Package heat dissipation with integrated carbon nanotube micro heat sink

Xiaojing Wang,Hongjun Liu,Jia Wang,Wen Zhang,Zongshuo Li
DOI: https://doi.org/10.1109/icept.2009.5270792
2009-08-01
Abstract:Micro-channel cooler is a very promising approach to meet the requirements of microelectronics package cooling. A lot of investigations about micro-channels have been undertaken in the past years. A silicon micro-channel can remove 790W/cm2 heat with a temperature rise of 71 degrees between the substrate and the coolant. The width and height of the silicon channel are 50um and 302um separately [1]. However, as the trends in the electronics industry moves towards higher packaging density, the high-pressure drop problem limits the performance of traditional silicon heat sink. Replacing the silicon fins with nanotube fins to enhance the thermal exchange rate between cooling liquid and substrate is one way to overcome this problem. Growing aligned nanotubes on the whole substrate is another one [2]. Carbon nanotubes (CNTs) are a new form of carbon which was discovered in 1991 by Iijima [3]. CNTs can be grown directly on the surface of silicon accurately according to predefined small-scale catalyst patterns normally transferred by standard photolithography processes. It was reported that nanotubes have an extremely high thermal conductivity over 3000W/(km) [4], so in the present work, the thermo physical properties of CNTs are obtained via molecular dynamics (MD) method. Fig. 1 shows one single-wall CNT simulation by LAMMPS. Several CNTs-fin heat sinks (1cm*1cm) with different fin width varied from 2um to 200um are tested (Fig. 2). Then the aligned CNTs heat sink is also simulated to compare with the long CNTs-fin model. The heat transfer coefficients and pressure drop of these structures are computed. Results indicate replacing the silicon-fin with CNTs-fin is a promising method and the long CNT-fin model shows better performance than the aligned CNTs for generating little pressure drop. Among the models, the best one is chosen for experiments in next work.
What problem does this paper attempt to address?