Guest User

Untitled

a guest
Nov 20th, 2026
239
0
Never
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
text 9.77 KB | None | 0 0
  1. I read the manuscript (940 pages) and checked the parts that carry the claimed deduction: the overview in the Introduction, Chapter 12 (the trace formula), Chapter 13 (zeros of Hecke L-functions) and Chapter 14.1 (the sanity checks). Yes -- there are flaws that prevent the document from being a proof as released. Some are textual, some are mathematical. The load-bearing ones are below.
  2.  
  3. 1. The chapter that would prove the result is a placeholder
  4.  
  5. Chapter 13 is the only place in the book where the location of zeros is supposed to be deduced. It is two pages long (pp. 521-522). Sections 13.1 ("Recollections on the Explicit Formulas") and 13.2 ("Holomorphic families of test functions") have headings and no text at all.
  6.  
  7. The central theorem is literally:
  8.  
  9. ▎ Theorem 13.3.1.2 (Divisor of L(s, χ)). ^2
  10. ▎ ^2 Insert statement.
  11.  
  12. and the footnote attached to the section reads:
  13.  
  14. ▎ ^3 State and prove the Main Theorems, also explaining how to deduce them by approximation φ -> e^{-ε|log(t)|} as Re(s) > 1.
  15.  
  16. The trace formula it restates is itself footnoted "Correct statement." (p. 522). So the manuscript, in its own words, does not contain the statement or the proof of the result on which everything else depends.
  17.  
  18. This is not isolated. Section 12.3, "Putting the pieces together: global to S-local deconstruction of the geometric kernel" (p. 514), consists of the heading, a footnote marker, and the footnote "Include section." The proof of Theorem 10.4.2.8 (p. 464) ends with "Return here and complete proof." Section 11.1 has "Return here." Section 1.3.4 has two "Finish!" footnotes, Sections 1.6.1 and
  19. 1.6.2 are empty, and the Introduction says it is unfinished. The claim in the acknowledgements that "the math is complete, and only filling in some writing remains" is not supported by the text: an unstated theorem is not missing prose.
  20.  
  21. 2. The bridge from a codimension-[F:Q] identity to zero location is missing -- and the manuscript admits using a theorem it does not state
  22.  
  23. The Introduction boasts twice (1.4.6, p. 30; 1.7.2, p. 35) that "Weil's Positivity Criterion is not used." But Chapter 13 and 14.1.2 say that Theorem 13.3.1.2 follows from the trace formula "combined with the Positivity Criterion (in its form from Theorem ??)" (p. 528), and again "through the Positivity Criterion of Weil, in its form of Theorem ??" (p. 533). "Theorem ??" is an unresolved
  24. LaTeX reference and is never stated anywhere in the book.
  25.  
  26. This matters mathematically, not just editorially. The trace formula is asserted only on the subspace A of test functions satisfying the [F:Q] vanishing conditions (13.3.1.2). An identity that holds only on a closed subspace of codimension d determines a functional only modulo the d-dimensional annihilator A-perp, which here is spanned by the moments φ -> M(φ)^(k)(0). The zero-location im
  27. claim says that a certain sum of functionals M(φ)(c), taken over all zeros in a strip, agrees with the sum over zeros on the line; the difference must lie in A-perp. A finite sum of functionals at distinct characters cannot lie in that finite-dimensional space, but an infinite sum over zeros can. Excluding that possibility is exactly the extra input that Weil positivity supplies. The paper
  28. invokes it but never states or proves it. Worse, the Introduction itself observes (1.7.2) that the self-convolutions used in Weil's criterion are the only test functions of that form satisfying φ(C_F^1) = 0, and that the only such function is zero. So the classical positivity criterion is vacuous on the test class for which the trace formula is claimed, and the replacement needed to close
  29. the gap is precisely the missing "Theorem ??."
  30.  
  31. 3. The "Main Theorem" is stated twice, and the two statements are inconsistent
  32.  
  33. Compare:
  34.  
  35. - Theorem 12.5.1.1 (p. 519): test functions φ in D(C_F) with φ(C_F^1) = 0; finite-place sum over all v in Σ_{F,f}; spectral sum over all zeros c with σ(c) - 1/2 in the interior of an interval I.
  36. - Theorem 13.3.1.1 (p. 522): test functions in the [F:Q]-codimensional subspace A defined by (13.3.1.2); finite-place sum over v in S_{F,f} only; spectral sum only over zeros with σ(c) = 1/2, with multiplicity capped by N_spec(c) = min{N(c), [F:Q]}.
  37.  
  38. These are different theorems. The spectral side of 13.3.1.1 already sums only over zeros on the critical line. If that statement is used as a premise to prove that all zeros lie on the critical line, as 13.3.1.2 would have to do, then the critical-line restriction is built into the input. (The cap min{N(c), [F:Q]} similarly presupposes the multiplicity bound that 14.1.2 presents as a
  39. consequence.)
  40.  
