silico-quantum

xtb-cluster-md

Run xTB (GFN-FF / GFN2-xTB) molecular dynamics for organic molecular clusters (e.g., anthracene/benzene), starting from random packed geometries; then generate animated GIFs (full trajectory, COM view, local clustered subset, bond-emphasized). Use when asked to simulate many-molecule stacking/aggregation dynamics, low-temperature “freezing/condensation” behavior, or to visualize xTB MD trajectories as animations.

silico-quantum 3 1 Updated 2mo ago

Resources

2
GitHub

Install

npx skillscat add silico-quantum/quantum-chem-skills/xtb-cluster-md

Install via the SkillsCat registry.

SKILL.md

xTB cluster MD + visualization

Quick workflow

1) Get a 3D monomer geometry (PubChem SDF)

Use PubChem PUG REST and save as molecule.sdf:

curl -L -o benzene.sdf "https://pubchem.ncbi.nlm.nih.gov/rest/pug/compound/name/benzene/SDF?record_type=3d"
curl -L -o anthracene.sdf "https://pubchem.ncbi.nlm.nih.gov/rest/pug/compound/name/anthracene/SDF?record_type=3d"

2) Build a random cluster XYZ (no RDKit/packmol needed)

python3 scripts/build_cluster.py --sdf benzene.sdf -n 96 --box 90 --min-com 6.0 --seed 1 -o benzene_N96_init.xyz

Tips:

  • If placement fails: increase --box or decrease --min-com.
  • For aromatics: --min-com ~ 6–8 Å is usually reasonable.

3) Run xTB MD (GFN-FF recommended for large N)

Create an xcontrol file, e.g. md.inp:

$md
   temp=279
   time=50.0
   step=1.0
   dump=50.0
   sdump=500.0
   nvt=1
   shake=1
   sccacc=2.0
   forcewrrestart=true
$end

Run:

xtb benzene_N96_init.xyz --gfnff --md -I md.inp --parallel 10 > md.log 2>&1

Notes:

  • dump=50 fs ⇒ each XYZ frame corresponds to 0.05 ps.
  • time=200 ps ⇒ about 4000 frames.

4) Make animations

(A) Full-system atom-level GIF (downsampled)

python3 scripts/make_atom_animation.py --traj xtb.trj -n 96 --nat-per-mol 12 --stride 20 --zoom 1.2 -o full.gif

(B) COM-only “motion overview” GIF

python3 scripts/make_animation.py --traj xtb.trj -n 96 --nat-per-mol 12 --stride 20 -o com.gif

(C) Local cluster (picked from last frame) with clearer bonds

Best for “final aggregation / hugging together”:

python3 scripts/make_local_cluster_animation.py \
  --traj xtb.trj \
  -n 96 --nat-per-mol 12 \
  --k 8 --dist minatom --drop-outlier --bonds \
  --pad 2.2 --stride 2 --max-frames 300 \
  -o local.gif

Local-cluster selection options:

  • --dist minatom (recommended): molecule–molecule distance = minimum atom–atom distance in the last frame.
  • --drop-outlier: after selecting the k-cluster, drop the farthest molecule from the cluster centroid (helps remove stragglers).
  • --drop-leftmost: drop leftmost by COM-x (useful for camera framing).

Included scripts

  • scripts/build_cluster.py: build N-molecule random cluster XYZ from PubChem SDF.
  • scripts/make_atom_animation.py: full atom-level GIF (colored by molecule index).
  • scripts/make_animation.py: COM-only 3D motion GIF.
  • scripts/make_local_cluster_animation.py: pick a “final cluster” from last frame and render only that subset; optional bond drawing.

Categories

Animation Code Gen
GitHub

Install

npx skillscat add silico-quantum/quantum-chem-skills/xtb-cluster-md

Install via the SkillsCat registry.

Recommended Skills