A stochastic generalization of renormalization-group transformation for continuous-time random walk processes is proposed. The renormalization consists in replacing the jump events from a randomly sized cluster by a single renormalized (i.e., overall) jump. The clustering of the jumps, followed by the corresponding transformation of the interjump time intervals, yields a new class of coupled continuous-time random walks which, applied to modeling of relaxation, lead to the general power-law properties usually fitted with the empirical Havriliak-Negami function.

• Received 4 February 2008

DOI:https://doi.org/10.1103/PhysRevE.78.011103