Power calculation for experiment with self-selection into information treatment

Question

I need help with the calculation of the minimum detectable difference between two within-group randomized treatments. In our experimental setup, individuals can self-select