Discussion Forum

Hi

what about standard harmonic domain sweep ?
if you want to hit any resonances (these must be low, thermal diffusion is a slow process, even in nano scale).
And do not forget to use "include inertial terms" in your main physics. This is ON by default, but it is often worth to turn it off for the first runs to ensure the model behaves, particularly if the conversion is slow and you are looking for optimal BC settings

--
Good luck
Ivar

Dear Mr. Kjelberg ,

Thank you for your time. I've checked "include inertial terms". Indeed it was quasi- static in the physics I'd been using.
According to your advice I've tried using a standard frequency- domain study. But so far I have no luck with that.
The displacement of the actuator is still zero and temperature is still temperature of the cantilever anchor (sink) over a frequency range.

A 1st order resonance frequency (f0) of the mechanical structure is about 8kHz. According to some articles I've found recently, the driving voltage frequency is supposed to be f0/2 or f0/3 to hit the resonance.

I'm really stuck and don't understand what's wrong.
Looking forward for your response.

And thanks again,
Olhas

Hi

my advice start with "quasi static", then once the model runs, adjust your initial conditions to something more reasonable (closer to the solution you have), Then only add the inertial terms and get the solver running again (often you need to use shorter time steps, at least until the transients settle, or you need to smoothen better any BC step or Dirac function.

I'm not used to think "oscillations" in thermal, as by essence a diffusion equation cannot generate "waves" that bounces arounds as for EM. But it's true that you can generate dying waves over a few times the heat diffusion length.

This is indeed more pronounced in MEMS and small scale devices than in the our more ususal and macroscopic world, so its probably posible, at least you should carefully optimise your shape and material choices to get a wave propagating far enough, and for that COMSOL shold be of great help.
8 kHz is a high frequency, the period is some 0.125 msec, the heat diffusivita alpha = k/rho/Cp of Aluminium is some 64 mm^2/s, steel only some 4 mm^2/s (see for example en.wikipedia.org/wiki/Thermal_diffusivity) so for 0.125 msec the decay length is some 3*sqrt(alpha*Dt) or about 250 um for Aluminium, slightly more form Si and less than 1/10th for other materials such as steel or SiO2. As you talk about Joule heating you get a rapid bulk T distribution, so this distance is probably a transverse thickness, but still small scale. For any distance much longer your "wave will be smeared out and of very low amplitude. By the way alpha of water at 25°C is only about 0.143 mm^2/s that is why it takes so long to heat up water (or only some 13 um at 8 kHz, as penetration depth)

--
Good luck
Ivar

Dear Mr. Kjelberg ,

Thank you for your time. I've checked "include inertial terms". Indeed it was quasi- static in the physics I'd been using.
According to your advice I've tried using a standard frequency- domain study. But so far I have no luck with that.
The displacement of the actuator is still zero and temperature is still temperature of the cantilever anchor (sink) over a frequency range.

A 1st order resonance frequency (f0) of the mechanical structure is about 8kHz. According to some articles I've found recently, the driving voltage frequency is supposed to be f0/2 or f0/3 to hit the resonance.

I'm really stuck and don't understand what's wrong.
Looking forward for your response.

And thanks again,
Olhas

Hi Olhas

I also met a similar problem. It seems we can only use transient solver. Have you made any progress on this simulation? Looking forward for your response.

James

Hey James,

Sorry it took me a while to response.
I used the 4.2a version and according to the documentation a frequency- domain analysis was not supported yet for that physics. So I ended up doing just a transient analysis in Comsol and using another software for a frequency domain one.

It would be probably worth to check the 4.3 version.

Please let me know f you have any luck with that or have another questions.

Regards,
Olzhas

Hi Olzhas,

I'm trying to build similar model of what you did in COMSOL but it is hard for me to simulate since this is my first time.

Can I have your sample model and ask some questions?

My email is yjkoh0307@gmail.com

Thanks

Hi,

I don't think I can share the model I've used.
Yeah, shoot the questions, I'll see how I could help.

Cheers,
Olzhas