{"id":"5698396b-1764-4110-bcc6-78dbf7238779","arxiv_id":"1908.03570","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"A function written as a sum of Laplacian eigenmodes with slowly growing coefficients is identically zero if it vanishes on a space-time cylinder whose duration exceeds the largest geodesic distance to the observation set.","lead":"This paper proves that a time-space signal built from the vibrating modes of a drum, if it vanishes on a small patch for long enough, must vanish everywhere. It also gives a unique-continuation statement for the wave equation: observing a solution on a small open set for a suitably long time determines the whole solution.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Final coefficient-vanishing step is unjustified: Hilbert-basis completeness does not cover s' coefficients; the gap is load-bearing but repairable via H^{-s} spectral theory.","rationale":"The reader's weakest assumption and my concern coincide. I considered other issues in the manuscript: the N=1 step appears to conflate a time shift with a spatial radius (it should set the time shift equal to the radius r, which is not stated), and the application of Lemma 2.7 requires spherical means to vanish for all centers in a neighborhood of x0, whereas the displayed computation gives the mean only at x0 (though the same argument with x∈ω repairs this). Both are real presentational gaps, but each has a straightforward repair and neither threatens the central claim itself. The final coefficient-injectivity step is more fundamental: it is the only place where the theorem's conclusion 'u≡0' is extracted from the established local vanishing, and the cited justification (Hilbert basis of H_0^1) genuinely does not apply to s′ coefficients. The H^{-s} spectral argument shows the underlying claim is true, so the appropriate verdict is unchanged: conditional acceptance, requiring the missing justification. I would not move to reject because the missing step is standard and no counterexample exists in the manuscript.","tokens_in":9942,"tokens_out":36421,"duration_ms":372419,"concrete_test":"Replace the final sentence of the proof with the following verification: given b_n∈s′ with ∑ b_n S_n=0 in D′(Ω), pick s with Σ|b_n|²λ_n^{-2s}<∞ and show the partial sums converge in H^{-s}(Ω). Since C_c^∞(Ω) is dense in H^s, the limiting H^{-s} element must be zero; pairing with S_m/λ_m^s yields b_m=0. If this computation cannot be carried through, the coefficient-vanishing step fails and Theorem 2.4 is unproved.","verdict_should_be":"UNCHANGED","load_bearing_attack":"At the end of the proof of Theorem 2.4 the author writes that, once u(t0,·) vanishes on a neighborhood of every point of Ω, 'the eigenvectors form a complete Hilbert basis of H_0^1(Ω)' suffices to conclude all a_n are zero. This is not a valid inference for the stated coefficient class. Hilbert-basis uniqueness applies to sequences in ℓ^2 (equivalently, to H_0^1 or L^2-valued sums); the assumption is only (a_n)∈s′, i.e. n^{-q}a_n∈ℓ^1 for some q, so the sum may define a distribution rather than an H_0^1 function. No theorem or reference is supplied showing that the eigenfunction expansion of a distribution with slowly growing coefficients is unique. This step is load-bearing: if a nontrivial s′-sequence satisfied ∑ b_n S_n=0 in D′(Ω), Theorem 2.4 would be false. The gap is probably repairable: for b∈s′, choose s so that Σ|b_n|²λ_n^{-2s}<∞; the series then converges in H^{-s}(Ω), where the normalized eigenfunctions form an orthonormal basis, and a zero distribution is the zero H^{-s} functional. But that argument uses a standard spectral representation absent from the paper, so the proof as written is incomplete.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies almost periodic distributions of the form u(t,x)=Σ a_n S_n(x)e^{iλ_nt} with (a_n) in the space s′ of slowly growing sequences and (λ_n²,S_n) the Dirichlet Laplacian eigenpairs on a bounded connected domain Ω. The main theorem asserts that if u vanishes on ω×(−T,T) for some nonempty open ω⊂Ω and T exceeds the supremum over P∈Ω of the geodesic distance from P to ω, then u≡0. The proof propagates local vanishing along polylines using spherical means and the Paley–Wiener theorem, and the paper derives a unique continuation property for the wave equation as an application. The geometric threshold T_max(Ω,ω) is explicit and parameter-free.","tokens_in":10165,"tokens_out":19185,"duration_ms":192664,"significance":"If