I tried to separately mesh the geometry.
Only the small part (1/2 radius, 1/3 thickness) is extremely fine mesh, the rest is regular mesh, see fig1.
The result looks similar, with shorter time for calculation.
Also, the contact area between the mirror and the spacer
is modeled as a constant temperature (@ 35 C)boundary condition.
The temperature profile across the surface,
nd the temperature difference along the depth due to RIN is plotted below.
The area under the curve for T,Z is 6.1e-9 [meter.Kelvin]
Expansion coeff for SIO2 is 5.1e-7. Thus, dL = 3.1e-15m.
Use df/f = dL/L, L =0.2035m, we get df = 4.5 Hz.
As for thin coatings, I tried swept mesh, which creates prisms of quadrilateral surface.
However, when I ran the simulation, there are warnings about
ill-precondition from the size of the coatings.
After a talk with Greg's SURF student, he wrote a code similar to what I'm doing.
I'll try MATLAB code this time.