{"id":"20f89fcd-24ce-4253-b1cf-6d5f90f29ede","arxiv_id":"2412.00780","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For the fractional Laplacian Schrödinger propagator, the paper obtains kernel bounds and dispersive/Strichartz estimates on real hyperbolic spaces and homogeneous trees, with no derivative loss on trees.","lead":"The paper proves dispersive and Strichartz estimates for the fractional Schrödinger equation on hyperbolic spaces and on homogeneous trees, for fractional powers α between 0 and 2. It shows that the derivative loss seen on Euclidean spaces persists on hyperbolic spaces but disappears on trees, where the propagator decays at the same rate for every α.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The small-time dispersive exponent stated in Theorem 1.1 and Theorem 4.7(i) for 0<α<1 is internally inconsistent with the kernel estimates and is contradicted at q=∞ by the nonzero t→0 limit of the kernel.","rationale":"The reader identified the omitted Strichartz TT* arguments as the weakest assumption and returned CONDITIONAL. My stress-test found a more direct and more serious problem: the small-time dispersive statement for 0<α<1, which is a central theorem of the paper, is false as written. The kernel estimates in Theorem 3.5 are detailed and appear coherent, and the homogeneous-tree section is independent and plausible, but the conversion from kernel bounds to dispersive bounds in §4.2 contains an exponent error. At q=∞ the claimed decay is faster than the kernel sup bound permits, and for explicit admissible parameters the kernel at r=0 has a nonzero limit as t→0, so no positive power decay can hold. The later Strichartz proof in (57) uses the corrected exponent βn/α, which indicates a repeated typo rather than a different strategy. Nevertheless, because Theorem 1.1, Theorem 4.7, and the admissibility region of Theorem 4.9 all rely on this exponent, the current preprint should not be accepted without correction. I recommend REJECT as the appropriate verdict for the version as written, with the expectation that correcting 2n-σ to 2n-2σ and rechecking the admissible ranges would resolve the objection.","tokens_in":32877,"tokens_out":38045,"duration_ms":339088,"concrete_test":"1) For n=3, α=1/2, σ=9/4, compute k_t^σ(0) from (13) and use dominated convergence to show lim_{t→0} k_t^σ(0)>0; this directly contradicts the asserted |t|^{-7.5} bound at q=∞. 2) Re-derive the interpolation step in §4.2 using the L^1→L^∞ estimate |k_t^σ| ≲ |t|^{-(n-σ)/α} from Theorem 3.5(ii) and the L^2→L^2 bound (51); verify that the resulting dispersive exponent is (1/2-1/q)(2n-2σ)/α. Then check that every use in Theorem 4.9, especially display (57), is consistent with this corrected exponent and not with the displayed (2n-σ)/α.","verdict_should_be":"REJECT","load_bearing_attack":"The problem is in the small-time estimates for 0<α<1. In Theorem 3.5(ii), the kernel k_t^σ of (-Δ)^{-σ/2} e^{it(-Δ)^{α/2}} satisfies |k_t^σ(r)| ≲ |t|^{-(n-σ)/α} in the relevant small-time small-scale regime. For a radial convolution operator, ||T||_{L^1→L^∞} = sup_r |k_t^σ(r)|, so the kernel estimate can yield at best decay |t|^{-(n-σ)/α} for the L^1→L^∞ norm. Theorem 4.7(i), however, asserts ||(-Δ)^{-(1/2-1/q)σ} e^{it(-Δ)^{α/2}}||_{L^{q'}→L^q} ≲ |t|^{-(1/2-1/q)(2n-σ)/α}; at q=∞ this is |t|^{-(2n-σ)/α}, which is strictly faster than |t|^{-(n-σ)/α}. This is impossible: for example, take n=3, α=1/2, σ=9/4. Then the integral defining k_t^σ(0) in (13) is absolutely convergent uniformly in t (the amplitude behaves like |λ|^{-1/4} at infinity), and by dominated convergence k_t^σ(0) tends to a positive constant as t→0. Hence the L^1→L^∞ norm cannot decay like |t|^{-7.5}. The correct exponent obtained by interpolating between (51) and the L^1→L^∞ bound is (1/2-1/q)·2(n-σ)/α = (1/2-1/q)(2n-2σ)/α. Internal confirmation: the Strichartz proof in display (57) uses |t-s|^{-(1/2-1/q)β n/α}, which is exactly the corrected formula when σ=(1-β/2)n, not the displayed (2n-σ)/α formula. Thus the small-time dispersive statement as written is false; the paper likely intends 2n-2σ, but the current statement is a load-bearing error.