Searcharxiv⌕ Search

arXiv subjects

D. Pellegrino

Publications and source records attributed to D. Pellegrino.

At least 37 records · Page 2Linked to original sources

Peano curves on topological vector spaces

The starting point of this paper is the existence of Peano curves, that is, continuous surjections mapping the unit interval onto the unit square. From this fact one can easily construct of a continuous surjection from the real line $\mathbb{R}$ to any Euclidean space $\mathbb{R}^n$. The algebraic structure of the set of these functions (as well as extensions to spaces with higher dimensions) is analyzed from the modern point of view of lineability, and large algebras are found within the families studied. We also investigate topological vector spaces that are continuous image of the real line, providing an optimal lineability result.

math.GN↗

On the real polynomial Bohnenblust-Hille inequality

It was recently proved by Bayart et al. that the complex polynomial Bohnenblust--Hille inequality is subexponential. We show that, for real scalars, this does no longer hold. Moreover, we show that, if $D_{\mathbb{R},m}$ stands for the real Bohnenblust--Hille constant for $m$-homogeneous polynomials, then $\displaystyle\lim\sup_{m}D_{\mathbb{R},m}^{1/m}=2.$

math.FA↗

On nonlinear absolutely summing maps

It is shown that the ultimate version of the nonlinear Pietsch Domination Theorem remains true, even in a stronger presentation, if one of its hypotheses is removed. More precisely, we show that no trace of (sub)-homogeneity assumption is needed. Some consequences are also presented.

math.FA↗

On the multilinear Bohnenblust--Hille constants: complex versus real case

The results of this note arise a rupture between the behavior of the real and complex best known constants for the multilinear Bohnenblust--Hille inequality; in one side, for real scalars, we show that new upper bounds for the real Bohnenblust--Hille inequality (the best up to now) can be obtained via a somewhat "chaotic" combinatorial approach, while in the complex case the combinatorial approach giving the best known constants seems to be fully controlled. We believe that the understanding of this fact is a challenging problem that may shed some new light to the subject. As a byproduct of our results we present new estimates for the constants of the Bohnenblust--Hille inequality as well as new closed formulas.

math.FA↗

A geometric technique to generate lower estimates for the constants in the Bohnenblust--Hille inequalities

The Bohnenblust--Hille (polynomial and multilinear) inequalities were proved in 1931 in order to solve Bohr's absolute convergence problem on Dirichlet series. Since then these inequalities have found applications in various fields of analysis and analytic number theory. The control of the constants involved is crucial for applications, as it became evident in a recent outstanding paper of Defant, Frerick, Ortega-Cerdá, Ouna\"ıes and Seip published in 2011. The present work is devoted to obtain lower estimates for the constants appearing in the Bohnenblust--Hille polynomial inequality and some of its variants. The technique that we introduce for this task is a combination of the Krein--Milman Theorem with a description of the geometry of the unit ball of polynomial spaces on $\ell^2_\infty$.

math.FA↗

On the optimal multilinear Bohnenblust--Hille constants

The upper estimates for the optimal constants of the multilinear Bohnenblust--Hille inequality obtained in [J. Funct. Anal. 264 (2013), 429--463] are here improved to: {0.1cm} {enumerate} For real scalars: $K_{n}\leq\sqrt{2}(n-1)^{0.526322}$. For complex scalars: $K_{n}\leq\frac{2}{\sqrtπ}(n-1)^{0.304975}$.{enumerate} {0.1cm} \noindent We also obtain sharper estimates for higher values of $n$. For instance, \[ K_{n}<1.30379(n-1) ^{0.526322}\] for real scalars and $n>2^{8}$ and \[ K_{n}<0.99137(n-1) ^{0.304975}\] for complex scalars and $n > 2^{15}.$

math.FA↗

On very non-linear subsets of continuous functions

In this paper we continue the study initiated by Gurariy and Quarta in 2004 on the existence of linear spaces formed, up to the null vector, by continuous functions that attain the maximum only at one point. Inserting a topological flavor to the subject, we prove that results already known for functions defined on certain subsets of R are actually true for functions on quite general topological spaces. In the line of the original results of Gurariy and Quarta, we prove that, depending on the desired dimension, such subspaces may exist or not.

math.FA↗

On the Bohnenblust-Hille inequality and a variant of Littlewood's 4/3 inequality

The search for sharp constants for inequalities of the type Littlewood's 4/3 and Bohnenblust-Hille, besides its pure mathematical interest, has shown unexpected applications in many different fields, such as Analytic Number Theory, Quantum Information Theory, or (for instance) in deep results on the $n$-dimensional Bohr radius. The recent estimates obtained for the multilinear Bohnenblust-Hille inequality (in the case of real scalars) have been recently used, as a crucial step, by A. Montanaro in order to solve problems in the theory of quantum XOR games. Here, among other results, we obtain new upper bounds for the Bohnenblust-Hille constants in the case of complex scalars. For bilinear forms, we obtain the optimal constants of variants of Littlewood's 4/3 inequality (in the case of real scalars) when the exponent 4/3 is replaced by any $r\geq4/3.$ As a consequence of our estimates we show that the optimal constants for the real case are always strictly greater than the constants for the complex case.

