Dual of 2D Fractional Fourier Transform Associated to It\^ O--Hermite Polynomials
Total Page:16
File Type:pdf, Size:1020Kb
DUAL OF 2D FRACTIONAL FOURIER TRANSFORM ASSOCIATED TO ITÔ–HERMITE POLYNOMIALS ABDELHADI BENAHMADI AND ALLAL GHANMI Dedicated to the memory of Professor Elhachmia Ait Benhaddou Abstract. A class of integral transforms, on the planar Gaussian Hilbert space with range in the weighted Bergman space on the bi-disk, is defined as the dual transforms of the 2d fractional Fourier transform associated with the Mehler function for Itô– Hermite polynomials. Some spectral properties of these transforms are investigated. Namely, we study their boundedness and identify their null spaces as well as their ranges. Such identification depends on the zeros set of Itô–Hermite polynomials. Moreover, the explicit expressions of their singular values are given and compactness and membership in p-Schatten class are studied. The relationship to specific fractional Hankel transforms is also established 1 Introduction The role played by the classical Mehler formula [14], +∞ tnH (x)H (y) 1 −t2(x2 + y2) + 2txy ! X n n = √ exp , (1.1) n 2 2 n=0 2 n! 1 − t 1 − t n x2 dn −x2 for the real Hermite polynomials Hn(x) := (−1) e dxn (e ), is well known in the literature [14, 18, 3, 12, 15, 17, 13, 8, 16]. Its complex analogues for Itô–Hermite ν polynomials Hm,n have been obtained in [19, 4, 10, 6] and have been employed in [6] ν to establish integral reproducing property for Hm,n by a like Fourier transform, and to provide a closed expression of the heat kernel for the Cauchy initial value problem for 2 2,ν −ν|z|2 a special magnetic Laplacian acting on the Hilbert space Lν(C) := L (C; e dλ). In [20], Zayed has used the one in (2.1) below to construct a non trivial 2d fractional Fourier transform Z ν ν −ν|ζ|2 Fu,vψ(ξ) = ψ(ζ)Ku,v(ζ; ξ)e dxdy; ζ = x + iy, (1.2) C ν whose eigenfunctions are the Itô–Hermite polynomials. Here Ku,v(ζ; ξ) stands for the kernel function ν ν n o arXiv:2005.08490v1 [math.CV] 18 May 2020 Kν (ζ; ξ) = exp −uv(|ζ|2 + |ξ|2) + uζξ + vζξ . (1.3) u,v π(1 − uv) 1 − uv In the present paper, we explore further applications in the context of integral 2 2 2 transforms and weighted Bergman spaces Bα,β(D ) on the bi-disk D = D × D; D = {z ∈ C, zz < 1}, defined as the Hilbert space of all analytic functions on D2 that are square integrable with respect to the measure 2 2 dµα,β(z, w) = ωα,β(|z| , |w| )dλ(z, w), (1.4) 2010 Mathematics Subject Classification. Primary 44A20; 30G35; 30H20 Secondary 47B38; 30D55. Key words and phrases. Weighted Bergman space on bi-disk; Itô–Hermite polynomials; Mehler formula; Singular values; p-Schatten class; Segal–Bargmann transform. 1 2 ABDELHADI BENAHMADI AND ALLAL GHANMI where the weight function is given by α β ωα,β(s, t) := (1 − s) (1 − t) ; α, β > −1, (1.5) and dλ denotes the standard Lebesgue measure. To this end we follow the scheme already applied in [7] to introduce and study the dual transforms of fractional Hankel transforms with ranges in weighted Bergman space on the disk. Mainly, we consider the family of integral transforms Z ν ν −ν|z|2 Rwf(u, v) = f(z)Ku,v(z; w)e dλ(z) (1.6) C on D2, labeled by ν > 0 and w ∈ C and seen as the dual transform of the 2d fractional ν ν Fourier transform in (1.2), Rwf(u, v) = Fu,vf(w). The aim in this paper concern identification of the null space and the range of the ν transforms Rw. We also study their boundedness and provide complete description of their compactness and membership in p-Schatten class. Our main results can be stated as follows ν 2 Theorem 1.1. The integral transform Rw is well defined and bounded from Lν(C) 2 2 into the weighted Bergman space Bα,β(D ) if and only if α > 0 and β > 0. The ν characterization of its null space Ker(Rw) depends on the zeros set of Itô–Hermite ν ν polynomials Hm,n. Namely, if Nw(H) = {(m, n); m, n = 0, 1, 