Found interesting reference on thermal conductivity of epoxy resin/AlN powder

http://cat.inist.fr/?aModele=afficheN&cpsidt=2640830
Titre du document / Document title

Thermal conductivity of epoxy resin filled with particulate aluminum nitride powderAuteur(s) / Author(s)

NAGAI Y. (1) ; LAI G.-C. (1) ; Affiliation(s) du ou des auteurs / Author(s) Affiliation(s)

(1) Research and Development Laboratory, Toyo Aluminium K. K., 4-1, Aioi-cho, Yao-shi, Osaka 581, JAPON
Résumé / Abstract

The thermal conductivity of an AlN- and an alumina-particulate-filled epoxy resin was investigated as a function of their volume content ratios. AlN composite exhibited thermal conductivity of 7.15 W/m.K at 68.5 vol% filler content but that of alumina composite was lower than 2.68 W/m.K even if the filler content was as much as 63.8 vol%. The thermal conductivity of the composites is strongly dependnt on that of the filler. In the AIN filler properties, the particle size of the filler used affected the thermal conductivity of composites, and composites with higher thermal conductivity could be obtained by using filler with large particle size. The thermal conductivity of composites is governed by the number of resin layers on the surface of an AlN particle. The composite thermal conductivities decreased with increasing the oxygen content of the filler. The thermal conductivity of an AlN particle depends on the oxygen content of the particle itself and the decrease of the composite thermal conductivity with the increase of the oxygen content of the filler is due to the increase of the aluminum oxide layer thickness produced on the surface of particles. In the thermal conductivity model, the measured thermal conductivity of AlN-ground-powder-filled composites disagreed with the ones calculated using Bruggeman's expression. However the measured values of alumina-spherical-powder-filled composites agreed with the calculated ones. It is considered that the results depended on the sphericity of the used powder.

Wow, I understood none of that article, other than O2 in the equation is bad. Im sure some of the modders here will find some of this info usefull

Seems a large particle size happens to have better thermal conductivity. Maybe they're proving my theory correct in that with larger particles may cause a thicker layer which technically is more insulative, but the larger particles provide a more direct path for heat flow since you're more likely to get a single layer sandwich of heat source-conducting particle-heatsink.

AlN is a good choice because it has a relatively high thermal conductivity of 175 W/mK but still being an electrical insulator. I believe as long as dielectric properties are not needed, the thermal conductivity of copper of almost 400 W/mK is better. Hence my preference for copper powder for bonding leds.

Uniform particle size is best, the ideal size is going to depend on how thick or thin you want to make your layer of thermal compound, I'm sure thermal epoxy manufacturers have done research on this. The thermal compound is just there to fill imperfections in the two surfaces you're joining so you want as thin a layer as possible.

Has anybody tried lapping their LEDs or heatsinks to get the thermal interface as smooth as possible and thus get best heat transfer?

Approx thermal conductivity for alumina is 33 W/m-K at 20 C (http://www.ceramics.nist.gov/srd/summary/scdaos.htm).

Advertised thermal conductivity for Arctic Alumina thermal compound is ">4 W/m-K". I don't see any advertised figure for the AA adhesive, but you'd think that the AA thermal compound would be an upper limit, and thus the thermal conductivity for the AA adhesive would be less than that for the AA compound. The paper that I found claimed a max 2.68 W/m-K for their alumina/epoxy mix "even if the filler content was as much as 63.8 vol%", which seems consistent with the above assumption.

This site says that the thermal conductivity of epoxy is 0.188 W/m-K:

http://www.efunda.com/materials/polymers/properties/polymer_datasheet.cfm?MajorID=epoxy&MinorID=8

If we assume a simple rule of mixtures for a 63.8-36.2 AA/epoxy mix, then we'd get a thermal conductivity of (>4*.638 + 0.188*.362) = >2.62 W/m-K, appearing remarkably consistent with the 2.68 W/m-K figure above.