{"id":"63ea237f-07f8-4bbf-9de1-f9a6bfa616a5","arxiv_id":"1908.02657","paper_version":1,"verdict":"REJECT","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The damped wave equation on the Heisenberg group has heat-like L2 decay: with L1 data the solution decays like (1+t)^(-Q/4), the horizontal gradient like (1+t)^(-Q/4-1/2), and the time derivative like (1+t)^(-Q/4-1).","lead":"This paper uses the group Fourier transform to derive L2 decay estimates for the damped wave equation on the Heisenberg group, for the solution, its horizontal gradient, and its time derivative. Its main new rates hold when the Cauchy data are also in L1, matching Euclidean formulas with the homogeneous dimension.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The L2-only estimates (5)-(6) omit the initial L2 norm that the paper's own low-frequency estimates produce; for low-frequency data the claimed decay rates fail.","rationale":"The paper's genuine contribution, the L1-L2 decay estimates (7)-(9), appears sound: the low-frequency analysis uses only the Riemann-Lebesgue bound (10) and the convergence of the series ∑ µ_k^-(n+1), the high-frequency part is exponential, and the derivative counting gives the stated powers of (1+t). The problem is the overclaim in the L2-only part of Theorem 1.1. The reader's identified false equality in §4.3 is real but not load-bearing, because the proof of (42) only needs the upper bound LHS ≲ ||(X_j ± iY_j)u0||_HS^2, which does hold. The load-bearing gap is in (5)-(6): the proofs themselves produce ||u0|| terms with the right t-decay, and the stronger statement without ||u0|| fails for low-frequency data, both in the Euclidean analogy and inside the Heisenberg group Fourier calculus used here. Thus the theorem should not be accepted as stated; the estimates could be repaired by adding ||u0|| to the right-hand sides of (5)-(6), which costs nothing for the subsequent applications since the L1 version is the main tool. Because the reader already recommended REJECT, my verdict is unchanged.","tokens_in":19082,"tokens_out":16961,"duration_ms":176083,"concrete_test":"Take n=1, u1=0, and choose u0 whose only nonzero Fourier component is û0(λ)_{0,0}, supported on a small interval [λ0,2λ0] with λ0<1/8 and normalized so ||u0||_L2=1; then ||∇H u0||_L2 ~ λ0^(1/2). Using the exact formula (26) with µ0=1 at t=1/λ0, the k=0 mode contributes ||∇H u(t)||_L2 ~ √λ0, while the right-hand side of (5) is (1+t)^-1/2 ||∇H u0|| ~ λ0. Since √λ0 / λ0 → ∞ as λ0→0, inequality (5) fails. The analogous computation using (35) checks (6).","verdict_should_be":"UNCHANGED","load_bearing_attack":"Theorem 1.1's L2-only bounds (5) and (6) are not supported by the proof. In Remarks 4 and 5 the low-frequency pieces are bounded by t^-2(||u0||^2+||u1||^2) and t^-1(||u0||^2+||u1||^2), respectively. To arrive at (5)-(6), one must convert those ||u0|| terms into ||∇H u0|| terms via an inequality of the form ||u0||_L2 ≲ ||∇H u0||_L2, but no such global Poincaré inequality holds on H^n: data concentrated at arbitrarily low group-Fourier frequencies have fixed L2 norm and vanishing horizontal gradient. The Euclidean analogue confirms the failure: with u0 of frequency ξ0→0, ||∇u(t)||_2 ~ |ξ0| e^{-|ξ0|^2 t}, so omitting ||u0|| is not harmless. The equality displayed after (42) is also false, as a one-component Hermite datum shows, but only the ≲ direction is needed for K_high, so that is a presentational flaw rather than the decisive gap.