A short proof of Deligne's bound for Frobenius eigenvalues, after Alpöge

Primarily AI-generated textHuman understanding: no partsmath.AG — Algebraic Geometrymath.NT — Number Theory

Contributed by Floris Alderliefste ↗

Version 1 / Nov 21, 2026 / CC0 1.0

Abstract

We write out in full the proof, sketched by L. Alpöge in a footnote of his note of 20 November 2026, of Deligne's bound for the eigenvalues of Frobenius. Let XX be a separated scheme of finite type over a finite field with qq elements, and let F\mathcal F be a constructible ℓ\ell-adic sheaf on XX whose Frobenius eigenvalues at each closed point xx have absolute value at most ∣κ(x)∣w/2|\kappa(x)|^{w/2}. Then the eigenvalues of Frobenius on Hci(X,F)H^i_c(X,\mathcal F) have absolute value at most q(w+i)/2q^{(w+i)/2}, and they have absolute value exactly q(w+i)/2q^{(w+i)/2} if XX is smooth and proper and F\mathcal F is lisse and pure of weight ww. The proof is short. Granted the formalism of étale cohomology (the trace formula, the Künneth formula, Poincaré duality and Artin's vanishing theorem) and a weak Lefschetz theorem for generic hyperplane sections of affine varieties, it takes less than six pages, and its only estimate is the Cauchy–Schwarz inequality for a character sum over a finite field. Lefschetz pencils, monodromy and Rankin's method are not used. The proof was produced by GPT-6 Astra (ChatGPT Pro mode) from that sketch. It was checked by Claude Opus 5.5 at effort high and, independently, by Claude Fable 5.2 at effort max, which also wrote this text. No human has checked it.

Author comments: This note writes out, with full proofs, the argument sketched in footnote 3 of L. Alpöge's note of 20 November 2026 (https://alpo.ge/remark.pdf); its ideas are his. The weak Lefschetz theorem it uses is Deligne's, from the appendix of N. Katz, Affine cohomological transforms, perversity, and monodromy, J. Amer. Math. Soc. 6 (1993). Checks: every step was checked by Claude Opus 5.5 at effort high and re-proved independently by Claude Fable 5.2 at effort max; no human has checked the proof. The prompt, the output of the run and the check records are at https://github.com/KokunoYumeto/grothendieck-montpellier-archive/tree/main/research/grh_20261120/deligne_sign_only

Provenance statement

Text: AI-authored; none of it was written by a human. The submitter takes responsibility for the veracity of the text. How it was produced. The argument is sketched in footnote 3 of L. Alpöge's note of 20 November 2026 (https://alpo.ge/remark.pdf). Claude Opus 5.5 (Anthropic), at effort max, put the question as a research prompt on 20 November 2026. GPT-6 Astra (OpenAI), in ChatGPT Pro mode, produced the proof on 21 November 2026, as Part I, Section 1 of the output of that run. Claude Opus 5.5, at effort max, wrote a first text from it on 21 November 2026. Claude Fable 5.2 (Anthropic), at effort max, wrote the present text on 21 November 2026, using that earlier text. How it was checked. All checks are of 21 November 2026. The run kept its own check record. Claude Opus 5.5, at effort high, checked every step of the proof (a targeted proof check). Claude Opus 5.5, at effort max, re-derived the character-sum lemma, the identity for the remainder with its twists and signs, and the exponents. Claude Fable 5.2, at effort max, without seeing those checks, re-proved every step and found no gap; it read the pages of N. Katz's article that the proof uses (J. Amer. Math. Soc. 6 (1993), pp. 152-158, 179-181 and 219-222) and re-proved the weak Lefschetz theorem independently. One correction came from that check and is adopted: the weak Lefschetz input is taken from Deligne's proof in Katz's appendix, and not from the printed proof of Katz's Corollary 3.4.1. The second checker is also the author of the present text. The foundational references (SGA 4, SGA 4 1/2, the Stacks Project) were not reopened. What the introduction says about the articles of Bombieri, Deligne, Kedlaya and Laumon is from memory. The two statements of Lafforgue's article and the introduction of Scholl's were read at the source by Claude Opus 5.5 at effort max. No human has checked the proof or the text, and nothing is formalized. Where the records are. The prompt, the output of the run and the check records are at https://github.com/KokunoYumeto/grothendieck-montpellier-archive/tree/main/research/grh_20261120/deligne_sign_only The text and accompanying work are dedicated to the public domain under CC0 1.0 Universal (https://creativecommons.org/publicdomain/zero/1.0/).

Tools used

OpenAI
GPT-6 Astra (ChatGPT Pro mode)
Anthropic
Claude Fable 5.2 (effort max)
Anthropic
Claude Opus 5.5 (effort high and max)

References

  1. L. Alpöge, note on the Riemann hypothesis, 20 November 2026, https://alpo.ge/remark.pdf.link
  2. E. Bombieri, Counting points on curves over finite fields (d'après S. A. Stepanov), Séminaire Bourbaki 1972/73, Exposé 430, Lecture Notes in Math. 383, Springer, 1974, 234--241.
  3. P. Deligne, La conjecture de Weil. I, Publ. Math. IHÉS 43 (1974), 273--307.
  4. P. Deligne, La conjecture de Weil. II, Publ. Math. IHÉS 52 (1980), 137--252.
  5. J.-P. Jouanolou, Théorèmes de Bertini et applications, Birkhäuser, 1983.
  6. N. M. Katz, Affine cohomological transforms, perversity, and monodromy, J. Amer. Math. Soc. 6 (1993), 149--222.
  7. K. S. Kedlaya, Fourier transforms and pp-adic ``Weil II'', Compos. Math. 142 (2006), 1426--1450.
  8. L. Lafforgue, Chtoucas de Drinfeld et correspondance de Langlands, Invent. Math. 147 (2002), 1--241.
  9. G. Laumon, Transformation de Fourier, constantes d'équations fonctionnelles et conjecture de Weil, Publ. Math. IHÉS 65 (1987), 131--210.
  10. A. J. Scholl, Hypersurfaces and the Weil conjectures, arXiv:0912.2073 (https://arxiv.org/abs/0912.2073).arXiv
Show all 14 referencesShow fewer references
  1. M. Artin, A. Grothendieck, J.-L. Verdier, Théorie des topos et cohomologie étale des schémas (SGA 4), Lecture Notes in Math. 269, 270, 305, Springer, 1972--1973.
  2. P. Deligne, Cohomologie étale (SGA 4124\frac12), Lecture Notes in Math. 569, Springer, 1977.
  3. The Stacks Project, https://stacks.math.columbia.edu/tag/00OY.link
  4. Transcription of A. Grothendieck's papers in the Montpellier archive, https://github.com/KokunoYumeto/grothendieck-montpellier-archive.link

Version history

  1. v1 · 10 pagesSubmitted by Floris AlderliefsteThis note writes out, with full proofs, the argument sketched in footnote 3 of L. Alpöge's note of 20 November 2026 (https://alpo.ge/remark.pdf); its ideas are his. The weak Lefschetz theorem it uses is Deligne's, from the appendix of N. Katz, Affine cohomological transforms, perversity, and monodromy, J. Amer. Math. Soc. 6 (1993). Checks: every step was checked by Claude Opus 5.5 at effort high and re-proved independently by Claude Fable 5.2 at effort max; no human has checked the proof. The prompt, the output of the run and the check records are at https://github.com/KokunoYumeto/grothendieck-montpellier-archive/tree/main/research/grh_20261120/deligne_sign_onlyCurrentNov 21, 2026