Square functions in the Hermite setting for functions with values in UMD spaces (Q466954): Difference between revisions

The authors consider the generalized Littlewood-Paley operator \(g_{\mathcal L, \sigma}\) associated with the Poisson semigroup for the Hermite operator \(\mathcal L(f)=\sum_{n\in \mathbb N^n} (2|k|+n)\langle f, h_k\rangle h_k\), where \(h_k\) is the orthonormal and complete system in \(L^2(\mathbb R^n)\) generated by the Hermite polynomials. As usual, \(-\mathcal L\) generates a semigroup of operators \((W_t^\mathcal L)_{t>0}\) given by \[ W_t^\mathcal L(f)=\sum_{n\in \mathbb N^n} e^{-(2|k|+n)t}\langle f, h_k\rangle h_k, \quad f\in L^2(\mathbb R^n), \] and one can consider the Poisson semigroup associated to \(\mathcal L\) given by means of the subordination formula \[ P_t^\mathcal L(f)=\frac{1}{\sqrt{4\pi}}\int_0^\infty u^{-3/2} e^{-t^2/4u}W_u^\mathcal L(f)\,du, \quad f\in L^2(\mathbb R^n). \] Hence the corresponding Littlewood-Paley \(g\)-function and the generalized version are defined by \[ g_{\mathcal L}(f)(x)=\left(\int_0^\infty |t \frac{\partial}{\partial t}P_t^{\mathcal L}(f)(x)|^2\frac{dt}{t}\right)^{1/2} \] and \[ g_{\mathcal L,\sigma}(f)(x)=\left(\int_0^\infty |t^{\sigma} \partial_t^{\sigma}P_t^{\mathcal L}(f)(x)|^2\frac{dt}{t}\right)^{1/2}, \] where the fractional derivative \(\partial_t^\sigma F(x,t)\) is defined in the sense of \textit{C. Segovia} and \textit{R. L. Wheeden} [J. Math. Mech. 19, 247--262 (1969; Zbl 0181.12403)]. One of the objectives of the paper is to analyze the behaviour of the operator \(g_{\mathcal L, \sigma}\) acting on vector-valued functions \(f\in L^p(\mathbb R^n, B)\) for a given Banach space \(B\). This is done modifying the operators, replacing the absolute-value by the norm, denoting now \[ G_{\mathcal L,\sigma, B}(f)(x)=\left(\int_0^\infty \|t^{\sigma} \partial_t^{\sigma}P_t^{\mathcal L}(f)(x)\|_B^2\frac{dt}{t}\right)^{1/2}, \quad f\in L^p(\mathbb R^n, B). \] These operators were already considered by \textit{T. Martínez} et al. [Adv. Math. 203, No. 2, 430--475 (2006; Zbl 1111.46008)], and more recently by \textit{J. L. Torrea} and \textit{C. Zhang} [Proc. R. Soc. Edinb., Sect. A, Math. 144, No. 3, 637--667 (2014; Zbl 1302.46008)] and the validity of classical results in the vector-valued setting holds only for Hilbert spaces. However, the authors consider another approach in the Hermite setting, inspired by the ideas in the papers by \textit{C. Kaiser} and \textit{L. Weis} [Stud. Math. 186, No. 2, 101--126 (2008; Zbl 1213.42145)] and by \textit{T. P. Hytönen} [Rev. Mat. Iberoam. 23, No. 3, 973--1009 (2007; Zbl 1213.42012)] which will lead to the equivalence of norms for the natural operators not only for Hilbert spaces, but for spaces with the UMD property. This is done by considering \(\gamma\)-radonifying operators. Recall that if \(H\) is a separable Hilbert space and \(B\) is a Banach space, an operator \(T:H\to B\) is \(\gamma\)-radonifying, denoted by \(T\in \gamma(H,B)\), whenever \(\sum_k \gamma_k T(e_k)\) converges in \(L^2(\Omega, B)\) for any orthonormal basis \((e_k)\), where \(\gamma_k\) are independent Gaussian random variables. Considering the Hilbert space \(H=L^2((0,\infty), \frac{dt}{t})\), the authors manage to define an operator \(\mathcal G_{\mathcal L+\alpha,\sigma, B}\) which maps vector-valued functions into functions with values in the space of \(\gamma\)-radonifying operators and show that, under the UMD assumption in \(B\), this operator is actually bounded from \(L^p(\mathbb R^n, B)\) into \(L^p(\mathbb R^n, \gamma(H,B))\) for a certain range of parameters. Also, weak-type \(L^1\) boundedness and boundedness on the Hardy space are proved.