Discussion Forum

Hi all, I'm pretty new to COMSOL, i'm trying to simulate the thermal stress response of a small Si [100] chip cooled down to 100mK, with a pulse heat source on a certain point/domain.

The model is as follows: the chip is 8x8mm2, 300microns thick, floating, connected to the thermal bath (kept at 100mK) by a certain number of bridges. The whole system is Si based only, that is because I want to cut the chip with its bridges directly on a silicon wafer. The number of bridges and their shape depends on the thermalization time of the chip, wich is fixed in around hundreds of microseconds.
The heat pulse I use is a step or rectangular function of 1ps, with a total power estimated of 10mW.

I made two different simulations: a 2D and a 3D version of the same problem. I'm encountering problems switching from one to the other, but that's a secondary problem, since I am first interested in understanding the thermal behaviour in a simpler way, so I will tell you of the 2D simulation.

Long story short, my biggest problem is to plot the whole time interval of my study, keeping a good resolution of the temperature rise. I put 2 probes: a point one to register the temperature rise of the point heat source, and a domain one to keep track of the thermalization of the chip. Since my pulse is 1ps wide and the complete thermalization should happens in hundreds of microseconds, I need to make a time dependent study with a good resolution over a wide range. COMSOL can't make a 1e-13 step simulation starting from 0 to 1e-3 (at least) as it would be a very large number of points (or at least it tells me this), so I tried to give a coarse resolution: 1e-6 seconds step from 0 to 1e-3 seconds. This simulate my behaviour quite good (I think), since I can see on my plot a pulse rise and decay and a bulk thermalization. Unfortunately the pulse rise temperature is not well reconstructed because of the bad time resolution (this means that the point rise temperature is higher as expected), probably due to a "bin" superposition in the very first steps.
I then tried to give the steps manually, changing the time step analysis after some orders of magnitude (i.e. 1e-13 step from 0 to 1e-12, then 1e-12 step until 1e-11 and so on). This works as long as I'm around my pulse, that is it seems to reconstruct my pulse shape and the relative temperature rise in a good way, but as I try to go further and analyse longer times the simulation goes bad, changing the initial temperature value of my pulse (instead of starting from 0.1K it starts higher).

Have you got any idea or suggestion on how to make my simulation works properly?
Thanks in advance and sorry for the wall text, hope to have been clear.
Dario