Papers

Snooker folklore holds that no single stroke can pocket all twenty-one object balls. We examine the claim in an explicit model: twenty-two equal disks on a rectangle with six pocket points, under constant rolling deceleration, frictionless collisions with normal restitution, and lossy specular cushions. We exhibit a configuration and a stroke, and prove in interval arithmetic that every stroke in an explicit box around it pockets all twenty-one object balls, so the set of successful strokes has positive measure. That set is a ribbon: every speed from 9.3213 to 12 m/s works, while the certified window of directions is 9.4813e-43 radians wide, a number explained by a first-order analysis of twenty near-grazing collisions. A second configuration, a near-frozen cluster, has a certified window 9.4e-4 radians wide, thirty-nine orders of magnitude wider; its certificate uses Taylor models of degree six. The strokes with a given list of events form a semialgebraic set, and the endpoints of the windows are algebraic numbers. For the regulation opening position we state the corresponding conjecture, and show that if it holds with a regular witness then a finite certified search establishes it.

A Lorentz boost acts on the canonical spin of a massive particle through a momentum-dependent Wigner rotation. We show that, for one fixed observer boost, reducing over an uncertain momentum can strictly contract every pairwise spin-state trace distance without producing a channel that is degradable from the less contracted one. For spin $1/2$, we first characterize the exact inversion-symmetric channel cone generated by a fixed Wigner angle and transverse momentum directions. Inside this cone lies the Pauli family $M_\alpha=\operatorname{diag}(1-\alpha,1-\alpha,1-2\alpha)$, $0\leq\alpha<1/2$. For $0<\alpha<\beta<1/2$, all trace distances between distinct spin states are strictly smaller after $M_\beta$ than after $M_\alpha$, yet the unique linear post-processing factor has a negative normalized Choi eigenvalue. We solve the optimization over all physical converters exactly: $\frac{1}{2}\inf_{\Lambda\in\mathrm{CPTP}}\|\Phi_\beta-\Lambda\circ\Phi_\alpha\|_\diamond=\frac{\alpha(\beta-\alpha)}{2-3\alpha}$, whereas the reverse deficiency is $\beta-\alpha$. Thus the identity dominates the family, while all positive-noise members are pairwise incomparable under CPTP post-processing. The ideal construction is realized as the narrow-packet limit of pure, normalizable five-component momentum states, and explicit perturbation and finite-shot tomography bounds certify an open set of examples. Separately, every nonidentity member fails embedding in a time-homogeneous Pauli-diagonal Lindblad semigroup. Hence ordering all unassisted spin distinguishabilities does not determine the quantum statistical post-processing order.

Unlike in finite dimensions, quantum information in continuous-variable systems has the peculiar feature that without imposing physical constraints, the sample complexity of state tomography can be unbounded. Remarkably, this is even the case for state-of-the-art protocols for learning Gaussian states, which have finite-dimensional descriptions: the best known rates scale with $\log \log E$, where $E$ is the energy of the system. We prove this is not an artifact of existing analyses, but a fundamental limitation of the measurements used. We show: (1) Any protocol that uses Gaussian measurements, even entangled or adaptively chosen ones, must incur a $\log \log E$ dependence. This answers an open question posed by a number of previous works. (2) There is a smooth tradeoff between the number of rounds of adaptivity and the energy dependence, and we give a matching protocol achieving this interpolated rate. (3) With highly entangled, non-Gaussian measurements, one can learn $n$-mode pure Gaussian states with $O(n^2 / \epsilon^2)$ samples, independent of $E$. This answers an open question posed by Chen et al. (4) A simple protocol based on the single-copy canonical phase POVM of Holevo and Helstrom learns single-mode pure Gaussian states with $O(1/\epsilon^2)$ samples, again independent of $E$. Our results clarify the role of energy in bosonic state tomography and shed new light on the intriguing interplay between adaptivity, entanglement, and magic in quantum learning.

