{"id":"907646cd-d513-4a55-9199-b72d3b6ceba7","arxiv_id":"2412.07920","paper_version":2,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"For sub-Laplacians on Metivier groups, L^p spectral multiplier estimates hold at the sharp regularity threshold s > d|1/p - 1/2| for a range of p.","lead":"This mathematics paper proves a sharp bound on how smooth a spectral multiplier must be for its operator to be bounded on L^p spaces, for sub-Laplacians on a broad class of two-step nilpotent Lie groups called Metivier groups. The result extends a previously known theorem from Heisenberg-type groups to a larger family and settles the optimal smoothness threshold in that range.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Theorem 1.1 rests on the unproved truncated restriction-type estimate Theorem 2.1 imported from the companion paper [Nie24b]; if that estimate fails, the central claim is unsupported.","rationale":"The reader's conditional verdict is well calibrated. The paper is carefully structured, and I found no clear internal contradiction in the main non-exceptional argument: the dimension numerology, the weighted Plancherel estimate, the dyadic decomposition, and the summation over ell are coherent, and the reduction via [COSY16] is standard. However, the proof is not self-contained at its core: Theorem 2.1 is cited from the author's companion preprint and is used in the most delicate steps, including the exceptional cases that require the full Cowling-Sikora norm. Without an independent verification of that theorem, the central claim cannot be considered fully established. This matches the reader's weakest-assumption analysis. I therefore keep the verdict unchanged at CONDITIONAL, with the caveat that acceptance should be contingent on confirmation of [Nie24b] Theorem 2.1 and its applicability exactly as stated.","tokens_in":39703,"tokens_out":32026,"duration_ms":305427,"concrete_test":"Inspect [Nie24b] and verify the proof of its Theorem 2.1 in full, checking specifically that the stated hypotheses (arbitrary two-step stratified group, 1 <= p <= min{p_d1, p_d2}, compact A subset of (0,infinity)) yield exactly (2.1) with the Cowling-Sikora norm ||F||_{2^ell,2}, uniformly in ell >= -ell_0 and with constants independent of F and ell. Then re-derive Step (5.b) of Proposition 7.1 for (d1,d2) = (8,6) and (8,7) using only the stated estimate; if the companion paper's proof requires any additional assumption, or if the exponent or the norm on the right-hand side of (2.1) differs, the exceptional-case arguments collapse and the theorem's stated range of p is not established.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing assumption is the externally imported Theorem 2.1 from [Nie24b]. The proof of Theorem 1.1 uses this truncated restriction-type estimate, not merely as a convenience, but as the essential engine in Step (4) and Step (5) of Proposition 7.1 and in the exceptional-case arguments of Section 8. In particular, the estimate\n\n||F(L)chi(2^ell U)||_{p to 2} <= C 2^{-ell d2(1/p - 1/2)} ||F||_2^{1-theta_p} ||F||_{2^ell,2}^{theta_p}\n\nmust hold uniformly in ell >= -ell_0 and with constants independent of F, including the cases (d1,d2) = (8,6) and (8,7), where the full Cowling-Sikora norm ||F||_{2^ell,2} is needed and cannot be replaced by the plain L2 norm. Neither Theorem 2.1 nor the supporting spectral-decomposition statements (Proposition 3.1, Proposition 3.2, Remark 2.2) are proved in this paper, and no machine-checked verification is supplied. If Theorem 2.1 fails, or if it holds only under additional hypotheses that are not checked here, then Theorem 1.1 is unsupported even in the main non-exceptional range. This is a verification-status concern rather than a demonstrated internal contradiction, but it is the single point on which the whole argument depends.