the main theorem is correct, it provides a clean time-space trade-off for unique determination of these almost periodic distributions from observations on an arbitrary open subset for a time longer than the maximal geodesic distance. The proof strategy—using spherical means to convert temporal vanishing into spatial vanishing and then propagating with a chain of balls—is coherent and largely self-contained, and the applications to wave unique continuation are natural. A notable strength is that the criteria involve no fitted parameters and depend only on the geometry of Ω and ω. However, the final coefficient-injectivity step is not justified as written and must be repaired before the theorem is established.","major_comments":[{"comment":"The inference from u(t0,·)=0 in D′(Ω) to a_n=0 for all n is not justified by the Hilbert-basis property of H_0^1(Ω). The Hilbert basis property gives uniqueness of coefficients for sums convergent in H_0^1 (or L²), i.e., for coefficient sequences in ℓ², whereas the assumption is only (a_n)∈s′. Since the series may define a distribution rather than an H_0^1 function, an additional spectral argument is needed; for example, one can choose s so large that Σ|a_n|²λ_n^{−2s}<∞ and work in H^{−s}(Ω), where the eigenfunctions form an orthonormal basis. The paper supplies no such argument or reference.","section":"Proof of Theorem 2.4, final paragraph"},{"comment":"Lemmas 2.5 and 2.7 are invoked although their hypothesis—the vanishing of the spherical mean for all centers x in a neighborhood of x0—has only been verified for the fixed center x0. The displayed computation tests S(x0,·) against θ_{N,r} and therefore yields the spherical mean at x0 only. To apply the lemmas one must first observe that, by choosing ε so that B_{2ε}(x0)⊂ω, the same computation can be repeated for every x∈B_ε(x0); as written this step is missing.","section":"Proof of Theorem 2.4, after Eqs. (2.12)–(2.13)"},{"comment":"The equation as printed is not correct. The left-hand side, 1/2(S(x0,t0+tc)+S(x0,−t0+tc)), is a time reflection and would produce cos(λ_nt0), while the right-hand side contains G1(rλ_n), which arises from a spatial spherical mean. The intended identity is 1/2(S(x0−r,tc)+S(x0+r,tc)) = Σ a_n K_φ(n,x0)G1(rλ_n)e^{iλ_ntc}, up to the normalization constants from θ_{1,r}. This notational confusion must be fixed because the N=1 case depends on it.","section":"Proof of Theorem 2.4, displayed equation after (2.13), N=1 case"}],"minor_comments":[{"comment":"The abstract restricts to N=1,2,3 while the introduction and Theorem 2.4 state N≥1; these should be harmonized.","section":"Abstract and Theorem 2.4"},{"comment":"The symbol T0 is introduced without definition and is used both as a spatial radius and as a time bound in (2.13); the radius should be named, say r0 or t0, and its use as a temporal shift should be explained.","section":"Proof of Theorem 2.4, paragraph before Eq. (2.13)"},{"comment":"The Fourier transform formula for J0(rλ) appears to have the support condition reversed: as written H(1−(r/ξ)²) is nonzero for |ξ|>r, while the text says the function is integrable on ]−r,r[. The intended formula likely has H(1−(ξ/r)²).","section":"Eq. (2.7)"},{"comment":"The boundary condition is written on [0,+∞)×∂Ω, but the time domain is ]−T,T[; it should be ]−T,T[×∂Ω.","section":"Section 3, Eq. (3.1)"},{"comment":"There are several typos: 'slowing growing' should be 'slowly growing'; 'fixed any ∈]−tP,tP[' is missing the variable t; and 'extended ... oddly' should be 'extended ... as an odd function'.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The paper fits the scope of math.AP and the main theorem is plausible. I recommend asking the author to supply the missing spectral uniqueness argument for s′ coefficients, to make explicit the verification of the hypotheses of Lemmas 2.5 and 2.7, and to correct the N=1 display. If these points are addressed, the theorem and its applications are likely to be acceptable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Short version: this is a real result with a repairable hole in the proof. The space-time uniqueness theorem (Theorem 2.4) is new as stated, and the spherical-mean/Paley-Wiener route is a sensible way to get it. The explicit geodesic threshold T_max is a clean addition, and the paper is self-contained against standard ingredients. No parameter fitting, no circularity, and the self-citations to [10] and [11] are appropriate.