The simulation setup looks good, but the memory requirements to run the simulation are very high. I would recommend moving up the x min boundary of the simulation region to reduce the volume of the simulation to help reduce memory requirements.
As well, because the structure looks like it is symmetric, you could also apply symmetry in the boundary conditions in the x and y directions to further reduce memory requirements:
You could also consider using the mesh override region just over the smaller structures, so that a coarser mesh can be used in the substrate and in the air above the structure which would also reduce memory requirements.
If you want to simulate the results for the structure under stress, then the physical deformation of the structure, and the change in refractive index due to applies stress are two effects that could be included in the simulation. However, depending on the particular case, you may not need to include the effects. For example, the stress and strain example here assumes that the change in the shape due to bending the structure is very small, so the deformation of the shape of the structure is not included in the simulation:
If the change in refractive index of the ZnO is small due to the applied stress, then it’s possible that you could exclude this effect. Unfortunately I don’t know the details of the calculation of the change in refractive index due to applied stress for ZnO with Von Mises stress, so it might require looking into it in more detail. The equations from the stress and strain example in the Knowledge base are for the particular case where the structure is bent with a specified bending radius R, so the equations in your case may be different, but the main idea of using the n,k import to import the refractive index over space for the structure with applied stress is the same.