{"id":"2abcc827-9bcf-4a01-a294-ce077ddcb29e","arxiv_id":"1908.04033","paper_version":1,"verdict":"ACCEPT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"The expected number of facets and the typical facet height of the convex hull of n uniform points on S^{d-1} are determined asymptotically in every regime where n and d tend to infinity.","lead":"This paper derives asymptotic formulas for the number and heights of the facets of the convex hull of n random points on a sphere in d dimensions, as n and d both grow. It identifies five growth regimes with different behavior, giving a complete high-dimensional picture of a basic random polytope.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"No significant objection identified: the central claims survive scrutiny, with only a non-central prefactor slip in Lemma 18.","rationale":"The reader's verdict of ACCEPT is consistent with my reading. The paper's central claims are asymptotic formulas for expected facet counts and typical facet heights in all regimes of (n,d). I re-derived the key steps in each regime: the dominated convergence analysis in Lemma 17, the Laplace estimates in Lemmas 18 and 21, the gamma-distribution coupling in Lemmas 26 and 27, and the total-variation argument in Theorem 5. I found no error that changes a stated theorem. The only concrete issue is in Lemma 18's endpoint case r<sqrt(2/pi), where the algebraic prefactor in the Laplace approximation appears off by a factor of (n-d)^2; however, the only theorem that uses that case, Theorem 2, relies on the ratio of integrals, and the exponential difference makes the polynomial prefactor irrelevant for the stated convergence in probability. The external integral representation (5) is the load-bearing input, but it comes from established literature and is consistent with the known fixed-dimension asymptotic constant; for d=3 it yields the exact deterministic count 2n-4, which provides a strong independent check. Thus I would keep the ACCEPT verdict unchanged.","tokens_in":31103,"tokens_out":54052,"duration_ms":454772,"concrete_test":"Evaluate Theorem 13's K_d for d=3 from equation (5) or from the printed formula: if the factor is 2^d, K_3=2, matching the exact deterministic facet count 2n-4 for a simplicial polyhedron on S^2; if the factor is 2d, K_3=1.5, contradicting Euler's formula. This check distinguishes a harmless OCR-level typo from a genuine error in the super-exponential constant.","verdict_should_be":"UNCHANGED","load_bearing_attack":"I found no load-bearing flaw in the central argument. The exact representation (5), cited from [8, 26], is the foundation, and the paper's use of it is consistent with known fixed-dimension limits (e.g., for d=3 the resulting K_d gives the deterministic count 2n-4 for facets of a simplicial polyhedron on S^2). Each regime proof was checked; the dominated-convergence and Laplace steps are standard and, where bounds are loose, the conclusions still follow. The only internal imperfection is algebraic: in Lemma 18, case r<sqrt(2/pi), the endpoint Laplace prefactor should be proportional to 1/(sqrt(d) (n-d) 2^(n-d)) rather than (n-d)/(sqrt(d) 2^(n-d)). This does not affect any central theorem, because Theorem 2's ratio is governed by the exponential order, and I[-1,1] is calculated from the correct r>r* case.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper studies the random polytope P_{n,d} formed as the convex hull of n i.i.d. uniform points on S^{d-1}, with n→∞ and d either fixed or tending to ∞. It identifies several asymptotic regimes for the pair (n,d) — sublinear, linear, subexponential, exponential, and super-exponential — and, in each regime, establishes the limiting distribution of the height of a typical facet, a tight interval for the heights of all facets, and asymptotic formulas for the expected number of facets. The results are derived from the exact integral representation (5) of the expected number of facets with height in an interval, which is cited from earlier work, followed by Laplace's method or dominated convergence. The paper also recovers and extends fixed-dimension results of Buchta–Müller–Tichy and Kabluchko–Thäle–Zaporozhets, and connects the facet-height asymptotics to the Hausdorff distance and to spherical Delaunay triangulations.","tokens_in":31253,"tokens_out":7516,"duration_ms":68874,"significance":"This is a significant contribution to high-dimensional stochastic geometry. The paper provides a systematic, regime-by-regime asymptotic analysis of facet heights and expected facet numbers for spherical random polytopes as d grows, with explicit constants and no free parameters. The derivation is rigorous and self-contained beyond the well-established integral formula (5), and the estimates are standard dominated-convergence and Laplace arguments. The consistency of