We describe the microstructure, shape and dynamics of growth of a droplet of martensite nucleating in a parent austenite during a solid-solid transformation, using a Landau theory written in terms of conventional affine, elastic deformations and {em
non-affine} degrees of freedom. Non-affineness, $phi$, serves as a source of strain incompatibility and screens long-ranged elastic interactions. It is produced wherever the local stress exceeds a threshold and anneals diffusively thereafter. A description in terms of $phi$ is inevitable when the separation between defect pairs, possibly generated during the course of the transformation, is small. Using a variational calculation, we find three types of stable solutions ({hv I}, {hv II} and {hv III}) for the structure of the product droplet depending on the scaled mobilities of $phi$ parallel and perpendicular to the parent-product interface and the stress threshold. In {hv I}, $phi$ is vanishingly small, {hv II} involves large $phi$ localized in regions of high stress within the parent-product interface and {hv III} where $phi$ completely wets the parent-product interface. While width $l$ and size $W$ of the twins follows $lproptosqrt{W}$ in solution {hv I}, this relation does not hold for {hv II} or {hv III}. We obtain a dynamical phase diagram featuring these solutions and argue that they represent specific microstructures such as twinned or dislocated martensites.
We study the nucleation dynamics of a model solid state transformation and the criterion for microstructure selection using a molecular dynamics (MD) simulation. Our simulations show a range of microstructures depending on the depth of quench. We clo
sely follow the dynamics of the solid and find that transient {em non-affine zones} (NAZ) are created at and evolve with the rapidly moving transformation front. The dynamics of these plastic regions determines the selection of microstructure. We formulate an {it elastoplastic model} which couples the elastic strain to the non-affine deformation, and recover all the qualitative features of the MD simulation. Using this model, we construct a dynamical phase diagram for microstructure selection, in addition to making definite testable predictions.
