{"id":"bedc9406-cc19-4e93-a06f-0b6898445d5b","arxiv_id":"2501.06957","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"On 3D manifolds close to constant negative curvature, wave solutions decay like t^{-1} and Schrödinger solutions like t^{-3/2}, with small metric and potential perturbations allowed.","lead":"This paper proves decay estimates for wave and Schrödinger equations on three-dimensional spaces that are small perturbations of the hyperbolic space. The result gives decay rates t^{-1} for waves and t^{-3/2} for Schrödinger solutions even when a small potential is added.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Cosine propagator bound in Theorem 1 is asserted but not proved in Section 5.4; tS_V in U(L1,L∞) does not control the time derivative C_V = ∂_t S_V.","rationale":"The reader's formal weakest_assumption is the L1(Γ) smallness of the curvature deviation, which is a plausible and clearly stated geometric hypothesis. However, the reader's rationale also flags the unproved cosine propagator bound in Section 5.4, and that is the most load-bearing concern about the central claim: Theorem 1's first estimate (10) is not established by the arguments provided. The inference from tS_V(t) ∈ U(L1,L∞) to the cosine L1→L∞ bound is not a routine consequence, because the cosine propagator is the time derivative of the sine propagator and the U-space estimate does not control derivatives. In the constant-curvature cases the paper supplies an explicit kernel (98) and performs the needed integration by parts; the analogous computation is absent for the nonconstant case. This is a specific, addressable gap rather than evidence that the theorem is false, so the appropriate verdict remains conditional, unchanged from the reader's assessment. My concern overlaps with the reader's rationale, hence partial agreement with the reader's stated weakest_assumption field.","tokens_in":31439,"tokens_out":8351,"duration_ms":83758,"concrete_test":"Re-derive the cosine bound in the nonconstant setting by setting C_V(t) = ∂_t S_V(t), inserting the convergent series (65) for S_V, and computing ∂_t S0(t) explicitly (the analogue of (98)). Check whether the δ' term can be bounded in B(W^{2,1}, L∞) by t^{-1} using the established estimates (41)–(42), (49), (57), (80)–(84) for a = det T. If the integration by parts produces a term requiring a bound on the tangential derivative of a^{-1/2} of size O(j(t)^{-1}) that is not among these estimates, then the inference in Section 5.4 is invalid. Alternatively, attempt the route C_V(t) = I − ∫_0^t H S_V(s) ds and show ∫_0^t ∥H S_V(s)f∥_{L∞} ds ≲ t^{-1}∥f∥_{W^{2,1}}; since H S_V(s) = −∂_s C_V(s), this is circular unless one differentiates the Duhamel series, which still requires the missing computation.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 5.4 (Proof of Theorem 1) states that tS_V(t) ∈ U(L1,L∞) implies the t^{-1} L1→L∞ decay estimate for the cosine propagator, same as (116), but no derivation is given. This does not follow from the stated U-space bound: C_V(t) = cos(t√H) = ∂_t S_V(t), and a bound on S_V(t) gives no control of its time derivative. In the constant-curvature cases (Sections 5.1–5.3), the cosine estimate is obtained from the explicit kernel (98), which contains a distributional δ' term (equivalently δ_{d=t}∂_r) and requires integration by parts against the amplitude a^{-1/2}; the needed estimates on ∇a and ∇²a are available from Proposition 6, but the paper never performs this computation in the nonconstant setting. Consequently the cosine estimate (10), a central claim of Theorem 1, is unproved as written. The same gap affects the exponential replacement t ↦ sinh(t(α0−δ)).","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves dispersive (L^1-to-L^\\infty) decay estimates for Schr\\\"odinger and wave equations on three-dimensional Riemannian manifolds, with the main theorem treating metrics whose curvature tensor is close to constant negative curvature in a geodesic-\\ell^1 sense, plus a small Kato-class scalar potential. The technical core is a construction of the sine propagator as a convergent series of approximate cone-supported measures, using Jacobi-field estimates and an algebra of integral kernels (the U-spaces). Model