We show that an isotropic random field on SU (2) is not necessarily isotropic as a random field on S 3 , although the two spaces can be identified. The ambiguity is due to the fact that the notion of isotropy on a group and on a sphere are different, the latter being much stronger. We show that any isotropic random field on S 3 is necessarily a superposition of uncorrelated random harmonic homogeneous polynomials, such that the one of degree d is necessarily a superposition of uncorrelated random spin weighted functions of every possible spin weight in the range {− d 2 , . . . , d 2 }, each of which is isotropic in the sense of SU (2). Moreover, for a random field of fixed degree, each spin weight appears with the same magnitude, in a sense to be specified.In addition we will give an overview of the theory of spin weighted functions and Wigner D-matrices, with the purpose of gathering together many different points of view and adding ours. As a byproduct of this survey we will prove some new properties of the Wigner matrices and a formula relating the operators ðð and the horizontal Laplacian of the Hopf fibration S 3 → S 2 , in the sense of [4].
Contents
MICHELE STECCONI
The Gaussian case 31References 321 There is some ambiguity in the literature, regarding the sign of s. We will justify our choice in Section 2.4.2 With this symbol we denote diffeomorphisms.3 If s ∈ Z, then T ⊗s is a true tensor power of the tangent bundle T S 2 = T ⊗1 and SO(3) is isomorphic to the frame orthonormal bundle of S 2 .