\n\nBut the proof has two load-bearing gaps. The final step concludes a_n = 0 from spatial vanishing on a neighborhood of every point using only the Hilbert-basis property of the eigenfunctions in H^1_0(Ω). That guarantees uniqueness for ℓ² coefficients, not for the s' sequences allowed here. The series may converge only as a distribution, so the inference is unjustified as written. The fix is standard: pick s so that Σ |a_n|² λ_n^{-2s} < ∞, work in H^{-s}, where the eigenfunctions are an unconditional basis, and zero distribution means zero functional. But that argument is absent, and the step is load-bearing. If a nontrivial s' sequence satisfied ∑ b_n S_n = 0 in D'(Ω), the theorem would be false.\n\nSecond gap: the N=1 case conflates time and space shifts. The paper writes S(x0, t0+tc) + S(x0, -t0+tc) to obtain G1(rλ_n), but the spatial spherical mean requires S(x0-r, tc) + S(x0+r, tc). The line as written is wrong; the intended fix is to apply the same Paley-Wiener convolution to the already-defined spatial mean (2.11). This is likely a typo, but it makes the N=1 proof incorrect as it stands.\n\nSmaller issues: the abstract says N=1,2,3 while the main text proves N≥1; the repeated-eigenvalue case is dispatched as \"essentially the same\" and would need more detail; and some interchanges of sums and integrals are justified by Monotone Convergence in a distribution-valued setting, which is sloppy. None of these are fatal.\n\nThe wave-equation application overlaps with known results (Ruiz; Bosi-Kurylev-Lassas; Joly-Laurent), so the novelty there is limited, but the corollary is a clean consequence of the main theorem.\n\nWho this is for: researchers in inverse problems and wave propagation who care about uniqueness from boundary or interior observations. The main theorem is likely true, but the write-up needs revision before acceptance. I would send it to a serious referee and expect major revision.","headline":"New space-time uniqueness theorem with an explicitly repairable gap in the final coefficient step; sound strategy, incomplete write-up.","tokens_in":69,"tokens_out":3900,"would_cite":false,"duration_ms":164135,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35L05","35B60","35P10","42A75"],"pacs":[],"model":"deepseek-v4-flash","headline":"Observing a wave on a small patch for longer than the maximal geodesic distance to it forces the wave and its initial data to vanish.","keywords":["almost periodic distributions","unique continuation","wave equation","spherical means","Dirichlet Laplacian","slowly growing sequences","geodesic distance","eigenfunction expansions"],"falsifier":"Find a nonzero slowly growing sequence (c_n) such that the distribution Σ c_n S_n vanishes identically on a nonempty open subset of Ω (or on all of Ω). For Ω = (0, π) with S_n = sin(nx), the sine coefficients of a zero distribution are all zero, so no counterexample exists there; for a general domain the question is open. A concrete search would take a domain, compute its Dirichlet eigenfunctions, and attempt to make partial sums Σ_{n=1}^N c_n S_n small on a chosen open patch while keeping the coefficients polynomially bounded and the sum nonzero elsewhere; if such coefficients persist as N → ∞, the final Hilbert-basis inference in Theorem 2.4 collapses.","tokens_in":9667,"feed_emoji":"🌊","tokens_out":10317,"duration_ms":114489,"temperature":0.7,"pith_summary":"The paper establishes a uniqueness theorem for space-time distributions built from Dirichlet Laplacian eigenfunctions: if such a distribution vanishes on an observation patch ω for a time interval longer than T_max(Ω, ω), the largest geodesic distance inside Ω from any point to ω, then it is identically zero. The threshold is purely geometric and independent of the coefficients. This yields a unique continuation property for the wave equation: zero observation of a wave on a small patch over a sufficiently long time forces the wave and its initial data to vanish. It also forces the source to vanish in a forced-wave problem with zero initial data.","feed_headline":"Silence on a patch long enough forces the whole wave to zero","feed_subtitle":"A time interval longer than the maximal geodesic distance to the observation set makes vanishing data uniquely decisive.","key_machinery":"The central object is the radial function G_N(rλ), the spherical-mean kernel of an eigenfunction: for fixed x, the spherical mean