{"id":"b762595b-11bf-40a0-bf87-993ce235f805","arxiv_id":"2608.19270","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"The authors extend Erdélyi's product formula for 1F1 to multivariate Φ2^(k), Ψ2^(k), and matrix-variate 1F1.","lead":"This note writes down new integral product formulas for multivariate and matrix-variate confluent hypergeometric functions, building on Erdélyi's classical result. A specialist reader gets a cleaner catalog of integral representations that may find use in multivariate statistics.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Propositions 3.3 and 3.4 hinge on the unproved interchange of a hypergeometric series with an integral over the noncompact contour L^k in Eq. (15); without a justification of that interchange, product formulas (14) and (16) are not established as written.","rationale":"The reader's weakest-assumption analysis correctly identifies the contour interchange in Eq. (15) as the most load-bearing unsupported step. My independent termwise check of (14) and (16) shows the formulas are algebraically consistent, but the paper's 'easy to verify' assertion skips the noncompact-contour convergence analysis. A standard domination argument using |t_i| ≥ 1/(k+1) on L^k likely repairs the gap, which is why I do not recommend a harsher verdict than the reader's CONDITIONAL. The other issues—the Prop. 4.1 hypothesis typo and the unstated matrix-argument symmetry convention—are genuine but secondary; they do not threaten the multivariate product formulas in the same way. In good faith, the central advance appears sound, but the proof as printed is incomplete at exactly the point the reader flagged.","tokens_in":7988,"tokens_out":35450,"duration_ms":311004,"concrete_test":"Prove the dominated-convergence bound S(z) = Σ_{m∈Z^k_{\\ge0}} |(h)_m/(g)_<m>| · (|z|/σ)^{<m>}/m! · ∫_{L^k} |t|^{-<m>} |dẽµ(h−1,g+k−<h>)(t)| < ∞ for some polydisc |z_i| < r, with σ = 1/(k+1) the common real part on L. If this estimate holds for all parameters satisfying the proposition's conditions, term-by-term integration in (15) is justified and the reader's concern is resolved; if it fails for a specific allowed (h,g), that parameter set is a counterexample to the proof as written and signals that (14) and (16) need additional hypotheses.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The weakest load-bearing step is the contour representation (15): e^{<z>} = ∫_{L^k} Φ2^(k)[h;g;z/t] dẽµ(h−1,g+k−<h>)(t), together with the analogous Ψ2 representation used in Proposition 3.4. Both are asserted by expanding the hypergeometric series and integrating term by term, but L^k is noncompact, so the interchange is not automatic. What must be shown is an absolutely convergent double series, such as finiteness of Σ_m |(h)_m/(g)_<m>| ∫_{L^k} |z/t|^m/m! |dẽµ(h−1,g+k−<h>)(t)| for z in a neighborhood, or an equivalent locally uniform domination of the integrand. The stated hypotheses ℜ(h−1)>0 and ℜ(g+k)>ℜ(<h>) make the base measure integrable, but the paper supplies no domination argument for the infinite series. Because Propositions 3.3 and 3.4 and the derived Whittaker formula all rely on these contour representations, a failure of the interchange for some allowed parameters would void (14) and (16). This is the least-secure part of the multivariate central claim; the misprinted hypothesis in Proposition 4.1, while real, is an independent and more easily corrected issue.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper revisits Erdélyi's 1939 product formula for the univariate confluent hypergeometric function and proposes two higher-dimensional generalizations. Section 2 gives two proofs of the univariate formula (4). Section 3 derives product formulas for the multivariate functions Φ_2^{(k)} and Ψ_2^{(k)}: formulas (9) and (10) use integrals over simplexes, while formulas (14) and (16) use contour integrals over L^k. A corollary product formula for multivariate Whittaker functions is also given. Section 4 states a matrix-variate analogue, Proposition 4.1, expressing the product of two matrix-variate 1F1's as a threefold matrix-beta integral. The derivations are reductions to standard integral representations of the exponential function, (2), (8), (12), (13), and (21), together with zonal-polynomial identities.","tokens_in":8284,"tokens_out":11618,"duration_ms":102728,"significance":"If the identified gaps are repaired, the paper provides clean, explicit product formulas for two classes