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper proves an Lp-spectral multiplier theorem under regularity s > d(1/p - 1/2) for sub-Laplacians on Métivier groups, for 1 < p ≤ p_{d1,d2} with p_{d1,d2} determined by the layer dimensions, together with a corresponding Bochner-Riesz summability result. The proof uses the dyadic reduction of Chen-Ouhabaz-Sikora-Yan, a truncated restriction-type estimate imported from the author's companion paper [Nie24b], a first-layer weighted Plancherel estimate proved in Section 6, and a case analysis of the exceptional dimension pairs (4,3), (8,6), (8,7). A separate second-layer weighted Plancherel estimate is developed in Section 8 for the case (4,3).","tokens_in":39982,"tokens_out":18034,"duration_ms":172242,"significance":"If correct, the result is a substantial advance: it reaches the sharp Euclidean-type regularity threshold for a broad class of two-step stratified groups, matching the lower bound of Martini-Müller-Nicolussi Golo, and it extends the earlier Heisenberg-type result to all Métivier groups except at two exceptional pairs where the range of p is reduced. The paper is clearly written, gives a transparent reduction strategy, and contains a careful dimension-numerology analysis via Radon-Hurwitz numbers. The main caveats are the heavy reliance on the unproved companion result Theorem 2.1 and a convergence gap in the endpoint of the (4,3) case; both are load-bearing for the stated theorem.","major_comments":[{"comment":"The truncated restriction-type estimate is imported from the author's companion paper [Nie24b] and is the essential engine of Step (4) and Step (5) of Proposition 7.1 and of the exceptional-case arguments in Section 8. Neither Theorem 2.1 nor Remark 2.2 nor Propositions 3.1 and 3.2 are proved in this paper, and no independent verification is supplied. If Theorem 2.1 fails, or if it requires hypotheses not checked here, Theorem 1.1 is unsupported even in the non-exceptional range. Please include a proof of the needed cases of Theorem 2.1, or cite a published version with the full proof and state explicitly which hypotheses are being used.","section":"Theorem 2.1; Eqs. (7.11), (7.23), (8.11)"},{"comment":"The endpoint p = 4/3 for (d1,d2) = (4,3) is not proved. In (8.11) the factor (R_ell/R)^{d1-d2-s-tilde q + 1} appears. For (4,3) and p = 4/3 one has q = 4, so the exponent is 1 - s-tilde q + 1 = 2 - 4 s-tilde. Since the admissible range is s-tilde > 1/2, this exponent is strictly negative for every admissible s-tilde; saying it 'gets arbitrarily close to zero' does not make the series over ell convergent, because a negative exponent makes the factor unbounded as ell tends to -infinity. Consequently (8.10) does not follow with a uniform constant, and Proposition 8.1 does not deliver the endpoint p = 4/3 stated in Theorem 1.1 for the (4,3) case. The argument appears to work for p < 4/3, but the endpoint requires an additional argument or a modified estimate.","section":"Section 8.1, Eq. (8.11)"}],"minor_comments":[{"comment":"The proof contains a typo ('Let ς(n) denote be the maximal number'), and the sufficiency direction is delegated to [Kap80] without explicitly noting that a Heisenberg type group is in particular a Métivier group; one sentence would make the equivalence fully transparent.","section":"Proposition 5.1"},{"comment":"The notation '≲_ι' is defined in (7.14) with a bound depending on R^{Cγ}, but the subscript ι is not used in later occurrences; a clearer name such as '≲_γ' would avoid confusion.","section":"Eq. (7.14)"},{"comment":"The special values p8,6 = 17/12 and p8,7 = 14/11 appear only inside the proof; including them in the statement of Proposition 7.1 would improve readability.","section":"Proposition 7.1"},{"comment":"There are several OCR/encoding artifacts in the displayed text (e.g., 'M´ etivier' in the header, non-ASCII characters in words such as 'specifically'); these should be corrected in the final version.","section":"Throughout"}],"recommendation":"major_revision","confidential_remarks":"The two major issues are independent. First, the proof leans essentially on the author's companion paper [Nie24b]; if the editor can confirm that [Nie24b] is accepted and its Theorem 2.1 is publicly available with proof, that concern is reduced but not eliminated, since the present paper should at least state the precise dependencies. Second, the (4,3) endpoint gap is concrete and mathematical; it is not a matter of presentation. If the endpoint can be repaired, the paper would still be of high interest."