Skip to content

Add systematic saddle-point search workflow - #26

Closed
ResPSnTY wants to merge 2 commits into
MagneticSimulation:masterfrom
ResPSnTY:feature/saddle-point-search
Closed

Add systematic saddle-point search workflow#26
ResPSnTY wants to merge 2 commits into
MagneticSimulation:masterfrom
ResPSnTY:feature/saddle-point-search

Conversation

@ResPSnTY

@ResPSnTY ResPSnTY commented Aug 1, 2026

Copy link
Copy Markdown

Summary

  • add an atomistic-first systematic saddle-point search workflow based on Hessian modes, minimum-mode following, and geodesic NEB
  • add a Spirit-aligned GNEB implementation with geodesic tangents/distances, climbing/falling/stationary images, VP and stereographic LBFGS-Atlas solvers, and optional GPU execution through the existing backend mechanism
  • add a concise skyrmion SPS example, public API documentation, CairoMakie path plotting, and CPU/CUDA formula-level tests
  • preserve the existing NEB implementation and all existing method signatures/default behavior

Output and units

  • per-transition energy.txt follows the legacy NEB table convention and stores absolute energies in joules
  • distance.txt stores dimensionless geodesic segment distances
  • transitions_energy.txt includes absolute joules plus relative joules/meV
  • plots show (E - E_min) / meV by default and label the axis Energy (meV), matching the current NEB documentation example
  • rejected searches are recorded in attempts.txt without candidate folders; only accepted transitions produce OVF/image folders

Validation

  • test/transition/test_transition.jl: CPU/CUDA Hessian 18/18, CPU/CUDA MMF/GNEB 38/38, automatic transition 17/17, plotting/axis units 5/5
  • test/transition/test_solvers.jl: CPU/CUDA solver and Spirit formula tests pass
  • existing test/test_neb.jl: 22/22 pass; src/neb/ has no diff
  • documentation build succeeds locally (apart from the pre-existing missing disk_llg.txt warning)
  • full Pkg.test() advances through the legacy suites but is blocked in the existing test/eigen/test_cubic.jl Enzyme dynamic-matrix compilation on Windows/Julia 1.12; the stack does not enter the transition code

References

@ww1g11
ww1g11 marked this pull request as ready for review August 2, 2026 06:38
@ResPSnTY ResPSnTY closed this by deleting the head repository Aug 2, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants