Mandelbrot cascades

Mandelbrot cascades are random measures on [0,1]2[0,1]^2 built by repeatedly splitting squares. Start with mass 11 on the unit square. At each level, split every square into 44 and multiply the mass in each new square by an independent copy of a random variable W≥0W \geq 0 with E(W)=1\mathbb{E}(W) = 1.

Here W=0W = 0 or W=2W = 2 each with probability pp, and W=1W = 1 otherwise. The case p=0.1p = 0.1 is the example in Chapter 5 of my PhD thesis. Darker squares carry more mass.

dim⁡Hsupp⁡μ=log⁡24(1−p)\dim_{\mathrm{H}} \operatorname{supp} \mu = \log_2 4(1-p)1.848
dim⁡Hμ=2−E(Wlog⁡W)log⁡2\dim_{\mathrm{H}} \mu = 2 - \frac{\mathbb{E}(W \log W)}{\log 2}1.800
dim⁡Fμ=2−log⁡2E(W2)\dim_{\mathrm{F}} \mu = 2 - \log_2 \mathbb{E}(W^2)1.737

Almost surely, conditioned on the cascade not dying out.

6 levels: 64 × 64 squares