Convergence of the Weierstrass-Mandelbrot process to fractional Brownian motion (Q2709706)

Let \(f_t\), \(t\in R\), be complex-valued functions defined on the real line, and let \((\xi_n,\eta_n)_{n\in Z}\) be two-dimensional random vectors. Motivated by Mandelbrot's idea that Weierstrass' classical nowhere differentiable continuous function can be modified and randomized to approximate fractional Brownian motion, the authors introduce the complex stochastic processes \(\sum_{n=-\infty}^{\infty}f_t(n/a) (\xi_n+i\eta_n)/\sqrt{a}\), \(t\in R\), and give several sufficient conditions on the functions \(f_t\) and the random vectors \((\xi_n,\eta_n)\) for their convergence in distribution, as \(a\to\infty\), to the stochastic process \(\int_Rf_t(u)(dB^1(u)+idB^2(u))\), \(t\in R\), where \(B^1(u), u\in R\), and \(B^2(u)\), \(u\in R\), are two independent standard Brownian motions. Several dependence structures for the \((\xi_n,\eta_n)\) sequence are considered (independence, strongly mixing sequences, martingale differences), and `convergence in distribution' means either `convergence of all finite-dimensional distributions' or `weak convergence in the function space \((C[0,\infty))\)'.