Comparative Analysis of Plasticity-based GND Density Estimation Methods in Crystal Plasticity Finite Element Models

In crystal plasticity finite element (CPFE) simulations, accurately quantifying geometrically necessary dislocations (GNDs) is critical for capturing strain gradients in polycrystals. We compare different methods for quantifying GNDs, all of which originate from the Nye tensor, which is computed as the curl of the plastic deformation gradient. The projection technique directly decomposes the Nye tensor onto individual screw and edge dislocation components to compute GNDs. This approach requires converting a nine-component Nye tensor into densities for a larger number of dislocation systems, a fundamentally underdetermined (non-unique) process, which is resolved using $L2$ minimization. In contrast, when employing CPFE analysis, one could directly compute dislocation densities on each slip system using shear gradients. Projection and slip gradient methods are compared with respect to their prediction of GNDs with changing grain size, strain, and grain neighborhoods, including multigrain junctions. Although these techniques match analytical GND densities for single slip, single crystal deformation, and are consistent with anticipated overall GND trends, we find that the GND densities from projection techniques are significantly lower than those predicted from CPFE-based slip gradients in polycrystals. A suggested improvement of only using the active dislocation systems in the projection technique almost entirely resolved this mismatch.

💡 Research Summary

This paper conducts a systematic comparison of three computational approaches for estimating geometrically necessary dislocation (GND) density within crystal plasticity finite element (CPFE) simulations. All methods originate from the Nye tensor, defined as the curl of the plastic deformation gradient (G = curl Fₚ). Because an FCC crystal possesses twelve slip systems (six edge and six screw) while the Nye tensor contains only nine independent components, mapping the tensor onto slip‑system densities is an underdetermined problem.

The first approach resolves the underdetermination by applying a Moore‑Penrose pseudoinverse and L₂‑norm minimization. The nine‑component Nye tensor Λ is expressed as Λ = A ρ_GND, where A encodes the Burgers vectors and line directions of all slip systems. Solving ρ_GND = Aᵀ(AAᵀ)⁻¹Λ yields a least‑squares distribution of GND across every slip system, regardless of whether that system is active in the current deformation.

The second approach, motivated by recent work, restricts the projection to only those slip systems that have experienced a non‑negligible amount of shear (γ > 10⁻⁸). In this “active‑system” formulation the matrix A is dynamically reduced, and the same pseudoinverse solution is applied. This prevents the artificial allocation of dislocation density to inactive systems.

The third approach bypasses projection entirely by using the slip‑gradient method. Starting from the small‑strain approximation Fₚ ≈ I + ∑γ_α m_α⊗n_α, the curl operation yields G = −∇×(∑γ_α m_α⊗n_α). For each slip system α, the relation ρ_α = (1/|b_α|)‖∇γ_α × n_α‖ is derived, providing a direct, physically transparent expression for the GND density on that system. The total GND is then the sum over all α.

Four numerical experiments validate the methods. (1) A 1 mm³ single crystal subjected to symmetric uniaxial tension (gradient‑free) and to a non‑uniform shear loading. The symmetric case yields zero GND for all methods, while the shear case produces non‑zero values that agree among methods. (2) A single‑slip beam (5 mm × 1 mm²) with an analytical solution ρ_GND = 7 812.5 mm⁻². Both the projection (L₂) and slip‑gradient approaches reproduce the analytical value within 0.5 % error. (3) Mesh‑convergence studies confirm that refinement reduces numerical noise and that both approaches converge to the same solution. (4) A polycrystalline RVE (5 µm cube, 64 grains, average equivalent sphere diameter 1.17 µm) is loaded in tension to 10 % strain. GND densities are extracted at 18 strain levels.

Key findings: In the polycrystal, the traditional L₂ projection systematically underestimates GND density by roughly 30 % compared with the slip‑gradient method. This discrepancy is traced to the inclusion of inactive slip systems in the projection, which dilutes the density assigned to active systems. When the projection is limited to active slip systems, the difference shrinks to less than 5 %, essentially eliminating the mismatch. All three methods capture the expected size‑effect trends: GND density increases as grain size decreases and as applied strain increases, consistent with Hall‑Petch scaling.

The study demonstrates that restricting the projection to active slip systems is both physically justified and computationally inexpensive, yielding GND estimates comparable to the more direct slip‑gradient approach. Moreover, the slip‑gradient formulation, being based on measurable shear gradients, offers a natural bridge to experimental techniques such as HR‑EBSD, T‑EBSD, and high‑energy diffraction microscopy. Future work should extend the comparison to large‑strain formulations, incorporate texture and anisotropic hardening, and explore multi‑slip interactions and temperature dependence.