[Paper Review] Universality of free homogeneous sums in every dimension

Rosaria Simone|arXiv (Cornell University)|Jan 7, 2014

Random Matrices and Applications21 references3 citations

TL;DR

This paper establishes a multidimensional invariance principle for free homogeneous sums based on Chebyshev polynomials of freely independent semicircular random variables. By combining the free Lindeberg method with the Fourth Moment Theorem, it proves universality of these sums not only for semicircular limits but also for free Poisson approximations across all dimensions.

ABSTRACT

We prove a general multidimensional invariance principle for a family of U-statistics based on freely independent non-commutative random variables of the type $U_n(S)$, where $U_n(x)$ is the $n$-th Chebyshev polynomial and $S$ is a standard semicircular element on a fixed $W^{\ast}$-probability space. As a consequence, we deduce that homogeneous sums based on random variables of this type are universal with respect to both semicircular and free Poisson approximations. Our results are stated in a general multidimensional setting and can be seen as a genuine extension of some recent findings by Deya and Nourdin; our techniques are based on the combination of the free Lindeberg method and the Fourth moment Theorem.

Motivation & Objective

• To establish a multidimensional invariance principle for homogeneous sums in free probability, extending classical universality results to non-commutative settings.
• To investigate whether the semicircular distribution is the only universal limit law for homogeneous sums in free probability, or if other laws—such as the free Poisson distribution—can also exhibit universality.
• To generalize previous unidimensional results on universality and the Fourth Moment Theorem to a full multidimensional framework involving both semicircular and free Poisson limits.
• To demonstrate that Chebyshev polynomials of the second kind generate universal homogeneous sums under free independence, regardless of the underlying non-Gaussian distribution of the input variables.

Proposed method

• Applies the free Lindeberg method to control the convergence of homogeneous sums in free probability spaces, replacing classical independence with free independence.
• Uses the Fourth Moment Theorem for free Wigner chaos to link convergence in distribution to convergence of fourth moments and vanishing of non-trivial contractions.
• Introduces Chebyshev sums as a generalization of homogeneous polynomials, defined via $ U_h(x) $, the $ h $-th Chebyshev polynomial of the second kind, to construct universal families of random variables.
• Employs mirror-symmetric kernels $ f_N $ on $ [N]^d $ to ensure self-adjointness and stability of moments in the limit.
• Applies Hölder-type inequalities and moment bounds (e.g., Lemma A.2 and Proposition A.1) to control higher-order moments and ensure tightness of the sequence.
• Combines the invariance principle with the convergence criteria from Theorems A.1 and A.2 (from [9] and [14]) to prove universality for both semicircular and free Poisson limits.

Experimental results

Research questions

• RQ1 Can the universality of semicircular limits for homogeneous sums be extended to a multidimensional setting in free probability?
• RQ2 Are there other universal limit laws beyond the semicircular distribution for homogeneous sums in free probability, such as the free Poisson distribution?
• RQ3 Do Chebyshev polynomial-based homogeneous sums exhibit invariance under different distributions of freely independent input variables, even when those inputs are non-Gaussian?
• RQ4 Can the convergence of homogeneous sums to a semicircular or free Poisson limit be characterized solely by moment conditions, such as vanishing contractions and fourth moment convergence?
• RQ5 Is the free Lindeberg method effective in proving universality results for multidimensional free homogeneous sums, particularly when the limit law is non-Gaussian?

Key findings

• The paper proves a multidimensional invariance principle for vectors of Chebyshev sums based on freely independent semicircular elements, showing that convergence to a semicircular limit is equivalent to convergence of fourth moments and vanishing of all non-trivial contractions.
• Chebyshev sums based on freely independent standard semicircular variables are universal with respect to both semicircular and free Poisson approximations, regardless of the distribution of the input variables.
• The convergence of homogeneous sums to a free Poisson limit is characterized by the vanishing of all non-trivial contractions and the convergence of the $ q/2 $-th contraction to the kernel, as per Theorem A.2.
• The paper establishes that for any sequence of freely independent, centered, unit-variance random variables $ ilde{X}_i $, the homogeneous sum $ Q_N( ilde{X}_1, ilde{X}_2, ilde{X}_N) $ converges in law to the same limit as $ Q_N(S_1, ilde{S}_2, ilde{S}_N) $, provided the kernels are symmetric and the fourth moment condition holds.
• The bound $ ig| ext{Inf}_i(f_N^{(h)}) ig| o 0 $ implies convergence of the $ h $-th component of the vector of Chebyshev sums, and this condition is sufficient for universality in the multidimensional case.
• The proof relies on the combination of the free Lindeberg method and the Fourth Moment Theorem, extending classical results from [10] and [16] to the free probability setting with explicit moment-based criteria.