of S_n over a sphere of radius r equals S_n(x) G_N(rλ_n), with G_1 = cos, G_2 = J_0 (the Bessel function of order 0), G_3 = sinc, and for N ≥ 4 an entire even power series. Because G_N(r·) has a compactly supported inverse Fourier transform, the identity expressing the time signal at a point can be transformed into an identity for spherical means at every radius. The mean-value lemmas (Lemmas 2.5 and 2.7) then turn vanishing of all spherical means around a point into vanishing on a full ball, and induction along a polyline of overlapping balls propagates the zero from the observation set to every point of Ω.","core_discovery":"Let Ω be open, bounded, and connected with continuous piecewise C∞ boundary, and let ($λ_n^{2}$, S_n) be the Dirichlet Laplacian eigenpairs normalized in $L^{2}$(Ω). For any open ω⊂Ω, define T_max(Ω, ω) = sup_{P∈Ω} gd(P, ω), the longest geodesic distance inside Ω from a point to the observation set. The theorem states: if u(t, x) = Σ a_n S_n(x) $e^{{i λ_n t}}$ with (a_n) a slowly growing sequence (in the space s′) and u|ω×]−T,T[ = 0 for some T > T_max, then u ≡ 0. The proof uses spherical means of eigenfunctions, which are shown to equal S_n(x) G_N(r λ_n), and a compactly supported distribution whose Fourier transform is the radial kernel G_N, to convert the long-time vanishing on ω into vanishing of spherical means on small balls. A mean-value lemma then propagates the zero ball-by-ball along any polyline from ω to an arbitrary point P ∈ Ω, giving u(P, t) = 0 near t = 0. The final step invokes completeness of the eigenbasis in $H_0^{1}$(Ω) to conclude that all coefficients vanish. Corollary 2.9 extends the result to two-sided frequencies, covering real cosine and sine waves, and two applications to the wave equation follow.","pith_inferences":["If the final Hilbert-basis step is made fully rigorous for slowly growing coefficient sequences, the same spherical-mean proof scheme should extend to other self-adjoint operators whose eigenfunctions have explicit spherical-mean kernels, such as the plate or biharmonic equation the author mentions; the paper does not pursue this.","The theorem supplies a sufficient time, not a necessary one. Testing numerically with finite sums on a domain shaped so that T_max is long (a geodesic shadow) could reveal whether shorter observation times still determine the distribution, or whether the threshold is sharp.","The mean-kernel argument is essentially a local unique-continuation device: it turns a global geometric condition into local propagation of zero. A natural extension would be to derive stability or quantitative estimates near T_max, though the paper does not do so.","The theorem as stated requires ω to be open because the proof uses test functions supported in ω. Whether an analogue holds for observation sets with empty interior (for example a point or a curve) is a separate question not settled here."],"forward_implications":["For the wave equation with Dirichlet boundary conditions and data (w0, v0) ∈ H_0^1(Ω) × L^2(Ω), vanishing on ω×]−T,T[ with T > T_max forces w0 = v0 = 0.","In the forced problem with zero initial data and source g(t)f(x), where g(0) ≠ 0 and f ∈ H^{−1}(Ω), vanishing on ω×]0,T[ forces f ≡ 0.","The same geometric threshold T_max works for any nonempty open observation set and depends only on Ω and ω, not on the coefficients or the particular frequencies.","Because coefficients are allowed to be slowly growing rather than absolutely summable, the result covers tempered distributions in time, a wider class than classical almost periodic functions.","Corollary 2.9 extends the uniqueness to sums containing both e^{iλ_n t} and e^{−iλ_n t}, which is exactly the form of real-valued wave solutions."],"supporting_citations":[{"why":"Supplies the mean-value identity (Lemma 2.6) used to propagate vanishing of spherical means from one ball to an adjacent ball for N ≥ 2.","marker":"[14]"},{"why":"Supplies the spherical-means framework used to represent eigenfunction means and to define the radial kernel G_N.","marker":"[7]"},{"why":"Defines the class of almost periodic distributions with slowly growing coefficient sequences and the earlier uniqueness analysis that this paper extends to the space-time setting.","marker":"[11]"},{"why":"Provides eigenfunction growth estimates used to justify interpreting the formal series u(·, x) as a tempered distribution.","marker":"[6]"},{"why":"Supplies the Weyl law for eigenvalue