# Sources for ζ and primes

Short working list for `zeta-space-and-primes.html`. Not a bibliography of the Riemann Hypothesis. Wikipedia is not used as an authority here. Nothing in this note is a claim that the Hypothesis has been proved.

## The landscape

- Bernhard Riemann, “Über die Anzahl der Primzahlen unter einer gegebenen Größe,” *Monatsberichte der Berliner Akademie*, 1859. The memoir: continuation, functional equation, explicit formula in outline, and the hypothesis.
- E. C. Titchmarsh, *The Theory of the Riemann Zeta-Function*, 2nd ed., revised by D. R. Heath-Brown, Oxford, 1986. Standard reference for the product, the strip, the explicit formula, and zero-free regions.
- H. M. Edwards, *Riemann’s Zeta Function*, Academic Press, 1974; Dover reprint. Reads the 1859 memoir closely.
- A. Ivić, *The Riemann Zeta-Function*, Wiley, 1985; Dover reprint.
- Enrico Bombieri, “Problems of the Millennium: the Riemann Hypothesis,” Clay Mathematics Institute. The official problem statement.
- J. Brian Conrey, “The Riemann Hypothesis,” *Notices of the American Mathematical Society* 50 (2003), 341–353. A survey. Not a proof.

## Partial results cited on the page

- Jacques Hadamard, and Charles de la Vallée Poussin, 1896: no zeros on Re(s) = 1, hence the prime number theorem.
- G. H. Hardy, 1914: infinitely many zeros on the critical line.
- Atle Selberg, 1942: a positive proportion on the critical line.
- Norman Levinson, “More than one third of zeros of Riemann’s zeta-function are on σ = 1/2,” *Advances in Mathematics* 13 (1974).
- J. B. Conrey, “More than two fifths of the zeros of the Riemann zeta function are on the critical line,” *Journal für die reine und angewandte Mathematik* 399 (1989).
- Kyle Pratt, Nicolas Robles, Alexandru Zaharescu, and Dirk Zeindler, “More than five-twelfths of the zeros of ζ are on the critical line,” *Research in the Mathematical Sciences* 7, article 2 (2020).
- I. M. Vinogradov and N. M. Korobov, independently, 1958: the wider zero-free region near Re(s) = 1. The shape on the page is the classical one. No record explicit constant is quoted.
- Helge von Koch, “Sur la distribution des nombres premiers,” *Acta Mathematica* 24 (1901): RH implies the square-root error in the prime-counting functions. The equivalence with ψ(x) = x + O(x^{1/2} log² x) is the form in Titchmarsh and Ingham.
- Harald Cramér, 1920: under RH, prime gaps are O(√p_n log p_n).

## Numerical and statistical evidence

- David J. Platt and Timothy S. Trudgian, “The Riemann hypothesis is true up to 3·10^{12},” *Bulletin of the London Mathematical Society* 53 (2021), 792–797. Rigorous verification to height 3,000,175,332,800, and simplicity of those zeros. A finite check.
- Andrew M. Odlyzko, “On the distribution of spacings between zeros of the zeta function,” *Mathematics of Computation* 48 (1987), 273–308. First 10^5 zeros, and 10^5 zeros near the 10^{12}-th. Later computations near height 10^{20} and above are his further numerical support, not a proof.
- Hugh L. Montgomery, “The pair correlation of zeros of the zeta function,” *Proceedings of Symposia in Pure Mathematics* 24, American Mathematical Society, 1973, 181–193. The pair-correlation conjecture. The GUE identification is Dyson’s observation, recorded in Odlyzko’s 1987 paper.
- Zeév Rudnick and Peter Sarnak, “Zeros of principal L-functions and random matrix theory,” *Duke Mathematical Journal* 81 (1996). Partial n-level correlations, for test functions with restricted Fourier support.

## 2026 preprints, not the stack this page rests on

- Levent Alpöge and Ralph Furman, arXiv:2608.13637 (August 2026).
- Youness Lamzouri, arXiv:2609.02882 (September 2026).

Both claim an unconditional proportion above 67% of nontrivial zeros simple and on the critical line. They are cited only as very recent claims. They are not a proof of the Riemann Hypothesis. The published bound the page treats as settled is still Pratt–Robles–Zaharescu–Zeindler, more than five twelfths.

## Ordinates and the 2 October 2026 figures

The eight imaginary parts on the page are rounded to six decimals. The first five were recomputed with `mpmath.zetazero` (mpmath 1.4.1) on 2 October 2026 PT. Real parts returned as 1/2, and |ζ| at each point was under 10^(−30). To six decimals: 14.134725, 21.022040, 25.010858, 30.424876, 32.935062. That matches the rounding previously cited from Odlyzko’s tables. Ordinates 6–8 in the table were not recomputed in this pass.

`img/critical-line-zeros.svg` plots those five values and their conjugates. Gaps printed on the figure are unrounded differences, rounded to two decimals.

`img/pi-vs-li.svg` plots π(x) from a sieve and li(x) from `mpmath.li` (Cauchy principal value), for x ≤ 200. Checkpoints: (10, π=4, li≈6.165600), (100, π=25, li≈30.126142), (200, π=46, li≈50.192171).

`img/explicit-formula-truncation.svg` plots the Chebyshev ψ(x), summed directly over prime powers, against the classical explicit formula truncated to those five conjugate pairs, for x ≤ 100. Sample residuals (ψ minus truncation): about 0.26 at x=10, 0.32 at x=50, −0.77 at x=100. Truncation is not the finished identity and not a proof of RH.

The hand-drawn plane sketch remains on the page and is still not those numbers. A GUE histogram was not drawn: no Odlyzko spacing file was loaded, and an invented histogram would be false data.

Checkpoint log: `research/figure-checkpoints.txt` in the atlas corpus (not served as a page).


## 3 October 2026 extension

Recomputed with mpmath 1.4.1, 20 digits. Checkpoint log: `research/figure-checkpoints-20261003.txt` in the atlas corpus.

- n = 1…20, all with real part 1/2 and |ζ| under 10^(−18). γ_20 = 77.144840 to six decimals. Being on the line for these twenty does not prove RH.
- π(1000) = 168. li(1000) ≈ 177.609658 (`mpmath.li`). Also π(500) = 95, li(500) ≈ 101.793872. Still far too short for the square-root error.
- Explicit formula, five pairs versus twenty. Signed residuals (ψ minus truncation), five then twenty: x=10, +0.260264 then −0.072965; x=20, +0.108085 then −0.071755; x=100, −0.773642 then −1.636707; x=200, +4.737728 then +2.319604. Not monotone. Imaginary part of the zero-sum stayed under 10^(−8).
- |ζ(1/2)| ≈ 1.460355. On a 0.05 grid plus the ordinates, the largest sample on [0, 50] is about 3.665 near t = 45.65. A sample peak, not a certified maximum.


## 4 October 2026 — signed error li − π

Sieve to 10 000; `mpmath.li` (principal value); sample every 5 from 2 to 10 000. Checkpoint log: `research/figure-checkpoints-20261004.txt`.

- π(1000) = 168, π(10 000) = 1229.
- li − π: 100 → ≈5.126142; 1000 → ≈9.609658; 5000 → ≈15.280840; 10 000 → ≈17.137216.
- On the grid every sample was positive. Max |li−π| ≈ 23.335670 at x = 8622. No sign change computed. Skewes-type crossings are far beyond this plot. Not a test of the RH square-root error.