math.FA↗

Subspaces of maximal dimension contained in $L_p(Ω) - \bigcup\limits_{q<p} L_q (Ω)$

Let $(Ω,Σ,μ)$ be a measure space and $1< p < +\infty$. In this paper we show that, under quite general conditions, the set $L_{p}(Ω) - \bigcup\limits_{1 \leq q < p}L_{q}(Ω)$ is maximal spaceable, that is, it contains (except for the null vector) a closed subspace $F$ of $L_{p}(Ω)$ such that $\dim(F) = \dim(L_{p}(Ω))$. We also show that if those conditions are not fulfilled, then even the larger set $L_p(Ω) - L_q(Ω)$, $1 \leq q < p$, may fail to be maximal spaceable. The aim of the results presented here is, among others, to generalize all the previous work (since the 1960's) related to the linear structure of the sets $L_{p}(Ω) - L_{q}(Ω)$ with $q < p$ and $L_{p}(Ω) - \bigcup\limits_{1 \leq q < p}L_{q}(Ω)$.

math.FA↗

When is the Haar measure a Pietsch measure for nonlinear mappings?

We show that, as in the linear case, the normalized Haar measure on a compact topological group $G$ is a Pietsch measure for nonlinear summing mappings on closed translation invariant subspaces of $C(G)$. This answers a question posed to the authors by J. Diestel. We also show that our result applies to several well-studied classes of nonlinear summing mappings. In the final section some problems are proposed.

math.FA↗

Estimates for the asymptotic behavior of the constants in the Bohnenblust--Hille inequality

A classical inequality due to H.F. Bohnenblust and E. Hille states that for every positive integer $n$ there is a constant $C_{n}>0$ so that $$(\sum\limits_{i_{1},...,i_{n}=1}^{N}|U(e_{i_{^{1}}},...,e_{i_{n}})|^{\frac{2n}{n+1}})^{\frac{n+1}{2n}}\leq C_{n}||U||$$ for every positive integer $N$ and every $n$-linear mapping $U:\ell_{\infty}^{N}\times...\times\ell_{\infty}^{N}\rightarrow\mathbb{C}$. The original estimates for those constants from Bohnenblust and Hille are $$C_{n}=n^{\frac{n+1}{2n}}2^{\frac{n-1}{2}}.$$ In this note we present explicit formulae for quite better constants, and calculate the asymptotic behavior of these estimates, completing recent results of the second and third authors. For example, we show that, if $C_{\mathbb{R},n}$ and $C_{\mathbb{C},n}$ denote (respectively) these estimates for the real and complex Bohnenblust--Hille inequality then, for every even positive integer $n$, $$\frac{C_{\mathbb{R},n}}{\sqrtπ} = \frac{C_{\mathbb{C},n}}{\sqrt{2}} = 2^{\frac{n+2}{8}}\cdot r_n$$ for a certain sequence $\{r_n\}$ which we estimate numerically to belong to the interval $(1,3/2)$ (the case $n$ odd is similar). Simultaneously, assuming that $\{r_n\}$ is in fact convergent, we also conclude that $$\displaystyle \lim_{n \rightarrow \infty} \frac{C_{\mathbb{R},n}}{C_{\mathbb{R},n-1}} = \displaystyle \lim_{n \rightarrow \infty} \frac{C_{\mathbb{C},n}}{C_{\mathbb{C},n-1}}= 2^{1/8}.$$

math.FA↗

$L_{p}[0,1] \setminus \bigcup\limits_{q>p} L_{q}[0,1]$ is spaceable for every $p>0$

In this short note we prove the result stated in the title; that is, for every $p>0$ there exists an infinite dimensional closed linear subspace of $L_{p}[0,1]$ every nonzero element of which does not belong to $\bigcup\limits_{q>p} L_{q}[0,1]$. This answers in the positive a question raised in 2010 by R. M. Aron on the spaceability of the above sets (for both, the Banach and quasi-Banach cases). We also complete some recent results from \cite{BDFP} for subsets of sequence spaces.

math.FA↗

Spaceability in Banach and quasi-Banach sequence spaces

Let $X$ be a Banach space. We prove that, for a large class of Banach or quasi-Banach spaces $E$ of $X$-valued sequences, the sets $E-\bigcup _{q\inΓ}\ell_{q}(X)$, where $Γ$ is any subset of $(0,\infty]$, and $E-c_{0}(X)$ contain closed infinite-dimensional subspaces of $E$ (if non-empty, of course). This result is applied in several particular cases and it is also shown that the same technique can be used to improve a result on the existence of spaces formed by norm-attaining linear operators.

math.FA↗

Dominated bilinear forms and 2-homogeneous polynomials

The main goal of this note is to establish a connection between the cotype of the Banach space X and the parameters r for which every 2-homogeneous polynomial on X is r-dominated. Let cotX be the infimum of the cotypes assumed by X and (cotX)* be its conjugate. The main result of this note asserts that if cotX > 2, then for every 1<= r < (cotX)* there exists a non-r-dominated 2-homogeneous polynomial on X.

math.FA↗

A note on lineability

In this note we answer a question concerning lineability of the set of non-absolutely summing operators.

math.FA↗