A variational multi-phase model for elastoplastic materials with microstructure evolution

A general model is formulated for elasto-plastic materials undergoing linear kinematic hardening to describe microstructure evolution associated with phase transformations. Using infinitesimal strain theory, the model is based on variational principles for inelastic materials. In our work we combine the so-called dissipation distance, which describes an immediate phase transition in time via an underlying probability matrix. In addition, the volume fractions of the newly emerging phases are represented by Young measures to obtain a time continuous microstructure evolution. The model is verified employing a two-dimensional benchmark test implemented by the Finite Element Method (FEM).

💡 Research Summary

**
The paper presents a comprehensive variational framework for elastoplastic materials that undergo linear kinematic hardening and experience phase transformations accompanied by microstructure evolution. Starting from classical single‑phase elastoplasticity, the authors introduce two thermodynamic potentials: the Helmholtz free energy Ψ, which depends on the total strain ε and internal variables (elastic strain, plastic strain, phase volume fractions), and the dissipation potential Δ, which captures irreversible processes. For a material capable of existing in k distinct phases, the free energy is defined as a volume‑fraction weighted average of the phase‑specific energies Ψ_i = ½(ε_i−ε_i^p):C_i:(ε_i−ε_i^p) + ½b_i‖ε_i^p‖² + c_i. The total free energy Ψ_tot = Σ_i λ_i Ψ_i respects the mass‑conservation constraint Σ_i λ_i = 1.

The dissipation potential is split into three contributions: (i) Δ_plast = Σ_i λ_i r_i‖·ε_i^p‖, representing rate‑independent plastic flow in each phase; (ii) Δ_trans = Σ_{i,j} g_{ij} D_i(ε_i^p, ε_j^p), describing instantaneous phase transitions via a “dissipation distance” D_i, which for the chosen linear dissipation yields D_i = r_i‖ε_j^p−ε_i^p‖; and (iii) Δ_reg, a viscous regularization term (½η_1 Σ_i λ_i‖·ε_i^p‖² + ½η_2 Σ_{i,j} g_{ij}²) that stabilizes the numerical scheme. The transition rates g_{ij} form a non‑negative, generally non‑symmetric matrix governing the stochastic evolution of phase fractions: λ̇_i = Σ_j (g_{ji}−g_{ij}).

To obtain a continuous description of the evolving microstructure, the authors employ Young measures. The internal variable distribution is represented as a discrete measure f(z)=Σ_i λ_i δ(z−z_i), where λ_i are the phase volume fractions and z_i the corresponding state variables. By minimizing the total free energy under the constraint Σ_i λ_i ε_i = ε, they derive effective macroscopic quantities: an effective stiffness tensor C_eff = (Σ_i λ_i C_i^{-1})^{-1} and an effective plastic strain ε_eff^p = Σ_i λ_i ε_i^p. The relaxed energy Ψ_rel(ε, ε_i^p, λ_i) = ½(ε−ε_eff^p):C_eff:(ε−ε_eff^p) + ½ Σ_i λ_i b_i‖ε_i^p‖² + Σ_i λ_i c_i then serves as the basis for the variational formulation.

The Lagrangian L = ȦΨ_rel + Δ_plast + Δ_trans + Δ_reg is minimized with respect to the rates of plastic strains (·ε_i^p) and transition rates (g_{ij}). Stationarity yields differential inclusions that lead to explicit evolution laws:

• Plastic flow: C_eff:(dev(ε)−ε_eff^p) − b_i ε_i^p − η_1·ε_i^p ∈ r_i sign(·ε_i^p), which can be recast into a familiar yield function ϕ_i = ‖C_eff:(dev(ε)−ε_eff^p) − b_i ε_i^p‖ − r_i.
• Phase transition: ∂Ψ_rel/∂λ_i − ∂Ψ_rel/∂λ_j − η_2 g_{ij} ∈ r_i‖ε_j^p−ε_i^p‖, giving a transition yield function ϕ_{ij} = ∂Ψ_rel/∂λ_i − ∂Ψ_rel/∂λ_j − r_i‖ε_j^p−ε_i^p‖.

The transition rates are computed as g_{ij} = η_2 max(ϕ_{ij},0), ensuring that at any instant only one direction of transition is active (g_{ij}=0 or g_{ji}=0). When a new phase appears (λ_i→0), its initial plastic strain ε_i^{p,ini} is chosen to maximize the transition driving force, i.e., ε_i^{p,ini}=arg max ϕ_{ij}.

Numerical implementation follows an incremental algorithm. Two benchmark problems validate the model. In a one‑dimensional cyclic tension test with three phases (initially λ_3=1), the model predicts immediate activation of λ_1 and λ_2 once the stress reaches the yield threshold, with the rate of increase controlled by the chosen material and viscosity parameters. The stress‑strain response exhibits distinct plateaus corresponding to phase activation, and the plastic strain evolution of the newly formed phases aligns with the theoretical predictions.

The second benchmark is a two‑dimensional compression of a square plate containing a circular hole. Using an unstructured mesh and the open‑source tools Gmsh and Julia‑Ferrite, the simulation captures stress concentration around the hole, where phase transformation initiates earlier than in the bulk. The evolving phase fractions modify the local stiffness, leading to pronounced localization of plastic deformation. Viscous parameters η_1 and η_2 regulate the smoothness of the transition and the convergence of the algorithm.

Overall, the paper successfully integrates variational principles, dissipation distance concepts, and Young measure relaxation to formulate a robust, thermodynamically consistent model for multi‑phase elastoplasticity with continuous microstructure evolution. The framework is versatile and can be extended to a wide range of materials—such as metallic glasses, amorphous silica, and complex composites—where phase transformations and microstructural patterning critically influence mechanical response. Future work may incorporate temperature effects, anisotropic hardening, or gradient‑enhanced regularizations to capture finer scale phenomena.