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies dispersive and Strichartz estimates for the fractional Schrödinger equation with Laplacian power α/2 on real hyperbolic spaces H^n and homogeneous trees T_Q. On H^n the authors derive detailed pointwise bounds for the radial kernel of (-Δ)^{-σ/2} e^{it(-Δ)^{α/2}}, splitting the analysis into the regimes 1<α<2 and 0<α<1, and they use these bounds to deduce L^{q'}→L^q dispersive estimates and Strichartz inequalities. On T_Q they exploit the compactness of the spectrum and a stationary-phase analysis to obtain a uniform (1+|t|)^{-3/2} decay with no derivative loss for all 0<α≤2. The paper closes with a short application to nonlinear Schrödinger well-posedness on homogeneous trees.","tokens_in":33327,"tokens_out":31377,"duration_ms":289175,"significance":"The paper targets a natural and previously incomplete problem: fractional-order dispersive estimates on negatively curved spaces. The kernel analysis for r>0 is detailed and uses standard tools (Harish-Chandra expansions, Stanton–Tomas asymptotics, van der Corput estimates), and the homogeneous-tree result is clean and striking, since it shows that the Euclidean Knapp-type derivative loss disappears in the discrete setting. If the statements for 0<α<1 can be corrected, the paper would be a useful contribution to the dispersive PDE literature. The explicit case-by-case kernel bounds are a strength, as they make the main technical work checkable.","major_comments":[{"comment":"The claimed pointwise bound at r=0 is not valid for the stated range. In (13), for r=0 and large λ, the integrand behaves like |λ|^{n-1-σ} e^{it(λ²+ρ²)^{α/2}}. Setting u=(λ²+ρ²)^{α/2}, the tail is comparable to ∫^{∞} u^{(n-σ)/α-1} e^{itu} du, which fails to converge as an improper integral when (n-σ)/α ≥ 1. This occurs inside the stated range n/2≤σ≤n, e.g. n=3, α=1/2, σ=9/4 gives (n-σ)/α = 3/2 and the tail integral ∫ u^{1/2} e^{iu} du diverges. Hence |k_t^σ(0)| is not finite in general, and the proof's bound |k_t^σ(0)| ≲ |t|^{-(n-σ)/α} cannot be correct. The dispersive theorem relying on this bound is therefore not justified at q=∞; the r=0 case should either be removed from the kernel theorem or handled as a distribution with a different argument.","section":"§3, Theorem 3.5(ii), Subcase 2.2.2"},{"comment":"The small-time exponent is internally inconsistent with the proof structure. Display (57) in the proof of Theorem 4.9 states the decay |t-s|^{-(1/2-1/q) β n/α} for the operator with σ(β,q)=(1/2-1/q)(1-β/2)n. This follows from a dispersive estimate of the form |t|^{-(1/2-1/q)·2(n-σ)/α} with σ=(1-β/2)n, not from the displayed exponent (1/2-1/q)(2n-σ)/α in Eq. (52). At q=∞, Eq. (52) would require an L^1→L^∞ bound with decay |t|^{-(2n-σ)/α}, which is already incompatible with the pointwise kernel bound |t|^{-(n-σ)/α} and is moreover unsupported by the r=0 divergence described in the previous comment. The small-time dispersive statements for 0<α<1 need to be corrected, and the q=∞ case must be restricted or removed unless a genuinely different argument is supplied.","section":"§4.2, Theorem 4.7(i), Eq. (52); Theorem 1.1"},{"comment":"The Strichartz theorems are asserted with the remark that proofs are omitted because the argument is 'standard'. This is not fully satisfactory here, because the admissible regions contain nonstandard features: the excluded endpoint (1/p,1/q,β)=(1/2,0,2n/α) in (55), the exceptional pairs in (59), and the reduction of the β-interval to [0,b_α] in Remark 4.10. After the dispersive exponent is corrected, the derivation of the lower bound in (56) from Young's inequality must be redone, and the Keel–Tao endpoint treatment should be verified against this corrected exponent. Please provide the TT* argument in enough detail to check these endpoint exclusions, or state precisely which parts of the cited arguments in [APV11, Thm. 6.3] and [AP14, Thm. 5.2] apply verbatim.","section":"§4, Theorem 4.5 and Theorem 4.9"}],"minor_comments":[{"comment":"The displayed identity has the wrong exponent: differentiating the right-hand side gives a factor (λ²+ρ²)^{α-2}, so the identity is false unless α=2. The intended identity is e^{it(λ²+ρ²)^{α/2}} = -i/(α t λ(λ²+ρ²)^{α/2-1}) ∂_λ e^{it(λ²+ρ²)^{α/2}}. The later amplitude formulas suggest that the authors used the correct identity in the body of the proof, but the displayed formula should be