diff --git a/lang/en/docs/index-guide.md b/lang/en/docs/index-guide.md index 0a59279f..60511383 100644 --- a/lang/en/docs/index-guide.md +++ b/lang/en/docs/index-guide.md @@ -45,7 +45,7 @@ Step-by-step recipes reproducing published work, one row per publication: the st | Jung et al. (2015)[^10] | 2D–2D interface | [Graphene / h-BN](tutorials/materials/specific/interface-2d-2d-graphene-boron-nitride.md) | Band structure, total energies (Coming Soon) | | Shan et al. (2011)[^11] | 3D–3D interface | [Cu / SiO2](tutorials/materials/specific/interface-3d-3d-copper-silicon-dioxide.md) | Band structure (Coming Soon) | | Kang et al. (2008)[^12] | 2D–3D interface | [Graphene / SiO2](tutorials/materials/specific/interface-2d-3d-graphene-silicon-dioxide.md) | Band structure (Coming Soon) | -| Dahal et al. (2014)[^13] | Interface optimization | [Graphene / Ni(111)](tutorials/materials/specific/optimization-interface-film-xy-position-graphene-nickel.md) | Total energies versus lateral shift, band structure (Coming Soon) | +| Dahal et al. (2014)[^13] | Interface optimization | [Graphene / Ni(111)](tutorials/materials/specific/optimization-interface-film-xy-position-graphene-nickel.md) | [Registry and separation](tutorials/materials/specific/optimization-interface-film-xy-position-graphene-nickel-simulation.md) | | Saidi et al. (2015)[^14] | Adatom island | [Pt on MoS2](tutorials/materials/specific/defect-point-adatom-island-molybdenum-disulfide-platinum.md) | Binding energy per Pt atom, density of states (Coming Soon) | | Aradi et al. (2007)[^15] | H-passivated nanowire | [Si](tutorials/materials/specific/passivation-edge-nanowire-silicon.md) | Band gap, density of states, formation energy (Coming Soon) | | Hansen et al. (1998)[^16] | H-passivated surface | [Si(100)](tutorials/materials/specific/passivation-surface-silicon.md) | Diffusion, reaction and desorption barriers (Coming Soon) | diff --git a/lang/en/docs/tutorials/materials/specific/optimization-interface-film-xy-position-graphene-nickel-simulation.md b/lang/en/docs/tutorials/materials/specific/optimization-interface-film-xy-position-graphene-nickel-simulation.md new file mode 100644 index 00000000..84caffa0 --- /dev/null +++ b/lang/en/docs/tutorials/materials/specific/optimization-interface-film-xy-position-graphene-nickel-simulation.md @@ -0,0 +1,182 @@ +--- +tags: + - graphene + - nickel + - interface + - registry + - adsorption + - work of adhesion + - relaxation + - machine-learned force field + - MACE + - C-2D-INT-Z + +hide: + - tags +# YAML header +render_macros: true +--- + +# Gr/Ni(111) Registry and Work of Adhesion + +## 1. Introduction + +This tutorial reproduces the structure and energetics of graphene on Ni(111) — which registry the +film adopts, how far it sits above the surface, and the work of adhesion of each arrangement — +using the interface created in the +[structure creation tutorial](optimization-interface-film-xy-position-graphene-nickel.md). + +!!!note "Manuscript" + **Arjun Dahal, Matthias Batzill**, + "Graphene-nickel interfaces: a review" Nanoscale, 6(5), 2548 (2014) + [DOI: 10.1039/c3nr05279f](https://doi.org/10.1039/c3nr05279f) [@Dahal2014] + +### 1.1. What is being reproduced + +Graphene and Ni(111) are lattice-matched to within a fraction of a percent, so the film locks into +a 1×1 registry. The review's section 2.1 collects the established structural facts: LEED I–V and +ion scattering identify the adsorbed structure as one carbon **atop** a first-layer Ni atom and the +other in the **fcc hollow**, 0.211 nm above the surface, with a 0.005 nm buckling in which the atop +carbon sits further out. The review's computed values come from Lahiri *et al.*, New J. Phys. 13, +025001 (2011) — open access, and the quantitative target here (its Table 1): + +| interface | work of adhesion (J/m²) | separation (Å) | +|---|---|---| +| fcc (atop + fcc hollow) | 0.81 | 2.16 | +| hcp (atop + hcp hollow) | 0.77 | 2.17 | +| hollow (fcc + hcp hollows) | 0.31 | 3.26 | + +The review's text quotes the hollow as 0.38 J/m²; the source paper's Table 1 says 0.31, and the +notebook targets the source. The bridge registry (Fig. 1d of the review) is not quantified in +either paper and is computed as an extra point beyond the published set. + +![The four registries of graphene on a close-packed metal surface](../../../images/tutorials/materials/optimization/optimization_interface_film_xy_position_graphene_nickel/0-figure-from-manuscript.webp "Registries of graphene on a close-packed metal surface") + +### 1.2. The published recipe, and why relaxation is not optional + +Lahiri *et al.