results are also given for constant-curvature S^3 and H^3, with and without potentials.","tokens_in":31664,"tokens_out":11925,"duration_ms":119256,"significance":"If the main theorem is correct, it is a substantial contribution: it would give global t^{-1} and t^{-3/2} decay on nonconstant negatively curved backgrounds under relatively explicit geometric smallness assumptions, and the exponential replacement by sinh(t(\\alpha_0-\\delta)) would be a strong hyperbolic-space-like result. The paper has real strengths: Proposition 4 gives a clean explicit sine-propagator identity for constant curvature; the contraction argument for Jacobi fields in Section 4.1 is detailed; the U-algebra framework in Proposition 6 is elegant and, notably, contains no fitted parameters—the smallness assumptions are hypotheses rather than quantities tuned to force the conclusions. The main weakness is that the cosine-propagator estimates, which are load-bearing for Theorems 1 and 3, are asserted rather than proved in the nonconstant and perturbed settings, and the one displayed Duhamel expansion used for the perturbed cosine propagator is algebraically incorrect as written.","major_comments":[{"comment":"The proof of Theorem 1 asserts that tS_V(t) \\in U(L^1,L^\\infty) \"implies the t^{-1} L^1 \\to L^\\infty decay estimate for the cosine propagator, same as (116)\". This implication is not established and is not a direct consequence of the stated U-space bound: C_V(t)=\\partial_t S_V(t), and an integrable bound on tS_V gives no control of its time derivative. In the constant-curvature cases the cosine estimate is obtained from the explicit distribution kernel (98), which contains a \\delta' term and requires the \\nabla a and \\nabla^2 a estimates supplied by Proposition 6; no analogue of that computation is carried out for the nonconstant metric. The same gap affects the claimed replacement of t by \\sinh(t(\\alpha_0-\\delta)) in (10), since only j_\\delta(t) S_V(t), not j_\\delta(t) C_V(t), is proved to lie in U(L^1,L^\\infty).","section":"Section 5.4, proof of Theorem 1"},{"comment":"The displayed expansion \\chi_{t\\ge0}C=(I+[\\chi S_0 V])^{-1}*[\\chi C_0] does not follow from differentiating (112). Since S=S_0*(I+V S_0)^{-1}, differentiation gives C=C_0*(I+V S_0)^{-1}+S_0*\\partial_t[(I+V S_0)^{-1}], and the second term is not accounted for in the series shown. The correct expansion should contain terms such as C_0*V S_0 and S_0*V C_0 in addition to S_0*V S_0*V C_0; omitting them means the cosine estimates (116) for the perturbed problem are not justified. This affects Theorem 3 and the constant-curvature-with-potential case, not just the general setting of Theorem 1.","section":"Section 5.3, Duhamel expansion for the cosine propagator"}],"minor_comments":[{"comment":"The formula S_0(t)=(4\\pi j(t))^{-1}\\delta_{d=|t|} is not correct on S^3 for all t: for |t|>\\pi the set {d=|t|} is empty, while the propagator is supported at distance 2\\pi-|t| with the sign given by the oddness relation S_0(t)=-S_0(2\\pi-t), which Section 5.1 later uses. Please state the formula on a fundamental period or with the appropriate antipodal identification.","section":"Proposition 4"},{"comment":"The cosine kernel formula (98) is stated without derivation; since the estimates (100)-(101) and (105)-(106) depend on its distributional structure, a short derivation from Proposition 4 or from the spectral decomposition would improve the readability and verifiability of the constant-curvature sections.","section":"Equation (98)"},{"comment":"The notation \\delta_{d=t} with t\\in(\\pi,2\\pi) is confusing because the sphere of radius t is empty; writing the support explicitly as \\delta_{d=2\\pi-t} would make the sign conventions in the cosine and sine estimates unambiguous.","section":"Section 5.1"},{"comment":"The estimate F[e^{it\\lambda^2}]\\lesssim t^{-1/2} is stated loosely: the oscillatory integral is a distribution and needs a stationary-phase justification, and the operator-valued Fourier transform notation should be clarified before the t^{-3/2} conclusion is drawn.","section":"Equation (109)"}],"recommendation":"major_revision","confidential_remarks":"The two major comments concern the two places where cosine decay is derived: the nonconstant proof in Section 5.4 and the perturbed constant-curvature proof in Section 5.3. Both are central to the advertised theorems. The sine-propagator construction and the U-algebra estimates appear sound and are the paper's genuine contribution, so I do not recommend rejection; however, the cosine estimates need either a correct derivation or a revision of the theorems to state only what is proved. The Section 5.3 derivative error in particular should be caught by the authors before resubmission."