corrected.","section":"§3.1, Eq. (26)"},{"comment":"In the sentence 'As first conclusion, we obtain |k_t^σ(0)| ≲ t^{-(n-σ)/α} when r=0 and t≥1', the condition 't≥1' is inconsistent with the subcase assumption 0<t<1; it should read 't<1'.","section":"§3.1.2, Subcase 2.2.2"},{"comment":"The displayed admissible set for the tree Strichartz estimates is typeset in a confusing way; the intended set should be stated cleanly, including the treatment of the endpoints (1/p,1/q)=(0,0) and (0,1/2) if they are included.","section":"§5, Theorem 5.6"}],"recommendation":"major_revision","confidential_remarks":"I recommend major revision rather than rejection because the homogeneous-tree results and the hyperbolic-space results for 1<α<2 appear to be based on a substantial and largely sound kernel analysis; the main problems concern the 0<α<1 small-time statements, where the pointwise kernel bound at r=0 is unjustified and the exponent in the dispersive theorem is inconsistent with the Strichartz proof. I found no citation or attribution concerns. The authors should be asked to correct the displayed statements and to provide enough detail on the Strichartz endpoint arguments to verify the final regions."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: despite the stress-test note, I don't think there's a load-bearing exponent error here. The notation in Theorem 1.1/4.7 is easy to misread: \"2 n−σ α\" means 2(n−σ)/α, not (2n−σ)/α. The proof of Theorem 4.9 confirms this: with σ=(1−β/2)n, the formula 2(n−σ)/α gives β n/α, which is exactly the exponent used in display (57). The stress-test also compares against the wrong smoothing: at q=∞ the operator in Theorem 4.7 has smoothing σ/2, so the relevant kernel is k_t^{σ/2}, not k_t^σ, and the kernel bound then gives exactly the stated decay. So the main dispersive statements are internally consistent.\n\nWhat the paper actually does: it completes the range 0<α<2 for fractional Schrödinger on H^n and on homogeneous trees, with the hard new part being 0<α<1 where the phase has two stationary points and a third-order degeneracy. The kernel estimates (Theorem 3.5) are proved in real detail, with careful subcase analysis using van der Corput and Harish-Chandra/Stanton-Tomas expansions. The tree result—no derivative loss for all 0<α≤2—is a clean contrast to the Euclidean Knapp loss. The paper is honest about what it delegates: Strichartz proofs are \"generally omitted\" and tree Strichartz is imported from prior work.\n\nSoft spots are real but proportionate. The omitted Strichartz arguments are the main one. For the endpoint cases in R_α and the interval [0,b_α] in Theorem 4.9, a referee needs to see that the standard TT*+Christ-Kiselev/Keel-Tao argument actually covers the stated admissible regions as written, especially the endpoint exclusions. The claimed extension to all rank-one symmetric spaces and Damek-Ricci spaces is asserted in two sentences; that's fine for a remark, not for a theorem statement. And the notation \"2 n−σ α\" should be cleaned up to avoid exactly the misreading the stress-test made.\n\nBottom line: this is a careful, substantive contribution to a mature program. It deserves peer review, and the main fixable issues are presentation and filling in standard arguments. I'd send it out.","headline":"The claimed exponent bug in the 0<α<1 regime doesn't survive a careful reading — the notation means 2(n−σ)/α, not (2n−σ)/α — and the paper is a solid, detailed completion of the fractional Schrödinger range on hyperbolic spaces and trees.","tokens_in":33902,"tokens_out":13824,"would_cite":true,"duration_ms":108593,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["35R11","22E30","35B45","35Q41","35R05","43A85","43A90"],"pacs":[],"model":"deepseek-v4-flash","headline":"Trees erase derivative loss in fractional Schrödinger estimates","keywords":["fractional Laplacian","Schrödinger equation","dispersive estimates","Strichartz estimates","hyperbolic spaces","homogeneous trees","Knapp phenomenon","oscillatory integrals"],"falsifier":"Evaluate the oscillatory integral (62) numerically on $T_Q$ for a small value of $\\alpha$, say $\\alpha=0.1$, over a grid of $(t,r)$ pairs, and check whether any pair violates $|k_t(r)| \\le C(1+|t|)^{-3/2}Q^{-r/2}$; a violation would falsify Theorem 5.3 and the