I checked the cavity length sensitivity to horizontal acceleration ( normal to the beam line axis). Unlike the result from vertical acceleration, the cavity length did not change smoothly with the position.  For the optimum point(30 degree, 1.18 inch apart), the displacement sensitivity due to horizontal acceleration is about a factor of 2 larger than that of vertical acceleration.

 Since both horizontal(H) and vertical(V) seismic noise on the table are comparable [psl:xxx], I want to make sure that there will be no serious displacement noise due to acceleration on vertical direciton.

On the COMSOL model, I fixed only one side of the support to push against horizontal acceleration (see pic). As the support can only push against the cavity, not pull. It should make more sense to use this boundary condition.

For the effect from H acceleration, I varied the angle from 12 to 42 degree, and distance from 0.8 to 1.2 inch. The displacement did not change smoothly with the support positions. So I could not tell which way I should choose for the optimum support. However, the displacement seems to be around 2x10^-10 inch for most of the positions. [the unit on the y axis of the plot should be inch per acceleration].

For the optimum support (1.18 Inch, 30 degree) the senstivity is about 1x10^-10 inch/ (m/s^2).