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The headline is simple: this is a genuine extension of the sharp p-specific spectral multiplier program from Heisenberg type groups to all M\\'etivier groups, and the proof is built around a genuinely new observation—the Radon-Hurwitz numerology says d1 > 3d2/2 for all but three exceptional dimension pairs, which makes a seemingly suboptimal restriction estimate good enough. If the companion paper [Nie24b] is correct, Theorem 1.1 is the sharp result for this class, matching the [MMNG23] lower bound up to the endpoint.\n\nThe paper itself is well written and careful. Section 6's first-layer weighted Plancherel estimate is proved in detail, and Proposition 8.7, which handles the exceptional (4,3) case via the special structure of so(4), is a solid piece of linear algebra. The author is explicit about what is imported and what is new. Credit where due: this is not a routine parameter scan.\n\nNow the soft spot, and it is load-bearing. The truncated restriction type estimate (Theorem 2.1) and the spectral decomposition machinery (Propositions 3.1, 3.2) come from [Nie24b], the author's own companion preprint. The present paper does not prove them, and the whole proof of Proposition 7.1—both the main range and the exceptional cases—depends on Theorem 2.1. The reader's conditional verdict is exactly right. This is not a demonstrated internal error; it is a verification-status problem. A referee would need access to [Nie24b] and enough time to check it. The stress-test note's concern holds up on reading: if Theorem 2.1 fails, or holds only under extra hypotheses, then Theorem 1.1 falls. I did not find any post-hoc fitting or circularity in the present paper, and the author's honesty about the dependence is a point in their favor.\n\nWho is this for? Anyone working on spectral multipliers, Bochner-Riesz, or sub-Riemannian geometry on nilpotent groups. It deserves a serious referee, but the referee report should be conditional on verification of the companion. I would not desk-reject it; I would send it out with a clear request to check [Nie24b]. If I worked in this area, I would cite it once the companion is accepted. I might not bring it to a general reading group until the external estimate is independently verified.","headline":"A genuinely new sharp spectral multiplier theorem for M\\'etivier groups, but the proof leans on an unproved restriction-type estimate from the author's companion paper, so the verdict is conditional rather than unconditional.","tokens_in":40527,"tokens_out":1791,"would_cite":true,"duration_ms":42802,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["42B15","22E25","22E30","43A85"],"pacs":[],"model":"deepseek-v4-flash","headline":"Métivier groups admit Lp spectral multiplier theorems with the sharp Euclidean regularity threshold.","keywords":["Métivier group","sub-Laplacian","spectral multiplier","Bochner–Riesz means","restriction type estimate","weighted Plancherel estimate","two-step stratified Lie group","Radon–Hurwitz number"],"falsifier":"Run the companion restriction-type estimate (Theorem 2.1) on a concrete two-step group, for instance the free two-step Lie group on four generators, with p close to min{p_{d1},p_{d2}} and a multiplier supported at one dyadic scale: a violation of inequality (2.1) would show the engine of the proof does not hold in the claimed generality. A second check is to search, on any Métivier group with dimensions outside {(4,3),(8,6),(8,7)}, for a bounded Borel F with ‖F‖_{$L^{2}$_s,sloc} < ∞ for some s > d(1/p − 1/2) whose Lp operator norm is infinite; such an example would disprove Theorem 1.1.","tokens_in":39483,"feed_emoji":"📐","tokens_out":8108,"duration_ms":68088,"temperature":0.7,"pith_summary":"The paper proves that every Métivier group—a two-step nilpotent Lie group whose Lie bracket is maximally non-degenerate—admits an Lp spectral multiplier theorem of Euclidean type. For a sub-Laplacian L, any bounded Borel function F with slightly more than d(1/p − 1/2) derivatives in a localized