SOF-8043: Gr/Ni(111) registry and separation simulation notebook

[VsevolodX]

Specific Example reproducing the graphene/Ni(111) registry energetics of Dahal & Batzill, Graphene–nickel interfaces: a review, Nanoscale 6, 2548 (2014) — SOF-8043, epic SOF-7994.

What it reproduces

claim	result	article
favourable registry	top_fcc	top-fcc (Fig. 1b)
chemisorption distance	2.01 Å	~2.1 Å (0.21 nm)
hollow registry	does not chemisorb	physisorbed only

Absolute adsorption energies are deliberately not compared — the manuscript's come from dispersion-corrected methods beyond semi-local DFT. The notebook compares the registry ordering and the separations, which transfer between methods.

How

Two tiers, so the notebook is useful in minutes and exact when it matters:

1. MACE-MP + D3, in the browser. The film is placed at top-fcc / top-hcp / bridge-top / hollow — sites measured from the substrate's own top three Ni layers, each registry labelled by where the second carbon sublattice actually lands — then scanned in z. A chemisorbing registry has two minima (chemisorbed ~2 Å, dispersion-bound further out), so the comparison reads the chemisorbed branch and compares each registry at its own minimum. Comparing all registries at one height misranks them.
2. Platform DFT. One total_energy job per registry at that geometry — PBE, nspin = 2 with a starting moment on Ni, smearing, 12×12×1 (multiple of 3 so K is sampled), vdw_corr = 'grimme-d3', identical cutoffs across all four. A default run submits one job and names the registries to activate; the Cypress feature activates all four.

Introduction.ipynb's Simulation column is filled for the Gr/Ni row.

Change to the structure notebook

It saved only the empirically-optimized variant; it now also saves the base interface, which the simulation notebook loads by name. The simulation notebook raises if that name is missing rather than silently falling back to Standata.

Verification

Fast tier is green end-to-end in JupyterLite, driven by the Cypress feature in the companion web-app PR: provenance, all four registries, all three checks ok.

The DFT tier is not yet verified — every job on the local stack errors with Account … did not sync to node …. That is an environment fault, not this notebook: mqueue-1's celery queue holds 3249 messages with 0 consumers, the RabbitMQ brokers are not clustered, and the cluster's workers connect to a proxy address that is none of them, so publisher and consumer land on different brokers. The stock Silicon total_energy.feature fails identically — its spec passes only because it asserts a label. Details in plan/log/platform.md.

Draft until the DFT tier runs.

🤖 Generated with Claude Code

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[coderabbitai]

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

• Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}