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the linear damped wave equation on the Heisenberg group H_n, namely ∂_t^2 u − Δ_H u + ∂_t u = 0, and derives L^2 decay estimates for the solution u, its horizontal gradient ∇_H u, and its time derivative ∂_t u. The main tool is the group Fourier transform combined with the Hermite spectral decomposition of the sub-Laplacian symbol, exactly in the spirit of WKB/phase-space analysis. Theorem 1.1 states L^2-only bounds (4)-(6) and, under additional L^1(H_n) regularity of the data, the heat-like decay bounds (7)-(9) with rates determined by the homogeneous dimension Q = 2n+2.","tokens_in":19259,"tokens_out":23874,"duration_ms":229704,"significance":"The L1-L2 estimates (7)-(9) are, if correct, a meaningful extension of Matsumura's Euclidean decay laws to a sub-Riemannian setting; the derivation is concrete, uses no fitted parameters, and the rates (1+t)^{-Q/4}, (1+t)^{-Q/4-1/2}, (1+t)^{-Q/4-1} match the expected diffusion asymptotics on H_n. The heavy use of Plancherel and Hermite functions is appropriate and the high-frequency part is handled by a clean exponential argument. However, the L2-only estimates (5)-(6), which the abstract advertises as a central goal, are false as stated; the proof in Remarks 4-5 does not establish them, and a concrete low-frequency datum shows the asserted rates fail. The false equality displayed after (42) is a further gap, although the needed inequality direction is true. Overall, the L1-L2 part appears defensible, but the main theorem as stated is not.","major_comments":[{"comment":"The displayed identity after (42) is false as written. For a datum u0 with a single Hermite component e_0, the left-hand side equals |λ| ‖u0(λ)e_0‖^2, while the right-hand side equals 4|λ| ‖u0(λ)e_0‖^2 (each of (X_j±iY_j)u0 contributes 2|λ|‖u0e_0‖^2). Only the ≲ direction is needed for the K_high estimate, so the argument can be repaired by replacing the equality with a two-sided inequality after inserting the appropriate constants. The author should correct this step explicitly, since the displayed equality is incorrect and used without comment.","section":"Theorem 1.1, (5)-(6); Remarks 4-5, Sections 4.2-4.3"}],"minor_comments":[{"comment":"The title contains a typo: 'He isenberg group' should be 'Heisenberg group', and the abstract has a similar spacing issue. A careful proofreading is needed throughout.","section":"Title and abstract"},{"comment":"The word 'liturature' should be 'literature'.","section":"Introduction, paragraph 1"},{"comment":"The sentence 'Note thatF (t,λ,k ) = ∂tG(t,λ,k )' is missing a space after 'that'. Please rewrite for clarity.","section":"Section 4.1, after (26)"},{"comment":"References [26] and [27] are preprints and [28] is an unpublished preprint of the author. If the final version is published, please update these entries; otherwise, the use of unpublished work for the Fujita-type critical exponent should be marked as forthcoming.","section":"References [26]-[28]"}],"recommendation":"major_revision","confidential_remarks":"The L1-L2 estimates (7)-(9) appear sound and are a plausible contribution, but the L2-only estimates (5)-(6) are false as stated, and the theorem currently overclaims. The fix (replacing ‖∇_H u0‖ with ‖u0‖ in (5)-(6)) is straightforward and within the paper's scope, and the false equality after (42) is also repairable. If the author declines to weaken the L2-only claims, the paper should be rejected; with the correction, it may become acceptable. I would ask the editor to require the author to state explicitly that the L2-only estimates with ‖∇_H u0‖ are false, and to verify the corrected statements by re-doing Remarks 4-5. The citation of the author's own forthcoming work is acceptable in context."