Preprint arXiv:2611.23682 · November 2026
Axion: A differentiable, GPU-native finite element framework, demonstrated on phase-field fracture
Axion (Autodifferentiable, XLA-compiled, Implicit, On-device Numerics) is a finite element framework written in JAX. The element kernels, Newton and Krylov iterations, bound-constrained updates, staggered coupling of fields and the loop over load increments are expressed as a single program that is just-in-time compiled and runs on the GPU without host intervention.
Consistent tangents come from forward-mode automatic differentiation of the element residual, and derivatives of the solution with respect to model parameters come from differentiating the same program. Operators are applied element by element without assembling a sparse matrix, octree refinement lives in fixed-size arrays, and a cell-based domain decomposition runs the same program on several GPUs. Phase-field fracture is the first module.
- 37.9×faster than FEniCSx, one GPU vs one CPU (2D tension benchmark)
- 92% / 96%strong / weak scaling on four GPUs
- 96M DOFconcrete gravity dam on a single GPU





