The more surface area, the more heat it would dissipate.
I have been joining the project on Cislunar Explorers group, a 2 3U CubeSats, responsible for the thermal Finite Element Analysis on the CubeSats. And I happened to watch the video (https://www.youtube.com/watch?v=E40PO_n8BpU&t=185s), I was inspired to do a simple Ansys simulation on the heat dissipation. In this video, Dr. J Class explains that the TSMC is going to use Water Cooling method to cool down the thermal effect of semiconductor chips.
According to the video, TSMC uses 12 different types of material in a single chip (TSMC SolC). Regardless of pure physical heat conduction, or air-cooled, water-cooled, or liquid nitrogen heat dissipation solutions, the heat-conducting medium contacts the packaging layer (die) in the chip and can only dissipate heat to the direct contact surface. This method will also put pressure on the top layer of the chip, because the heat of the entire chip has to be conducted from the silicon chip to the internal thermally conductive material, and then to Die, and finally concentrated through the thermally conductive medium to the radiator to dissipate. In the latest conference, TSMC demonstrated research on on-chip water cooling as a new heat dissipation solution involving the direct integration of water channels into the chip design, which excites many computer users fantasizing the micro conduit’s existence on chips.
Let’s review the three types of heat transfer: conduction, convection and radiation.
Conduction is the process by heat energy is transmitted through collisions between neighboring atoms or molecules.
Convection is the transfer of energy through fluid or air.
Radiation is the emission or transmission of energy in the form of waves or particles through space or through a material medium. The equation for calculating is through Stefan-Boltzmann law.
As can be seen from the above equation, convection is influenced by the exposed surface area, therefore in this article I would like to see the thermal dissipation effect of fins structure as they are widely used in semiconductor.
Since there are many layers in the chips, I simplified the structure to a single layer which I use the material as aluminum alloy. While the heat dissipation effect might be reduced, we can still expect to see some heat dissipation after the simulation.
After the engineering data, we have to plot the geometry of the structure in the SpaceClaim.
In Ansys Mechanical, I set the temperature as ambient temperature (22 degree Celsius) and the heat flux from logic chip to the structure as 90000 W/cm2. The contact type would be bonded.
3. Result and conclusion
After the initial simulation, the maximum and minimum are both on the top of the chips and for the single layer and fins structure they share the same temperature.
While we expect that the temperature after heat flux of the two structures should be different, in this simulation the value for heat flux is too large and, comparatively, the structure for both are too small (125 x 70 x 9.5 mm). After the heat flux exert on the structure whole upper structure experience the same effect and does not cause significant result. Therefore, I’ll put together article two to observe more on this effect!
For the following articles, I will also another structure that includes the water in between the fins structure as comparison.
Hope you enjoy reading this article!
- The heat flux of the chips in this article “Cooling Limits for High Heat Flux Chips in Servers” (https://nanoheat.stanford.edu/projects/cooling-limits-high-heat-flux-chips-servers), it set the heat flux as 1000 W/cm2. But in this simulation we would like a exaggerated result, therefore we times 90 on it.