429 1963;Hill 1963). The Hashin-Shtrikman bounds are based on the assumption that each phase in the composite is in its minimum energy state and is described by a positive defi nite energy function. Therefore, this approach does not apply to materials in the vicinity of phase transformations (for elastic instabilities, see Salje 1993). In addition, more specifi c schemes have been developed. The shear-lag approach has seen a multitude of versions, which are all best justifi ed for fi brous inclusions rather than in case of rocks where the individual mineral grains have
LETTER ABSTRACTThe elastic properties of ceramics and a mix of minerals can be calculated in a simple way including some properties of the interfaces between grains. While the Hashin Shtrikman averaging method has bounds, which are strictly correct for many realistic physical conditions, one fi nds empirical data which lie outside these bonds. One possible interpretation is that the scaling of the volume proportions in mineral assemblies is not truly represented by the nominal volume proportion f of each phase. It is argued that interfacial effects scale with f(1-f). In the case of certain assemblies, it is shown that the scaling is entirely with f 2 rather than f. For intermediate cases, a more realistic scaling replaces f in the relevant averaging schemes by f(1-S) + S f 2 where S weighs the effect of the non-linearities in the volume expansion of the averaging schemes.