Likely melt transport processes vary substantially depending on the geotherm. In the adiabatically convecting mantle, reaction between ascending melt and mantle minerals leads to an increase in melt mass. This creates additional porosity along crystal grain boundaries, locally increasing permeability and enhancing melt flux, which in turn increases porosity still more. On a local scale, formation of dissolution channels is more rapid than viscous compaction. The result is a coalescing network of porous channels which focus melt flow. Dunites within the mantle section of ophiolites probably represent relict dissolution channels formed in this way.

Within rising mantle undergoing decompression melting beneath spreading ridges, porosity generally increases with decreasing depth. As a result, downstream decreases in permeability are unlikely, so that development of magmatic overpressure and resulting hydrofracture are also unlikely. Similarly, depending on the pattern of solid mantle upwelling, shear stresses are likely to be small so that proposed melt focusing mechanisms resulting from shear stresses in excess of ~ 1 MPa are unlikely.

In the shallow mantle where rising melt begins to cool conductively, crystallization in pore space will reduce permeability. This is enhanced when reactions between melt and mantle minerals decrease melt mass. If melt flux is low and cooling is slow, then the result is diffuse porous flow. Even where an organized channel network is originally present, this process will produce a random permeability structure instead. If melt flux is large and cooling is fast, melt will accumulate beneath a permeability barrier, and melt overpressure may develop within a region extending for a “compaction length” below the barrier. Hydrofractures extending through the permeability barrier are likely to result. These most likely originate within pre-existing areas of relatively high porosity. Pyroxenite and gabbro dikes in the mantle section of ophiolites probably are formed in this way.