Convergence of complex multiplicative cascades (Q990377)

Let \(u=u_1\cdots u_n\in\mathcal A^n\) be a word of length \(n\) obtained from the alphabet \(\mathcal A=\{0,\dots,b-1\}\ni u_i\) for some integer \(b\geq2\). Let \(W(u)=(W_0(u),\dots,W_{b-1}(u))\) be i.i.d. complex-valued random vectors for every word \(u\in\bigcup_{n\geq0}\mathcal A^n\). Now write \(u\in[0,1]\) as its \(b\)-adic expansion \(u=\sum_{k\geq1}u_kb^{-k}\) and let \(u|k-1=u_1\cdots u_{k-1}\). The object of study is the sequence of random functions \[ F_n(t)=\int_0^tb^n\prod_{k=1}^nW_{u_k}(u|k-1)\,du \] under the general condition \(\mathbb E[\sum_{i=0}^{b-1}W_i]=1\). The authors give sufficient conditions for almost sure uniform convergence to nontrivial statistically selfsimilar limits and investigate whether the limit is mono- or multifractal. Under strong assumptions on the moments of \(W\), the limit is shown to be monofractal. If the sufficient conditions for almost sure uniform convergence do not hold, the limit is either trivial or diverges almost surely. In this case, the authors provide natural normalizations such that, under suitable conditions, functional weak convergence to a Brownian motion in multifractal time takes place.