Growth by shear and by diffusional processes, both taking place predominantly by means of ledge mechanisms, are reviewed for the purpose of distinguishing critically between them at the atomic, microscopic, and macroscopic levels. At the atomic level, diffusional growth is described as individual, poorly coordinated, thermally activated jumps occurring in the manner of biased random walk, whereas growth by shear is taken to be tightly coordinated "glide" of atoms to sites in the product phase which are "predestined" to within the radius of a shuffle. Obedience to the invariant plane strain (IPS) surface relief effect and the transformation crystallography prescribed by the phenomenological theory of martensite is shown to be an unsatisfactory means of distinguishing between these two fundamentally different atomic growth mechanisms. In substitutional alloys, continuous differences in composition and in long-range order (LRO) from the earliest stages of growth onward are concluded to be the most useful phenomenological approach to achieving differentiation. At a more fundamental level, however, the details of interphase boundary structure are the primary determinant of the operative mechanism (when the driving force for growth is sufficient to permit either to occur). In the presence of a stacking sequence change across the boundary, terraces of ledges are immobile irrespective of their structural details during diffusional growth. Kinks on the risers of superledges are probably the primary sites for diffusional transfer of atoms across interphase boundaries. In martensitic transformations, on the other hand, terraces containing edge dislocations in glide orientation or pure screw dislocations are mobile and accomplish the lattice invariant deformation (LID), though probably only after being overrun by a transformation dislocation. Risers associated with transformation dislocations are also mobile and cause the crystal structure change during growth by shear. The successes achieved by the invariant line (IL) component of the phenomenological theory of martensite in predicting precipitate needle growth directions and precipitate plate habit planes (Dahmen and co-workers) are here ascribed to the rate of ledge formation usually being a minimum at an interface containing an IL, primarily because nuclei formed sympathetically at this boundary orientation are likely to have the highest edge energies. Since martensite plate broad faces also contain the IL, the ability of the phenomenological theory to predict the habit plane and the orientation relationships of both precipitate and martensite plates is no longer surprising. The IPS relief effect at a free surface can be generated by precipitate plates when growth ledges are generated predominantly on only one broad face and only one of several crystallographically equivalent Burgers vectors of growth ledges is operative. Both pReferences probably result from larger reductions in transformation strain energy for the particular geometry with which a given plate intercepts the free surface. Precipitate morphology often differs significantly from that of martensite even if precipitates are plate-shaped and can readily differ very greatly. Whereas martensite morphology is determined by the need to minimize shear strain energy, that of precipitates derives from the more flexible base of the interphase boundary orientation-dependence of the reciprocal of the average intergrowth ledge spacing, as modified by both the orientation-dependence of interkink spacing on growth ledge risers and the spacing/ height ratio dependence of diffusion field overlap upon growth kinetics.
ASJC Scopus subject areas