# Nonlinear Wave Limits of Reversible Finite-State Cellular Automata

Zhenyu Sun, Northeastern University at Qinhuangdao

## Build the paper

Run from this directory with pdfLaTeX, BibTeX, and latexmk installed:

```sh
latexmk -pdf -interaction=nonstopmode -halt-on-error manuscript.tex
```

The seven PDF figures, five TeX tables, bibliography, and required Springer
class/style files are included. They allow compilation without running the
experiments. The manuscript PDF and generated bibliography are not included.

## Reproduce the calculations

Numerical source files are under `anc/`, as required by the repository.
Prepare the project layout used by the manuscript commands once:

```sh
cp -R anc/experiment experiment
cp requirements.txt requirements-lock.txt experiment/
cp anc/make_manuscript_tables.py .
```

Use Python 3.12 or 3.13. From the same directory, run:

```sh
python -m pip install -r experiment/requirements-lock.txt
python experiment/reproduce.py
python experiment/verify_outputs.py
python experiment/verify_results.py
python make_manuscript_tables.py
latexmk -pdf -interaction=nonstopmode -halt-on-error manuscript.tex
```

The YAML configuration supplies the exact rational model parameters.
The complete run generates all integer and continuum trajectories, summaries,
tables, and figures from scratch. No precomputed result file is required.
Generated trajectories, reports, caches, and build logs are omitted from this
source archive.

Plot generation uses LaTeX and `dvipng`, with the `lmodern`, `type1cm`, and
`type1ec` packages. Plot text and mathematics use Latin Modern, matching the paper.
`--no-plots` skips plotting; `--only tests` runs the arithmetic checks.
Other selections are `convergence`, `spectroscopy`, `collision`, `source_check`,
`reference_check`, and `causal`. Run `collision` before `source_check`, whose
comparison uses the generated collision trajectories.

Results are written to `experiment/results/`. NPZ files store numeric arrays
without Python pickles. Integer runs save `Q`, `P`, `fields`, and `times` at every
time level; their JSON companions record model parameters and diagnostics.
Continuum reference files save fields and velocities. The generated
`environment.json` records software versions and source hashes.
The two checking programs replay saved trajectories and independently check
selected forces, normalizations, grid restrictions, and frequency fits.

## Use of AI-assisted tools

ChatGPT was used for proofreading and language editing.