Simulation blow up at initial step

[Peng Deng]

Dear Dedalus users,

I am simulating the nondimensional three-layer quasi-geostrophic potential vorticity equation within a doubly periodic domain in dedalus3. 

 The solver could be built but the simulation blows up after the first step. I guess the problem stems from the initial condition. 

The initial condition I want to prescribe is the infinitesimal random noise, I use the 'fill_random' method as in the Rayleigh-Benard convection example in dedalus3 website. I could not find other references to enforce random perturbations in dedalus3.

The simulation script and the log file are attached.

Could someone give some suggestions?

Best,

Peng