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The L1-L2 decay estimates (7)-(9) are the real content of this paper, and they look right. The Heisenberg group with the sub-Laplacian is the natural setting, and the homogeneous dimension Q=2n+2 enters exactly as in the Euclidean analogue. The proof is a serious adaptation of Matsumura's phase-space analysis: group Fourier transform, Hermite functions, low/high frequency splitting, and careful eigenvalue-series convergence. That part deserves credit and will be useful to people working on semilinear damped waves on sub-Riemannian structures.\n\nThe soft spots are concentrated in the L2-only estimates (5) and (6). They are stated with only ||∇_H u0|| + ||u1|| on the right, but the proofs in Remarks 4 and 5 actually produce low-frequency bounds with ||u0||. To remove that term you would need a global Poincaré inequality on H_n, which does not hold: data concentrated at low group-Fourier frequencies have fixed L2 norm and arbitrarily small horizontal gradient. The Euclidean analogue already shows the problem. Take u0 with Fourier frequency ε, unit L2 norm, ||∇u0||=ε. Then ||∇u(t)|| behaves like εe^{-ε^2t}, which for ε << t^{-1/2} is much larger than (1+t)^{-1/2}ε. So (5) and (6) fail as stated. The fix is straightforward: put ||u0|| back into the right-hand sides, as the paper's own intermediate inequalities already allow.\n\nThere is also a false displayed equality after (42). The sum of weighted Fourier coefficients is comparable, not equal, to twice the sum of the Hilbert-Schmidt norms of (X_j±iY_j)u0. The proof only needs the ≲ direction, so this is a presentational error rather than a gap in the gradient estimates, but it should be corrected.\n\nThe citation pattern is fine; [21] is used as a benchmark, not as a crutch, and the self-citations concern applications rather than the proof. No circularity.\n\nWho is this for? People working on decay estimates for evolution equations on Lie groups and on Fujita-type exponents for semilinear damped waves. The main L1-L2 result is likely correct and worth having. I would send the paper to a serious referee, but with the clear expectation of major revision: the overclaimed L2-only estimates must be fixed and the false equality repaired. After that, this should be a solid contribution.","headline":"The L1-L2 decay estimates for the damped wave on H_n are a genuine and useful new result, but Theorem 1.1's L2-only estimates (5)-(6) are overstated and false as written.","tokens_in":19833,"tokens_out":7704,"would_cite":true,"duration_ms":76732,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35L10","35R03","58J45","33C45","43A30","43A80"],"pacs":[],"model":"deepseek-v4-flash","headline":"For the damped wave equation on the Heisenberg group, L2 norms and their derivatives satisfy heat-like decay rates with homogeneous dimension Q=2n+2.","keywords":["damped wave equation","decay estimates","Heisenberg group","group Fourier transform","Hermite functions","sub-Laplacian","homogeneous dimension","L1-L2 estimates"],"falsifier":"Test the identity after (42) on data whose group Fourier transform at one fixed $\\lambda$ is the rank-one operator $(\\cdot, e_0)e_0$. The left-hand side equals $|\\lambda| |\\hat{u}_{0,0}|^2$, while the right-hand side equals $4|\\lambda| |\\hat{u}_{0,0}|^2$, so the asserted equality fails; this calculation settles that the derivation as written is invalid, while leaving the rates themselves to be established through the two-sided inequality.","tokens_in":18836,"feed_emoji":"📉","tokens_out":14761,"duration_ms":147281,"temperature":0.7,"pith_summary":"This paper claims and argues that the damped wave equation $\\partial_t^2 u - \\Delta_H u + \\partial_t u = 0$ on the Heisenberg group $\\mathbf{H}_n$ obeys heat-like $L^2$ decay laws, with the homogeneous dimension $Q = 2n+2$ playing the role that the Euclidean dimension plays in the classical setting. For data in $H^1 \\times L^2$, the horizontal gradient decays as $(1+t)^{-1/2}$ and the time