growth, used to place the frequencies λ_n and the admissible coefficient space.","marker":"[5]"},{"why":"Supplies the power-series solution for the radial kernel G_N for N ≥ 4, needed to define the kernel in all dimensions.","marker":"[13]"},{"why":"Provides a prior unique continuation result for the wave operator that the wave-equation application extends or complements.","marker":"[12]"},{"why":"Gives a related inverse wave-source uniqueness result used in deriving the forced-wave application with source g(t)f(x).","marker":"[3]"}],"fun_headline_variants":["Wave dies if silent on a patch longer than the maximum travel time","Observe zero on a patch long enough, the wave is identically zero","Threshold time from geodesic distance forces wave to vanish","Silence beyond the geodesic bound makes the wave zero","A wave silent too long on a patch must be zero everywhere"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The final step of the proof assumes that if the distribution Σ a_n S_n vanishes in a neighborhood of every point of Ω then every coefficient a_n is zero, even though the coefficients are only slowly growing and the cited completeness of the eigenfunctions is a Hilbert-basis statement for square-summable coefficients; the paper supplies no separate theorem for this distributional case.","fun_headline_variants_meta":{"raw":{"variants":["Wave dies if silent on a patch longer than the maximum travel time","Observe zero on a patch long enough, the wave is identically zero","Threshold time from geodesic distance forces wave to vanish","Silence beyond the geodesic bound makes the wave zero","A wave silent too long on a patch must be zero everywhere"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000545,"raw_usage":{"total_tokens":2671,"prompt_tokens":1072,"completion_tokens":1599,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":688,"completion_tokens_details":{"reasoning_tokens":1512}},"tokens_in":688,"tokens_out":1599,"duration_ms":13585,"temperature":1.0,"reasoning_tokens":1512,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T14:17:12.709458+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Find a nonzero slowly growing sequence (c_n) such that the distribution Σ c_n S_n vanishes identically on a nonempty open subset of Ω (or on all of Ω). For Ω = (0, π) with S_n = sin(nx), the sine coefficients of a zero distribution are all zero, so no counterexample exists there; for a general domain the question is open. A concrete search would take a domain, compute its Dirichlet eigenfunctions, and attempt to make partial sums Σ_{n=1}^N c_n S_n small on a chosen open patch while keeping the coefficients polynomially bounded and the sum nonzero elsewhere; if such coefficients persist as N → ∞, the final Hilbert-basis inference in Theorem 2.4 collapses.","supporting_citations":[{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the mean-value identity (Lemma 2.6) used to propagate vanishing of spherical means from one ball to an adjacent ball for N ≥ 2."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the spherical-means framework used to represent eigenfunction means and to define the radial kernel G_N."},{"cited_title":"1, 135–152","cited_arxiv_id":null,"evidence_quote":"Defines the class of almost periodic distributions with slowly growing coefficient sequences and the earlier uniqueness analysis that this paper extends to the space-time setting."},{"cited_title":"Grebenkov and Binh Nguyen, Geometrical structure of Laplacian eigen- functions, SIAM Review 55 (2013), no","cited_arxiv_id":null,"evidence_quote":"Provides eigenfunction growth estimates used to justify interpreting the formal series u(·, x) as a tempered distribution."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the Weyl law for eigenvalue growth, used to place the frequencies λ_n and the admissible coefficient space."},{"cited_title":"140, Amer","cited_arxiv_id":null,"evidence_quote":"Supplies the power-series solution for the radial kernel G_N for N ≥ 4, needed to define the kernel in all dimensions."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides a prior unique continuation result for the wave operator that the wave-equation application extends or complements."},{"cited_title":"9-10, 2055–2069","cited_arxiv_id":null,"evidence_quote":"Gives a related inverse wave-source uniqueness result used in deriving the forced-wave application with source g(t)f(x)."}],"review_version":1}