Compatibility of Martensitic Microstructures in Polycrystals

Technical Summary: Compatibility of Martensitic Microstructures in Polycrystals

Problem Statement
This paper investigates the kinematic compatibility of martensitic microstructures across grain boundaries in polycrystals. While single-crystal microstructures arise from the requirement of compatibility between symmetry-related martensite variants and the austenite phase, polycrystals introduce an additional constraint: the microstructure in each grain must be compatible with the microstructure in neighboring grains across their boundaries. The authors focus on the case where the high-temperature austenite possesses cubic symmetry and the material undergoes a transformation to lower-symmetry phases (specifically cubic-to-tetragonal and cubic-to-orthorhombic). The central question is whether zero-energy microstructures (those minimizing the free energy functional) can exist that are compatible across arbitrary grain boundaries, and if so, what complexity (order of laminates) is required.

Methodology
The study employs the nonlinear elasticity model for martensitic phase transformations. The free energy density $\psi$ is minimized on a set of energy wells $K = \bigcup_{i=1}^N SO(3)U^i$, where $U^i$ are the transformation strains.

1. Reduction to Planar Boundaries: The authors first demonstrate that the local compatibility problem at a grain boundary can be reduced, under regularity assumptions, to the problem of a planar interface separating two grains with a relative rotation $R \in SO(3)$ and a normal vector $m$.
2. Zero-Order Laminates (Constant Gradients): The compatibility of constant deformation gradients $A$ and $B$ on either side of the boundary is analyzed using the Hadamard jump condition ($A - B = a \otimes m$). This involves checking for rank-one connections between the rotated energy wells.
3. First-Order Laminates (Simple Laminates): The authors investigate whether replacing constant gradients with simple laminates (first-order microstructures) on both sides of the boundary allows for compatibility. This leads to a system of equations involving volume fractions, rotations, and amplitude vectors.
4. Symbolic Computation: For the cubic-to-tetragonal case, the compatibility condition for simple laminates is reduced to the vanishing of a product of Dixon resultants, $F(m, R)$. The authors utilize computer-assisted symbolic computation (MATLAB) to prove that this function is not identically zero.
5. Taylor Set Analysis: The paper defines the "Taylor set" $S(K) = \bigcap_{R \in SO(3)} K_{qc}R$, where $K_{qc}$ is the quasiconvexification of the energy wells. This set characterizes macroscopic deformation gradients that correspond to zero-energy microstructures regardless of grain geometry or rotation. The authors derive new upper bounds for this set using geometric properties, minors relations, and convexity arguments.

Key Contributions and Results

• Incompatibility of Constant Gradients: For cubic-to-tetragonal transformations, constant gradients can be compatible across a grain boundary even if the relative rotation $R$ is not in the cubic group $P_{24}$, provided two eigenvalues of the transformation strain are equal. However, for cubic-to-orthorhombic transformations (distinct eigenvalues), compatibility for constant gradients fails for a dense open set of relative rotations.
• Incompatibility of Simple Laminates: The primary result for cubic-to-tetragonal transformations (Theorem 8) is that the set of grain boundary normals $m$ and relative rotations $R$ allowing compatibility via simple laminates on both sides is a closed set of measure zero in the manifold $S^2 \times SO(3)$. This is proven by showing that the compatibility condition reduces to a system of nine scalar equations for eight unknowns, and the resulting algebraic condition (the vanishing of a product of resultants) is not satisfied generically. A similar, slightly weaker result (holding for almost all deformation parameters) is established for cubic-to-orthorhombic transformations (Theorem 10).
• Necessity of Higher-Order Laminates: The measure-zero result for simple laminates suggests that higher-order laminates (e.g., double laminates) are generically required to achieve compatibility at grain boundaries. The authors conjecture that compatibility can be achieved for generic $(m, R)$ if one side of the boundary is a simple laminate and the other is a double laminate (or more generally, if the sum of the orders of laminates on both sides is at least 5).
• Taylor Set Bounds: The paper establishes new upper bounds for the Taylor set $S(K)$ for cubic-to-tetragonal and cubic-to-orthorhombic transformations.
  • For cubic-to-tetragonal, if $D = \text{diag}(v_1, v_2, v_3) \in \Delta(K)$, then $v_1 v_2 v_3 = \eta_1^2 \eta_2$ and the diagonal entries satisfy specific inequalities involving $\eta_1$ and $\eta_2$ (Theorem 19).
  • These bounds imply a characterization of positive diagonal matrices in $K_{qc}$ (Theorem 18), recovering a result by Peigney [37] and providing a simple independent proof.
  • The bounds generalize previous results by Bhattacharya & Kohn [15, 16] from the geometrically linearized theory to the nonlinear setting.

Significance and Claims
The paper claims to explain the observation of complex, higher-order laminated microstructures in polycrystals (such as in BaTiO$_3$ and Ru$_{50}$Nb$_{50}$) as a necessary consequence of kinematic compatibility constraints. The results suggest that simple microstructures are insufficient to accommodate the geometric mismatch at grain boundaries for generic orientations.

The definition and analysis of the Taylor set provide a rigorous framework for identifying macroscopic deformations that are energetically admissible in any polycrystalline configuration, independent of specific grain shapes or rotations. The derived upper bounds refine the understanding of the "recoverable strain" in polycrystals.

The authors remain modest regarding the complexity of the required microstructures, noting that while the Taylor set guarantees the existence of zero-energy microstructures, the order of laminates required to realize them in practice may be unrealistically high. They suggest that future work could focus on developing approximate models using low-order laminates to capture correlations between grains and extending the analysis to include slip and applied tension.