Angular Geometry of Multivariate Mellin Integrals: Ray Evaluation, Convergence Polytopes, and Singularity Integrability
Year: 2026
Type: preprint
Abstract: We study **ray-valued multivariate Mellin integrals**\[\mathcal M_{\boldsymbol\theta}[f](\mathbf s)=\int_{(0,\infty)^n}\prod_{j=1}^n t_j^{s_j-1}f(e^{i\theta_1}t_1,\dots,e^{i\theta_n}t_n)\,d\mathbf t,\]in which the integration variables \(t_j\) remain positive real, while the phase parameters \(\boldsymbol\theta\in\mathbb R^n\) only rotate the rays on which the kernel \(f\) is evaluated. This setting differs from classical contour integration over complex rays: no Jacobian factor occurs in the definition, and the role of the phases is entirely absorbed into the evaluation of the kernel. **The first group of results establishes the gauge structure.** Under linear phase rotation, and assuming that the boundary terms in the integration by parts vanish, we obtain \(\partial_{\theta_j}\mathcal M_{\boldsymbol\theta}[f]=-is_j\mathcal M_{\boldsymbol\theta}[f]\) (Theorem 1.10). Consequently, the gauge transform \(\widehat{\mathcal M}_{\boldsymbol\theta}=e^{i\sum_j\theta_js_j}\mathcal M_{\boldsymbol\theta}\) is constant on each connected component of the feasible angle set, and reduces to the classical Mellin integral on the component containing \(\mathbf 0\). Thus the differentiation rule and the global phase factor induced directly by linear phase rotation are gauge artifacts; genuinely new content can reside only in the **geometry of the domain of convergence** and in the **geometry of the feasible angle set**. **The second group of results develops the geometry of non-product kernels.** For \(f(\mathbf t)=F(\sum_j t_j)\), we establish an **exact angular averaging formula** (Theorem 3.7), which reduces the \(n\)-fold integral exactly to an \((n-1)\)-fold weighted average over the simplex together with a one-dimensional rotated Mellin integral:\[\mathcal M_{\boldsymbol\theta}[f](\mathbf s)=\int_\Delta\prod_j u_j^{s_j-1}|w_{\boldsymbol\theta}(\mathbf u)|^{-\Sigma}\mathcal M_{\arg w_{\boldsymbol\theta}(\mathbf u)}[F](\Sigma)\,d\mathbf u,\]where \(w_{\boldsymbol\theta}(\mathbf u)=\sum_j u_je^{i\theta_j}\) and \(\Delta\) denotes the open simplex. Under the nondegenerate-cone condition \(c(\boldsymbol\theta)=\operatorname{dist}(0,\operatorname{conv}\{e^{i\theta_j}\})>0\) and uniform power-law hypotheses, we prove the **convergence polytope theorem** (Theorem 3.14): the multivariate domain of convergence is\[\mathcal C_{\boldsymbol\theta}(f)=\{\sigma_j>0\ \forall j,\ \alpha<\Sigma<\beta\},\]a set defined by at most \(n+2\) half-spaces; in general it is not a product of strips in the individual variables (Corollary 3.18). **The third group of results provides an angular integrability criterion and a classification of singularities.** The **angular integrability criterion** (Theorem 4.5) characterizes \(\boldsymbol\theta\in\Theta(f)\) precisely by the condition that \(\sigma_j>0\) for all \(j\) and that \(A_{\arg w}\) be integrable against the pushforward of the simplex measure; this clarifies that "the cone avoids the singularities" is a sufficient but not a necessary condition. The **angular smoothing theorem for singularities** (Theorem 4.17) describes local integrability at an isolated singularity \(z_0\) lying inside the cone: setting \(L_{\boldsymbol\theta}(\mathbf t)=\sum_j e^{i\theta_j}t_j\), \(k=\operatorname{rank}_{\mathbb R}L_{\boldsymbol\theta}\in\{1,2\}\) and \(|F(z)|\asymp|z-z_0|^{-\gamma}\), local integrability holds if and only if \(\gamma<k\). **A simple pole is integrable when \(k=2\), but not when \(k=1\)**—this is the essential distinction between the multivariate and univariate settings. Furthermore, non-product behaviour is governed by three mechanisms: fatal crossing of an isolated obstructing direction with \(\gamma\ge2\); boundary contact at an isolated obstructing direction with \(\gamma<2\); and open regions of obstructing directions (Propositions 4.25, 4.26 and Remark 4.30). The product case is the exception rather than the rule. **Applications and numerics.** We give three closed forms for kernels of type \(F(\sum t_j)\) (Theorems 5.3–5.5), tables of kernels, integral representations of multiple zeta kernels carrying \(\prod\Gamma\) factors (Theorems 5.16, 5.18), and a numerical scheme based on the angular averaging formula. The \(\gamma<k\) integrability principle of Theorem 4.17 is complementary to sector decomposition for Feynman integrals and to amoeba/Newton polytope theory: within the \(F(\sum t_j)\) structure we obtain explicit polytopes and an angular decomposition, and we treat non-rational kernels and the integrability of singularities.
Author: NI, Chuangao
Language: English
Subfield: Applied Mathematics
Field: Mathematics
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7225998057