of multivariate confluent hypergeometric functions and for the matrix-variate 1F1. The use of Oshima's contour transformations to obtain Φ_2-type and Ψ_2-type product formulas is an interesting and potentially useful technique, and the paper is self-contained up to standard references. The formulas are concrete and falsifiable, and the univariate and matrix-variate cases are checked against known integral representations. The main weakness is that two load-bearing steps are asserted rather than proved: the noncompact contour interchange behind Propositions 3.3 and 3.4, and the well-definedness of the matrix-variate integrand in (19). The Whittaker corollary also has an incorrect parameter condition. These issues are local and repairable, so the central ideas appear sound.","major_comments":[{"comment":"The contour formula e^{⟨z⟩} = ∫_{L^k} Φ_2^{(k)}[h;g;z/t] dẽµ(h-1,g+k-⟨h⟩)(t) is asserted by termwise integration, but L^k is noncompact and no domination argument is supplied. This interchange is load-bearing because formulas (14) and (16) depend on it. The stated hypotheses ℜ(h-1)>0 and ℜ(g+k)>ℜ(⟨h⟩) likely suffice: on L one has |t| ≥ (k+1)^{-1}, and the ratio |(h)_m/(g)_{⟨m⟩}| grows at most polynomially, so an absolutely convergent double-series estimate can be written down. However, the paper does not provide it, and the proof is therefore incomplete as written.","section":"Sec. 3, Eq. (15), Props. 3.3 and 3.4"},{"comment":"The hypothesis states min{ℜ(a),ℜ(b),ℜ(g),ℜ(c-a),ℜ(b-d),ℜ(h-g)} > (m-1)/2. But the measure dµ_{b,d-b} in the integrand requires ℜ(d-b) > (m-1)/2, not ℜ(b-d) > (m-1)/2. With the printed sign, the measure dµ_{b,d-b} is not covered by (18), and Proposition 4.1 is not stated under valid conditions.","section":"Sec. 4, Prop. 4.1"},{"comment":"The integrand 1F1(h/g; (U^{1/2}XU^{1/2}+V^{1/2}YV^{1/2})T) has a matrix argument W T that is not symmetric in general, because the positive definite matrices W and T need not commute. Since 1F1 of matrix argument is defined via zonal polynomials on symmetric matrices, the expression is not well-defined without an additional convention. The proof implicitly uses the identity Cκ(W T)=Cκ(W^{1/2} T W^{1/2}), which follows from the fact that W T and W^{1/2} T W^{1/2} have the same eigenvalues. This convention should be stated explicitly in the manuscript.","section":"Sec. 4, Eq. (19)"},{"comment":"The stated parameter conditions for the Whittaker product formula are incomplete. Applying (16) to the Whittaker representation requires the t-measure dµ(2λ+1, -⟨λ⟩-κ-k/2) to satisfy the condition ℜ(g)>ℜ(⟨h⟩), i.e., ℜ(κ)<-ℜ(⟨λ⟩)-k/2. The printed condition ℜ(κ)<min{0, k/2-ℜ(⟨ν⟩), k/2-ℜ(⟨β⟩)} does not imply this estimate. As written, the corollary is stated for a parameter range that may violate the hypotheses of Proposition 3.4.","section":"Sec. 3, Whittaker corollary"}],"minor_comments":[{"comment":"There are small typos, such as \"Erdélyi' s\" in the abstract and inconsistent spacing in the notation Φ(k)2, Ψ(k)2; these should be cleaned up.","section":"Abstract and Section 1"},{"comment":"The statement that (12) 'can be easily verified by expanding' is acceptable because the integration is over the compact simplex and the series is locally uniformly convergent, but a short justification would improve the exposition.","section":"Sec. 3, after Eq. (12)"},{"comment":"In the definition of the matrix beta measure, the positivity condition is written as min{ℜ(a),ℜ(c-a)} > (m-1)/2, which is correct only up to the usual convention that the real parts are positive; this is fine, but the subsequent use in Proposition 4.1 should match the signs exactly.","section":"Sec. 4, Eq. (18)"},{"comment":"References [10] and [13] are self-citations to unpublished or very recent work; this is acceptable since they are used only for notation and motivation, but the authors might note that the present results do not depend on those papers.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"The core formulas appear correct and the paper is a reasonable short note for math.CA, but the missing contour-interchange proof and the incorrect parameter signs need to be fixed before publication. The issues are local rather than fatal: a domination argument for Eq. (15) can be supplied, and the corrections to Proposition 4.1 and the Whittaker corollary are straightforward. I recommend major revision rather than rejection. The paper's novelty is modest, but the explicit formulas and the use of Oshima's transformations are sufficiently useful for a note in this field."