polynomial wave-function treatment and the coefficient formulas for 2D and 3D grids that the paper extends to the heat equation."}],"review_version":1}