We show that for every $\eta>0$ and sufficiently large $n$, there exists a poset of size $2^{(1+\eta)n/2}$ containing all the $n$-element posets as induced subposets. This improves a recent result of Bastide, Groenland and Nenadov. Our proof provides a labeling scheme preserving transitivity, inspired by the Boolean lattice. Among other tools, we use the Szemerédi Regularity Lemma.

We prove a lower bound for the Birkhoff min-max invariant of a Riemannian sphere in terms of the min-max width of its embedded circles. The main tool is a method to induce a sweepout by pairs of points in an embedded circle from a given sweepout of the sphere by closed curves, so that the points of each pair of the induced sweepout lie close to two curves of the ambient sweepout that are close to each other. Moreover, considering a non-compact complete Riemannian manifold, we relate upper bounds for the min-max width of its embedded circles to the existence of continuous functions from the manifold to finite graphs with level sets that have uniformly bounded diameters, thus giving an estimate for its 1-width in Urysohn's quantitative dimension theory. In the specific case of the Euclidean plane, we bound from below the classical width of a simple closed curve by its min-max width. Finally, we prove related rigidity results characterizing the round spheres among Zoll spheres.

We give a streamlined proof of the fact that generically, isospectral hyperbolic surfaces are isometric. We also prove some versions of this result allowing for quasi-Fuchsian groups or considering the simple length spectrum.

Data assimilation estimates the state of a dynamical system from model forecasts and incoming observations. Many observation mechanisms, however, are many-to-one, implicit, non-smooth, or accessible only through simulation, and need not provide the residual structures or likelihood guidance required by existing ensemble filters. We introduce implicit data assimilation, in which the analysis law is defined as an energy tilt of the forecast distribution. We then propose the Ensemble Controlled-flow Filter (EnCF), which realizes this update through a stochastic controlled flow and learns the observation-dependent control by adjoint matching from terminal energy gradients. For simulator-defined observations, EnCF-LF learns a surrogate conditional energy from samples and applies the same controlled-flow solver. We prove ideal exactness, derive a one-step error decomposition, and establish non-accumulation of local errors under filter stability. Numerical results show that Kalman-type filters remain preferable for smooth additive-Gaussian observations, while the proposed methods are better suited to non-Gaussian, many-to-one, multimodal, and implicit observation models.

This paper investigates preservation and definability for the equality-free fluted fragment over finite relational signatures. We first prove an equirank homomorphism preservation theorem: every homomorphism-preserved fluted sentence has an existential-positive fluted equivalent of no greater quantifier rank. We then refute Purdy's claimed Los–Tarski preservation theorem for the fluted fragment. Our counterexample is a rank-three sentence over one binary relation that is preserved under extensions but has no existential fluted equivalent. Finally, a two-variable counterexample over a unary base signature shows that weak and ordinary Beth definability both fail. All three results hold over all structures and over finite structures.

Lower numerical precision reduces storage and memory traffic but raises the perturbation floor. We study rank compensation: reinvesting saved memory in a larger approximation rank. For matrices, the singular-value error identity yields a directly testable sufficient condition requiring the additional singular component to offset the perturbation from storing the rank-augmented approximation in lower precision. On ten SuiteSparse matrices, all 100 truncation-dominated configurations (50 FP32 and 50 FP16) are certified non-increases and strict accuracy wins, with mean error ratio $0.963$ and storage ratios $58.8\%$ and $29.4\%$ relative to the FP64 baseline. FP16 failures occur only in tail-rank stress tests near the perturbation floor. At the largest resident matrix-application batch, compensated FP32 and FP16 achieve geometric-mean A100 speedups of $1.28\times$ and $2.12\times$; neither accelerates the smallest batch. For Tensor-Train (TT) approximation, we give a conditional a posteriori extension based on the measured truncation gain and rounded-core perturbation. Across three-way and six-way synthetic tests, FP32 and FP16 achieve combined accuracy-memory wins in 10 of 20 and 14 of 20 trials. On public hyperspectral tensors and FROSTT top-active subtensors, the corresponding counts are 44 of 60 and 54 of 60; four FP16 Salinas-A tail-stress cases fail. No certified TT case exceeds the FP64 error beyond numerical tolerance. Reconstruction of six public tensors yields geometric-mean compensated speedups of $1.38\times$ (FP32) and $1.94\times$ (FP16). Timings cover resident downstream kernels, not factorization, transfers, or end-to-end acceleration.

