Convection Performance of Nanofluids for Electronics Cooling

Modern microelectronic systems generate a large amount of heat which must be transferred out of the system without excessive temperature rise. Conventional forced air convection and microchannel cooling plates have reached their performance limits. Nanofluids are proposed as an innovative way to solve the problem. A nanofluid is nanoscale solid particles dispersed in a traditional heat transfer liquid. Some studies show an anomalous increase in the thermal conductivity for stationary nanofluids. However, there are only few previous studies on the convection heat transfer rate and viscosity of nanofluids. Both convection and stationary measurements of the thermal conductivity are widely scattered, so it is not known if nanofluids will be useful in electronic cooling applications. We characterized the particle size distribution and conducted convective heat transfer experiments for both oxide and CNT nanofluids. The oxide nanofluids contain spherical oxide nanoparticles such as Al2O3, CuO and ZnO in deionized water and CNT nanofluids contain multiwalled carbon nanotubes in deionized water. The nanofluids were prepared by ultrasonicating the mixture. Dynamic Light Scattering (DLS) system and Scanning Electron Microscopy (SEM) images were employed for the characterization of the nanofluids. The DLS system measured the mean particle diameter of the spherical nanoparticles. From the DLS result, we found numerous oxide nanoparticles were agglomerated in the nanofluids. The agglomerated particles were clearly observed in the SEM images. Various efforts were made to separate the agglomerated nanoparticles including long sonication times and centrifuging to remove the largest particles. The pH of the nanofluids was widely varied and surfactants were added. Unfortunately, none of these efforts were effective in fully deagglomerating the oxide nanoparticles. The convection experiment was conducted with a 200mum hydraulic diameter MEMs fabricated microchannel. Heaters and temperature sensors were distributed along the bottom of the micro-channel. More comprehensive convection data were obtained with a 500mum diameter stainless steel tube. The tube wall was heated electrically producing a constant heat flux boundary condition and an infrared camera was used to measure the outside tube wall temperature. A full numerical conjugate analysis inferred the effective thermal conductivity increase of the nanofluids. The results showed that there is no anomalous thermal conductivity increase for the oxide nanofluids, while there was a large increase for a certain CNT nanofluids. The traditional effective medium theory failed to predict the nanofluid thermal conductivity. The viscosity data were acquired by measuring the pressure drop in the tube. The viscosity of nanofluids increased much more rapidly than predicted by Batchelor's viscosity model. A simple analysis showed that CNT nanofluids would be effective in a parallel channels cooling application. However, the viscosity increased so rapidly with increasing particle loading that volume percentages of CNTs are limited to less than 0.2% in practical systems.