  41. In addition, the interval I in 12.5.1.1 is never quantified in the statement, although the left-hand side is manifestly independent of I. If the identity really held for every interval I, then for compactly supported φ (where the spectral sum is absolutely convergent) the interval function I -> Σ_{zeros in strip} N(c)M(φ)(c) would be constant, which is already essentially the Riemann
  42. hypothesis for φ in that class. That is far stronger than what the proof delivers: the proof only covers one chosen I ("For I = [-α,α]", p. 520, with α coming from the operator, not from the theorem). So the statement is at best under-specified.
  43.  
  44. 4. The final limiting step is asserted, not justified -- and the trace-class estimate fails at exactly that step
  45.  
  46. The deduction of GRH in 1.4.5/1.8.5 (pp. 29-30, 55) does not use a compactly supported test function. It applies the trace formula to the family φ_z(t) = t^{-z} for t > 1, zero for t < 1, with φ_z(1) = 1/2 (p. 28), approximated by compactly supported h_{ε,θ,z}. Three things are asserted without proof:
  47.  
  48. - that the trace formula passes to the limit ε -> 0, θ -> ∞ on both sides;
  49. - that the resulting spectral sum is holomorphic on C_{>a};
  50. - that the limit of the geometric side is L_f'/L_f(z+1/2).
  51.  
  52. The first is the serious one quantitatively. The nuclearity proof (Theorem 10.4.2.8, estimate (10.4.2.34), p. 464) bounds the nuclear norm by a sum over the zeros of |M(φ)^(i)(...)|. For compactly supported φ, M(φ) decays exponentially in the imaginary direction, so the sum converges against N(T) ~ T log T. In the limit, M(φ_z)(c) = 1/(z - (c-1/2)), so the corresponding sum is Σ_{zeros in
  53. strip} 1/|z - (c-1/2)|, which diverges like Σ log T. The limiting operator is therefore not trace class, the spectral series is only conditionally convergent, and the equality of the two sides in the limit needs an ordering/renormalization argument. The text invokes Montel's theorem to get a locally uniform limit (p. 29), but Montel gives a subsequence and holomorphy; it does not by itself
  54. identify the trace at the limit with Σ N(c)/(z - (c-1/2)). The sentence "This is, ultimately, the main consequence of the Trace Formula. Everything else follows from this" (p. 29) hides all of this in one line.
  55.  
  56. There is also a mismatch of hypotheses: the family φ_z is not in D(C_F) and does not satisfy φ(C_F^1) = 0 (its value on |g| = 1 is 1/2, and it has non-compact support). The author acknowledges that the condition cannot be removed "before GRH is proved" (1.4.6, p. 30). The limiting argument may be repairable, but it is not written.
  57.  
  58. 5. A concrete error in the sanity check
  59.  
  60. Proposition 14.1.2.3 (p. 531) claims that GRH plus Global Langlands implies the multiplicity bound N(α) ≤ [F:Q]. The proof inflates χ to the Galois closure F', induces from W_{F'} to W_Q, and computes dim Ind_{W_{F'}}^{W_Q}(χ') = [F':Q] -- the proof states this -- and then concludes that this is "a finite set of [F:Q] elements." For F'/F nontrivial these differ (F = Q(∛2): [F:Q] = 3, [F':Q]
  61. = 6). The written argument gives the bound [F':Q], not [F:Q]. It can be repaired by inducing directly from W_F, but as written the check does not establish what it claims. Moreover, under the assumed GSH the zeros of each constituent are simple, so the bound is essentially a counting of constituents; it is not independent evidence for the trace formula.
  62.  
  63. 6. A structural remark
  64.  
  65. The "trace formula" is not a trace formula in the usual sense. Its spectral side is built into the definition of the geometric kernel: (12.5.1.4) computes ∫k(x,x) as minus the spectral sum "by design of the global geometric kernel ... and by Theorem 10.4.2.8" (p. 520). The content is the S-local computation plus the claim that all orbital pieces for places outside S vanish. That is a
  66. legitimate way to organize an explicit-formula proof, but it means the theorem carries no independent information about zeros beyond what the chosen Hilbert completion admits. That is why the whole argument hinges on the unproved positivity/codimension step.
  67.  
  68. Bottom line
  69.  
  70. There is a coherent strategy here -- a strip-truncated explicit formula obtained by trace-class methods on Hardy-Sobolev completions, with the strip widths used as a parameter. But the released manuscript does not prove the claimed result:
  71.  
  72. 1. the theorem and its proof are absent (Chapter 13 is a skeleton);
  73. 2. the one extra ingredient needed to pass from a codimension-[F:Q] identity to zero location is an unstated "Theorem ??";
  74. 3. the two statements of the main theorem are not equivalent;
  75. 4. the limiting argument that produces L_f'/L_f is asserted, and at the level of the trace-norm estimate it fails;
  76. 5. the central reduction (Section 12.3) is empty, and several other proof-critical places are marked "complete proof" or "Return here."
  77.  
  78. So, as a proof of GRH (or even of "GRH for non-central zeros"), the document does not close. It is a very long draft with a plausible architecture and an unfinished core. The strongest single sentence summary: the proof depends on a positivity criterion that is invoked twice and stated nowhere, in a chapter that says "Insert statement."
  79.  
  80. ✻ Brewed for 9m 12s · done 6:39 PM
Advertisement
Add Comment
Please, Sign In to add comment