* state their method plainly: **LDA**, because "GGA does not provide an adequate +description of Ni–graphene bonding" for this interface; spin-polarized throughout; and **geometry +relaxation** with the bottom substrate layers fixed. The buckling is itself one of the published +numbers, and no rigid placement can produce a buckling — so every result in this tutorial comes +from a relaxed structure, and rigid scans are used only to bracket the starting separations. + +## 2. Prerequisites + +Run the [structure creation tutorial](optimization-interface-film-xy-position-graphene-nickel.md) +first. Its notebook builds the Gr/Ni(111) interface and saves it into the `uploads` folder as +`Graphene_Nickel_interface`; the simulation notebook loads it back by exactly that name and stops +if it is missing. The reduced cell is the 1×1 match: 2 carbon and 4 nickel atoms. + +## 3. What is calculated + +Two tiers, both relaxed: + +1. **Fast tier — MACE-MP + D3, in the browser.** Each registry is placed (the surface sites are + measured from the substrate's own top layers, and each registry label is re-verified after + relaxation, so a structure that slid into a neighbouring registry cannot be reported under the + wrong name), bracketed by a rigid scan, then relaxed with the bottom substrate layers fixed — + the paper's scheme. Same-cell relaxed references (bare Ni slab, free-standing graphene) turn the + energies into works of adhesion: `W = [E(slab) + E(graphene) − E(interface)] / A`. + + **The fast tier is expected to fail the energetic targets, and says so.** MACE-MP is PBE-trained, + and PBE-level physics is exactly what the paper rejects for this interface: chemisorption comes + out several times too weak. What the fast tier is good for is the geometry survey — the E(z) + curves, the two-minimum structure, and the dispersion-bound hollow, whose work of adhesion it + nearly matches (0.30 vs 0.31 J/m²). Its comparison table prints against the paper's values with + pass/fail per check and an honest per-tier verdict line. + +2. **Precise tier — the paper's LDA on the platform.** One relaxation + total-energy job per + selected registry, starting from the MACE-relaxed geometry, plus the two same-cell references — + LDA (`pz`, GBRV ultrasoft — the platform carries the LDA set for both Ni and C), spin-polarized, + **no dispersion correction**, matching the paper: LDA binds this interface unaided, which is the + stated reason its authors chose it. This tier carries the reproduction claim. + +| registry | Fig. 1 | carbon sublattices | published target | +|---|---|---|---| +| `atop_fcc` | (b) | atop + fcc hollow | 0.81 J/m² at 2.16 Å, favourable | +| `atop_hcp` | (c) | atop + hcp hollow | 0.77 J/m² at 2.17 Å | +| `hollow` | (a) | fcc + hcp hollows | 0.31 J/m² at 3.26 Å — dispersion-bound | +| `bridge` | (d) | C–C bond straddling a first-layer Ni | beyond the published set | + +## 4. Calculation parameters + +| | fast tier | precise tier | Lahiri et al. | +|---|---|---|---| +| Method | MACE-MP-0 (large, float64) + D3 | LDA (`pz`), GBRV ultrasoft | LDA, all-electron LCAO (DMol) | +| Spin | via training data | collinear, moment started at 0.7 μB on Ni | spin-polarized (bulk Ni: 0.56 μB) | +| Relaxation | BFGS, bottom 2 Ni layers fixed | platform relaxation + total energy | bottom 2 of 5 Ni layers fixed | +| Cutoffs | — | 40 / 200 Ry (GBRV's published pair) | all-electron | +| k-grid | — | 12×12×1 (multiple of 3, so K is on the mesh) | converged, not stated | +| Smearing | — | Marzari-Vanderbilt cold, `degauss = 0.01` Ry | not stated | +| Dispersion | D3 | none — matching the paper | none | + +Stated divergences from the paper: the slab is the structure tutorial's 4 Ni layers rather than 5; +the vacuum is 20 Å rather than 90; the platform relaxation cannot hold the bottom layers fixed +(the fast tier can, and does); plane-wave pseudopotentials rather than all-electron LCAO. The SCF +convergence settings (cold smearing, `local-TF` mixing, `mixing_beta = 0.2`, 200 iterations) exist +because the