paper-tutorial-vsc-water

name	paper_tutorial_vsc_water
description	Dual-purpose tutorial to reproduce manuscript liquid-water vibrational strong-coupling spectra and CPU scaling (single-mode cavity plus Meep Bragg resonator), and to adapt the same procedure to related VSC systems by retuning cavity/resonator parameters, size ladders, and execution settings.

Paper Tutorial: Water VSC Reproduction (Figure 3a-3c)

Use this skill for the scoped paper-tutorial reproduction request targeting vibrational strong coupling in water (request alias: Fig.4 scope, manuscript mapping: Figure 3 VSC family).

Direct Simulation Strategy

• Runtime work must happen in projects/YYYY-MM-DD-<scope>/.
• Recommended scope slug for this tutorial: vsc-water.
• Copy this skill's assets into the run folder and execute two coupled branches:
  • implementation_2025/vsc_water_singlemode: Figure 3a / fig_001 single-mode cavity spectra.
  • implementation_2025/vsc_water_maxwellmd: Figure 3b / fig_002 Meep-Bragg spectra.
• Reuse timing outputs from both branches to build Figure 3c / fig_003.
• Keep the publication parameter anchors:
  • dt_fs=0.4, steps=2.5e5, cavity marker near 3550 cm^-1.
  • size ladder size in {1,2,4,8}.
  • coupling/current rescaling with system size: 1/size^1.5.
• Acceptance is trend-level agreement for LP/UP splitting and scaling behavior, not bitwise trajectory identity.

Minimal Execution Recipes

1. Create a dated run workspace and copy tutorial assets:

RUN_DIR="projects/YYYY-MM-DD-vsc-water"  # projects/YYYY-MM-DD-<scope>/
mkdir -p "$RUN_DIR"
cp -R skills/paper_tutorial_vsc_water/assets/implementation_2025 "$RUN_DIR/"

2. Run single-mode size scan and collect timing stats:

cd "$RUN_DIR/implementation_2025/vsc_water_singlemode"
bash run_vsc_scan_size.sh
bash get_time_statistics.sh
bash get_lammps_statistics.sh

3. Run Meep-Bragg size scan and collect timing stats:

cd "$RUN_DIR/implementation_2025/vsc_water_maxwellmd"
bash run_vsc_scan_size.sh
bash get_time_statistics.sh
bash get_lammps_statistics.sh

4. Build the publication panel (fig3_vsc.pdf):

cd "$RUN_DIR/implementation_2025/plotting"
python plot_fig3_vsc.py

5. Artifact gate (all figure families):

for size in 1 2 4 8; do
  test -f "$RUN_DIR/implementation_2025/vsc_water_singlemode/size_${size}_cube/vsc_water_maxwellmd_data.npz"
  test -f "$RUN_DIR/implementation_2025/vsc_water_maxwellmd/size_${size}_cube/vsc_water_maxwellmd_data.npz"
  test -f "$RUN_DIR/implementation_2025/vsc_water_singlemode/size_${size}_cube/time_count.txt"
  test -f "$RUN_DIR/implementation_2025/vsc_water_singlemode/size_${size}_cube/lammps_count.txt"
  test -f "$RUN_DIR/implementation_2025/vsc_water_maxwellmd/size_${size}_cube/time_count.txt"
  test -f "$RUN_DIR/implementation_2025/vsc_water_maxwellmd/size_${size}_cube/lammps_count.txt"
done
test -f "$RUN_DIR/implementation_2025/plotting/fig3_vsc.pdf"

Figure Routing

• fig_001: Reproduce liquid-water VSC in a single-mode cavity, testing LP/UP splitting around 3550 cm^-1 and size-rescaled coupling invariance; playbook: skills/paper_tutorial_vsc_water/playbooks/fig_001_singlemode_vsc.md.
• fig_002: Reproduce Meep+LAMMPS VSC in a 1D Bragg resonator, including Bragg transmission context and librational/O-H splitting behavior; playbook: skills/paper_tutorial_vsc_water/playbooks/fig_002_meep_bragg_vsc.md.
• fig_003: Reproduce stepping-time scaling versus driver cores and compare MaxwellLink+LAMMPS against standalone LAMMPS trends; playbook: skills/paper_tutorial_vsc_water/playbooks/fig_003_scaling.md.

Beyond Manuscript Exploration

• Sweep single-mode damping (damping_au) while fixing frequency_au and baseline size ladder to probe linewidth-splitting tradeoffs.
• Sweep Bragg resonator discretization (resolution) and monitor whether LP/UP positions remain stable within tolerance.
• Test shorter production windows (steps) for cost reduction, then validate that qualitative splitting order is preserved before claiming equivalence.
• Keep dt_fs close to 0.4 unless revalidated.
• Preserve coupling normalization with system size (1/size^1.5) for fair scaling comparisons.
• Do not mix timing data from different queue/node policies in one scaling fit.