sidmkit

A transparent toolkit for self-interacting dark matter (SIDM) micro→macro work:

• Yukawa / dark-photon style self-interaction cross sections (Born / classical / Hulthén / partial-wave)
• Velocity averaging (Maxwellian relative-speed baseline) for ⟨σ/m⟩ and ⟨σ v⟩/m
• Curated summary constraint sets (from reviews) for fast sanity checks
• A simple halo-level mapping via the interaction radius r1 (Γ(r1) t_age = 1)
• Lightweight likelihood scaffolds + a tiny Metropolis MCMC for runnable end-to-end demos
• Benchmarks & regression numbers to catch numerical/unit regressions

sidmkit is designed to be easy to audit and extend. The included constraints and likelihoods are starting points, not publication-grade analyses.

---

Install

git clone
cd sidmkit
python -m venv .venv
source .venv/bin/activate
pip install -U pip
pip install -e .

---

CLI overview

sidmkit --help
sidmkit sigma --help
sidmkit avg --help
sidmkit constraints --help
sidmkit halo --help
sidmkit likelihood --help
sidmkit infer --help
sidmkit benchmark --help
sidmkit validate --help

---

End-to-end pipeline

This is the recommended reproducible workflow.

0) Run internal benchmarks (sanity + regression)

sidmkit benchmark
# optional slow partial-wave checks:
sidmkit benchmark --slow

1) Compute σ/m(v) and save a curve

Example model:

• mχ = 10 GeV
• m_med = 0.05 GeV (50 MeV)
• α = 0.01
• attractive Yukawa

sidmkit sigma --mchi 10 --mmed 0.05 --alpha 0.01 --potential attractive \
  --vmin 1 --vmax 5000 --n 200 --csv sigma_curve.csv

2) Do the velocity average (this is upgrade #1)

Compute ⟨σ v⟩/m for a Maxwellian relative-speed distribution with σ1D = 50 km/s:

sidmkit avg --mchi 10 --mmed 0.05 --alpha 0.01 --potential attractive \
  --sigma1d 50 --moment 1

• moment=0 → ⟨σ/m⟩
• moment=1 → ⟨σ v⟩/m

3) Check curated constraint sets (upgrade #2)

List available sets:

sidmkit constraints --list-sets --mchi 10 --mmed 0.05 --alpha 0.01

Evaluate a modern summary set (example: Adhikari+ review Table I):

sidmkit constraints --set literature_table1_summary --mchi 10 --mmed 0.05 --alpha 0.01

Or evaluate the classic Tulin & Yu Table I set:

sidmkit constraints --set literature_table1_compilation --mchi 10 --mmed 0.05 --alpha 0.01

4) Halo-level mapping: compute r1 for an NFW halo (upgrade #3)

Example: MW-like halo (M200=1e12 Msun, c200=10), age 10 Gyr:

sidmkit halo --mchi 10 --mmed 0.05 --alpha 0.01 \
  --m200 1e12 --c200 10 --tage 10 --vmode jeans

This solves for r1 where Γ(r1) t_age = 1 using:

• NFW density profile,
• isotropic Jeans velocity dispersion,
• Maxwellian relative-speed averaging for ⟨σ v⟩/m.

5) Dataset-style likelihood + inference (upgrades #4 and #5)

We ship two “starter” likelihoods:

1. Point constraints dataset (JSON): reproducible, auditable, summary-level.
2. Toy rotation-curve likelihood (CSV): demonstrates micro→macro → V_circ(r).

5a) Point-constraints likelihood

Example dataset file included:

• examples/datasets/kaplinghat2016_summary_plus_bullet.json

Evaluate log-likelihood at a model point:

sidmkit likelihood --kind points --data examples/datasets/kaplinghat2016_summary_plus_bullet.json \
  --mchi 15 --mmed 0.017 --alpha 0.00729927 --potential attractive

Run a small Metropolis MCMC:

sidmkit infer --data examples/datasets/kaplinghat2016_summary_plus_bullet.json \
  --potential attractive --method auto --n-steps 4000 --seed 0

5b) Toy rotation-curve likelihood (demo)

Example dataset file included:

• examples/datasets/toy_rotation_curve.csv

Evaluate the rotation-curve log-likelihood (you must specify a halo M200,c200):

sidmkit likelihood --kind rotation_curve --data examples/datasets/toy_rotation_curve.csv \
  --m200 1e11 --c200 12 --tage 10 \
  --mchi 10 --mmed 0.05 --alpha 0.01 --potential attractive

Be critical: the rotation-curve model is deliberately simple (isothermal core matched to NFW at r1). It exists to show the mechanics of a forward model, not to replace full analyses.

---

Validation hooks (upgrade #4)

Validate against an external curve (your digitized data)

If you digitize a paper’s σ/m(v) curve into a CSV with columns: v_km_s, sigma_over_m_cm2_g, you can compare:

sidmkit validate --target curve --reference your_curve.csv \
  --mchi 10 --mmed 0.05 --alpha 0.01 --potential attractive

This is a non-authoritative regression test using a small set of auto-digitized points:

sidmkit validate --target fig13 --mchi 15 --mmed 0.017 --alpha 0.00729927 --potential attractive

---

SPARC rotation-curve batch fits (phenomenological baseline)

sidmkit also ships a self-contained batch fitter for the SPARC *_rotmod.dat files. It fits simple halo profiles (NFW, Burkert) plus stellar mass-to-light parameters.

Install plotting extra (recommended):

pip install -e ".[plot]"

1) Batch fit in chunks (recommended)

python -m sidmkit.sparc_batch batch   --inputs path/to/Rotmod_LTG path/to/Rotmod_ETG   --outdir outputs/sparc_chunks/chunk_0   --skip 0 --limit 25   --plots --plot-format png

Run additional chunks by increasing --skip (e.g. 25, 50, ...). Use --resume to make reruns idempotent.

2) Merge chunk summaries

python -m sidmkit.sparc_batch merge   --inputs outputs/sparc_chunks/chunk_*/summary.json   --out outputs/sparc_all_summary.json

3) Population report

python -m sidmkit.sparc_batch report   --summary-json outputs/sparc_all_summary.json   --outdir outputs/sparc_report

These are halo-profile fits (baseline model comparison), not a full SIDM microphysics → core-size inference of SPARC.

---

Python quickstart

python examples/quickstart.py

---

License

MIT