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: the sine-propagator and Schrödinger parts look like real progress. The cosine propagator bound in Theorem 1 is not proved in the text; Section 5.4 jumps from tS_V in U(L1,L∞) to the t^{-1} cosine estimate, which doesn't follow because cos = ∂_t sin. That's a genuine gap, not a cosmetic issue.\n\nWhat's new: the non-constant negative curvature setting with small potential. The geometric machinery—Jacobi fields, L1(Γ) conditions, the U-algebra—is carefully developed. Proposition 6 is detailed and plausible: the error estimates, the scattering construction of T∞, and the bootstrapping all line up. If the sine and Schrödinger parts are correct (and they look like they are), this is the first L1→L∞ decay for Schrödinger on such manifolds, which goes beyond existing Strichartz results.\n\nSoft spots, in order. First and load-bearing: the cosine estimate (10) in Theorem 1 is unproved. Section 5.4's sentence \"The latter implies...\" is the entire argument. In the constant-curvature case (Section 5.2), the cosine estimate uses the explicit kernel (98) with a δ' term and integration by parts; that computation is never done for nonconstant curvature, and there's no reason it should be automatic from the U-space bound. The exponential replacement carries the same gap. This is addressable, but it's real work and it's missing. Second: the metric regularity assumptions are inconsistent between the introduction (two derivatives) and the abstract/theorem (four derivatives, via ∇^2 Rm). Minor, but should be fixed. Third: the constant-curvature parts overlap with known results; the paper acknowledges this, so it's not a major issue.\n\nWho is this for? People working on dispersive estimates on curved manifolds, especially negative curvature. The sine and Schrödinger results are probably citable once the gap is fixed, but as is, I wouldn't rely on Theorem 1's cosine bound. This paper deserves a serious referee—it's exactly the kind of work where an expert can help the author close a specific gap. I'd send it out rather than desk reject, with a clear request for a proof of the cosine estimate.","headline":"Strong sine-propagator and Schrödinger results on perturbed hyperbolic manifolds, but the cosine bound in Theorem 1 is asserted without proof; referee should demand a real argument.","tokens_in":32155,"tokens_out":5094,"would_cite":false,"duration_ms":51962,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35L05","35Q41","35B40","53B20","58J37","58J45"],"pacs":[],"model":"deepseek-v4-flash","headline":"The paper claims that on 3D manifolds with curvature close to constant negative curvature, the wave propagator decays as $t^{-1}$ and the Schrödinger propagator as $|t|^{-3/2}$, even after adding a small scalar potential.","keywords":["dispersive estimates","Schrödinger equation","wave equation","Riemannian manifolds","constant negative curvature","Jacobi fields","Kato class","propagator series"],"falsifier":"Integrate the Jacobi equation numerically along a single long geodesic in a 3D metric built so that $\\int |Rm-Rm_0|\\,ds = \\epsilon/2$ but the curvature deviation is concentrated in a long interval; the paper's estimates require $|J(r)-j(r)e|\\lesssim \\epsilon j(r)$ throughout. If, before the curvature integral reaches $\\epsilon$, the deviation exceeds $C\\epsilon j(r)$ for any reasonable constant $C$, the contraction step in Section 4.1 fails and the central claim would be false for that family of metrics.","tokens_in":31249,"feed_emoji":"🌊","tokens_out":7716,"duration_ms":71152,"temperature":0.7,"pith_summary":"This paper tries to show that the basic decay laws for waves and Schrödinger solutions—$t^{-1}$ for the