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Dear colleague,\n\nThis is a short, honest special-functions note. The genuinely new content is a handful of product formulas: multivariate Φ2 and Ψ2 analogues of Erdélyi's 1939 integral, and a matrix-variate version. The proofs are mostly routine—they combine known integral representations (Dirichlet, Exton, Oshima's contour measure, zonal polynomial integrals) with the same 'coupling' trick Erdélyi used. That is not a criticism: the statements do not appear in the literature, and the derivations are transparent enough that a specialist can verify the algebra in an afternoon.\n\nThe strongest part is Section 4. The matrix-variate result (19) is derived cleanly from the standard zonal polynomial integral, and the only real issue there is a typo in the hypothesis: it says ℜ(b−d) > (m−1)/2 but the measure dµ_{b,d−b} needs ℜ(d−b) > (m−1)/2. Obvious once you look. Also, the matrix argument in (19) is a product of non-commuting symmetric matrices; as written it is not symmetric, so the zonal polynomial expansion requires the usual convention Cκ(XT)=Cκ(X^{1/2}TX^{1/2}). The paper never states this, but it is a standard convention.\n\nThe soft spot that actually bothers me is in Propositions 3.3 and 3.4. The contour representation (15), e^{<z>} = ∫_{L^k} Φ2[h;g;z/t] dẽµ(...)(t), is asserted with 'easy to verify' after expanding Φ2 and integrating term by term. But L^k is noncompact, and the paper gives no domination argument for that interchange. The stated hypotheses make the base measure integrable, but the series in the integrand is infinite and needs uniform control. Without that, (14) and (16) are not established as written. This may well be fixable—the analogous Ψ2 representation on the simplex is fine—but the proof as printed has a hole. The stress-test note is right on this point.\n\nAlso minor: the abstract promises a 'useful connection related to fractional calculus,' but I don't see it in the text; the only mention is that Oshima's transformations are related to Riemann-Liouville operators. The paper should either include that connection or drop it from the abstract.\n\nAll in all: the central formulas are plausible, the errors are fixable, and the contribution is useful for people who work with multivariate confluent hypergeometrics. It is not a high-impact paper, but it is a serious, non-crackpot extension. I would send it to a referee specializing in hypergeometric functions, not desk-reject it. The referee should ask for a proof of the contour interchange and the typo fixes.\n\nRecommendation: engage, with expected minor-to-moderate revision.","headline":"A clean but modest extension of Erdélyi's product formula; the formulas look right, with one genuinely missing convergence justification in the contour cases.","tokens_in":8785,"tokens_out":3961,"would_cite":false,"duration_ms":35873,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"pith_extraction":{"msc":["33C15","33C65"],"pacs":[],"model":"deepseek-v4-flash","headline":"An old product formula for ${}_1F_1$ now covers multivariate $\\Phi_2^{(k)}$, $\\Psi_2^{(k)}$, and matrix-variate confluent hypergeometric functions.","keywords":["confluent hypergeometric function","multivariate Whittaker function","product formula","zonal polynomial","matrix argument","Humbert functions Φ2 and Ψ2","Erdélyi product formula","fractional calculus"],"falsifier":"Evaluate identity (15) for $k=1$ numerically: with $z=1$, $h=3/2$, $g=5/2$, compute the right-hand side by quadrature over the vertical line $L=\\{1/2+is: s\\in\\mathbb{R}\\}$, truncating the ${}_1F_1$ series with a rigorous tail bound; if the truncated sum and integral do not commute to within the error tolerance, the term-by-term interchange fails and formula (14) cannot hold as stated.","tokens_in":7775,"feed_emoji":"🧮","tokens_out":13804,"duration_ms":103934,"temperature":0.7,"pith_summary":"This note extends Erdélyi's 1939 product formula for the univariate confluent hypergeometric function ${}_1F_1$ to two higher-dimensional families: the multivariate Humbert functions $\\Phi_2^{(k)}$ and $\\Psi_2^{(k)}$, and the confluent hypergeometric function of a matrix argument. In the multivariate setting the product of two $\\Phi_2^{(k)}$'s is written as an integral of a ${}_1F_1$ over a simplex product and a one-dimensional $\\beta$ law, or as an integral of a $\\Psi_2^{(k)}$ over three simplices, with contour-integral analogues for both $\\Phi_2^{(k)}$ and $\\Psi_2^{(k)}$. For matrices, the product of two matrix-variate ${}_1F_1$'s is shown to equal a threefold matrix-$\\beta$ integral of a third ${}_1F_1$. These identities matter because product formulas are the standard bridge from special functions to convolution structures, addition formulas, and bounds for orthogonal polynomials, and the matrix-argument case feeds directly into multivariate statistical distribution theory. A product formula for Humbert's multivariate Whittaker function and a connection to fractional calculus are recorded as corollaries.","feed_headline":"Erdélyi's product formula extends to many variables and matrices","feed_subtitle":"Products of multivariate and matrix hypergeometric functions become one integral, opening convolution identities.","key_machinery":"The machinery is a family of coupling identities that rewrite the exponential as an integral of a confluent hypergeometric function: $e^z = \\int_0^1 {}_1F_1(g;h;zt)\\,d\\mu_{h,g-h}(t)$, its simplex analogue $e^{\\langle z\\rangle} = \\int_{E^k} \\Psi_2^{(k)}[g;h;z\\circ t]\\,d\\mu(h,g-\\langle h\\rangle)(t)$, the contour analogue $e^{\\langle z\\rangle} = \\int_{L^k} \\Phi_2^{(k)}[h;g;z/t]\\,d\\widetilde{\\mu}(h-1,g+k-\\langle h\\rangle)(t)$, and the matrix-argument version $\\mathrm{etr}(X) = \\int_{O<T<I} {}_1F_1(h;g;XT)\\,d\\mu^{(m)}_{g,h-g}(T)$. Substituting one of these into the double integral obtained from the Euler representations of the two factors turns the product into a single integral of a hypergeometric function. The parameter bookkeeping is carried by $\\beta$ and Dirichlet measures in the simplex cases and by the complex measure $\\widetilde{\\mu}$ on the contour $L^k$ in the other cases, with the matrix case resting on the zonal polynomial integral $\\int_{O<X<I} C_\\kappa(XT)\\,d\\mu^{(m)}_{a,b}(T) = \\frac{[a]_\\kappa}{[a+b]_\\kappa} C_\\kappa(X)$ and the identity $\\sum_{\\langle\\kappa\\rangle=\\ell} C_\\kappa(X) = (\\mathrm{tr}\\,X)^\\ell$.","core_discovery":"The central claim is that formulas (9), (10), (14), (16), and (19) hold under the stated parameter conditions, each expressing a product of two hypergeometric functions as an integral of one hypergeometric function against a product of $\\beta$ or Dirichlet measures, with the complex measure $\\widetilde{\\mu}$ supported on $L^k$ in the contour cases. Formula (19) is representative: for positive definite $X,Y \\in \\mathrm{Sym}_m(\\mathbb{R})$ and parameters with $\\min\\{\\Re(a),\\Re(b),\\Re(g),\\Re(c-a),\\Re(b-d),\\Re(h-g)\\} > \\tfrac12(m-1)$, the product ${}_1F_1(a;c;X)\\,{}_1F_1(b;d;Y)$ equals a triple integral over matrix $\\beta$ measures of ${}_1F_1\\bigl(h;g;(U^{1/2}XU^{1/2}+V^{1/2}YV^{1/2})T\\bigr)$. The multivariate formulas similarly couple two $\\Phi_2^{(k)}$ or $\\Psi_2^{(k)}$ factors through a one-dimensional or simplex integral of a ${}_1F_1$ or $\\Psi_2^{(k)}$. The paper presents two proofs of the original Erdélyi formula and adopts the more instructive one, inserting an exponential representation of $e^z$ into the Euler integral representation of the two factors, and the Whittaker function product formula follows by specializing (16).","pith_inferences":["The same coupling scheme should yield product formulas for other Humbert or Appell-type series, for example $\\Phi_1$, $\\Phi_3$, or the two-variable Appell functions, by writing their Euler kernels as exponentials and applying identity (8); this is a direct test of the method's scope.","Applying the matrix-argument argument to the matrix-variate $\\Phi_2^{(k)}$ should give a threefold or fourfold matrix-Dirichlet integral; the authors explicitly leave that extension to the reader, so it is a natural check on the method.","If the term-by-term interchange over $L^k$ can be justified, formulas (14) and (16) may extend by analytic continuation to a larger parameter region, and the boundary $\\Re(g+k)=\\Re(\\langle h\\rangle)$ would mark the true domain."],"forward_implications":["For $\\Phi_2^{(k)}$ and $\\Psi_2^{(k)}$, the product of two functions with different parameter sets becomes a single hypergeometric function integrated against an explicit measure, giving ready-made convolution kernels for multivariate Laguerre and Whittaker expansions.","From (16) the paper derives a product formula for Humbert's multivariate Whittaker function $M_{\\kappa,\\mu}$, expressing the