“…In the exceptional case λ 2 = 0 the decrease is not exponential (G(ζ * 2 ) = 0) but a power law; one can check that for a Gaussian probability distribution of ζ 2 , P G ∝ √ te −ζ 2 2 t/2D , the statistical moments of f are proportional to 1/ √ t for any n. The values of S (−λ 2 ), S(−λ 2 ), as well as the whole shape of S, are determined by the statistics of velocity gradients. One can show (by means of the techniques developed in [17,20,21]) that for isotropic A ij (t), the possible value of S (−λ 2 ) is restricted by the boundaries |S (−λ 2 )| < 3/2. Thus, for these processes saturation of f n happens already at n < 3/2, so for all integer n ≥ 2 the n-order moments decrease with the same exponent.…”