the results with known fixed-dimension limits and with Gaussian analogues, such as Theorem 12 matching the form of results in [9], gives additional confidence in the conclusions. The paper also demonstrates the versatility of the exact integral representation (5) as a tool for dimension-dependent asymptotics.","major_comments":[],"minor_comments":[{"comment":"In the case r < sqrt(2/π), the endpoint Laplace prefactor in (15) and (16) should be proportional to 1/[2^{n-d} (n-d) sqrt(d) |sqrt(2/π)-r|], not (n-d)/[2^{n-d} sqrt(d) |sqrt(2/π)-r|]; the current expression is too large by a factor (n-d)^2. This error does not affect the proofs of Theorems 2 and 9 because those conclusions depend on the exponential order and on the r > sqrt(2/π) case, but the lemma statement as written is incorrect and should be corrected.","section":"§6.1.1, Lemma 18"},{"comment":"After defining A1 = [0,b_{n,d}] and A2 = [b_{n,d},∞), the sentence \"Therefore, for any [a,b] ⊆ A2\" appears in the middle of the argument for A1; it should read A1, since the stronger inequality (37) is claimed for subsets of A1.","section":"§6.2.4, Proof of Theorem 5"},{"comment":"The phrase \"For k ∈ N\" introduces a symbol k that is never used; the random variable is X_{d-1}, so this phrase should be removed or replaced with a sentence that does not introduce unused notation.","section":"Theorem 5"},{"comment":"The displayed formula for h2 is garbled by the typesetting (\"/radicaltp /radicalvertex /radicalvertex\"); it should be h2 = sqrt(1 - (r2 d/n)^{2(d+1)/(d-1)^2}) as defined later in (28).","section":"Theorem 8, equation (1)"}],"recommendation":"minor_revision","confidential_remarks":"The manuscript is technically sound and the main theorems are not affected by the noted issues. The error in Lemma 18 is local and should be fixed before publication, but it does not undermine the central claims. The reliance on the exact integral representation (5) from [8,26] is reasonable and appropriately cited. I recommend minor revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Bottom line: this paper does something nobody did before—it gives the expected number of facets and the typical facet height for the convex hull of uniform points on the sphere in every regime where both n and d grow. The fixed-dimensional cases were known, and so was the Gaussian analogue; the spherical high-dimensional picture was missing. The regime classification is clean: sublinear, linear, subexponential, exponential, super exponential, each with matching asymptotic formulas.\n\nThe proofs are built on the exact integral representation (5), so there are no free parameters and the asymptotics are deduced, not fitted. The slow regimes use Laplace's method and dominated convergence, done carefully. The fast regimes use a Gamma random variable approximation (Lemma 26) that is nice and uniform. I checked the exponent-level algebra and the constants in the central theorems; they hold up.\n\nSoft spots, in proportion. First, there is a real but minor algebraic slip in Lemma 18, case r < sqrt(2/pi): the endpoint Laplace prefactor should be 1/(sqrt(d)(n-d)2^(n-d)), not (n-d)/(sqrt(d)2^(n-d)). This does not touch any central theorem—Theorem 2's ratio is decided by exponential order, and I[-1,1] comes from the correct interior-maximum case—but it should be fixed. Second, the proof of Theorem 5 is a bit compressed around the \"arguments similar as in the proof of Theorem 13\" step; I would ask the authors to expand that before publication. Third, the paper leans on representation (5) from Kabluchko–Thäle–Zaporozhets and Bonnet et al. That is standard and reasonable input, and I see no circularity.\n\nAudience: stochastic geometry and high-dimensional probability, plus anyone using spherical random configurations for coherence or compressed sensing. It deserves a serious referee. I would bring it to my reading group and would cite it.\n\nRecommendation: send it to peer review. Expect minor revision, not major surgery.","headline":"Complete asymptotic regime classification for spherical random polytopes; solid and citable, with one minor algebraic slip that does not affect the main results.","tokens_in":31770,"tokens_out":4208,"would_cite":true,"duration_ms":40103,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60D05","52A22","60F05"],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper derives explicit asymptotic formulas for the number and heights of the facets of the convex hull of $n$ uniform points on the sphere $S^{d-1}$, across every regime of $n$ and $d$.","keywords":["random polytopes","spherical random points","convex hull","facet heights","expected number of facets","high-dimensional asymptotics","stochastic geometry"],"falsifier":"For an increasing sequence such as $d_j=j^2$ and $n_j=d_j+5j$ (so $(n-d)/\\sqrt{d}\\to 5$ while $n-d\\ll d$), evaluate the integral (5) by numerical quadrature and compare its logarithm with the asymptotic from Theorem 9, $F[-1,1]=\\binom{n}{d}2^{d-n+1}e^{(n-d)^2/(\\pi d)+o(1)}$. A persistent relative mismatch in the exponent would refute the paper's central claim; the same check can be repeated for one sequence in each regime.","tokens_in":30895,"feed_emoji":"📐","tokens_out":15686,"duration_ms":150695,"temperature":0.7,"pith_summary":"This paper works out, in every asymptotic regime of $n$ (number of points) and $d$ (dimension), what the facets of the convex hull of $n$ uniform points on the unit sphere $S^{d-1}$ look like. It gives explicit asymptotic formulas for the expected number of facets and for the distribution of the typical facet height, showing that behavior splits into slow regimes, subexponential, exponential, and super-exponential regimes. If correct, these formulas give a complete classification of facet statistics for spherical random polytopes as $n\\to\\infty$ and $d$ either fixed or growing, and they extend known fixed-dimension results to high dimension. The results come from analyzing one exact integral representation, not from simulation.","feed_headline":"Explicit asymptotics for facets of spherical random polytopes","feed_subtitle":"Expected facet counts and typical heights are pinned down from near-linear to super-exponential growth.","key_machinery":"The engine is the exact integral representation of the expected number of facets with height in $[h_1,h_2]$, namely $F[h_1,h_2]=\\binom{n}{d}2c_{(d^2-2d-1)/2}\\int_{h_1}^{h_2}(1-h^2)^{(d^2-2d-1)/2}\\left(c_{(d-3)/2}\\int_{-1}^{h}(1-s^2)^{(d-3)/2}ds\\right)^{n-d}dh$. The inner integral is a distribution function that becomes a rescaled normal CDF as $d$ grows, so the integrand is approximated by $\\exp(d f_\\rho(r))$ with $f_\\rho(r)=\\rho\\ln\\Phi(r)-r^2/2$. The standard asymptotic method for integrals with a single dominating peak then turns the integrals into explicit asymptotics. A Gamma substitution and tail bounds isolate the typical height in the fast regimes, giving total variation convergence to a Gamma variable when $\\ln n\\gg d\\ln d$.","core_discovery":"The paper's central claim is that the expected number of facets $F[-1,1]$ and the typical facet height $H_{\\mathrm{typ}}$ are governed by a handful of explicit formulas depending on how $\\ln n$ compares with $d$. In the slow sub-linear regime $n-d\\ll d$, $F[-1,1]=\\binom{n}{d}2^{d-n+1}e^{(n-d)^2/(\\pi d)+o(1)}$. In the exponential regime $(\\ln n)/d\\to\\rho$, $F[-1,1]=[2\\pi(e^{2\\rho}-1)d(1+o(1))]^{(d-1)/2}$, and the typical height converges in probability to $\\sqrt{1-e^{-2\\rho}}$. In the super-exponential regime $\\ln n\\gg d$, the expected number grows as $n K_d h_*^{d-1}$ with $h_*=\\sqrt{1-d^{3/(d-1)}n^{-2/(d-1)}}$. The same analysis identifies nearly deterministic bounds $[h_1,h_2]$ that contain all facet heights with probability tending to one.","pith_inferences":["The paper gives first-order asymptotics only; the same integral representation could be expanded to yield finite-$n$ corrections and rates of convergence in each regime.","Because the starting integral formula exists for the broader class of beta polytopes, the regime classification and the main formulas may extend with modified constants to beta-distributed points, a testable reading of the cited formulas.","The agreement between these regimes and the known angle-extremum regimes for $n$ random spherical vectors suggests facet heights and angular coherence are two views of the same high-dimensional geometry; a sharper finite-$n$ correspondence between minimum angle and maximum facet height could be checked directly from the hull of $n$ points.","The threshold near $\\rho\\approx3.4$ for negative-height facets in expectation suggests a large-deviation regime for the origin being outside the hull that is not captured by the probability threshold alone."],"forward_implications":["In the exponential regime $(\\ln n)/d\\to\\rho$, the expected facet count grows as $[2\\pi(e^{2\\rho}-1)d]^{(d-1)/2}$, so the polytope has super-exponentially many facets in $d$ while typical facet height stays bounded away from $0$ and $1$.","In the linear regime $n-d=\\rho d+o(d)$, the classical threshold puts the origin inside the hull with probability going to one for $\\rho>1$, but the expected number of negative-height facets still diverges for $\\rho$ up to about $3.4$, an intermediate regime the paper identifies.","The Hausdorff distance from the polytope to the ball converges in probability to $1-\\sqrt{1-e^{-2\\rho}}$ in the exponential regime, to $0$ in the super-exponential regime, and to $1$ when $\\ln n\\ll d$.","When $\\ln n\\gg d\\ln d$, the rescaled quantity $n\\Gamma(d/2)(2\\sqrt{\\pi}\\Gamma((d+1)/2))^{-1}(1-H_{\\mathrm{typ}}^2)^{(d-1)/2}$ converges in