product $M_{\\kappa,\\nu}(x)M_{\\kappa,\\beta}(y)$ as an integral of a single $M_{\\kappa,\\lambda}$ times an explicit prefactor.","The matrix-variate formula (19) holds for positive definite matrix arguments under the analogue $\\min\\{\\Re(a),\\Re(b),\\Re(g),\\Re(c-a),\\Re(b-d),\\Re(h-g)\\} > \\tfrac12(m-1)$ of the univariate parameter conditions, providing the matrix-argument product identity used in random matrix distribution theory.","The contour formulas (14) and (16) cover parameter ranges where the simplex-based formulas (9) and (10) do not apply, provided the stated dominance conditions such as $\\Re(g+k)>\\Re(\\langle h\\rangle)$ hold."],"supporting_citations":[{"why":"The 1939 Erdélyi product formula that the paper generalizes; its two proofs set the template for all extensions.","marker":"[8]"},{"why":"Supplies the Euler-type integral representation of $\\Phi_2^{(k)}$ used at the start of Propositions 3.1 and 3.3.","marker":"[7]"},{"why":"Supplies the exponential representation (12) of $e^{\\langle z\\rangle}$ in terms of $\\Psi_2^{(k)}$, the coupling identity behind (10) and (16).","marker":"[9]"},{"why":"Introduces the contour $L^k$ and the moment relation (13) that generate the contour product formulas (14) and (16).","marker":"[17]"},{"why":"Provides the matrix-variate Euler integral for ${}_1F_1$ and the zonal polynomial facts used to prove (19).","marker":"[16]"},{"why":"Provides the zonal polynomial integral identity (Theorem 7.2.10) that verifies the matrix coupling identity (21).","marker":"[15]"}],"fun_headline_variants":["Erdelyi product formula goes multivariate and matrix","Multivariate and matrix product formulas for confluent hypergeometrics","Products of multivariate and matrix hypergeometrics as one integral","Erdelyi's product formula extended to matrix and multivariate cases","Hypergeometric products now expressed as one integral"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the series defining $\\Phi_2^{(k)}$ and $\\Psi_2^{(k)}$ may be integrated term by term over the noncompact contour $L^k$ in formulas (14) and (16); if that interchange of sum and integral fails for the stated parameter ranges, those two product formulas are not established.","fun_headline_variants_meta":{"raw":{"variants":["Erdelyi product formula goes multivariate and matrix","Multivariate and matrix product formulas for confluent hypergeometrics","Products of multivariate and matrix hypergeometrics as one integral","Erdelyi's product formula extended to matrix and multivariate cases","Hypergeometric products now expressed as one integral"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000968,"raw_usage":{"total_tokens":4103,"prompt_tokens":918,"completion_tokens":3185,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":534,"completion_tokens_details":{"reasoning_tokens":3103}},"tokens_in":534,"tokens_out":3185,"duration_ms":21061,"temperature":1.0,"reasoning_tokens":3103,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-27T20:10:15.826694+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Evaluate identity (15) for $k=1$ numerically: with $z=1$, $h=3/2$, $g=5/2$, compute the right-hand side by quadrature over the vertical line $L=\\{1/2+is: s\\in\\mathbb{R}\\}$, truncating the ${}_1F_1$ series with a rigorous tail bound; if the truncated sum and integral do not commute to within the error tolerance, the term-by-term interchange fails and formula (14) cannot hold as stated.","supporting_citations":[{"cited_title":"Erdélyi, An integral representation for the product of two Whittaker functions, J","cited_arxiv_id":null,"evidence_quote":"The 1939 Erdélyi product formula that the paper generalizes; its two proofs set the template for all extensions."},{"cited_title":"Exton, Multiple Hypergeometric Functions and Applications, Halsted Press, London (1976)","cited_arxiv_id":null,"evidence_quote":"Supplies the exponential representation (12) of $e^{\\langle z\\rangle}$ in terms of $\\Psi_2^{(k)}$, the coupling identity behind (10) and (16)."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Introduces the contour $L^k$ and the moment relation (13) that generate the contour product formulas (14) and (16)."},{"cited_title":"Muirhead, Aspects of Multivariate Statistical Theory, John Wiley & Sons, New York, 1982","cited_arxiv_id":null,"evidence_quote":"Provides the zonal polynomial integral identity (Theorem 7.2.10) that verifies the matrix coupling identity (21)."}],"review_version":1}