derivative as $(1+t)^{-1}$; if the data also lie in $L^1$, the solution decays as $(1+t)^{-Q/4}$, the gradient as $(1+t)^{-Q/4-1/2}$, and the time derivative as $(1+t)^{-Q/4-1}$. The proof proceeds through the group Fourier transform and Hermite functions, diagonalizing the sub-Laplacian mode by mode and splitting frequencies into a low region with heat-like polynomial decay and a high region with exponential decay. These rates match, with $Q$ in place of $n$, the known Euclidean estimates that the paper sets out to generalize.","feed_headline":"Heisenberg damped waves decay at heat-like rates","feed_subtitle":"On the Heisenberg group, L1 data give heat-like decay; gradients and time derivatives fall faster.","key_machinery":"The key machinery is the group Fourier transform of $\\mathbf{H}_n$ paired with the Hermite basis of $L^2(\\mathbb{R}^n)$. Under a Schr\\\"odinger representation $\\pi_\\lambda$, the sub-Laplacian becomes $d\\pi_\\lambda(\\Delta_H) = -|\\lambda| H_w$, where $H_w$ is the harmonic oscillator with Hermite eigenfunctions $e_k$ and eigenvalues $\\mu_k = 2|k| + n$. Each Fourier coefficient $(\\hat{u}(t,\\lambda)e_k, e_\\ell)_{L^2}$ solves the scalar damped oscillator equation $\\partial_t^2 v + \\partial_t v + \\mu_k |\\lambda| v = 0$. The proof splits $|\\lambda|$ about the threshold $1/(8\\mu_k)$: for small $|\\lambda|$, the factor $e^{-\\mu_k |\\lambda| t}$ together with the bound $\\|\\hat{u}(\\lambda)\\| \\le \\|u\\|_{L^1}$ and the convergence of $\\sum_k \\mu_k^{-(n+1)}$ yields the polynomial heat-like decay; for large $|\\lambda|$, uniform exponential decay takes over. Plancherel's formula with measure $c_n |\\lambda|^n d\\lambda$ converts these mode-by-mode estimates into the stated $L^2(\\mathbf{H}_n)$ bounds.","core_discovery":"The central assertion is that the heat-like decay law of the Euclidean damped wave equation extends to the Heisenberg group after replacing the Euclidean dimension $n$ by the homogeneous dimension $Q = 2n+2$. Theorem 1.1 states that, for data in $H^1(\\mathbf{H}_n) \\times L^2(\\mathbf{H}_n)$, the solution satisfies $\\|u(t,\\cdot)\\|_{L^2} \\lesssim \\|u_0\\|_{L^2} + \\|u_1\\|_{L^2}$, while $\\|\\nabla_H u(t,\\cdot)\\|_{L^2} \\lesssim (1+t)^{-1/2} (\\|\\nabla_H u_0\\|_{L^2} + \\|u_1\\|_{L^2})$ and $\\|\\partial_t u(t,\\cdot)\\|_{L^2} \\lesssim (1+t)^{-1} (\\|\\nabla_H u_0\\|_{L^2} + \\|u_1\\|_{L^2})$. With additional $L^1$ regularity of the data, those rates improve to $(1+t)^{-Q/4}$, $(1+t)^{-Q/4-1/2}$, and $(1+t)^{-Q/4-1}$, respectively. The argument is a phase-space analysis: the group Fourier transform sends the sub-Laplacian to $|\\lambda|$ times a harmonic oscillator, Hermite functions diagonalize that oscillator, and each resulting scalar equation is a damped oscillator whose low-frequency part behaves like a heat kernel and whose high-frequency part decays exponentially.","pith_inferences":["Editorial inference: because only a two-sided comparison holds at the step after (42), the theorem's rates are probably still true with a larger constant; the proof requires an inequality in place of the displayed equality, not a conceptual change.","Editorial inference: if these decay rates are optimal, they should produce a Fujita-type critical exponent $1 + 2/Q$ for the semilinear damped wave equation with power nonlinearity on $\\mathbf{H}_n$, matching the heat equation's critical exponent on the group; the paper's announced application points in this direction.","Editorial inference: the low-frequency/$L^1$ splitting should extend to other stratified Lie groups with an explicit Plancherel formula and Hermite-type eigenbases, yielding the same $Q$-based exponents whenever such bases are available."],"forward_implications":["If Theorem 1.1 is