Discussion Forum

Hello

I find solution, in my model there are no boundaries with fixed temperature condition so comsol didn't find a temperature distribution.

Bye
Luca

Hi

Well I notice too that often by describing my problem to someone else, or preparing it for the forum, I end up find the issue myself.

That is why I find it usefull for me to keep an eye on the questions and problems coming up on the forum.

Good luck
Ivar

Hi Ivar,

I have one question regarding a jolue heating problem. I'm simulating the 3D contact between a Carbon nanotube cylinder (I've modified already the suitable properties for it), and a Gold Substrate. It's a pretty simple model with joule heating interface, the current densities seem fairly logical. But the point is that the model is overheating, reaching temperatures as high as the sun surface, i.e, of the order of 10^6 (I'm sure you wouldn't buy a switch that heats like that). Obviously, this is completely nonsense. I don't have a clue what the problem is, cause basically I followed the same steps as in the busbar example; I defined Heat flux for all boundaries but the two electrodes; I defined them; Ground and E. Potential. I defined a heat transfer coefficient typical for air; 10 W/(m^2*K). Do you have any clue of what could be the problem?

Really I ran out of ideas. It should be a pretty easy problem to solve, but it hasn't .

Thanks

Hi

Well COMSOL allows you to plot most variables involved, so you could study the heat exchange values check that the integrated values are correct, as it's easy to get something wrong when handling densities. But take a look at the warnings and explanations in the KB (Knowledge Base) about flux calculations particularly in diffusion equation driven systems such as in HT.

So far HT has always given me correct values so I trust it, but indeed a few times I have found bizarre results, until I discover that I had missed a conversion value, entered an absolute value instead of a density etc,

One comment, for me convection in calm air is rather in the 2 to 5 (W/m^2/K), and if you scale down to MEMS sizes I'm not sure this value remains or should also be scaled with the dimensions. But this is not going to coool down your "sun"

--
Good luck
Ivar

Hi Ivar,

I checked the busbar example and made a simple example similar to the busbar. I´m aware that dimensions can change heat transference, but the temperatures I´m getting are by far illogical. Here I attach the CNT model I did, and the simpler model, both are reaching very high temperatures, and for the simple one, I have even changed the size of the structure and still get the same results.

Please Ivar, help me debug it, It´s a pretty simple model so It won´t take too much time for you to take a look at it.

All your help It´s greatly appreciated.

Hi

crazy temp or crazy power ?

I have two - three comments.

1) If you want convective cooling, then use the correct BC. With Heat Flux BC you type the "inflow power", so you are adding heat, not extracting any heat.

2) secondary issue probably not even used, but still: why strain reference temp "0" K while all other default values are at default 293 K ?

3) with a conductivity of >5E7 S/m and 5 V applied , and a few micron size items, suspect that the power densities are rather high (i find>2380 A/mm^2)

--
Good luck
Ivar

Hi Ivar,

I was going through the examples of comsol for joule heating and heat transfer and suddenly realized that they used very small voltages for the potential boundary condition. I changed the voltage to the values I´m using (about 5V) , ran the simulation and I got temperatures of the order of 10^5 at the busbar and MEMS Jolue heating example. So, I obviously deduced that there´s a great voltage dependence in the heating of the problem. So I checked the equations for heat transfer coefficient in natural convection and found out that of course, as the heat source of the problem is determined by the resistive losses, and resistive losses are directly proportional to V^2, If I rise the voltage from a mV to V, it means that either the Temperature difference or the heat transfer coefficient should be increasing by 10^6 as Newton´s equation states:

q=h*A*deltaT

, that without taking into account the effect of the size of the geometry. So, if I want reasonable temperature deltas I need to find the proper heat transfer coefficient, which should be obviously very high. Right now I{m looking for some papers about convection at nanoscales, so I can justify what I´m doing. What do you think about using very high heat transfer coefficients?

I appreciate any feedback for this topic.

Thanks

Hi

indeed, Joules law and its link to the resistivity is important and remains valid even in small scale.

One important issue with small scale is the surface to volume ration that is radically different from our macroscopic world. This has effects on certain physical effects and the energy balances/ratio change (but total energy calulations remains valid ;).

The volume scales as R^3, while the surface as R^2 so surface effects become predominant when scaling down. But i.e. convective cooling , that is a pure surface effect, also changes in the nano scale when we come down to atomic level. But this does not give me the impression one should change the value of a convection heat exchange, as it is already expressed as a "density".
It's worth to look up the literature, and find you how the different thermal effects scale, under "+/- normal" conditions, but does "normal" conditions apply also for very small scale ? (p, T certainly, but again other indirect scale effects ? I would have too look up too, but that is not on my path of today ... ;)

As a general rule you should always check your results by comparing them with some simple hand calculations, to ensure you have the right order of magnitude



--
Good luck
Ivar

Hi again Ivar,

I spoke with a friend of mine who knows a lot of nanochemistry and he told me that indeed things can get very hot at nanoscales. However, he compared data with some plasmon papers he's read and temperature never got that hot. So he suggested that as temperatures are so high, maybe I shouldn't understimate the heat transfered to ambient by radiation. So I put some radiation boundary conditions, but the model didn't converge, that's logical, because radiation is of the order of T^4, so if T was already high, the radiation is simply ridiculous.

I've been through all the Joule heating and MEMS examples, but they conveniently use very low voltages as inputs. For example in the MEMS Joule heating example, they do a parametric sweep between 3 mV and 10mV. Reaching a maximum of 792K. However if I change the voltage to 200mV, the temperatures rise up to 1.2*10^5 K. I did some hand calculations and maybe the temperature isn't that bad. However, all the papers I've checked while doing this project have used voltages as high as 48V and It seems ther isn't any supernova reported in any of them. I know there should be really high temperatures when you reach voltages as high as 70V and there must be superheating effects like the one in the photo attached, but It seems that they can live with voltages of 5-20 V.

Do you know someone with similar issues or anyone at comsol that knows how to dela with it?

Thanl you

The photo