wave propagator and $|t|^{-3/2}$ for Schrödinger—are not special to flat or hyperbolic space, but survive on any three-dimensional Riemannian manifold whose curvature tensor is a small perturbation of constant negative curvature. The main result, Theorem 1, states that under such a geometric smallness condition, together with a small scalar potential in a global Kato class, the cosine propagator decays like $t^{-1}$ in $L^1\\to L^\\infty$ and the Schrödinger propagator like $|t|^{-3/2}$. If true, dispersive behavior is stable under small geometric perturbations, so decay-based arguments for nonlinear equations can be run on these curved backgrounds. The paper also proves matching estimates on the sphere $\\mathbb{S}^3$ and hyperbolic space $\\mathbb{H}^3$, where the sphere shows the expected periodic refocusing with $1/\\sin t$ factors.","feed_headline":"Near-hyperbolic 3D manifolds keep hyperbolic decay rates","feed_subtitle":"Small curvature perturbations leave the t^{-1} wave rate and t^{-3/2} Schrödinger rate intact.","key_machinery":"The load-bearing object is the sine propagator series. On constant-curvature space the sine propagator is exactly $(4\\pi j(t))^{-1}\\delta_{d(x_0,x)=t}$, with $j(t)=\\sinh(\\alpha_0 t)/\\alpha_0$ on $\\mathbb{H}^3$ and $j(t)=\\sin t$ on $\\mathbb{S}^3$. For a perturbed metric the paper writes the approximate propagator $S_0(t)=(4\\pi\\sqrt{a})^{-1}\\delta_{d=t}$, where $a=J_1\\wedge J_2$ is the area element from two Jacobi fields along geodesics; the Laplacian of $a^{-1/2}$ produces an error term $E$ supported on the light cone, and the true propagator is $S=S_0+S_0*E+S_0*E*E+\\cdots$ in the convolution algebroid $U(X,Y)$ whose norm tests time-integrated operator bounds between Banach lattices. Conditions (24), (26), and (27) make the contraction estimate $|J-j\\,e|\\lesssim \\epsilon j$ and the error bound $\\|E\\|_{U(L^1)}\\lesssim \\epsilon$ work, so the series converges and the derivation identity $tS=(tS_0)*(I-E)^{-1}+S_0*(I-E)^{-1}*(tE)*(I-E)^{-1}$ converts free $t^{-1}$ decay into perturbed $t^{-1}$ decay. The same algebra with exponential weights produces the $\\sinh$ improvement.","core_discovery":"The central claim is that on a simply connected 3D manifold whose Riemann curvature tensor satisfies the geodesic integrability conditions (24), (26), and (27)—the $L^1$ norm along every geodesic of $Rm-Rm_0$, $\\nabla Rm$, and $\\nabla^2 Rm$ is smaller than a tiny $\\epsilon$, with corresponding global $L^1$ smallness—and with a small Kato-class potential $V$, the cosine propagator satisfies $\\|\\cos(t\\sqrt{H})f\\|_{L^\\infty} \\lesssim t^{-1}\\|f\\|_{W^{2,1}}$ and the Schrödinger propagator satisfies $\\|e^{itH}f\\|_{L^\\infty} \\lesssim |t|^{-3/2}\\|f\\|_{L^1}$. When $V$ lies in the modified Kato class $\\widetilde{K}$, the polynomial $t$ can be replaced by $\\sinh(t(\\alpha_0-\\delta))$ for any $\\delta>0$. The mechanism is direct and geometric: the sine propagator is approximated by the explicit surface measure built from Jacobi fields, the error is shown to be small in a convolution algebroid, and the true propagator is obtained by iterating this error in a convergent series. A byproduct is the absence of embedded eigenvalues and threshold resonances for small potentials.","pith_inferences":["If the result is correct, the natural next test is whether the geodesic-by-geodesic condition (24) can be relaxed to an averaged $L^1$ condition, since the paper itself uses global $L^1$ smallness to sharpen the geodesic bounds; one could engineer a metric where the geodesic supremum is slightly large but the average is small and see whether decay still holds.","The same Jacobi-field contraction may extend to asymptotically hyperbolic or non-positive curvature settings in three dimensions, while positive curvature would require a different ansatz because conjugate points destroy the surface-measure propagator; the paper explicitly flags positive curvature as future work.","A concrete numerical check of the paper's estimates is to integrate the Jacobi equation along a long geodesic in a perturbed metric and compare $a=J_1\\wedge J_2$ with $j^2$: the estimates predict $|a-j^2|\\lesssim \\epsilon j^2$ and $|\\partial_r a-2jj'|\\lesssim \\epsilon jj'$, which