A finite quantum reference token with group-valued misalignment induces a random-unitary channel, but optimal degradation is generally an optimization over all quantum post-processings. For a unitary representation U of a finite group G, we prove that the following conditions are equivalent: U contains every irreducible type; one ancilla-assisted input has an orthonormal G-orbit; signed group measures embed isometrically into channels in diamond norm; and, for every pair of noise laws p,q, $\inf_{\Lambda\in\mathrm{CPTP}} \frac{1}{2}\|\Phi_q^U-\Lambda\circ\Phi_p^U\|_\diamond =\min_{r\in\mathcal{P}(G)}\frac{1}{2}\|q-r*p\|_1$. Thus representation completeness is the exact carrier condition for universal reduction of quantum post-processing to classical convolution. We determine the minimum ancilla dimensions for an orthogonal orbit and for an invariant calibration seed. For an incomplete carrier, with visible Plancherel dimension S(U), we derive the exact conditional-expectation distance $\frac{1}{2}\|\operatorname{id}-\Phi_u^U\|_\diamond=1-1/S(U)$ and an explicit quantum–classical deficiency gap. For irreducible carriers the deficiency is obtained in closed form; the faithful two-dimensional representation of $S_3$ yields an exact ten-percent reduction relative to classical convolution. We also characterize law identifiability through the conjugation representation, provide finite linear programs and decision witnesses, and establish both a finite-dimensional obstruction and stable visible-band reconstruction for infinite compact groups. Deterministic ancillary code reproduces the finite-group examples and numerical regression checks.

Foundational work by Avramov and Buchweitz, as well as by Huneke and Jorgensen, established symmetry results for the vanishing of Ext over local complete intersections and AB-rings, respectively. Later, several authors studied symmetry in the vanishing of Ext over local Gorenstein rings under additional homological assumptions on the modules involved. In this work, we provide a generalized symmetry in the vanishing of Ext for homologically finite complexes over a Noetherian ring with a dualizing complex, under additional homological conditions on the complexes involved. The results in this paper extend and unify several symmetry results on the vanishing of Ext previously proved over local Gorenstein rings and show that the dualizing complex has been hidden in these results.

Comparing two noisy quantum reference frames as statistical experiments depends on the dimension of the ancillary memory available to the decision procedure. For finite-dimensional channels A and B with invertible A, we show that exact simulation of all measurements assisted by an r-dimensional ancilla is equivalent to r-positivity of the unique factor Gamma = BA^{-1}. The hierarchy can be realized by physical channel pairs: every unital, trace-preserving map that is k-positive but not (k+1)-positive embeds as the factor between the channels D_a and Gamma composed with D_a on an exact interval determined by the smallest Choi eigenvalue. For depolarizing source and target channels D_a and D_b, including negative and singular source parameters, the phase boundary is $\mathcal{D}_a \succeq_r \mathcal{D}_b \Longleftrightarrow -1/(dr-1) \leq b/a \leq 1$ for $a\neq 0$. We derive closed formulas for the restricted level-r deficiency and for the distance to every physical post-processing, $\delta_{\mathrm{phys}}(\mathcal{D}_b\mid\mathcal{D}_a)=(1-1/d^2)\operatorname{dist}(b,I_a)$, where $I_a=\operatorname{conv}\{a,-a/(d^2-1)\}$. The largest physical conversion cost hidden from all tests through level k is $(d-k)/[d(d^2-1)]$. An untouched m-level spectator changes the first detecting external level from k+1 to $\lfloor k/m\rfloor+1$. A transpose–depolarizing construction shows that the separation is not confined to depolarizing factors. The results quantify the distinction between ancilla-restricted statistical simulation and implementation by a single quantum channel.