correct, $L^1(\\mathbf{H}_n)$ data produce the same decay rates as the heat semigroup on $\\mathbf{H}_n$: the solution decays as $(1+t)^{-Q/4}$ with $Q = 2n+2$.","In the $L^1$-improved regime, the time derivative decays faster than the horizontal gradient by a factor $(1+t)^{-1/2}$, matching the pattern of Euclidean damped waves.","Even without $L^1$ data, the horizontal gradient and time derivative decay polynomially, at rates $(1+t)^{-1/2}$ and $(1+t)^{-1}$, with constants depending only on $H^1 \\times L^2$ norms.","The vertical derivative $Tu$ satisfies a decay estimate with the same rate as $\\partial_t u$, at the price of requiring $T^{1/2} u_1 \\in L^2$, as stated in Section 5."],"supporting_citations":[{"why":"Supplies the Euclidean $L^1$-$L^2$ decay estimates that Theorem 1.1 generalizes, along with the low-frequency/$L^2$ splitting logic.","marker":"[21]"},{"why":"Introduces the projection of Fourier-transformed damped wave equations onto eigenbases for the sub-Laplacian on graded groups, the approach this paper adapts to the Heisenberg group.","marker":"[30]"},{"why":"Provides the group Fourier transform formalism and the Plancherel formula for $\\mathbf{H}_n$ used throughout the proof.","marker":"[11]"},{"why":"Gives the Hermite function orthonormal basis and harmonic oscillator eigenvalues used in Proposition 3.1 and in the mode-by-mode estimates.","marker":"[25]"}],"fun_headline_variants":["Heisenberg damped waves decay with homogeneous dimension Q=2n+2","L1 data gives heat-like decay on Heisenberg damped waves","Damped waves on Heisenberg: heat decay rate uses Q=2n+2","Heat-like decay on Heisenberg group: L1 data speeds it up"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof's gradient estimate depends on the equality displayed after (42), which identifies a weighted sum of Fourier coefficients of $u_0$ with twice the sum of Hilbert\\,--\\,Schmidt norms of the Fourier transforms of $(X_j + iY_j)u_0$ and $(X_j - iY_j)u_0$; that equality is false as stated, and only a comparable two-sided inequality is available, so the derivation needs an extra constant.","fun_headline_variants_meta":{"raw":{"variants":["Heisenberg damped waves decay with homogeneous dimension Q=2n+2","L1 data gives heat-like decay on Heisenberg damped waves","Damped waves on Heisenberg: heat decay rate uses Q=2n+2","Heat-like decay on Heisenberg group: L1 data speeds it up"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001291,"raw_usage":{"total_tokens":5288,"prompt_tokens":976,"completion_tokens":4312,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":592,"completion_tokens_details":{"reasoning_tokens":4230}},"tokens_in":592,"tokens_out":4312,"duration_ms":33452,"temperature":1.0,"reasoning_tokens":4230,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:41:46.288722+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Test the identity after (42) on data whose group Fourier transform at one fixed $\\lambda$ is the rank-one operator $(\\cdot, e_0)e_0$. The left-hand side equals $|\\lambda| |\\hat{u}_{0,0}|^2$, while the right-hand side equals $4|\\lambda| |\\hat{u}_{0,0}|^2$, so the asserted equality fails; this calculation settles that the derivation as written is invalid, while leaving the rates themselves to be established through the two-sided inequality.","supporting_citations":[{"cited_title":"Math., vol","cited_arxiv_id":null,"evidence_quote":"Provides the group Fourier transform formalism and the Plancherel formula for $\\mathbf{H}_n$ used throughout the proof."},{"cited_title":"Pseudo Diff","cited_arxiv_id":null,"evidence_quote":"Gives the Hermite function orthonormal basis and harmonic oscillator eigenvalues used in Proposition 3.1 and in the mode-by-mode estimates."}],"review_version":1}