Ancient caloric functions and parabolic frequency on graphs

Archana Mohandas; Tang-Kai Lee

REVIEW 3 major objections 5 minor 4 references

Ancient caloric functions and parabolic frequency on graphs

T0 review · 3 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read This paper proves that on any strip-type graph — a weighted graph crossed with the integer line — ancient caloric functions of polynomial growth must vanish identically, and uses a monotone parabolic frequency to prove backward uniqueness…

desk verdict The strip nonexistence theorem is unproven because the reverse Poincaré inequality drops boundary edges that Lemma 3.1 needs; the parabolic frequency monotonicity is new and sound. read the letter →

arxiv 2412.15145 v1 pith:FKQM6CP4 submitted 2024-12-19 math.AP math.CO

Tang-Kai Lee , Archana Mohandas This is my paper · ORCID

classification math.APmath.CO MSC 35R0239A1231C2005C6335B40

keywords ancientcaloricfunctionsparabolicfrequencyweightedgraphsstrip-typereversePoincaréinequalitybackwarduniquenesspolynomialgrowthdiscreteheatequation

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper asks whether the discrete heat equation on an infinite strip-shaped graph can have ancient solutions — solutions defined for all negative times — that grow at most polynomially, in the sense $|u(x,t)| \le C(1 + d(x_0,x) + \sqrt{-t})^d$. It answers no for strip-type graphs: if $G = G_0 \square \mathbb{Z}$ and the solution is supported on $W_0 \times \mathbb{Z}$ with $W_0$ a finite proper non-empty set of cross-section vertices and a Dirichlet-type condition outside, then any such ancient caloric function of polynomial growth is identically zero. The same frequency-based machinery shows that on a finite graph an ancient caloric function of polynomial growth must be a time-independent harmonic function, and it yields a backward uniqueness theorem: a heat-equation solution with bounded potential that vanishes at one time vanishes at all earlier times. A reverse Poincaré inequality on graph cylinders converts this rigidity into an exponential-growth contradiction that polynomial growth cannot survive.

What carries the argument

Two devices carry the argument. The first is a reverse Poincaré inequality (Proposition 3.3): for solutions of the Dirichlet problem (1.2), the $L^2$ mass over a parabolic cylinder $Q_r$ is controlled by the mass difference between a larger cylinder $Q_R$ and $Q_r$ through the factor $C/(R-r)^2$; iterating $r \mapsto r + r_0$ forces the mass to grow like $e^k$ as the radius grows linearly, which contradicts polynomial growth. This inequality is built from the positivity of the first Dirichlet eigenvalue of the finite subgraph $W_0$, applied slice by slice in $\mathbb{Z}$. The second is a discrete parabolic frequency $U(t) = D(t)/I(t) = -\sum_{x\in V} |\nabla u|^2(x)\mu(x) / \sum_{x\in V} u^2(x)\mu(x)$, the graph counterpart of the manifold frequency for heat equations, which obeys $U' \ge 2C^2(U-1)$ whenever $|(\partial_t - \Delta)u| \le C(t)(|u| + |\nabla u|)$. Monotonicity of the frequency yields a Harnack-type lower bound for $I(t)$ that implies the backward uniqueness theorem: vanishing at the final time forces vanishing on the whole interval.

What would settle it

Take the smallest strip: $G_0$ has two vertices $a,b$ joined by one edge of weight $1$, and $W_0=\{a\}$, so $W = \{a\}\times\mathbb{Z}$. On the active row the equation reduces to $\partial_t u_z = \frac{1}{4}(u_{z+1}+u_{z-1}) - \frac{3}{4}u_z + c u_z$. The constant mode $u_z \equiv 1$ is a static solution exactly when $c = \frac{1}{4}$; solving this recurrence explicitly for any $|c| < \frac{1}{4}$ and finding a nonzero polynomial-growth ancient solution would contradict Theorem 1.1, and the calculation also fixes the sharp threshold in this example.

