I used COMSOL to simulate temperature change caused by 10^-4 fluctuation from the main DC power.

That is the temperature change in the substrate due to power change from Pin to Pin + Pin*10^-4.  Pin is 10mW.

This should give us the upper limit of the expected frequency noise.

The area under the plot of Temperature vs Depth is 1.9 * 10^-8 [meter*Kelvin].

thermal coefficient for substrate is 0.51 *10^-6

thus dL is (0.51*10^-6) * (1.9*10^-8) ~ 10^-14.

Use dL/L  = df/f ;  where L is the cavity length = 0.2035m,  f = c/ lambda.

df = 15 Hz. (effect from thermal expansion 1 mirror in the cavity only)