Asymptotic uncorrelation for Mexican needlets (Q1043934): Difference between revisions

The author considers expression of the form \[ \beta_{t,x} =\sum_{n=1}^\infty\sum_{m=-n}^{n} f(t^2\lambda_n)a_{n,m} Y^m_n(x), \] where \(f\) is a Schwartz function on \([0,\infty)\) with \(f(0)=0\), for example \(f(s)= s^re^{-s}\), \(Y^m_l\) are the spherical harmonics on \(S^2\), \(\lambda_n=n(n+1)\) are the eigenvalues of the Laplacian associated with \(Y^m_n\) and \(a_{n,m}\) are considered to be random variables satisfying \(\overline{a}_{n,m}= (-1)^ma_{n,-m}\). In applications the coefficients \(a_{n,m}\) arise as inner products \(\langle F,Y^m_n\rangle\) where \(F\in L^2(S^2)\) has second order isotropy. The main result in the paper is an estimate which shows that the correlation between \(\beta_{t,x}\) and \(\beta_{t,t}\) goes to \(0\) as \(t\) tends to zero for fixed different points \(x\) and \(y\).