Watch

Extended reading notes

Core claim

The central claim is Theorem 1.1: let $G_0$ be a locally finite connected weighted graph, let $G = G_0 \square \mathbb{Z}$ be its Cartesian product with the integer line, and let $W_0$ be a finite proper non-empty subset of $V_0$. If $u : V \times \mathbb{R}_{\le 0} \to \mathbb{R}$ is an ancient solution of $\partial_t u = \Delta u + c(x)u$ supported on $W = W_0 \times \mathbb{Z}$, satisfies $u = 0$ on $(V \setminus W) \times \mathbb{R}_{\le 0}$, has $|c| \le \varepsilon_0(G)$, and obeys the polynomial growth bound above, then $u \equiv 0$. In the authors' words, such strips admit no non-trivial ancient caloric functions of polynomial growth; the assumption that $W_0$ is proper is necessary, as the paper's one-line example of a static linear solution on $\{v_0\} \times \mathbb{Z}$ shows. The paper also proves that on finite graphs the analogous statement is stronger: polynomial-growth ancient caloric functions are exactly the time-independent harmonic functions, and it establishes backward uniqueness on general weighted graphs under the assumption that the solution and its time derivative lie in $W^{1,2}_\mu$ at each time slice.

Load-bearing premise

The proof depends on a reverse Poincaré inequality: the squared $L^2$ mass over the strip is controlled by squared differences along edges with both endpoints inside the strip, while the edges that cross from $W_0$ into its complement are dropped as a boundary term; if that internal control fails, the exponential-growth contradiction collapses.

Editorial extensions

If this is right

Theorem 1.1 implies that a strip-type graph $G_0 \square \mathbb{Z}$ with a finite proper cross-section $W_0$ admits no non-zero harmonic functions of polynomial growth, because a harmonic function is a static ancient caloric function.
On any finite connected weighted graph, every ancient caloric function of polynomial growth is harmonic and constant in time; the space of such solutions is one-dimensional in the connected case.
Backward uniqueness holds for the continuous-time heat equation on weighted graphs with bounded potential, while the paper notes the discrete-time heat equation does not enjoy the same property: a non-trivial solution on the two-vertex graph $K_2$ can vanish after finite time.
The equality case of the frequency monotonicity characterizes the sharp solutions: if $U' \equiv 0$ for a heat solution, then $\Delta u = (U/2)u$ and $u(x,t) = e^{U(t-a)/2}u(x,a)$, so constant-frequency solutions are exactly single-eigenfunction exponential solutions.
The same Dirichlet-energy estimate extends to more general operators under a stronger assumption, so the backward uniqueness property is not specific to the pure heat equation.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The frequency machinery is likely to yield discrete unique-continuation or Landis-type decay estimates on general weighted graphs: a heat solution that decays faster than every exponential on large balls should be identically zero, a statement the paper does not make.
Replacing the factor $\mathbb{Z}$ by $\mathbb{Z}^m$ is a natural test: the reverse Poincaré iteration would probably still force exponential growth of cylinder masses when $W_0$ is finite, but the paper's cutoff uses the one-dimensional structure of $\mathbb{Z}$ and would need a new argument.
The explicit threshold seen in the smallest ladder example — a static solution appears exactly at a critical potential equal to the spectral gap — suggests $\varepsilon_0(G)$ is quantitative, namely a spectral gap of the strip, and extracting its value from the Dirichlet eigenvalue of $W_0$ would make the theorem effective.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

The paper has a real gap in its main theorem, but it also contains a genuinely useful new tool. The parabolic frequency monotonicity for weighted graphs (Theorem 5.1, Corollaries 5.5 and 5.8) is derived cleanly and appears correct. That part alone justifies a look. The finite-graph ancient caloric classification (Theorem 4.1) is standard but nicely written.

The problem is in Section 3. Proposition 3.3 is the load-bearing step for the strip nonexistence result, and its proof does not work. At (3.11), the authors lower-bound the internal Dirichlet energy over W0 × W0 slices by applying Lemma 3.1. Lemma 3.1, via Grigor'yan, requires the full gradient on the graph, including edges from W0 to V0 \ W0. Those boundary edges are absent from the W × W sum. They were deliberately dropped in (3.5) because they are non-positive in the energy identity. The inequality used in (3.11) is false in general: on G0 a single edge, W0 one vertex, and u constant on W0 × Z, the internal Dirichlet energy is zero but the L2 mass is positive. That particular u is not a solution to the small-potential equation, so it doesn't directly falsify Proposition 3.3, but it shows the cited Lemma 3.1 cannot justify the step. The exponential growth contradiction in Section 3.2 relies entirely on this estimate, so Theorem 1.1 is unproven as written.