derivative-free dispersive claim of Corollary 5.5.","tokens_in":32671,"feed_emoji":"🌳","tokens_out":9227,"duration_ms":180379,"temperature":0.7,"pith_summary":"The paper proves dispersive and Strichartz estimates for the fractional Schrödinger equation $i\\partial_t u + (-\\Delta)^{\\alpha/2}u = F$ on real hyperbolic spaces $\\mathbb{H}^n$ and homogeneous trees $T_Q$, for $0<\\alpha<2$. On $\\mathbb{H}^n$ the estimates require a smoothing loss $\\sigma$ before measuring the solution, mirroring the Knapp derivative loss familiar on Euclidean space. On $T_Q$ the local-in-time analysis becomes trivial, so the loss disappears: the propagator maps $\\ell^{q'}$ to $\\ell^{\\tilde q}$ with decay $(1+|t|)^{-3/2}$ for every $\\alpha\\in(0,2]$, with no fractional Laplacian inserted. These bounds matter because they give the sharp decay mechanism of the fractional Schrödinger flow on curved and discrete geometries, and they open the way to nonlinear well-posedness results without the smoothing penalty that complicates the continuous case.","feed_headline":"No derivative loss on trees for fractional Schrödinger equations","feed_subtitle":"The tree propagator decays like (1+|t|)^(-3/2) with no smoothing, unlike hyperbolic and Euclidean spaces.","key_machinery":"The load-bearing object is the radial convolution kernel $k_t^\\sigma(r) = \\int e^{it(\\lambda^2+\\rho^2)^{\\alpha/2}}\\varphi_\\lambda(r)\\,|c(\\lambda)|^{-2}(\\lambda^2+\\rho^2)^{-\\sigma/2}\\,d\\lambda$, written as an oscillatory integral against the spherical function and the Plancherel density. The machine that drives the argument is the phase $\\psi_{r/t}(\\lambda)=(\\lambda^2+\\rho^2)^{\\alpha/2}-(r/t)\\lambda$: for $1<\\alpha<2$ it has one stationary point and strictly convex phase, while for $0<\\alpha<1$ the function $\\theta(\\lambda)=\\alpha\\lambda(\\lambda^2+\\rho^2)^{\\alpha/2-1}$ is unimodal, forcing two stationary points, a degeneracy at $\\lambda_0=\\rho\\sqrt{1-\\alpha}$, and a third-order Van der Corput analysis. The proof combines the Harish-Chandra large-scale expansion of the spherical function, the Stanton-Tomas small-scale expansion, and repeated integration by parts in the different $r/t$ regimes, yielding Theorem 3.5; dispersive estimates then follow by interpolation and the $TT^*$ argument, with the Kunze-Stein phenomenon converting radial kernel bounds into $L^{q'}\\to L^q$ bounds.","core_discovery":"The central claim is a complete set of pointwise bounds for the kernel $k_t^\\sigma(r)$ of $(-\\Delta)^{-\\sigma/2}e^{it(-\\Delta)^{\\alpha/2}}$, obtained by stationary-phase analysis of the oscillatory integral whose phase is $\\psi(\\lambda) = (\\lambda^2+\\rho^2)^{\\alpha/2} - (r/t)\\lambda$. The bounds split into large-scale ($r\\ge 1$) and small-scale ($r\\le 1$) regimes, with a further split between $1<\\alpha<2$, where the phase has a single stationary point, and $0<\\alpha<1$, where the phase derivative is unimodal and produces two stationary points plus a degenerate third-order point at $\\lambda_0=\\rho\\sqrt{1-\\alpha}$. From these kernel bounds the paper derives $L^{q'}\\to L^q$ dispersive estimates: on $\\mathbb{H}^n$ small-time decay $|t|^{-(1/2-1/q)m}$ with $m=\\max\\{2(n-\\sigma)/\\alpha, (n-2\\sigma)/(\\alpha-1)\\}$ and large-time decay $|t|^{-3/2}$, and on $T_Q$ the uniform decay $(1+|t|)^{-3/2}$ without any smoothing factor. The tree statement is the distinctive conclusion: whereas Euclidean and hyperbolic spaces suffer a Knapp-type derivative loss, homogeneous trees do not.","pith_inferences":["The mechanism suggests that the Knapp derivative loss is tied to small-scale continuous geometry: any graph whose small-time evolution is trivial in this sense might admit derivative-free dispersive bounds of the same form, so the tree result may extend to more general hyperbolic graphs.","On $\\mathbb{H}^n$, the paper leaves open whether restricting initial data to be radial in the geodesic distance removes the loss for all $\\alpha$; testing this would require re-running the kernel analysis with radial data, which the present argument does not do.","Sending $\\alpha\\to 1$ in the tree estimates should recover the half-wave decay on $T_Q$ as a limiting case, while on $\\mathbb{H}^n$ the excluded $\\alpha=1$ case would need a separate phase