Canonical ensemble Monte Carlo simulation on phase separation of alloy systems

In Nickel-based alloys, the γ’-Ni3Al phase coherently precipitates from the supersaturated γ-Ni phase via the isothermal aging heat treatment. I think whether I could apply the cluster expansion combined with canonical ensemble Monte Carlo to simulate such process via icet and mchammer. In this way, I could track the fraction of γ’ phase, distributions of alloying solutes in two phases, the degree of order to study the mechanism of such dynamical evolution.
I have tried to apply CEMC via icet and mchammer to realize it. First, I enumerate the structures of Ni-Al systems and calculated total energies by VASP. Next, I set up a cluster expansion model by icet. Finally, I carry out canonical ensemble Monte Carlo simulations combined with constructed CE model with random initial configurations. I obtain the following results: (a) 0 step; (b) 1,000,000 steps; (c) 5,000,000 steps.

My questions are:

  1. Is it reasonable to apply above CEMC method to study the precipitation process of γ’ phase?
  2. The new version of icet could separate long-ranged strain contributions. However, can it study the evolution of systems with different incoherent phase interface.

Hi @peng_hu,

  1. The challenge with coherent phase coexistence as studied with cluster expansions is that if there is a siginificant difference in lattice parameter between the two phases, there will be strain. This strain is long ranged in the sense that it is large also far from the coherent interface. Since the cluster expansion typically has a cutoff of a few ångström, the model will not capture this strain except possibly close to the interface. The thermodynamics you probe will thus in some respects be more closely related to incoherent phase coexistence, except of course close to the interface. Whether your approach is still reasonable in your case is hard to say and up to you to figure out. It depends on what aspects you are interested in, how accurate the model needs to be, how much strain there is, and how important that strain is.
  2. The constituent strain module is primarily meant for studying coherent interfaces. It effectively makes the cluster expansion long ranged, such that the shortcomings alluded to above can be overcome. Here’s an example of how it can be used: [2110.07883] Quantitative predictions of thermodynamic hysteresis: Temperature-dependent character of the phase transition in Pd-H