{"id":"ce80981e-3bdc-4b51-b198-780c34190de4","arxiv_id":"1908.00736","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":7.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"Explicit series and Hankel determinant formulas are derived for the distribution of the maximal height of N non-intersecting Bessel paths.","lead":"This paper derives exact formulas for the probability that N non-intersecting Bessel paths, starting at a point and ending at the origin, stay below a height M. The result generalizes known Brownian motion maximum distributions and connects the problem to multiple orthogonal polynomials.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Lemma 2 is false as stated, and the proof of Theorem 1 depends on it; the central derivation is unsound.","rationale":"The paper's strongest claim is the exact closed-form maximum distribution in Theorems 1 and 2. The proof route is: Karlin-McGregor formula (3.3), then asymptotics of q and q_M in Proposition 1, then determinant evaluations. The reader's weakest-assumption analysis focused on the Karlin-McGregor coalescing limit and on interchanging limits with the infinite Bessel-zero series. Those are legitimate rigor gaps, but the more load-bearing defect is internal: Lemma 2 is a false algebraic statement. A concrete N=2 example with f(z)=z^2 gives an identically zero determinant while the claimed leading term is nonzero; the proof of Lemma 2 drops the permutation sum when ordering the exponents k_j. Since Eq. (4.13) is derived from Lemma 2 and is used to obtain the asymptotics of q_M in Proposition 1, the proof of Theorem 1 is unsound as written. This is not a disagreement with consensus or a stylistic issue; it is a demonstrable error in the argument. The final formulas do pass the Brownian special-case checks (α=±1/2) and the N=1 check against Pitman-Yor, so I am not asserting the theorems are false. But the current manuscript does not establish them rigorously, and the reader's identified gaps are not the only obstruction. The verdict should move from CONDITIONAL to UNVERDICTED pending a rewritten proof that replaces Lemma 2 with the correct confluent-determinant expansion (or otherwise justifies the asymptotics of q_M independently).","tokens_in":16394,"tokens_out":22247,"duration_ms":224800,"concrete_test":"The one decisive check is to evaluate Lemma 2 for N=2, a=1, f(z)=z^2, y=(1,2): the exact determinant det_{i,j}[(x_i y_j)^2] is identically 0, whereas the lemma's claimed leading term is 8(x_2−x_1) (nonzero). If this check produces the claimed nonzero term, Lemma 2 is refuted; then recompute Eq. (4.13) with the correct confluent determinant det_{i,j}[ y_j^{i−1} J_α^{(i−1)}(a y_j)/(i−1)! ] and trace whether (2.2) can still be recovered via Cauchy-Binet over the Bessel-zero index n.","verdict_should_be":"UNVERDICTED","load_bearing_attack":"Section 3.2, Lemma 2: for a smooth f, det[f(x_i y_j)] as x→a is claimed to equal (∏_j f^{(j-1)}(a y_{N+1-j})/Γ(j)) Δ(x−a)Δ(y){1+O(|x−a|)}. This is not correct. Counterexample with N=2, a=1, f(z)=z^2: the matrix entries are (x_i y_j)^2, so the determinant is identically zero for every x, while the claimed right-hand side equals a^2 y_2^2 · 2a y_1 · (y_2−y_1)(x_2−x_1), which is nonzero for y=(1,2). The proof's error is the step 'we order the index k=(k_1,...,k_N) such that {k_j} is decreasing'; this drops the sum over permutations assigning derivative orders to columns. The correct leading term is Δ(x−a)·det_{i,j}[ y_j^{i−1} f^{(i−1)}(a y_j)/(i−1)! ] plus higher order, a confluent determinant, not a product. This false lemma is used directly in Eq. (4.13) to evaluate det[J_α(x_{n_j}/M x_i)] in Proposition 1, and Proposition 1 is the input to the proof of Theorem 1 in §4.2. Hence the derivation of the central formulas (2.2) and (2.4) is not valid as written. The final formulas may be repairable by using the correct confluent determinant and a Cauchy-Binet summation, but that repair is absent from the manuscript.","agreement_with_reader":"disagree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper considers an ensemble of N non-intersecting Bessel paths of order α, all starting at a≥0 and conditioned to end at 0 at time 1. The authors state two theorems: Theorem 1 gives the probability that the maximum of the outermost path stays below M as a determinant involving a sum over zeros of the Bessel function J_α (Eqs. (2.2) and (2.4)); Theorem 2 rewrites these probabilities as Hankel determinants associated with discrete (multiple) orthogonal polynomials (Eqs. (2.16) and (2.17)). The derivation proceeds via the Karlin-McGregor formula, a spectral expansion for the transition density with an absorbing wall at M, and asymptotic evaluations of the resulting determinants. Special cases α=-1/2 and α=1/2 are shown to reproduce the known non-intersecting Brownian motion formulas, and the N=1 case reproduces a result of Pitman-Yor.","tokens_in":16694,"tokens_out":12170,"duration_ms":100398,"significance":"If the stated formulas are correct, they provide the first explicit finite-N closed forms for the maximum-height distribution of non-intersecting Bessel paths, and the connection to discrete multiple orthogonal polynomials is a useful bridge to Riemann-Hilbert asymptotics. The paper is self-contained in the sense that it introduces no fitted parameters, and the agreement of the α=±1/2 special cases with the known Brownian formulas is a strong nontrivial check. The determinant representations in Theorem 2 are concrete and potentially computable. However, the validity of the proof is the critical issue.","major_comments":[{"comment":"Lemma 2 is false as stated. For N=2, a=1, y=(1,2), and f(z)=z^2, the left-hand side det_{i,j}[f(x_i y_j)] is identically zero for all x, while the right-hand side equals 8(x_2-x_1), which is nonzero. The flaw is in the proof: after applying Lemma 1, the step 'we order the index k=(k_1,...,k_N) such that {k_j} is decreasing' discards the summation over permutations that assign derivative orders to columns; for generic f the correct leading term is Δ(x-a) det_{i,j}[ y_j^{i-1} f^{(i-1)}(a y_j)/(i-1)! ] plus higher-order terms, not a product. Since this lemma is used directly in Eq. (4.13) of Proposition 1 to evaluate det[J_α(x_{n_j}/M x_i)], and Proposition 1 is the input to the proof of Theorem 1 in §4.2, the derivation of the central formulas (2.2) and (2.4) is not valid as written. The final formulas may be repairable, but the proof must be reworked.","section":"Section 3.2, Lemma 2 (Eq. (3.17))"},{"comment":"The asymptotic expansions (4.3)-(4.7) are obtained by taking the limits x→a and y→0 inside the infinite sums over Bessel zeros and inside the Taylor-series representation of the modified Bessel function I_α, without a uniform-convergence or dominated-convergence justification. For example, in the derivation of (4.4), the series in (4.2) over n∈N^N is interchanged with the limit x→a and y→0; the terms contain oscillatory factors J_α(x_{n,α}/M ·) and the exponential e^{-x_{n,α}^2/(2M^2)}, and near y=0 the behavior J_α(x_{n,α} y/M) ∼ (x_{n,α} y/(2M))^α/Γ(α+1) makes the convergence non-uniform in y for α<0. This gap is load-bearing because Proposition 1 provides the main asymptotic estimates that feed Theorem 1.","section":"Section 4.1, Proposition 1 (Eqs. (4.3)–(4.7))"},{"comment":"The Karlin-McGregor formula is invoked for the coalescing limit x→a, y→0, but the standard form of the formula requires distinct starting and ending points. The authors cite [13] for this limiting procedure, but no proof is given that the ratio of determinants converges to the non-intersecting probability in the presence of the singular boundary at 0. This is a gap in the derivation of the central identity (3.3); although the agreement of the final formulas with the known Brownian special cases is reassuring, it does not by itself supply the missing justification.","section":"Section 3.1, Eq. (3.3)"}],"minor_comments":[{"comment":"The set N0 = {0,1,2,...} is used in the proof but is not defined before the lemma; please define it in the statement or in a preliminary remark.","section":"Section 3.2, proof of Lemma 2"},{"comment":"The row-reduction argument leading to (4.26) is described only in words; in particular, the parity properties of the polynomials P, Q, P~, Q~ in Lemma 3 are asserted but not explicitly used to justify the elimination of the J_α terms at each step. A more explicit induction would improve readability.","section":"Section 4.3, proof of Theorem 2"}],"recommendation":"major_revision","confidential_remarks":"The false Lemma 2 is a serious defect that must be fixed before the paper can be accepted. Given that the special cases α=±1/2 reproduce known results, I suspect the final formulas are correct, but the current proof is not. I would recommend asking the authors to supply a corrected proof of the asymptotic determinant evaluations, ideally with a rigorous treatment of the limiting interchanges. The paper's connection to multiple orthogonal polynomials is potentially valuable."