analysis since the convexity dichotomy collapses.","Numerically probing the kernel near the degenerate point $\\lambda_0$ for $0<\\alpha<1$ could indicate whether the $r^{-1/3}e^{-\\rho r}$ term arising from the third-order stationary point is optimal or whether cancellations improve it."],"forward_implications":["On $\\mathbb{H}^n$, the dispersive estimates yield Strichartz inequalities whose admissible region $R_\\alpha$ is strictly larger than the Euclidean one, for both $1<\\alpha<2$ (Theorem 4.5) and $0<\\alpha<1$ (Theorem 4.9).","On $T_Q$, the Strichartz estimates hold for every admissible pair in the full square $[0,1/2]^2$ and every $0<\\alpha\\le 2$, with no smoothing parameter.","The spherical-function analysis transfers to all rank-one symmetric spaces and Damek-Ricci spaces, as the authors note that all formulas extend to those settings.","For the fractional nonlinear Schrödinger equation on $T_Q$ with a power nonlinearity, the paper states local well-posedness for arbitrary $\\ell^2$ data and global well-posedness for small data, following the template of the $\\alpha=2$ case.","On trees the half-wave case $\\alpha=1$ needs no separate treatment, in contrast to the continuous setting where it is excluded."],"supporting_citations":[{"why":"Shows that fractional Schrödinger Strichartz estimates on Euclidean space lose derivatives by the Knapp phenomenon, and that radial data remove the loss; this is the baseline the paper compares against.","marker":"[GW14]"},{"why":"Supplies the endpoint $TT^*$ trick that the paper invokes, without proof, to turn dispersive estimates into Strichartz estimates.","marker":"[KT98]"},{"why":"Gives the kernel and Strichartz estimates for $\\alpha=2$ on $\\mathbb{H}^n$ that the present work extends to fractional powers.","marker":"[AP09]"},{"why":"Provides the Schrödinger method on Damek-Ricci spaces and the Kunze-Stein lemma used here to pass from radial kernel bounds to $L^{q'}\\to L^q$ dispersive bounds.","marker":"[APV11]"},{"why":"Provides the small-scale asymptotic expansion of spherical functions that underlies the small-spatial-scale kernel analysis.","marker":"[ST78]"},{"why":"The Christ-Kiselev lemma invoked for the non-endpoint Strichartz estimates in the omitted $TT^*$ proofs.","marker":"[CK01]"},{"why":"Establishes the $\\alpha=2$ Schrödinger dispersive, Strichartz, and well-posedness results on homogeneous trees that the tree section extends to every $\\alpha\\in(0,2]$.","marker":"[Edd13a]"},{"why":"Combines the Stanton-Tomas expansion into the small-scale formula (9) used to estimate the kernel near $r=0$.","marker":"[Ion00]"}],"fun_headline_variants":["Trees dodge Knapp loss in fractional Schrödinger decay","Tree kernels decay uniformly: (1+|t|)^(-3/2), no smoothing","Unlike Euclidean and hyperbolic, trees show no Schrödinger loss","Fractional Schrödinger on trees: uniform decay without smoothing"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The Strichartz theorems assume that the standard $TT^*$ argument with the Christ-Kiselev and Keel-Tao endpoint tricks applies verbatim to the stated admissible regions, but that step is stated without proof.","fun_headline_variants_meta":{"raw":{"variants":["Trees dodge Knapp loss in fractional Schrödinger decay","Tree kernels decay uniformly: (1+|t|)^(-3/2), no smoothing","Unlike Euclidean and hyperbolic, trees show no Schrödinger loss","Fractional Schrödinger on trees: uniform decay without smoothing"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.002005,"raw_usage":{"total_tokens":7806,"prompt_tokens":911,"completion_tokens":6895,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":527,"completion_tokens_details":{"reasoning_tokens":6821}},"tokens_in":527,"tokens_out":6895,"duration_ms":45946,"temperature":1.0,"reasoning_tokens":6821,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T05:03:20.167560+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Evaluate the oscillatory integral (62) numerically on $T_Q$ for a small value of $\\alpha$, say $\\alpha=0.1$, over a grid of $(t,r)$ pairs, and check whether any pair violates $|k_t(r)| \\le C(1+|t|)^{-3/2}Q^{-r/2}$; a violation would falsify Theorem 5.3 and the derivative-free dispersive claim of Corollary 5.5.","supporting_citations":[],"review_version":1}