Malliavin smoothness of the Rosenblatt process

Laurent Loosveldt, Yassine Nachit, Ivan Nourdin, Ciprian Tudor · 2025

Read Malliavin smoothness of the Rosenblatt process by Laurent Loosveldt, Yassine Nachit, Ivan Nourdin, Ciprian Tudor (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-10299.txt.

Introduction

Laurent Loosveldt, Yassine Nachit, Ivan Nourdin, and Ciprian Tudor study the smoothness of probability densities for the Rosenblatt process, a non-Gaussian process arising from long-range dependence. Working inside Malliavin calculus, they prove that finite-dimensional vectors of the process are nondegenerate in the Malliavin sense. As a consequence, their densities are Schwartz functions: smooth and rapidly decreasing at infinity.

The proof establishes that the determinant of the Malliavin matrix has finite moments of every negative order, exploiting the special structure of variables in the second Wiener chaos. The authors also derive exponential-type upper bounds for the partial derivatives of these densities. Smooth density results of this kind underpin limit theorems and statistical procedures for strongly dependent data.

Catalog record

CREATORLaurent Loosveldt, Yassine Nachit, Ivan Nourdin, Ciprian Tudor
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE31 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2511-10299

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Malliavin smoothness of the Rosenblatt process Laurent Loosveldt, Yassine Nachit, Ivan Nourdin, Ciprian Tudor Abstract We investigate the smoothness of the densities of the finite-dimensional distributions of the Rosenblatt process. Within the Malliavin calculus framework, we prove that Rosenblatt random vectors are nondegenerate in the Malliavin sense. As a consequence, their densities belong to the Schwartz space of rapidly decreasing smooth functions. The proof relies on establishing the existence of all negative moments of the determinant of the Malliavin matrix, exploiting the specific structure of random variables in the second Wiener chaos. In addition, we derive exponential-type upper bounds for the partial derivatives of the densities of the finite-dimensional distributions of the Rosenblatt process. Keywords: Rosenblatt process; Malliavin calculus; nondegeneracy; Bouleau–Hirsch criterion; density of finite-dimensional distributions. 2020 AMS Classification Numbers: 60G22; 60G18; 60H07; 60F05. 1 Introduction Let F=\left(F_{1},\ldots,F_{m}\right) be a random vector whose components are differentiable in the Malliavin sense. Denote by \Gamma_{F} its Malliavin matrix, whose entries are given by \Gamma_{F}(i,j)=\langle DF_{i},DF_{j}\rangle_{\mathfrak{H}} , 1\leq i,j\leq m , where D denotes the Malliavin derivative with respect to an isonormal process (W(h),h\in\mathfrak{H}) on a real and separable Hilbert space \mathfrak{H} . A classical result in Malliavin calculus (the Bouleau-Hirsch criterion ) states that if \Gamma_{F} is almost surely invertible, then F admits a density with respect to the Lebesgue measure on \mathbb{R}^{m} . The random vector F is said to be nondegenerate if its components are infinitely Malliavin differentiable and if \left(\det\Gamma_{F}\right)^{-1}\in L^{p}(\Omega) for every p\geq 1 . In this case, its density is smooth and belongs to the Schwartz space of rapidly decreasing C^{\infty} functions on \mathbb{R}^{m} . Hence, obtaining a smooth density is tightly linked to the existence of negative moments of the Malliavin derivatives of the components. When the components of F lie in a fixed Wiener chaos, the above result can be strengthened. It was shown in [ 6 ] that in this setting the almost sure invertibility of \Gamma_{F} is actually equivalent to the absolute continuity of the law of F . Moreover, [ 8 ] proves that a pair of multiple stochastic integrals fails to admit a density if and only if its components are proportional, mirroring the Gaussian case. In the recent work [ 4 ] we applied the Bouleau-Hirsch criterion to establish the absolute continuity of finite-dimensional distributions of Hermite processes. Let q\geq 1 be an integer and (Z^{H,q}_{t},t\in\mathbb{R}) a Hermite process of order q with self-similarity index H\in\left(\frac{1}{2},1\right) . Recall that Z^{H,q} is an H -self-similar process with stationary increments and long memory, living in the q th Wiener chaos. The class of Hermite processes includes fractional Brownian motion ( q=1 ), which is Gaussian, and the Rosenblatt process ( q=2 ), which belongs to the second Wiener chaos. For any m\geq 1 and any grid 0<t_{1},...<t_{m} , [ 4 ] shows that the determinant of the Malliavin matrix of the Hermite vector Z=\left(Z^{H,q}_{t_{1}},\ldots,Z^{H,q}_{t_{m}}\right) is almost surely strictly positive; therefore, Z admits a density. The objective of the present work is to go a step further and investigate the smoothness of this density. As previously mentioned, smoothness is deeply linked to nondegeneracy, and in particular to the existence of negative moments of the determinant of the Malliavin matrix. While establishing these negative moments appears challenging for general q , we succeed in proving nondegeneracy for q=2 , that is for Rosenblatt vectors. This relies crucially on the specific structure of random variables in the second Wiener chaos, which can be represented as weighted sums of independent centered chi-square random variables. Our first main result concerns the smoothness of the density of increments of the Rosenblatt process. Theorem 1 Let (Z_{t},t\in\mathbb{R}) be a Rosenblatt process with self-similarity index H\in\left(\frac{1}{2},1\right) . Let m\geq 1 and 0=t_{0}<t_{1}<...