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"This paper has a genuine new result: an exact distribution for the maximal height of N non-intersecting Bessel paths, for general alpha and a>0. The checks against the known Brownian limits (alpha = ±1/2) and the N=1 Pitman–Yor result are reassuring, and the connection to multiple orthogonal polynomials is a nice touch. The problem is that the central proof is unsound. Lemma 2 in Section 3.2 is false as stated. For N=2, a=1, f(z)=z^2, the determinant det[f(x_i y_j)] is identically zero for all x and y, while the claimed asymptotic right-hand side is 2y_1 y_2^2 (x_2−x_1)(y_2−y_1), which is nonzero for y=(1,2). The error is in the step where the authors ‘order the index’ and drop the permutation sum; the correct leading term is a confluent determinant, Δ(x−a)·det_{i,j}[y_j^{i−1} f^{(i−1)}(a y_j)/(i−1)!], not a product. This is not a peripheral issue. Equation (4.13) uses Lemma 2 directly to evaluate the determinant of Bessel functions appearing in q_M, and that feeds Proposition 1, which is the input to Theorem 1. The derivation of (2.2) and (2.4) therefore does not go through as written. Theorem 2 inherits the problem because it relies on Theorem 1. The final formulas might well be correct—the special-case checks are strong evidence—and the overall strategy (Karlin–McGregor, spectral expansion, then a Cauchy–Binet summation) is standard enough that a repair is plausible. But the repair is not in the manuscript. There is also a secondary unaddressed point about the coalescing limit in (3.3) and the interchange of limits with the infinite Bessel-zero sums; that is a lesser concern because the same pattern appears in the Brownian cases, but it should be mentioned. I would not reject the paper out of hand. Send it to a competent referee; the referee will likely catch the same flaw, and the authors need to fix Lemma 2 and redo Proposition 1. As it stands, this is not acceptable for publication.","headline":"The formulas are likely new and may even be true, but the proof of the main theorem rests on a false determinant lemma, so the paper needs major revision before it is publishable.","tokens_in":17244,"tokens_out":7735,"would_cite":false,"duration_ms":70284,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":["60J65","33C47"],"pacs":[],"model":"deepseek-v4-flash","headline":"For N non-intersecting Bessel paths, the probability that the outermost path stays below a level M is given exactly by a determinant series over the zeros of the Bessel function J_α.","keywords":["non-intersecting Bessel paths","maximum distribution","Hankel determinant","multiple orthogonal polynomials","Bessel function zeros","Karlin-McGregor formula","Bessel process","Brownian motion"],"falsifier":"A decisive check is to simulate two independent Bessel processes with α = 1 started at the same small a, conditioned to end at 0 and not to collide, and compare the empirical frequency of max < M with formula (2.2); a mismatch beyond sampling error would refute the formula, as would any numerical disagreement between the α = 1/2 reduction of (2.4) and the classical Brownian-excursion formula (1.2).","tokens_in":16175,"feed_emoji":"📐","tokens_out":9557,"duration_ms":90437,"temperature":0.7,"pith_summary":"This paper derives exact closed-form expressions for the probability that the outermost of N non-intersecting Bessel paths stays below a level M over the whole time interval [0,1], with all paths starting at a ≥ 0 and conditioned to end at 0. The answer is a constant times the determinant of an N×N matrix whose entries are infinite series over the positive zeros of the Bessel function J_α; when a = 0 the expression is a similar series with powers of the same zeros. The same probabilities are recast as Hankel determinants of moments of multiple discrete orthogonal polynomials for a > 0, and of discrete orthogonal polynomials for a = 0. Because Bessel paths are the radial distances of multidimensional Brownian motion, these formulas extend the known maximum-height results for non-intersecting Brownian excursions and