2, ··· ; Hm,n(w, w) = 0}, ν then Ker(Rw) is a vector space spanned as ν ν Ker(Rw) = Span{Hm,n;(m, n) ∈ Nw(H)}. ν 2 2 2 Theorem 1.2. For α, β > 0, the operator Rw : Lν(C) −→ Bα,β(D ) is compact and its singular values are given by νπΓ(α + 1)Γ(β + 1) !1/2 sν,α,β(w) = |Hν (w, w)|. m,n νm+nΓ(α + m + 2)Γ(β + n + 2) m,n Moreover, it belongs to the p-Schatten class for every p > max(2/(α + 1); 2/(β + 1)). The proof of Theorem 1.1 is contained in Propositions 3.1, 3.2, 3.3 and 3.5 presented in Section 3, while the one of Theorem 1.2 is given in Section 4. The next section is devoted to some preliminaries concerning weighted Bergman space on the bi-disk and Itô–Hermite polynomials. We conclude the paper by discussing the close connection of ν Rw to the fractional Hankel transforms √ α/2 ! 2 2 2ν u Z ∞ 2ν uv −ν(x +uvy ) ν,α 1−uv Hu,v (f)(y) = xf(x)Iα xy e dr, (1.7) 1 − uv v 0 1 − uv where Iα denotes the modified Bessel function [1, p.222]. 2 Theorem 1.3. Let f ∈ Lν(C) and gk the associated Fourier coefficients. Then, ν,α ν Hu,v (gk), for varying integer k, are the fractional Fourier coefficients of Fu,vf, the fractional Fourier transform of f. 2 Preliminaries 2 2 For fixed reals α, β > −1, the weighted Bergman space Bα,β(D ) is a closed subspace 2 2 2 of the Hilbert space Lα,β(D ) := L (D × D; dµα,β) endowed with the scaler product Z hf, giα,β = f(z, w)g(z, w)dµα,β(z, w). D2 DUAL OF 2D FRFT ASSOCIATED TO ITÔ–HERMITE POLYNOMIALS 3 2 2 We denote by k·kα,β the associated norm. An orthonormal basis of Bα,β(D ) is given by α,β α,β −1/2 ϕm,n := γm,n em,n m n α,β where em,n(z, w) := z w and γm,n is its square norm given by π2Γ(α + 1)Γ(β + 1)m!n! γα,β := = ke k2 . m,n Γ(α + m + 2)Γ(β + n + 2) m,n α,β 2 2 Subsequently, the sequential characterization of Bα,β(D ) is given by ∞ ∞ 2 2 X X α,β 2 Bα,β(D ) = am;nem,n; γm,n|am;n| < ∞ . m,n=0 m,n=0 2 2 2 2 Accordingly, B0,0(D ) is identified to be the Hardy space on the bi-disk, while B−1,−1(D ) 2 2 2 is the classical Bergman space on D . The reproducing kernel of Bα,β(D ) is given by (α + 1)(β + 1) K ((u, v); (z, w)) = . α,β π2(1 − uz)α+2(1 − vw)β+2 2 2 Thus Bα,β(D ) is obtained as the Bergman projection (α + 1)(β + 1) Z ϕ(z, w) P (ϕ)(u, v) = dµα,β(z, w) π2 D2 (1 − uz)α+2(1 − vw)β+2 2 2 of Lα,β(D ). Another realization is by means of the unitary two-dimensional second Bargmann transform Z ! 1 α β sw + tz − (s + t) Bα,βϕ(z, w) = α+1 β+1 s t exp ϕ(s, t)dsdt (1 − z) (1 − w) R+2 (1 − z)(1 − w) 2 +2 α β −x−y acting on the Hilbert space L (R ; x y e dxdy). The kernel function of Bα,β ap- pears as the tensor product of two copies of the kernel function of the standard one- dimensional second Bargmann transform [2, p. 203]. However, it can be seen as the bilinear generating function involving the product of generalized Laguerre polynomials (α) (β) Lm (s)Ln (t). 2 2 In the sequel, we will provide interesting realization of specific subspaces of Bα,β(D ) by invoking the complex Mehler function 2 0 2 0 ! 0 1 −uv(ν|z| + ν |w| ) + ν uzw + νvzw Kν;ν (z, w) = exp , (2.1) u,v 1 − uv 1 − uv 0 ν with ν, ν > 0 and u, v ∈ D, associated to Itô–Hermite polynomials Hm,n defined on the complex plane C by [11, 9, 5] ∂m+n Hν (z, z) = (−1)m+neνzz e−νzz . (2.2) m,n ∂zm∂zn ν 2 To this end, let recall that the Hm,n form an orthogonal basis of Lν(C) and that the Mehler function in (2.1) can be expanded in terms of normalized Itô–Hermite polyno- mials 1/2 ν 0 ψν := Hν (2.3) m,n πνm+nm!n! m,n as [6] ∞ ν;ν0 X m n ν ν Ku,v (z, w) := u v ψm,n(z)ψm,n(w). (2.4) m,n=0 4 ABDELHADI BENAHMADI AND ALLAL GHANMI For ν = ν0 = 1, this is exactly the one announced by Wünsche [19] and proved later by Ismail in [10, Theorem 3.3] as a specific case of his Kibble–Slepian formula [10, Theorem 1.1]. ν 3 Basic properties of Rw We begin by observing that the kernel function in (1.3) reads in terms of the one in (2.1) as ν Kν (z, w) = Kν;ν(z, w).