Sobolev sense gives a bounded operator F(L) on Lp, for p in a range determined by the layer dimensions. The regularity order is optimal up to a boundary case, matching the general necessary condition for sub-Laplacians on smooth manifolds. The proof exploits a truncated restriction-type estimate that looks suboptimal, but becomes effective after a weighted Plancherel estimate uses the fact that the first layer of a Métivier group is typically much larger than its center, so the topological dimension d rather than the homogeneous dimension controls the regularity.","feed_headline":"Métivier groups get sharp Euclidean-type multiplier bounds","feed_subtitle":"A sub-Laplacian multiplier needs only s > d(1/p − 1/2) regularity, matching the Euclidean threshold.","key_machinery":"The engine is the truncated restriction type estimate of Theorem 2.1, taken from the companion paper [Nie24b]: for a two-step stratified group, ‖F(L)χ(2^ℓ U)‖_{p→2} ≤ C $2^{{−ℓ d2(1/p − 1/2)}}$ ‖F‖_{2^ℓ,2}, with a norm introduced by Cowling and Sikora. This is combined with the first-layer weighted Plancherel estimate of Proposition 6.1, which shows the convolution kernel of F(L)χ(2^ℓ U) satisfies an |x|^α-weighted L2 bound with the multiplier measured only in L2. The two estimates together let the proof treat the convolution kernel as essentially supported on balls of size 2^ℓ × $R^{2}$ rather than R × $R^{2}$, turning the homogeneous dimension Q into the topological dimension d = d1 + d2 in the required Sobolev order. The numerology that makes the sums converge is Proposition 5.1: for Métivier groups, d1 > 3d2/2 except for (d1,d2) ∈ {(4,3),(8,6),(8,7)}, a fact derived from Radon–Hurwitz bounds on vector fields on spheres.","core_discovery":"The main theorem states that if G is a Métivier group with first layer dimension d1 and second layer dimension d2, and p lies between 1 and p_{d1,d2} (the Stein–Tomas exponent of $R^{{d2}}$, with the exceptional values p_{8,6}=17/12 and p_{8,7}=14/11), then every bounded Borel function F with ‖F‖_{$L^{2}$_s,sloc} < ∞ for some s > d(1/p − 1/2) defines a bounded multiplier F(L) on Lp(G), with operator norm controlled by that Sobolev norm. The same condition yields uniform Lp bounds for the Bochner–Riesz means (1−tL)^δ_+ whenever δ > d(1/p − 1/2) − 1/2. The necessary condition of [MMNG23] shows the regularity order cannot be lowered, up to the endpoint. This is the first sharp p-specific spectral multiplier theorem covering the full class of Métivier groups, which strictly contains Heisenberg type groups, and it does so without any rotation-invariance.","pith_inferences":["Editorial inference: the role of Radon–Hurwitz numbers suggests that any two-step group whose first layer is sufficiently large compared to its center may admit the same sharp p-specific multiplier theorem; the bottleneck is likely not the restriction estimate but the availability of weighted Plancherel control on the second layer.","Editorial inference: in the exceptional cases (8,6) and (8,7), the paper's range is smaller than what the Heisenberg-type result gives; Remark 8.5 indicates that a full second-layer weighted Plancherel estimate of the form (8.12) would raise the range to the full Stein–Tomas exponent for those dimensions, a concrete route to improvement.","Editorial inference: the methods of Section 8 could be tested on direct products of Heisenberg type groups, where second-layer weighted Plancherel estimates are known, to see whether the full range 1 ≤ p ≤ 2(d2+1)/(d2+3) is achieved beyond Métivier groups.","Editorial inference: since the truncated restriction estimate is valid for all two-step stratified groups, the paper's real content is the interaction between the Cowling–Sikora norm's dyadic localization and a large first layer; this suggests a general principle that topological dimension can replace homogeneous dimension in the regularity order whenever the bracket structure has enough room in t"],"forward_implications":["Bochner–Riesz means (1−tL)^δ_+ are uniformly bounded on Lp(G) for δ > d(1/p − 1/2) − 1/2 whenever 1 ≤ p ≤ p_{d1,d2}, matching the order of the Bochner–Riesz conjecture in