reflected Brownian motions, and they give a direct starting point for asymptotics as N grows.","feed_headline":"Exact formula gives max height of non-intersecting Bessel paths","feed_subtitle":"For any order α and any N, the probability is a determinant over Bessel zeros; Brownian limits are recovered.","key_machinery":"The engine of the proof is the Karlin-McGregor ratio P = lim_{x→a, y→0} q_M(x,y)/q(x,y), where q and q_M are determinants of one-particle transition densities and q_M is built from the eigenfunction expansion of the Bessel generator on (0,M) with an absorbing wall at M; the eigenfunctions are Bessel functions evaluated at the zeros x_{n,α}. The asymptotic evaluation of the ratio when all starting points coalesce to a and all ending points to 0 is carried out with a determinant identity (Lemma 1) and a Schur-function expansion (Lemma 2) that extract the leading powers of Δ(x−a)Δ(y²) or Δ(x²)Δ(y²). The conversion to Hankel determinants is done by row operations using Bessel recurrence relations (Lemma 3) that replace the higher derivatives $J_α^{{(i−1)}}$(a x/M) with combinations of J_α and J_{α+1}, leaving the moment matrices of the discrete weights w_1, w_2, and w̃.","core_discovery":"On its own terms, the paper establishes Theorem 1: for α > −1, N ≥ 1, M ≥ a, and a > 0, the probability that max_{0<t<1} b_N(t) < M equals c_N(α) $M^{{−N(3N+2α+1)/2}}$ times the determinant of the matrix whose (i,j) entry is ∑_{n=1}^∞ (−1)^{i−1} x_{n,α}^{i+2j+α−3} $J_α^{{(i−1)}}$(a x_{n,α}/M) $e^{{−x_{n,α}}$^2/($2M^{2}$)} / J_{α+1}^2(x_{n,α}), where x_{n,α} are the positive zeros of J_α and c_N(α) is an explicit constant. For a = 0, the formula is the analogous determinant (2.4) with powers x_{n,α}^{2i+2j+2α−4}. Theorem 2 rewrites both probabilities as Hankel determinants of moments of the discrete weights w_1, w_2, and w̃, so the distribution is expressed through multiple discrete orthogonal polynomials or discrete orthogonal polynomials. The paper also shows that the α = ±1/2 cases reproduce the known non-intersecting Brownian wall formulas, and that the single-path case N = 1, a = 0 reproduces Pitman and Yor's law of the Bessel bridge maximum.","pith_inferences":["Editorial inference: the same q_M/q determinant-ratio method should extend to joint survival probabilities for several levels, because the spectral expansion in Bessel zeros is the only model-dependent input and the determinant identities do not use the specific level M.","Editorial inference: since squared Bessel processes are squared norms of multidimensional Brownian paths, the formula should translate into a statement about the radial components of N non-colliding d-dimensional Brownian particles confined to a ball of radius M, possibly connecting to Wishart-type eigenvalue statistics.","Editorial inference: a natural testable extension is to let both N and α scale together; the exact finite-N series could be compared against large-deviation asymptotics before the conjectured GOE limit sets in.","Editorial inference: the two starting-point cases a > 0 and a = 0 have different orthogonal-polynomial structures, so the transition as a → 0 may reveal a discrete-weight phenomenon not visible in the Brownian limits."],"forward_implications":["For every finite N and every α > −1, the maximum-height distribution is now an explicit series, so probabilities and their derivatives can be evaluated numerically without simulating the conditioned paths.","In the special cases α = 1/2 and α = −1/2, the formulas reduce to the known maximum distributions for non-intersecting Brownian excursions and reflected Brownian motions, so the Bessel model is an exact interpolation between those two classical wall models.","The Hankel-determinant writing brings the problem into the range of steepest-descent and orthogonal-polynomial asymptotics; the authors expect, under suitable scaling, convergence to the Tracy-Widom distribution of the Gaussian orthogonal ensemble as N → ∞.","The N = 1, a = 0 reduction agrees with Pitman and Yor's known closed form, which anchors the general formula to a single-particle result.","The formulas respect the natural large-M limit, tending to 1 as the ceiling M grows, which is consistent with the probability interpretation of the expression."],"supporting_citations":[{"why":"Supplies the Karlin-McGregor formula in an affine Weyl alcove that is