fix(phd-bib-dedupe): remove 7 BibTeX duplicates + fix unclosed brace on coxeter1973regular

[gHashTag]

What this PR does

Removes 7 duplicate BibTeX entries from docs/phd/bibliography.bib and fixes one structural bug that would cause tectonic / bibtex to error at compile.

R5-honest hygiene work split off from Phase 3.5 LB audit. Sister PR is #615 (3.5 LB FULL PASS).

Changes

Brace fix (CRITICAL)

coxeter1973regular at line 1832 was missing its closing }. The note field's brace closed on line 1839, but the entry's outer brace never did — the BibTeX parser would have either errored out at compile time or silently swallowed everything until the next @<type>{. Inserted } before the % --- LB-Springer top-up … comment block.

Duplicates removed (kept first occurrence)

key	dup line	rationale
coxeter1973regular	1601	verbatim duplicate of line 1591
kepler_harmonices	1611	1619 Linz original; line 34 keeps 1997 Aiton/Duncan/Field translation (richer biblio)
coxeter1973regular	1638	verbatim duplicate of line 1591
binet_formula	1655	line 98 has properly LaTeX-encoded accents ({\'e})
weil_number_theory	1664	line 107 keeps the 1995 3rd edition (more recent printing)
codata2022	1672	line 376 keeps 2018 RMP version with full author names
coxeter1973regular	1832	the malformed entry; same content as kept line 1591

R11 publisher-balance impact

Gate	Before (main, 212 entries)	After this PR (205 entries)	After merge with #615 (208 entries)
Springer ≥ 25 %	52 (24.53 %) ❌	51 (24.88 %) ❌	53 (25.48 %) ✅
MIT/CUP/Ox ≥ 15 %	31 (14.62 %) ❌	31 (15.12 %) ✅	33 (15.87 %) ✅
arXiv-only ≤ 20 %	5 (2.36 %) ✅	5 (2.44 %) ✅	5 (2.40 %) ✅

R5 honesty disclosure

This PR alone does not restore R11-Springer ≥ 25 %. It must merge together with or after #615, which adds 2 verified Springer entries (lee_smooth_manifolds, kanerva_hdc_2009) and 1 verified MIT/CUP entry (strang_linear_algebra).

The Springer dip (52 → 51) is honest: removing the duplicate weil_number_theory correctly reduces the unique-key Springer count by one. The dup was inflating the share artificially.

How to verify locally

git checkout feat/phd-bib-dedupe
cd /path/to/trios
python3 - <<'PY'
import re
bib = open("docs/phd/bibliography.bib").read()
entries = re.findall(r"^@\w+\{([^,]+),", bib, re.M)
from collections import Counter
dups = [k for k, c in Counter(entries).items() if c > 1]
assert not dups, f"Still duplicate: {dups}"
print(f"OK: {len(entries)} entries, 0 duplicates")
PY

R-rules tracking

• ✅ R1 — no .py / .sh files added (the verification snippet is documentation)
• ✅ R5 — every removal preserves the canonical first occurrence; honest disclosure of post-PR balance
• ✅ R11 — combined with feat(phd-phase3-rules-audit-3-1): App.H anchor + Phase 3 R-RULES audit report (3.1/3.3/3.4/3.6/3.8 PASS) #615, all publisher gates clear with margin
• N/A R2 — not a per-chapter PR; no feat/phd-chNN branch needed
• N/A R3, R7, R12, R14 — bibliography hygiene only

Anchor

φ² + φ⁻² = 3 · DOI 10.5281/zenodo.19227877 · defense T-37d (2026-06-15).

[agent=phase3-bib-dedupe]

[gHashTag]

🛑 SUPERSEDED — closing as part of PhD T-33 cascade cleanup

main HEAD уже содержит 295 BibTeX entries vs PR #618 имеет 205 — этот PR стал регрессией (потерял бы 90+ entries). Bib dedupe для 7 дубликатов давно поглощён более крупными bib-обновлениями в main.

Honest R5 verdict (verified 2026-05-12T17:53Z)

PR содержит 155–176 PhD-файлов в diff, но почти все эти изменения либо уже в main, либо стали regression относительно более поздних обогащений.

What we kept

Любое уникальное содержимое было экстрагировано и применено хирургически. See PR #750 (Theorem 35.13 surgically re-applied from #593).

Anchor: phi^2 + phi^-2 = 3
Closing as superseded per PHD-STATUS-RVR-003 (Document ID).

[agent=perplexity-computer phd-cascade-cleanup user-approved]