Atomic mechanisms of diffusional nucleation and growth and comparisons with their counterparts in shear transformations

Abstract

An integrated overview is presented of a viewpoint on the present understanding of nucleation and growth mechanisms in both diffusional and shear (martensitic) transformations. Special emphasis is placed on the roles played by the anisotropy of interphase boundary structure and energy and also upon elastic shear strain energy in both types of transformation. Even though diffusional nucleation is based on random statistical fluctuations, use of the time reversal principle shows that interfacial energy anisotropy leads to accurately reproducible orientation relationships and hence to partially or fully coherent boundaries, even when nucleation at a grain boundary requires an irrational orientation relationship to obtain. Since the fully coherent boundary areas separating most linear misfit compensating defects are wholly immobile during diffusional growth because of the improbability of moving substitutional atoms even temporarily into interstitial sites under conditions normally encountered, partially and fully coherent interphase boundaries should be immovable without the intervention of growth ledges. These ledges, however, must be heavily kinked and usually irregular in both spacing and path if they, too, are not to be similarly trapped. On the other hand, the large shear strain energy usually associated with martensite requires that its formation be initiated through a process which avoids the activation barrier associated with nucleation, perhaps by the Olson-Cohen matrix dislocation rearrangement mechanism. During growth, certain ledges on martensite plates serve as transformation dislocations and perform the crystal structure change (Bain strain). However, the terraces between these ledges in martensite (unlike those present during diffusional growth) are also mobile during non-fcc/hcp transformations; glissile dislocations on these terraces perform the lattice invariant deformation. Growth ledges operative during both diffusional and shear growth probably migrate by means of kink mechanisms. However, diffusional kinks appear to be nonconservative and sessile (and therefore resist immediate transmission of elastic shear strain energy), whereas those associated with martensitic growth must be conservative and glissile (and fully transmit such strain energy). The broad faces of both diffusionally and martensitically formed plates contain an invariant line, as emphasized by Dahmen and Weatherly. However, in the diffusional case, minimization of growth ledge formation kinetics seems to be the main role thereby played, whereas in martensitic growth, the main purpose of such an interface is to minimize elastic shear strain energy. The latter minimization requires that martensite forms as plates (or perhaps as laths) enclosed by a pair of invariant line-containing interfaces. During diffusional transformations, on the other hand, other interfaces at which growth ledge formation kinetics are not too much faster than those at the invariant line interface can also comprise a significant portion of the interfacial area, thereby leading to the formation of other, quite different morphologies, such as intragranular idiomorphs and grain boundary allotriomorphs. Critical problems remaining unsolved in diffusional transformations include calculation of critical nucleus shapes when the crystal structures of the two phases are significantly different, highly accurate calculation of the energies of the interphase boundaries thus formed, and direct observation of atomic scale kinks on the risers of growth ledges by means of a yet-to-be-invented three-dimensional (3-D) atomic-resolution form of transmission electron microscopy. Experimental identification and characterization of transformation dislocations and experimental testing of “nucleation” mechanisms are now of special importance in fundamental studies of martensitic transformations.