A few smaller things: the function space W^{1,2}_μ is defined as square summability only, with no gradient term, which is inconsistent with the frequency arguments; and the proof of Theorem 1.4 should make explicit which summability of |∇u| is needed. These are minor and likely fixable.

My overall sense: the frequency part is solid and could be extracted as a separate contribution. The strip-theorem gap is serious but not obviously fatal—a repair that keeps the boundary term or uses the equation to control it might work. The paper deserves peer review rather than desk rejection, but the referee should be sent to the reverse Poincaré estimate with a clear warning.

Referee Report

3 major / 5 minor

Summary. The paper studies ancient caloric functions on weighted graphs, with a focus on strip-type graphs of the form G0 □ Z. The main result (Theorem 1.1) asserts that on such a graph, if W = W0 × Z for a finite proper subset W0 of V0, then any ancient solution supported on W with a Dirichlet-type boundary condition, polynomial growth, and a sufficiently small potential c must vanish identically. The proof is based on a reverse Poincaré inequality (Proposition 3.3) that is derived from a weighted energy estimate using Green's formula and a Dirichlet eigenvalue bound. The paper also proves a backward uniqueness result (Theorem 1.4) via a parabolic frequency monotonicity formula, and a finite-graph counterpart (Theorems 1.3 and 4.1) asserting that nontrivial ancient solutions on finite graphs have exponential growth in time.

Significance. If the main theorem were correct, it would provide a discrete analogue of results of Hang–Lin and Gui on exponential growth of solutions in Euclidean strips, and would be a meaningful contribution to the study of ancient caloric functions on graphs. The parabolic frequency method is potentially useful and the finite-graph result is a natural complement. However, the proof of the central claim contains a load-bearing error in the reverse Poincaré inequality, and the backward uniqueness section has a functional-setting gap. As a result, the paper's main conclusions are not established in the present form.

major comments (3)

[§3.1, Eq. (3.11)] The reverse Poincaré inequality in (3.11) is invalid. Lemma 3.1 requires the full graph gradient on V0, including edges from W0 to V0\W0, to control the L2 mass on W0. However, the left-hand side of (3.11) contains only the internal sum over W0×W0. For a function constant on W0, the internal gradient vanishes while the L2 mass is positive, so the claimed inequality fails. For instance, when G0 is a single edge and W0 is one of its vertices, a function identically 1 on W0 has zero internal gradient but positive L2 mass. Since this inequality is the mechanism that produces the exponential growth contradiction in §3.2, Theorem 1.1 is not proven.
[§3.1, Eqs. (3.5)–(3.7)] The boundary term in Green's formula (3.5) is dropped. This boundary term is exactly the contribution from edges between W0 and V0\W0 that Lemma 3.1 would need. Dropping it removes the only source of these edges in the subsequent estimates, and keeping it would only add a nonnegative term to the right-hand side, which does not help control the L2 mass by the internal Dirichlet energy. Thus the error cannot be repaired by restoring this term; a different argument would be needed.
[§5.1] The function space W^{1,2}_μ is defined by square summability only: u ∈ W^{1,2}_μ(V) if ∑ u²μ < ∞. However, the frequency function U(t) is defined through D(t) = -∑ |∇u|² μ, which requires finite Dirichlet energy. Square summability does not imply finite gradient energy on infinite graphs. Consequently, Theorems 5.1 and Corollary 5.5 are not justified under the stated hypotheses, and this affects the validity of Theorem 1.4 and hence the finite-graph conclusion Theorem 1.3.

minor comments (5)

[Title and Abstract] There are typos in the title and abstract: 'P ARABOLIC' and 'GRAPH S' should be 'parabolic' and 'graphs'.
[Introduction, §1] The remark that uniqueness properties do not hold for the heat equation with discrete time appears twice, verbatim, in the introduction.
[§3.1, Eq. (3.8)] The notation in (3.2) defines ∫_{Q_r} w using a function w, but in (3.8) the letter w is also used for the edge weight, which is confusing. A different symbol for the integrand would improve readability.
[§3.1, Eq. (3.9)] The combination of the sums over W and V in the display following (3.9) is algebraically incorrect: the term 4Cε∑_{x,y∈V} φ²|∇u|²w is over all of V×V, while the term -1/4∑_{x,y∈W} φ²|∇u|²w is over W×W, so they cannot be directly merged into a single coefficient on the W-sum. The intended estimate can be salvaged by dropping the nonnegative boundary contribution, but as written the step is misleading.
[References] The reference [CCM95] appears in the bibliography but does not seem to be cited in the text.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the paper's frequency monotonicity, reverse Poincaré lemma, and eigenvalue arguments are all derived from external standard results and internal inequalities, not from the theorem being proved.

