Phase 2: New planet discovery pipeline

[humancto]

Summary

Phase 2 infrastructure for discovering new exoplanets on stars that TESS has never flagged.

• python/download_sector_bulk.py — Downloads ALL light curves from a TESS sector via astroquery.mast, filters out known TOIs/cTOIs, optionally filters for FGK dwarf stars. Supports resume and saves a reproducibility manifest (targets.json).
• python/flag_discoveries.py — Cross-references BLS candidates against 3 catalogs (TOI, cTOI, confirmed exoplanets) and classifies each as NEW, KNOWN_TOI, KNOWN_PLANET, or LOW_SCORE. Planet-sized (Rp/Rs < 0.3) new candidates are highlighted for ExoFOP cTOI submission.
• Makefile targets: make phase2 PHASE2_SECTOR=56 runs the full pipeline (download → BLS → validate → flag)
• 22 new tests (18 discovery flagging + 4 bulk download) — total now 124 (70 Rust + 54 Python)
• CLAUDE.md updated with Phase 2 pipeline docs, architecture diagram, and sector selection strategy

Key insight

Phase 1 re-validated existing TOIs (stars TESS already flagged). Phase 2 downloads ALL stars from a sector and runs BLS on unstudied stars — any transit detection on a non-TOI star is potentially a genuinely new planet candidate.

Initial results (sector 56, SPOC, 174 stars)

• 161 BLS detections (SNR >= 6.0)
• 87 flagged as NEW (not in TOI/cTOI catalogs)
• 0 planet-sized (Rp/Rs < 0.3) — expected for SPOC targets where planets are already TOIs
• 1 marginal: TIC 219791066 (Rp/Rs=0.3079, Score=90) — needs deeper analysis
• Next: run on QLP data (160k stars/sector) where real discoveries live

Phase 2 workflow

MAST query (all sector stars)
  → Filter out TOIs/cTOIs
  → Download non-TOI light curves
  → Rust BLS search (same engine)
  → Rust validation (tests 1-3, 5; test 4 skipped for non-TOIs)
  → Discovery flagging (cross-ref 3 catalogs)
  → NEW candidates → ExoFOP cTOI submission

Test plan

• 22 new Python tests pass
• 70 Rust tests pass (no changes to Rust code)
• Full pipeline e2e: download → BLS → validate → flag on sector 56
• Discovery report generated with correct classification
• QLP data test (larger star count, higher discovery potential)

🤖 Generated with Claude Code

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, you can upgrade your account or add credits to your account and enable them for code reviews in your settings.