total variation to a gamma variable with shape parameter $d-1$, which also recovers the distribution of circumscribed cap radii in spherical Delaunay triangulations.","The same formulas answer the three questions posed: the distribution of the typical facet, a tight range containing all facet heights, and the expected number of facets, in every regime considered."],"supporting_citations":[{"why":"Supplies the exact integral representation (5) of the expected number of facets, the starting point of every proof.","marker":"[26]"},{"why":"Provides the detailed derivation of that integral formula and the facet-probability interpretation used throughout.","marker":"[8]"},{"why":"Gives the high-dimensional Gaussian-polytope facet formulas whose structure the slow and fast regimes are compared with.","marker":"[9]"},{"why":"Establishes the fixed-dimension asymptotic for the expected number of facets that Theorems 9--13 extend.","marker":"[10]"},{"why":"Identifies the same $n,d$ regimes for pairwise angles among random spherical vectors, the geometric counterpart of facet heights.","marker":"[11]"},{"why":"Supplies the threshold for the origin being outside the hull used in Remark 1 to interpret negative-height facets.","marker":"[38]"},{"why":"Connects facet heights to circumscribed cap radii in random Delaunay triangulations, recovered by Theorem 5.","marker":"[17]"}],"fun_headline_variants":["Exact facet counts and heights for spherical random polytopes","Spherical polytope facets: explicit formulas in all regimes","From near-linear to super-exponential: facet growth formulas","Random polytope facet counts and heights: explicit asymptotics","Sharp formulas for facet heights and counts of random polytopes"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"All of the paper's conclusions rest on a single previously derived exact formula for the expected number of facets of the spherical hull, which the paper uses as its starting point and never re-derives.","fun_headline_variants_meta":{"raw":{"variants":["Exact facet counts and heights for spherical random polytopes","Spherical polytope facets: explicit formulas in all regimes","From near-linear to super-exponential: facet growth formulas","Random polytope facet counts and heights: explicit asymptotics","Sharp formulas for facet heights and counts of random polytopes"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000658,"raw_usage":{"total_tokens":2963,"prompt_tokens":851,"completion_tokens":2112,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":467,"completion_tokens_details":{"reasoning_tokens":2027}},"tokens_in":467,"tokens_out":2112,"duration_ms":16141,"temperature":1.0,"reasoning_tokens":2027,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T13:55:40.061761+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"For an increasing sequence such as $d_j=j^2$ and $n_j=d_j+5j$ (so $(n-d)/\\sqrt{d}\\to 5$ while $n-d\\ll d$), evaluate the integral (5) by numerical quadrature and compare its logarithm with the asymptotic from Theorem 9, $F[-1,1]=\\binom{n}{d}2^{d-n+1}e^{(n-d)^2/(\\pi d)+o(1)}$. A persistent relative mismatch in the exponent would refute the paper's central claim; the same check can be repeated for one sequence in each regime.","supporting_citations":[{"cited_title":"Beta polytopes and poisson polyhedra: f-vectors and angles","cited_arxiv_id":null,"evidence_quote":"Supplies the exact integral representation (5) of the expected number of facets, the starting point of every proof."},{"cited_title":"Monotonic- ity of facet numbers of random convex hulls","cited_arxiv_id":null,"evidence_quote":"Provides the detailed derivation of that integral formula and the facet-probability interpretation used throughout."},{"cited_title":"Facets of high-dimensional Gaussian polytopes","cited_arxiv_id":"1808.01431","evidence_quote":"Gives the high-dimensional Gaussian-polytope facet formulas whose structure the slow and fast regimes are compared with."},{"cited_title":"M¨ uller, and Robert F","cited_arxiv_id":null,"evidence_quote":"Establishes the fixed-dimension asymptotic for the expected number of facets that Theorems 9--13 extend."},{"cited_title":"Distribution of angles in random packing on spheres","cited_arxiv_id":null,"evidence_quote":"Identifies the same $n,d$ regimes for pairwise angles among random spherical vectors, the geometric counterpart of facet heights."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Supplies the threshold for the origin being outside the hull used in Remark 1 to interpret negative-height facets."},{"cited_title":"Random inscribed p olytopes have similar radius functions as Poisson-Delaunay mosaics","cited_arxiv_id":null,"evidence_quote":"Connects facet heights to circumscribed cap radii in random Delaunay triangulations, recovered by Theorem 5."}],"review_version":1}