<t_{m} . Consider the random vector Z_{\pi}=\left(Z_{t_{1}}-Z_{t_{0}},\ldots,Z_{t_{m}}-Z_{t_{m-1}}\right), (1) Then Z_{\pi} admits a density belonging to the Schwartz space \mathcal{S}(\mathbb{R}^{m}) of rapidly decreasing smooth functions. As a consequence, for all m\geq 1 and for all 0\leq s_{1}<...<s_{m} , the vector (Z_{s_{1}},\ldots,Z_{s_{m}}) also has a density in \mathcal{S}(\mathbb{R}^{m}) , see Corollary 1 . The proof of Theorem 1 relies on the classical Malliavin calculus theorem asserting that nondegenerate random vectors possess densities in the Schwartz space. Our second main result provides exponential-type bounds for the partial derivatives of the density of the finite-dimensional distributions of the Rosenblatt process. For m\in\mathbb{N}^{\ast} and n=(n_{1},...,n_{m})\in\mathbb{N}^{m} , set \partial_{x}^{n}=\prod_{j=1}^{m}\left(\frac{\partial}{\partial x_{j}}\right)^{n_{j}}. Theorem 2 Let (Z_{t},t\in\mathbb{R}) be a Rosenblatt process with self-similarity index H\in\left(\frac{1}{2},1\right) . For any m\geq 1 and n=(n_{1},...,n_{m})\in\mathbb{N}^{m} , there exists C>0 (depending possibly on m and n ) such that for any grid 0=t_{0}<t_{1}<...<t_{m} , the density p_{\pi} of the vector Z_{\pi} defined in ( 1 ) satisfies \left|\partial_{x}^{n}p_{\pi}(x)\right|\leq C\prod_{j=1}^{m}(t_{j}-t_{j-1})^{-H(1+n_{j})}e^{-c\frac{x_{j}}{(t_{j}-t_{j-1})^{H}}}, with c=c(m,q)>0 , for every x=(x_{1},...,x_{m}) such that x_{j}\geq 2(t_{j}-t_{j-1})^{H} , j=1,...,m. The proof of Theorem 2 is based on a Malliavin integration by parts formula expressing the partial derivatives of p_{\pi} (see relation ( 11 ) in Proposition 2 below) and on a precise control of the resulting terms using Malliavin calculus and tail estimates for random variables in Wiener chaos. The paper is organized as follows. Section 2 provides the necessary preliminaries on Malliavin calculus, Wiener chaos and the Rosenblatt process. Section 3 contains the proof of the smoothness of the density of Rosenblatt vectors. Section 4 is devoted to the analysis of the regularity of the partial derivatives of this density. 2 Preliminaries In this preliminary section, we introduce the basic notions of Malliavin calculus and Wiener chaos that will be used throughout the paper. We also present the Rosenblatt process and recall its main properties. 2.1 Wiener chaos, multiple stochastic integrals and Malliavin derivative We refer to the classical references [ 7 ] and [ 5 ] for a comprehensive exposition of Wiener chaos and Malliavin calculus. Let B=(B_{t},t\in\mathbb{R}) be a two-sided Brownian motion defined on a complete probability space (\Omega,\mathcal{F},P) . Set \mathfrak{H}=L^{2}(\mathbb{R}) , and for each h\in\mathfrak{H} define B(h)=\int_{\mathbb{R}}h(s)\,dB_{s}, the Wiener integral of h with respect to B . The family \{B(h):h\in\mathfrak{H}\} forms an isonormal Gaussian process, that is, a centered Gaussian family satisfying \mathbf{E}[B(h)B(g)]=\langle h,g\rangle_{\mathfrak{H}}\qquad\text{for all }h,g\in\mathfrak{H}. We now recall the definition of the Wiener chaos generated by this isonormal process. Let (H_{q})_{q\geq 0} denote the Hermite polynomials, defined recursively by H_{0}(x)=1,\qquad H_{1}(x)=x,\qquad H_{q+1}(x)=xH_{q}(x)-qH_{q-1}(x). For q\geq 1 , the q th Wiener chaos \mathcal{H}_{q} is the closed linear span in L^{2}(\Omega) of \{H_{q}(B(h)):\,h\in\mathfrak{H},\ \|h\|_{\mathfrak{H}}=1\}. In particular, \mathcal{H}_{1} is the Gaussian chaos generated by B(h) , while \mathcal{H}_{q} is non-Gaussian for every q\geq 2 . For q\geq 1 and h\in\mathfrak{H} with \|h\|_{\mathfrak{H}}=1 , we set I_{q}(h^{\otimes q}):=H_{q}(B(h)). If \|h\|\neq 1 , we instead define I_{q}(h^{\otimes q}):=\|h\|_{\mathfrak{H}}^{q}\,H_{q}\!\left(B\!\left(\frac{h}{\|h\|_{\mathfrak{H}}}\right)\right). By polarization, one extends the definition to elementary tensors h_{1}\otimes\cdots\otimes h_{q}\in\mathfrak{H}^{\otimes q} , and then by linearity and density to any symmetric kernel f\in\mathfrak{H}^{\odot q} . This yields a linear isometry I_{q}:\mathfrak{H}^{\odot q}\to L^{2}(\Omega), satisfying \mathbf{E}[I_{p}(f)I_{q}(g)]=\begin{cases}p!\,\langle f,g\rangle_{\mathfrak{H}^{\otimes p}},&p=q,\\[2.84526pt] 0,&p\neq q,\end{cases} for all f\in\mathfrak{H}^{\odot p} and g\in\mathfrak{H}^{\odot q} . One then obtains \mathcal{H}_{q}=\{I_{q}(f):f\in\mathfrak{H}^{\odot q}\},\qquad q\geq 1, that is, the q th Wiener chaos is precisely the image of \mathfrak{H}^{\odot q} under I_{q} . We now introduce the Malliavin derivative with respect to B . Let \mathcal{S} denote the class of smooth cylindrical random variables of the form F=f(B(g_{1}),\dots,B(g_{n})), where n\geq 1 ,

Read the complete text