platform defaults stop at "convergence NOT achieved after 100 iterations" on this +spin-polarized metal slab, with the energy oscillating in its fourth decimal — charge sloshing. + +## 5. Step-by-step instructions + +### 5.1. Create the structure + +Run the [structure creation notebook](optimization-interface-film-xy-position-graphene-nickel.md). +It saves `Graphene_Nickel_interface` into `uploads`. + +### 5.2. Open the simulation notebook + +``` +other/materials_designer/specific_examples/optimization_interface_film_xy_position_graphene_nickel_SIMULATION.ipynb +``` + +### 5.3. Run the fast tier + +*Run* > *Run All Cells*. Sections 2–4 need no platform account: they load the interface, derive and +verify the registries, relax each one with MACE, and print the comparison against Lahiri Table 1 — +including the honest `[MACE tier]` verdict. + +### 5.4. Run the precise tier + +Section 5 authenticates and submits, per selected registry, a relaxation + total-energy job at the +paper's LDA, plus the two reference jobs. A default run selects one registry — three jobs. Leaving +`DFT_REGISTRY_NAMES` **empty** skips the platform tier entirely; the automated test does exactly +that, because relaxation jobs outlast what a browser test may wait for. + +### 5.5. Read the verdict + +The final cell restates the published targets and prints one verdict per tier: + +``` +Reproduces Lahiri et al. Table 1 [MACE tier]: no +Reproduces Lahiri et al. Table 1 [DFT tier]: yes +``` + +The fast tier failing its energetic checks is the physics working as documented, not a bug — see +section 3. The DFT-tier line appears once the selected registries and both references have +finished. + +## 6. Troubleshooting + +If a registry's rigid scan finds no bracketed minimum, widen the scan window. If every registry +comes back physisorbed-only in the fast tier, check `MACE_MODEL` and `MACE_DEFAULT_DTYPE` — the +medium/float32 combination misses the chemisorbed minimum entirely. The first MACE call downloads +the foundation model; later runs use the cache. If a platform job stops at "convergence NOT +achieved", the smearing/mixing block in the parameters cell is the knob — those settings exist +precisely because the defaults do not converge this slab. + +## 7. Interactive JupyterLite notebook + +The notebook below runs the fast tier and, when registries are selected, the platform tier. +Select *Run* > *Run All Cells*. + +{% with origin_url=config.extra.jupyterlite.origin_url_lab %} +{% with notebooks_path_root=config.extra.jupyterlite.notebooks_path_root %} +{% with notebook_name='specific_examples/optimization_interface_film_xy_position_graphene_nickel_SIMULATION.ipynb' %} +{% include 'jupyterlite_embed.html' %} +{% endwith %} +{% endwith %} +{% endwith %} + + +## 8. References diff --git a/mkdocs-guide.yml b/mkdocs-guide.yml index 31c6c13c..81e2cf0d 100644 --- a/mkdocs-guide.yml +++ b/mkdocs-guide.yml @@ -215,6 +215,7 @@ nav: - Grain Boundary in Cu (FCC): tutorials/materials/specific/defect-planar-grain-boundary-3d-fcc-metals-copper.md - Grain Boundary (2D) in h-BN: tutorials/materials/specific/defect-planar-grain-boundary-2d-boron-nitride.md - Gr/Ni(111) Interface Optimization: tutorials/materials/specific/optimization-interface-film-xy-position-graphene-nickel.md + - Gr/Ni(111) Registry and Separation: tutorials/materials/specific/optimization-interface-film-xy-position-graphene-nickel-simulation.md - Pt Adatoms Island on MoS2: tutorials/materials/specific/defect-point-adatom-island-molybdenum-disulfide-platinum.md # 2. Simulations diff --git a/mkdocs.yml b/mkdocs.yml index 17e81bb4..c2dbe2d1 100644 --- a/mkdocs.yml +++ b/mkdocs.yml @@ -175,6 +175,7 @@ nav: - Grain Boundary in FCC Metals (Copper): tutorials/materials/specific/defect-planar-grain-boundary-3d-fcc-metals-copper.md - Grain Boundary (2D) in h-BN: tutorials/materials/specific/defect-planar-grain-boundary-2d-boron-nitride.md - Gr/Ni(111) Interface Optimization: tutorials/materials/specific/optimization-interface-film-xy-position-graphene-nickel.md + - Gr/Ni(111) Registry and Separation: tutorials/materials/specific/optimization-interface-film-xy-position-graphene-nickel-simulation.md - Pt Adatoms Island on MoS2: tutorials/materials/specific/defect-point-adatom-island-molybdenum-disulfide-platinum.md # COMMON UI COMPONENTS