We show that the monoidal Grothendieck construction can be applied to recover the known structure theorem for Clifford monoids, which states that they correspond to functors from a semilattice into the category of groups. Furthermore, we capture the category of Clifford monoids itself as a Grothendieck construction of the functor sending a semilattice L to the functor category [L, Grp] and use this to construct a number of factorisation systems on this category. Finally, we prove a general result on taking monoids in a monoidal fibration and apply it to establish a correspondence between inverse semirings and lax monoidal functors from an idempotent semiring into the category of abelian groups.

Strict complementarity of a primal-dual pair of optimal solutions is fundamental in the numerical analysis of semidefinite programs (SDPs). Strict complementarity drives the convergence behavior of interior point algorithms. When it fails, pathological examples show a striking gap between two error measures of approximate solutions. The first of these is the forward or "true" error, i.e., the distance to the optimal solution set. The second is the less useful backward error, measured by the constraint violation. We first characterize the lack of strict complementarity in SDPs via a simple normal form. Our normal form has three key features: (i) it is obtained using elementary row operations and rotations; (ii) it makes the lack of strict complementarity evident; and (iii) it lets us construct any such SDP by a simple algorithm. A variant of our generating algorithm allows us to construct any SDP that fails strict complementarity but satisfies Slater's condition on both the primal and dual sides. Thus, we parametrize the data of all SDPs that lack strict complementarity in a manner similar to how the Jordan normal form parametrizes square matrices with given eigenvalue structure. Next, we precisely characterize when the singularity degree of an SDP equals the number of constraints – a result that underlies our generating algorithms and that we believe is of independent interest. We construct and share a set of SDPs that lack strict complementarity and present a detailed computational study. We find that forward-backward error gaps are quite common: in many small SDPs (with matrix order $\leq 20$), the forward ("true") error exceeds the backward error by up to seven orders of magnitude. Further, in several data sets the forward and backward errors are inversely correlated. In other words, the worse the "true" forward error is, the harder it is to detect.

We show that a simply connected symplectic manifold admitting a complete nonpositively curved compatible metric is symplectomorphic to R^{2n} with its standard symplectic form. This answers the "symplectic Hadamard question" of McDuff-Salamon in the affirmative.

Stochastic convex optimization is a classical problem with well-understood guarantees under first-order feedback. In contrast, for zero-order optimization with noisy function evaluations, a logarithmic gap has persisted between known upper bounds and the $\Omega(1/\sqrt{T})$ lower bound, even in the one-dimensional case. In this work, we study the problem of minimizing a convex function $f : [0,1] \to [0,1]$ using a zero-order oracle with subGaussian noise. We propose a computationally efficient algorithm that achieves the optimal $O(1/\sqrt{T})$ convergence rate, matching the lower bound. The result closes the existing gap in one dimension, providing the first sharp rate guarantee in this setting.

We study universally null sets in forcing extensions and the cardinal invariants of their ideal $\mathcal{UN}$. We prove that the Miller model contains a nonmeager universally null set of cardinality $\aleph_2=\mathfrak{c}$, answering a question of Brendle and Larson. The proof uses preservation of universal nullity by countable support iterations of Miller forcing and shows that the set of Miller reals added during the iteration is universally null. We also prove that every universally null set in the $\mathbb{PT}_{f,g}$ model has cardinality at most $\aleph_1$, and show that there is consistently a forcing destructible universally null set. For the cardinal invariants, we show that $|X|<\operatorname{cof}(\mathcal{UN})$ for every $X\in\mathcal{UN}$, and that $\operatorname{add}(\mathcal{N})=\mathfrak{c}$ implies $\operatorname{cof}(\mathcal{UN})=\mathfrak{d}_{\mathfrak{c}}$. Finally, we prove the consistency of $\operatorname{add}(\mathcal{UN})<\operatorname{cov}(\mathcal{UN}) <\operatorname{non}(\mathcal{UN})<\operatorname{cof}(\mathcal{UN})$.