full rationale

The derivation chain in this paper is self-contained and does not reduce any central claim to its own inputs. Theorem 1.1 rests on Proposition 3.3, whose proof uses Lemma 3.1 and Green's formula. Lemma 3.1 is proved by citing [Gri18, Theorem 4.5], an external textbook result on the positivity of the first Dirichlet eigenvalue; its conclusion is not equivalent to Theorem 1.1. Green's formula is also cited to [Gri18, Theorem 2.1]. The frequency-based backward uniqueness result (Theorem 1.4) is proved in Section 5 from algebraic identities and Cauchy-Schwarz, following the general method of Colding-Minicozzi but with the discrete estimates derived in the paper. Theorem 4.1 uses standard finite-dimensional Jordan normal form and the elementary fact that the heat semigroup gives the solution; invoking Theorem 1.4 there is a convenience, not a circular premise, and Theorem 1.4 is independently proved. The only self-citations, [BHL24] and [BK23], appear as contextual references for geometric flows and are not load-bearing. The paper also explicitly notes a limitation for discrete-time heat equations, which is a genuine caveat rather than a circular step. The reviewer-flag concern about the application of Lemma 3.1 to internal Dirichlet energy in (3.11) is a possible mathematical correctness gap, not a circularity: even if the reverse Poincaré estimate were unjustified as written, the alleged flaw is a mismatch of hypotheses, not the conclusion being assumed. No fitted parameter is renamed as a prediction, no uniqueness theorem is imported from the authors' prior work to forbid alternatives, and no known result is merely relabeled. The central conclusions are derived from stated inequalities and external standard facts, so the circularity score is 0.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

No fitted constants appear; epsilon0 is chosen explicitly from the Poincare constant. The paper relies on standard external results for graph Laplacians and Poincare inequalities. No new physical or mathematical entities are postulated.

assumptions (5)

standard math Lemma 3.1: positive first Dirichlet eigenvalue for finite proper subsets of a weighted graph, quoted from [Gri18, Theorem 4.5].
Used in Proposition 3.3 to convert Dirichlet energy into L2 mass on each slice W0; the proof of the reverse Poincare inequality depends on this external result.
domain assumption All graphs are locally finite, connected, undirected weighted graphs.
Assumed throughout by the authors; ensures the Laplacian, vertex weights, and Green's formula are well-defined.
domain assumption u and ∂t u belong to W^{1,2}_mu on each time slice with finite Dirichlet energy D.
The frequency U = D/I and Theorems 5.1 and 1.4 require finite gradient energy, but W^{1,2}_mu is defined only by square summability, so gradient regularity is a hidden assumption.
standard math Green's formula, Proposition 2.4, for functions with one compactly supported factor.
Taken from [Gri18]; used in the integration by parts that produces the reverse Poincare estimate.
standard math For finite connected weighted graphs, eigenvalues of the graph Laplacian lie in [-2,0].
Used in Theorem 4.1 to assert exponential growth or decay of matrix exponential solutions.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Ancient caloric functions and parabolic frequency on graphs." pith.science (2026). https://pith.science/paper/FKQM6CP4

@misc{pith2026241215145,
  author       = {Pith},
  title        = {Pith review of: Ancient caloric functions and parabolic frequency on graphs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/FKQM6CP4}},
  note         = {Machine review of arXiv:2412.15145}
}

abstract

We study ancient solutions to discrete heat equations on some weighted graphs. On a graph of the form of a product with $\bb Z,$ we show that there are no non-trivial ancient solutions with polynomial growth. This result is parallel to the case of finite graphs, which is also discussed. Along the way, we prove a backward uniqueness result for solutions with appropriate decaying rate based on a monotonicity formula of parabolic frequency.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

4 extracted references · 4 canonical work pages

[1]

[Alm79] F. J. Almgren Jr., Dirichlet’s problem for multiple valued functions and the r egularity of mass minimizing integral currents, Minimal submanifolds and geodesics (Proc. Japan-United S tates Sem., Tokyo, 1977), North-Holland, Amsterdam-New York, 1979, pp. 1–6. MR574247 [BHL24] J. Baldauf, P. T. Ho, and T.-K. Lee, Parabolic Frequency for the Mean Cu...

work page 2024
[21]

Kleiner, A new proof of Gromov’s theorem on groups of polynomial growt h, J

MR4565037 [Kle10] B. Kleiner, A new proof of Gromov’s theorem on groups of polynomial growt h, J. Amer. Math. Soc. 23 (2010), no. 3, 815–829. MR2629989 [Li97] P. Li, Harmonic sections of polynomial growth , Math. Res. Lett. 4 (1997), no. 1, 35–44. MR1432808 [LT89] P. Li and L.-F. Tam, Linear growth harmonic functions on a complete manifold , J. Diﬀerentia...

work page 2010
[2018]

Hang and F

MR3822363 [HL99] F. Hang and F. Lin, Exponential growth solutions of elliptic equations , Acta Math. Sin. (Engl. Ser.) 15 (1999), no. 4, 525–534. MR1756781 [Hua11] B. Hua, Harmonic functions of polynomial growth on singular spaces with nonnegative Ricci curvature , Proc. Amer. Math. Soc. 139 (2011), no. 6, 2191–2205. MR2775397 [Hua22] , Dimensional bounds...

work page 1999
[4039]

Hua and J

MR4391882 [HJ15] B. Hua and J. Jost, Polynomial growth harmonic functions on groups of polynomi al volume growth , Math. Z. 280 (2015), no. 1-2, 551–567. MR3343919 [HJLJ13] B. Hua, J. Jost, and X. Li-Jost, Polynomial growth harmonic functions on ﬁnitely generated abelian groups, Ann. Global Anal. Geom. 44 (2013), no. 4, 417–432. MR3132083 [Hua20] X.-T. Hu...

work page 2015

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.

[1] [1]

[Alm79] F. J. Almgren Jr., Dirichlet’s problem for multiple valued functions and the r egularity of mass minimizing integral currents, Minimal submanifolds and geodesics (Proc. Japan-United S tates Sem., Tokyo, 1977), North-Holland, Amsterdam-New York, 1979, pp. 1–6. MR574247 [BHL24] J. Baldauf, P. T. Ho, and T.-K. Lee, Parabolic Frequency for the Mean Cu...

work page 2024

[2] [21]

Kleiner, A new proof of Gromov’s theorem on groups of polynomial growt h, J

MR4565037 [Kle10] B. Kleiner, A new proof of Gromov’s theorem on groups of polynomial growt h, J. Amer. Math. Soc. 23 (2010), no. 3, 815–829. MR2629989 [Li97] P. Li, Harmonic sections of polynomial growth , Math. Res. Lett. 4 (1997), no. 1, 35–44. MR1432808 [LT89] P. Li and L.-F. Tam, Linear growth harmonic functions on a complete manifold , J. Diﬀerentia...

work page 2010

[3] [2018]

Hang and F

MR3822363 [HL99] F. Hang and F. Lin, Exponential growth solutions of elliptic equations , Acta Math. Sin. (Engl. Ser.) 15 (1999), no. 4, 525–534. MR1756781 [Hua11] B. Hua, Harmonic functions of polynomial growth on singular spaces with nonnegative Ricci curvature , Proc. Amer. Math. Soc. 139 (2011), no. 6, 2191–2205. MR2775397 [Hua22] , Dimensional bounds...

work page 1999

[4] [4039]

Hua and J

MR4391882 [HJ15] B. Hua and J. Jost, Polynomial growth harmonic functions on groups of polynomi al volume growth , Math. Z. 280 (2015), no. 1-2, 551–567. MR3343919 [HJLJ13] B. Hua, J. Jost, and X. Li-Jost, Polynomial growth harmonic functions on ﬁnitely generated abelian groups, Ann. Global Anal. Geom. 44 (2013), no. 4, 417–432. MR3132083 [Hua20] X.-T. Hu...

work page 2015