REVIEW 3 major objections 6 minor 6 references
Exact solution of the Heat Equation for initial polynomials or splines
T0 review · 3 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash
Pith's one-line read 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.
desk verdict 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. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
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.
What would settle it
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.
Extended reading notes
Core claim
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.
Load-bearing premise
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.
Editorial extensions
If this is right
- 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.
Reading between the lines
- 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.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
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.
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 (3)
- [§2.2, Eq. (11)] 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.
- [§2.4] 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.
- [§2.5.2] 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.
minor comments (6)
- [§2.1] The word 'indeces' in the remark at the end of the section should be 'indices'.
- [§3.3.2] The cross-reference 'Fig.2(a)' in the paragraph after Eq. (24) should be 'Fig.4(a)'.
- [§4] The word 'paralleliped' should be 'parallelepiped'.
- [§4.1.1] 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.
- [§6] The phrase 'tailored to to a rectangular mesh' contains a duplicated 'to'.
- [§2.2] The notation remark '[The symbols p, q, r are used to denote indeces, never powers.]' is unnecessary and slightly confusing; consider removing it.
Circularity Check
No significant circularity: Eq. (7) is derived by direct integration by parts with no fitted parameters; self-citation [6] is non-load-bearing; the false recurrence (11) is a correctness defect, not a circular step.
full rationale
The central derivation is self-contained. Eq. (7) follows from repeated integration by parts of the standard convolution solution (3): each step replaces ∫χ_p·u^(p+1) by ∫χ_(p+1)·u^p plus boundary terms (6), terminating because u^(n+1) ≡ 0; the coefficients C_i^p are jumps of the initial data and nothing is fitted, so the solution is not assumed. No uniqueness theorem is invoked and no ansatz is imported by citation. The only self-citations are [6] (Andrews, Eur. Phys. J. Plus 2017): in §3.3.2 and §4.3 it is cited for finite-difference-style short cuts for the regular-mesh coefficients C^(1,1) and C^(1,1,1) (the formulas are also stated in the text), and in §5 for a remark on interference asymptotics; none of these supports a load-bearing premise of Eqs. (7), (21), or (29). Per the reviewing rule I flag an omitted/incorrect proof that is a correctness risk, not circularity: §2.2 asserts “It follows that the required sequence of integrals of the propagator is … p χp(x,t) = x χp−1(x,t) + 2κt χp−2(x,t)” under definition (5) χ_(p+1)(x,t) := ∫_0^x χ_p(x′,t)dx′. At p=1, x=0 the left side is χ_1(0,t)=0 while the right side is 2κt·χ(0,t)=√(κt/π)≠0, so the recurrence is false as stated; the lower-limit-0 antiderivatives acquire endpoint corrections. Relatedly, §2.4's claim that “each χ_p(x,t) also satisfies the heat equation” fails for p=1, since (∂_t−κ∂²_x)χ_1 = −κχ(0,t); the combination in (7) survives because Σ_i C_i^1=0 for compact support, but that cancellation is neither stated nor proved. The §2.5.2 triangular example also carries an incorrect factor ½ and omits the 1/a normalization. None of these defects makes the derivation reduce to its own inputs, so the circularity score is low (1), flagged as a correctness risk rather than as circularity.
Assumptions & free parameters
assumptions (5)
- domain assumption The heat equation with constant diffusivity on an infinite homogeneous medium, with the fundamental solution (2) and superposition (3) as the unique solution
- domain assumption The initial state is exactly piecewise polynomial, of finite degree, zero outside a finite region, and in 2D/3D all segment boundaries lie on axis-aligned rectangles/cuboids
- standard math Taylor's theorem applied to each polynomial segment yields the sign-function representations (14), (19), (26)
- standard math Each antiderivative χ_p of the heat propagator satisfies the heat equation and ∂_x χ_p = χ_{p-1}
- ad hoc to paper The recurrence relation (11) correctly evaluates the repeated integrals χ_p
Cite this review
Pith. "Pith review of Exact solution of the Heat Equation for initial polynomials or splines." pith.science (2026). https://pith.science/paper/VZWS5WWP
@misc{pith2026241115169,
author = {Pith},
title = {Pith review of: Exact solution of the Heat Equation for initial polynomials or splines},
year = {2026},
howpublished = {\url{https://pith.science/paper/VZWS5WWP}},
note = {Machine review of arXiv:2411.15169}
}
read the original abstract
The exact evolution in time and space of a distribution of the temperature (or density of diffusing matter) in an isotropic homogeneous medium is determined where the initial distribution is described by a piecewise polynomial. In two dimensions, the boundaries of each polynomial must lie on a grid of lines parallel to the axes, while in three dimensions the boundaries must lie on planes perpendicular to the axes. The distribution at any position and later time is expressed as a finite linear combination of Gaussians and Error Functions. The underlying theory is developed in detail for one, two, and three dimensional space, and illustrative examples are examined.
Figures
Figures from the paper (4 more)
Reference graph
Works this paper leans on
-
[1]
Encyclopedia of Mathematics, encyclopediaofmath.org
The thermal-conductance equation. Encyclopedia of Mathematics, encyclopediaofmath.org
-
[2]
Encyclopedia of Mathematics, encyclopediaofmath.org
The Stone-Weierstrass theorem. Encyclopedia of Mathematics, encyclopediaofmath.org
-
[3]
Abramowitz, M. and Stegun I.A. (eds.): Handbook of Mathematical Functions. National Bureau of Standards, Washington DC (1964). 7.2.3
work page 1964
-
[4]
arXiv:1706.10034v3, doi.org/10.48550/arXiv.1706.10034 (2017)
V´ azquez, J.L.: Asymptotic behaviour methods for the Heat Equation. arXiv:1706.10034v3, doi.org/10.48550/arXiv.1706.10034 (2017)
- [5]
-
[6]
Andrews, M.: The evolution of piecewise polynomial wave functions. Eur. Phys. J. Plus 132(1), (2017), Article 1 of Issue 1, pp14 doi 10.11.1140/epjp/i2017-11280-8 14
work page 2017
Reviewed August 12, 2026 · model on record in the stance chip above.
Discussion (0). Continue with ORCID to comment.