is directly testable.","For applications, the exponential-in-$\\sinh$ rate suggests that very large-time behavior is governed by the constant-curvature background, so model-space or effective one-dimensional reductions may be justified for nonlinear problems on these manifolds."],"forward_implications":["The $t^{-3/2}$ Schrödinger and $t^{-1}$ wave decay estimates imply the usual Strichartz estimates on the perturbed manifold, so nonlinear well-posedness results on hyperbolic space transfer to these curved backgrounds.","For small potentials in the modified Kato space, the decay is exponential in $\\sinh(t(\\alpha_0-\\delta))$, giving a quantitative rate independent of the potential except through its size.","On $\\mathbb{S}^3$, the wave propagators obey $1/\\sin t$ estimates, consistent with periodic refocusing and discrete spectrum; the paper derives spectral projection bounds from these estimates.","The convergent Born/Duhamel series rules out embedded eigenvalues and threshold resonances for small potentials, since a resonance would create a pole the convergent series cannot have.","The estimates are new even in some constant-curvature cases, providing $L^p$ decay for shifted wave and Schrödinger equations on $\\mathbb{H}^3$ and $\\mathbb{S}^3$ through one geometric proof."],"supporting_citations":[{"why":"Supplies the global Kato class in which the smallness of the potential $V$ is measured and gives the kernel bounds used to control the potential.","marker":"[RodSch]"},{"why":"Introduces the $U(X,Y)$ algebroid and the approach of summing convolutions of error terms to estimate perturbed propagators.","marker":"[BecGol1]"},{"why":"Provides the companion method for passing from small to large potentials, cited by the paper as the way to extend the present small-potential results.","marker":"[BecGol2]"},{"why":"Serves as the model result for nonlinear waves on 3D hyperbolic space that the current paper's decay estimates extend and improve.","marker":"[MetTay]"},{"why":"Supplies the global parametrices and dispersive estimates for variable-coefficient wave equations that motivate the integrability-along-geodesics assumptions used here.","marker":"[MetTat]"}],"fun_headline_variants":["Hyperbolic decay survives small metric tweaks","Small curvature bumps don't change decay rates","Bending space slightly? Decay rates hold","3D manifolds: perturbed, but decay rates intact","Hyperbolic geometry locks in decay exponents"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The whole proof rests on the assumption that along every infinite geodesic the total amount by which the curvature tensor differs from constant negative curvature, together with the totals of its first two derivatives, is smaller than one tiny epsilon; if some geodesic accumulates even a moderate amount of curvature deviation, the contraction estimate for Jacobi fields and the convergence of the propagator series may fail.","fun_headline_variants_meta":{"raw":{"variants":["Hyperbolic decay survives small metric tweaks","Small curvature bumps don't change decay rates","Bending space slightly? Decay rates hold","3D manifolds: perturbed, but decay rates intact","Hyperbolic geometry locks in decay exponents"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000338,"raw_usage":{"total_tokens":1888,"prompt_tokens":984,"completion_tokens":904,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":600,"completion_tokens_details":{"reasoning_tokens":834}},"tokens_in":600,"tokens_out":904,"duration_ms":7474,"temperature":1.0,"reasoning_tokens":834,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T20:50:37.254217+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Integrate the Jacobi equation numerically along a single long geodesic in a 3D metric built so that $\\int |Rm-Rm_0|\\,ds = \\epsilon/2$ but the curvature deviation is concentrated in a long interval; the paper's estimates require $|J(r)-j(r)e|\\lesssim \\epsilon j(r)$ throughout. If, before the curvature integral reaches $\\epsilon$, the deviation exceeds $C\\epsilon j(r)$ for any reasonable constant $C$, the contraction step in Section 4.1 fails and the central claim would be false for that family of metrics.","supporting_citations":[],"review_version":1}