We construct a separable Banach space $X$ with a Schauder basis such that $B_X$ is not a uniformly continuous retract of $B_{X^{**}}$. Consequently, $X$ is not a uniformly continuous retract of $X^{**}$ and hence is not a Lipschitz retract of $X^{**}$.

How large can a set of mutually orthogonal quantum Latin squares (MOQLS) get? We show that a set of n - 2 MOQLS of order n is necessarily classical and construct large non-classical sets of MOQLS of orders that are prime powers, improving both the previously known lower and upper bounds.

We present MXsim v0.1: a MATLAB library for simulating the OCP MX block floating-point arithmetic and formats. The library is built on top of a custom-precision floating-point simulator, CPFloat, which allows custom variants of non-block IEEE 754 floating-point formats to be simulated. MXsim is a MATLAB simulator, and is targeted at mixed-precision algorithm development community to perform experimentation in OCP MX block floating-point arithmetic by using a simple and well-documented interface in MATLAB, which is especially useful when access to the latest NVIDIA and AMD GPUs is not available. Furthermore, due to full customisation of exponents, precisions of input and output formats, and the precision of the accumulator within the dot product operation, MXsim allows experimenting with custom variants of the OCP MX block floating point for research. Finally, since the bit-wise numerical behavior of OCP MX formats is not documented by GPU vendors, and the precision and rounding are not prescribed by the OCP specification itself, MXsim allows experiments with various possible settings in order to determine the effects on mixed-precision numerical algorithms.

A dynamical system is said to be non-autonomous when the differential equations describing its evolution depends explicitly on time. Among the various geometric approaches to investigate such systems, the cosymplectic formulation provides a natural framework that extends symplectic geometry to time-dependent Hamiltonians systems. However, preserving the associated geometric structures under numerical discretization remains a challenging problem: standard integrators, such as explicit Euler schemes, generally fail to conserve the cosymplectic volume or the underlying Poisson structure. In this work we propose a geometric method for the discretization of non-autonomous Hamiltonian systems based on generalized canonical transformations. The approach constructs a symplectomorphism on the extended phase space $T^*(Q \times \mathbb{R})$ whose projection onto $T^*Q \times \mathbb{R}$ defines a structure-preserving discrete flow. We show that this formulation guarantees the preservation of key invariants, including the volume form, the Poisson bracket, and the symplectic structure on each time fiber.

For an integer $s \ge 2$, a subset $S \subseteq [n]$ is {\em $s$-stable} if $\min \{j - i, n + i - j\}\ge s$ for every $i,j \in S$ with $i<j$. Denote the set of all $s$-stable subsets of size $k$ of $[n]$ by $\binom{[n]}{k}_{s\text{-stable}}$. Schrijver proved in 1978 that whenever $n\ge 2k$, the chromatic number of the Kneser graph $\mathrm{KG}\big( \binom{[n]}{k}_{2\text{-stable}}\big)$ is $n - 2k +2$. Generalizing this result, Meunier conjectured in 2011 that $\chi\left( \mathrm{KG}\big( \binom{[n]}{k}_{s\text{-stable}} \big) \right)= n - sk +s$ for all $n\ge sk$. This conjecture was previously proven for all even $s$, for $s \ge 4$ and large enough $n$, and for $k=2$. We prove the conjecture in the cases $s=3$ and $n$ large enough, or $k=s=3$. To this end, we prove versions of the Hilton-Milner theorem for $s$-stable sets. We also present a topological approach towards Meunier's conjecture.