this subelliptic setting.","The regularity threshold s > d(1/p − 1/2) is optimal up to the endpoint, so no sharper Euclidean-type L2-based multiplier theorem can hold in general on Métivier groups.","For s > d/2, the operator F(L) is of weak type (1,1) and bounded on all Lp with 1 < p < ∞, independently of the dimensions d1 and d2.","The theorem reduces to the Heisenberg-type result of [Nie24a] when G is of Heisenberg type, with slightly weaker p-ranges only in the exceptional dimensions (8,6) and (8,7) where the new result still covers all Métivier groups.","The proof gives a concrete route to sharp p-specific multiplier theorems for the whole class of Métivier groups, showing that the earlier Heisenberg-type restriction is not essential."],"supporting_citations":[{"why":"Supplies the truncated restriction type estimate of Theorem 2.1, the black-box engine used in every step of the spectral multiplier proof.","marker":"[Nie24b]"},{"why":"Provides the reduction of spectral multiplier estimates to dyadic-frequency multipliers via Proposition 4.2 and Corollary 4.5.","marker":"[COSY16]"},{"why":"Introduces the Cowling–Sikora norm ‖·‖_{M,2} and the bound (2.2) that connects it to classical Sobolev norms.","marker":"[CS01]"},{"why":"Gives the spectral decomposition of the matrices J_µ and the second-layer weighted Plancherel estimates used in the exceptional-case analysis of Section 8.","marker":"[MM14b]"},{"why":"Proves the general necessary condition s ≥ d|1/p − 1/2| that establishes sharpness of the regularity threshold up to the endpoint.","marker":"[MMNG23]"},{"why":"Provides the bound on the number of linear independent vector fields on spheres, which underlies the Radon–Hurwitz numerology for Métivier group dimensions.","marker":"[Ada62]"},{"why":"Gives the equivalence between d2 < ρ_RH(d1) and the existence of Métivier/Heisenberg-type groups, used in the proof of Proposition 5.1.","marker":"[Kap80]"},{"why":"Establishes the sharp Heisenberg-type spectral multiplier theorem that this paper generalizes and whose methods are adapted here.","marker":"[Nie24a]"}],"fun_headline_variants":["Sharp Lp multiplier theorem for all Métivier groups","Métivier groups: Euclidean multiplier regularity is optimal","Sub-Laplacian multipliers on Métivier groups: sharp condition","First sharp multiplier result for full Métivier class","Métivier groups get optimal multiplier regularity"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The proof's load-bearing premise is that the truncated restriction type estimate of Theorem 2.1, imported as a black box from the companion paper, holds for every two-step stratified Lie group; if that estimate fails, the argument gives no Lp bounds.","fun_headline_variants_meta":{"raw":{"variants":["Sharp Lp multiplier theorem for all Métivier groups","Métivier groups: Euclidean multiplier regularity is optimal","Sub-Laplacian multipliers on Métivier groups: sharp condition","First sharp multiplier result for full Métivier class","Métivier groups get optimal multiplier regularity"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000605,"raw_usage":{"total_tokens":2790,"prompt_tokens":882,"completion_tokens":1908,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":498,"completion_tokens_details":{"reasoning_tokens":1842}},"tokens_in":498,"tokens_out":1908,"duration_ms":13903,"temperature":1.0,"reasoning_tokens":1842,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T18:25:13.086751+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Run the companion restriction-type estimate (Theorem 2.1) on a concrete two-step group, for instance the free two-step Lie group on four generators, with p close to min{p_{d1},p_{d2}} and a multiplier supported at one dyadic scale: a violation of inequality (2.1) would show the engine of the proof does not hold in the claimed generality. A second check is to search, on any Métivier group with dimensions outside {(4,3),(8,6),(8,7)}, for a bounded Borel F with ‖F‖_{$L^{2}$_s,sloc} < ∞ for some s > d(1/p − 1/2) whose Lp operator norm is infinite; such an example would disprove Theorem 1.1.","supporting_citations":[],"review_version":1}