pyiron / pyiron_workflow_atomistics / 25965430370
90%

DEFAULT BRANCH: main
Ran
Jobs 1
Files 55
Run time 1min
Badge
Embed ▾
README BADGES
x

If you need to use a raster PNG badge, change the '.svg' to '.png' in the link

Markdown

Textile

RDoc

HTML

Rst
coverage: 90.079%. Remained the same
25965430370

push

github

web-flow 
docs(notebooks): overlay EMT/EAM/GRACE/MACE κ(T) on FCC Al (#59)

* docs(notebooks): overlay EMT/EAM/GRACE/MACE κ(T) on FCC Al

Replaces the single-Cu/EMT phonon thermal conductivity demo with an
FCC-Al sweep across four force engines. Each calculator runs in its own
conda env via a small subprocess driver (notebooks/_kappa_driver.py) and
pickles essential PhononOutput fields back to the notebook for plotting:

  - EMT       (test_pyiron_workflow_atomistics)
  - EAM Zhou  (test_pyiron_workflow_atomistics, Al_zhou.eam.alloy)
  - GRACE-1L-OAM (grace env, TF foundation model on GPU)
  - MACE-MP small (mace env, torch foundation model on GPU, float64)

The subprocess split avoids the cu12/cu13 wheel conflict between
tensorpotential and mace-torch on Blackwell GPUs (sm_120). The driver
forces a headless matplotlib backend so jupyter-spawned subprocesses
don't trip over matplotlib-inline when mace transitively imports
matplotlib in an env that lacks it.

The notebook produces three overlay figures: κ(T), phonon
band-structure + DOS, and Helmholtz F(T)/S(T)/Cv(T). κ(T) shows the
expected 1/T anharmonic-scattering decay; the foundation models cluster
near the (under-converged) lattice-only κ ≈ 5-10 W/(m·K) range for Al,
while EAM-Zhou over-predicts and EMT is the toy baseline.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

* docs(notebooks): optimise lattice per potential, zero F(T) at ZPE

Each calculator now runs `optimise_cubic_lattice_parameter` (from
`pyiron_workflow_atomistics.physics.bulk`) before the phonon macro, so
κ(T), bands+DOS and F(T)/S/Cv are computed at the potential's own
mechanical equilibrium rather than a single shared `a=4.05 Å`.

Adds T=0 to the temperature grid; phono3py returns κ(0)=0 which the
notebook filters out, and phonopy's harmonic free energy at T=0 supplies
ZPE per potential. F(T) is now plotted as F(T)−F(0 K) so the curves
share a 0-K reference; ZPE is surfaced in the summary table and legend.

Headli... (continued)

3305 of 3669 relevant lines covered (90.08%)

0.9 hits per line

Jobs
ID Job ID Ran Files Coverage
1 25965430370.1 55
90.08
 GitHub Action Run
Source Files on build 25965430370