Let $\left(M, \mathcal{O}_M \right)$ be a smooth algebraic variety over field $\kappa$ of characteristic $0$ with an algebraic symplectic form $\omega$, or a complex manifold with a holomorphic form $\omega$. Furthermore, let $E$ be a vector bundle over $\left(M, \mathcal{O}_M \right)$ and $\mathcal{O}_{\hbar}$ a deformation quantization of $\mathcal{O}_M$ compatible with $\omega$. Assuming that $E$ possesses a deformation quantization to order $\hbar^k$ we consider the problem of extending it to order $\hbar^\ell$ for $\ell > k$, and establish triviality of an obstruction class as a necessary condition for this extension to exist. Furthermore, in the case $\ell \le 2k+1$, we prove that this condition is also sufficient.

We show the Randomized Hamiltonian Monte Carlo (RHMC) algorithm has accelerated mixing time guarantees for sampling from log-concave probability distributions. RHMC proceeds by repeatedly simulating the continuous-time Hamiltonian dynamics for some random integration times, and resetting the velocity to be an independent Gaussian random variable between each simulation. We show that when the target distribution is log-concave and satisfies an $\alpha$-Talagrand inequality (for example, if the target distribution is $\alpha$-strongly log-concave), if we use a random integration time from either the triangular or the exponential distribution with mean $\Theta(\alpha^{-1/2})$, then RHMC converges exponentially fast in KL divergence, and the total integration time to reach error $\varepsilon$ in KL divergence scales as $O(\alpha^{-1/2} \log(\varepsilon^{-1}))$. We also show that when the target distribution is log-concave, if we use a sequence of random integration times from the triangular distribution with exponentially increasing means, then the total integration time to reach error $\varepsilon$ in KL divergence scales as $O(\varepsilon^{-1/2})$. Our analysis relies on a bound on the average KL divergence along Hamiltonian dynamics, which is inspired by an analogous result on accelerated optimization methods based on Hamiltonian dynamics.

One might expect that finite relaxed elastic energy rules out diffuse singularities in the derivative, leaving only absolutely continuous and jump parts. This is suggested by the role of $SBV$ in free-discontinuity problems and by interpreting jumps as vertical segments of limiting graphs. We show that it fails for the relaxed one-dimensional Willmore energy. We construct a continuous function $u\in BV((0,1))$ with $D^c u\neq0$ and $\overline{\mathcal{W}}(u)<\infty $, so finite relaxed Willmore energy does not imply $u\in SBV((0,1))$. The idea is to concentrate the Cantor part exactly where the absolutely continuous slope blows up. There the singular diffuse measure meets the blow-up condition of the relaxation theorem, while the weighted curvature term stays integrable. Geometrically, the example shows that Cantor parts of $BV$-graph derivatives need not be intrinsic singularities of the underlying curve. The graph has an arc-length parametrization of class $C^1\cap W^{2,2}$, and a suitable rotation turns it into a Lipschitz graph whose derivative has no singular part. The construction also rescales to make the relaxed Willmore energy arbitrarily small, and it extends to relaxed $L^p$-curvature energies for all $p>1$.

A chemotaxis-driven multiphase multispecies diffusion system, arising in the formation of vascular-like structures, is analyzed. The model couples porous-medium-type cross-diffusion equations for the volume fractions of the cellular components with multispecies Keller-Segel equations governing the chemoattractant concentrations, posed in a bounded domain with no-flux boundary conditions. The system is derived within a multiphase framework based on mass and force balance laws, together with a characterization of the mixture pressure gradient, which follows from the volume-filling constraint. The existence of a global weak solution, the weak-strong uniqueness property, the exponential decay to the constant steady state, and the vanishing diffusion limit are established. The existence proof extends the boundedness-by-entropy method to solution codomains that are bounded in some directions only, while the other results are based on various entropy estimates and uniform dissipation bounds.