used as equation (3.3), the starting point of the whole derivation.","marker":"[13]"},{"why":"Gives the Bessel transition probabilities (1.9)-(1.10) and the diffusion setup used to define the process.","marker":"[3]"},{"why":"Provides the baseline maximum-height formula for Brownian excursions that the Bessel result generalizes and reproduces at α = 1/2.","marker":"[24]"},{"why":"Supplies the N = 1, a = 0 law of the maximum of a Bessel bridge used in Remark 2 as a consistency check.","marker":"[21]"},{"why":"Shows how such maximal distributions can be written as Hankel determinants of discrete orthogonal polynomials and analyzed by steepest-descent methods, serving as the template for Theorem 2.","marker":"[17]"},{"why":"Derives the Brownian motion maximum distribution by the Karlin-McGregor approach, the proof strategy the paper adapts.","marker":"[14]"},{"why":"Defines the multiple discrete orthogonal polynomials whose moment Hankel determinants appear in Theorem 2 for a > 0.","marker":"[2]"}],"fun_headline_variants":["Exact max-height law for N non-intersecting Bessel paths","Non-intersecting Bessel paths: exact max height formula","Hankel determinant gives max height for non-intersecting Bessel paths","Max height via Hankel determinant for non-intersecting Bessel paths"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The argument assumes that the Karlin-McGregor determinant ratio (3.3) remains valid for Bessel processes with all N paths starting at the same point and ending at the same point, even though the Bessel process has a singular boundary at the origin and the infinite series over Bessel zeros is interchanged with the coalescing limits.","fun_headline_variants_meta":{"raw":{"variants":["Exact max-height law for N non-intersecting Bessel paths","Non-intersecting Bessel paths: exact max height formula","Hankel determinant gives max height for non-intersecting Bessel paths","Max height via Hankel determinant for non-intersecting Bessel paths"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.001749,"raw_usage":{"total_tokens":6902,"prompt_tokens":934,"completion_tokens":5968,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":550,"completion_tokens_details":{"reasoning_tokens":5890}},"tokens_in":550,"tokens_out":5968,"duration_ms":40958,"temperature":1.0,"reasoning_tokens":5890,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T15:34:25.630938+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"A decisive check is to simulate two independent Bessel processes with α = 1 started at the same small a, conditioned to end at 0 and not to collide, and compare the empirical frequency of max < M with formula (2.2); a mismatch beyond sampling error would refute the formula, as would any numerical disagreement between the α = 1/2 reduction of (2.4) and the classical Brownian-excursion formula (1.2).","supporting_citations":[{"cited_title":"Katori and H","cited_arxiv_id":null,"evidence_quote":"Supplies the Karlin-McGregor formula in an affine Weyl alcove that is used as equation (3.3), the starting point of the whole derivation."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Gives the Bessel transition probabilities (1.9)-(1.10) and the diffusion setup used to define the process."},{"cited_title":"Schehr, S","cited_arxiv_id":null,"evidence_quote":"Provides the baseline maximum-height formula for Brownian excursions that the Bessel result generalizes and reproduces at α = 1/2."},{"cited_title":"Pitman and M","cited_arxiv_id":null,"evidence_quote":"Supplies the N = 1, a = 0 law of the maximum of a Bessel bridge used in Remark 2 as a consistency check."},{"cited_title":"Liechty, Nonintersecting Brownian motions on the half-line and discrete Gaussian orthogonal polynomials, J","cited_arxiv_id":null,"evidence_quote":"Shows how such maximal distributions can be written as Hankel determinants of discrete orthogonal polynomials and analyzed by steepest-descent methods, serving as the template for Theorem 2."},{"cited_title":"Kobayashi, M","cited_arxiv_id":null,"evidence_quote":"Derives the Brownian motion maximum distribution by the Karlin-McGregor approach, the proof strategy the paper adapts."},{"cited_title":"Arves´ u, J","cited_arxiv_id":null,"evidence_quote":"Defines the multiple discrete orthogonal polynomials whose moment Hankel determinants appear in Theorem 2 for a > 0."}],"review_version":1}