We classify the moduli spaces of representations of subspace quivers which are Fano fourfolds, under a natural assumption on the dimension vector. These moduli spaces can also be described as GIT quotients of products of Grassmannians by the diagonal action of a projective linear group, and there are exactly four of them. They are rational, of pure Hodge-Tate type, infinitesimally rigid, and have finite automorphism groups, with Picard ranks 5, 6, 6 and 7, making them interesting examples in the classification of Fano fourfolds of large Picard rank, as they are not toric or products. Two are known varieties: Manivel's Segre cousin of the Segre cubic 3-fold, and the Fano model of the blowup of $\mathbb{P}^4$ in six points. The other two appear to be new: one is an involution surface bundle over $\mathbb{P}^2$, and the other is a "Segre cousin once-removed", whose geometry closely parallels that of the Segre cousin. Using techniques from quiver moduli, which we survey, we describe the geometry of all four fourfolds in detail.

A claw-free cubic graph is a cubic graph with no induced subgraph isomorphic to $K_{1,3}$. The zero forcing process begins with an initial set $S$ of colored vertices. At each step, a colored vertex with exactly one uncolored neighbor forces that neighbor to become colored. If repeated applications of this rule color every vertex of $G$, then $S$ is called a zero forcing set. The minimum cardinality of a zero forcing set is the zero forcing number, denoted by $Z(G)$. In this paper, we answer three open questions posed by Davila and Henning concerning upper bounds on the zero forcing number of claw-free cubic graphs. We characterize the connected claw-free cubic graphs satisfying $Z(G)=\alpha(G)+1$, where $\alpha(G)$ is the independence number. In addition, we establish the improved upper bound $Z(G)\leq \frac{T}{2}+D+2$ for claw-free cubic graphs with Hamiltonian contraction multigraphs, where $D$ is the number of diamonds and $T$ is the number of triangles in $G$.

Generating functions are among the most important analytical tools in the theory of orthogonal polynomials, providing a unified framework for deriving structural identities, asymptotic expansions, and zero distributions. However, despite the extensive development of discrete Sobolev orthogonal polynomials, no general generating-function theory has been available for the Sobolev-type Charlier and Meixner families associated with arbitrary-order forward differences $j\geq 1$ and an exterior mass point $\alpha<0$. In this paper, we develop the first unified generating-function framework for these families. Starting from explicit connection formulas, we derive generating functions for the Sobolev-type polynomials and their iterated forward differences, which serve as the main analytical tool for establishing new Mehler–Heine formulas. The resulting asymptotic analysis shows that an exterior Sobolev perturbation generates exactly one exceptional zero converging to the mass point, while the remaining zeros preserve the classical asymptotic distribution. Moreover, the limiting Mehler–Heine functions are independent of both the Sobolev mass parameter and the order of the forward difference operator, revealing a universality phenomenon for higher-order discrete Sobolev perturbations. These results considerably extend the analytical theory of discrete Sobolev orthogonal polynomials and establish a direct connection between generating functions, Mehler–Heine asymptotics, and the asymptotic distribution of the zeros for the discrete Sobolev-type Charlier and Meixner families.

Tensegrity form-finding and physical property prediction are fundamental problems in structural mechanics, which aim to determine equilibrium configurations and internal force distributions. These problems are challenging due to strong nonlinearity arising from the coupling between geometry and forces, and the need to satisfy equilibrium, stability, and structural constraints. This paper proposes an energy-based learning approach for clustered tensegrity form finding and physical property prediction. The proposed approach incorporates total potential energy minimization and constitutive relations into the training objective, enabling the prediction of equilibrium nodal configurations and the reconstruction of physical quantities such as member forces and force densities. By integrating energy-based physical losses directly into the learning process, the method promotes physical consistency while combining data-driven learning with physics-based constraints. Numerical experiments on tensegrity prism and lander structures demonstrate accurate prediction of equilibrium configurations and internal forces across different training-data ratios, indicating the potential of the proposed approach for nonlinear tensegrity form finding and structural analysis.