Zeros on the Wall $\sigma = \tfrac{1}{2}$ — The Arithmetic Attractor
This chapter is about a single number — one-half — and why the prime numbers seem to insist on it. Everything technical that follows serves one plain idea, so it is worth stating the idea before any symbol appears.
The Riemann zeta function is the object that encodes the primes. It carries a built-in symmetry, known since Riemann himself: a kind of mirror that swaps a value on the left of a vertical strip with its partner an equal distance to the right. Fold the strip along its centre and the two halves land on one another. The single line the fold leaves fixed — the crease — is the vertical line at one-half. Every other line has a distinct partner; only the centre line is its own reflection.
In the previous chapter the primes were drawn as a twisting geometric structure — a "contact form" — living on a flat space, and flatness was the problem: nothing in a flat sheet singles out the centre. This chapter curves the space using precisely that mirror symmetry, so the centre line stops being an arbitrary choice and becomes the one geometrically special place — the direct analogue of the unit circle the book's smooth toy model (Chapter 10) was built around. The Riemann Hypothesis, in this picture, becomes the statement that the zeros of the zeta function are pulled onto that crease.
The sections that follow make this precise in three steps. §12.1 writes the mirror down as the zeta function's functional equation. §12.2 shows that this symmetry is a contactomorphism — a transformation that preserves the geometric structure rather than distorting it — and that its only fixed locus is the centre line. §12.3 recasts the Riemann Hypothesis as the claim that this fixed line attracts the zeros, echoing how the toy model's unit circle attracts every nearby orbit. One honest boundary holds throughout: this chapter reformulates known mathematics in geometric language — it does not prove the Riemann Hypothesis. Whether the fixed line is genuinely an attractor is deferred to the Global Positivity Theorem of Chapter 14.
Chapter 11 built the arithmetic contact form $\alpha_{\mathrm{arith}} = c(\sigma,t)\,d\tilde U - g(\sigma,t)\,d\tilde V$ on the flat extended phase space $\mathbb{R}^3_{(U,V,t)}$. The form was shown to be non-integrable — the planes twist, driven by the prime frequencies. But the phase space was flat, and nothing in the flat geometry distinguished $\sigma = \tfrac{1}{2}$ from any other value of $\sigma$.
Chapter 12 curves the space. The curvature comes from a symmetry that the Riemann zeta function has always possessed: the functional equation $\zeta(s) = \chi(s)\zeta(1-s)$. This symmetry pairs $\sigma$ with $1-\sigma$ and identifies a unique fixed locus: the critical line $\sigma = \tfrac{1}{2}$, where $s = 1-s$.
The Riemann zeta function satisfies the functional equation:
In the zeta phase portrait of Ch 11: the trajectory at $\sigma = 0.7$ and the trajectory at $\sigma = 0.3$ are mirror images of each other (up to the $\chi$ factor). The diagram below shows this mirror pairing. Drag the $\sigma$ slider to see how the trajectory at $\sigma$ mirrors the trajectory at $1-\sigma$. At $\sigma = \tfrac{1}{2}$ the trajectory is its own mirror.
A contactomorphism of a contact manifold $(M, \ker\alpha)$ is a diffeomorphism $\phi: M \to M$ that preserves the contact structure: $\phi^*\alpha = f\alpha$ for some nonzero function $f$. It maps contact planes to contact planes — the geometric structure is respected.
The derivation is the logarithmic derivative of $\zeta(s) = \chi(s)\zeta(1-s)$, taken together with the fact that $c$ is even in $t$ and $g$ is odd in $t$ — both immediate from $c$ and $g$ being the cosine and sine coefficients of $\textstyle\sum_n \Lambda(n) n^{-\sigma}$. Confirmed numerically to thirty digits at eight points, with $\sigma \in \{0.3,\,0.5,\,0.8,\,1.1,\,1.5,\,2.3\}$ and $t$ from $0.7$ to $25$, inside the critical strip and outside it.
The critical line is distinguished by this, and not by analogy. The constraint $g(\sigma,t) = g(1-\sigma,t)$ holds identically on $\sigma = \tfrac{1}{2}$ and on no other vertical line. Chapter 10 supplies the picture this instantiates — a unit circle that attracts because the geometry gives it no alternative — but the mechanism here is arithmetic: it is the gamma factor of the functional equation becoming real exactly on the wall.
That identification is the one worth pausing on, because it is not internal to this framework. $\vartheta$ is the phase of the completed zeta function, and the Riemann–von Mangoldt formula reads $N(T) = \vartheta(T)/\pi + 1 + S(T)$ — the smooth part of the count of zeros up to height $T$. So the $d\tilde U$ coefficient of $\alpha_{\mathrm{arith}}$, restricted to the critical wall, is exactly the density of the zero-counting function, and its integral along the wall counts the zeros. The contact form is not analogous to classical analytic number theory here; on that line it is classical analytic number theory, written in a different alphabet.
And the two coefficients divide the labour completely. At a zero $\rho = \tfrac{1}{2}+i\gamma$ the logarithmic derivative has a simple pole, and along the wall $s - \rho = i(t-\gamma)$ is purely imaginary — so the pole falls entirely into $g$ and not at all into $c$. Numerically, $g(\tfrac12,\gamma+\delta) \approx -1/\delta$ ($-1000.08$ at $\delta = 10^{-3}$, $-10000.08$ at $10^{-4}$), while $c$ converges quietly to $\vartheta'(\gamma)$. On the critical line, $c$ is prime-free, smooth, and counts; $g$ carries every pole, one per zero. §12.6's claim that each zero is a fold singularity now has a precise form: the singularity is a simple pole in the $d\tilde V$ component, and the $d\tilde U$ component is analytic across it.
The revised conjecture remains well-posed and remains independent of the Riemann Hypothesis: its truth or falsity would be a concrete contribution to arithmetic contact geometry, and Result 12.1 has already converted part of it from a question into an identity.
To be deposited. The framework of Chapters 11–12, together with the results of this section, is set out in The Riemann Hypothesis as Non-Integrability of an Arithmetic Contact Structure on the Adele Class Space, DOI reserved: 10.5281/zenodo.22179684 — preprint, not peer reviewed; the record is not yet published, so the link will not resolve until it is. Its §4.6 carries a status-of-claims table separating what is proved, what is machine-checked, what is classical, and what is numerical only.
Machine-checked, and how far. The Lean file is ZetaReflection.lean. As audited on 2 September 2026 it carries eleven theorems, ten of them proved on [propext, Classical.choice, Quot.sound] and nothing else, and one admitted.
Proved: lseries_vonMangoldt_eq_neg_Zlog, that $c$ and $g$ are the real and imaginary parts of $-\zeta'/\zeta$ — the step that ties this chapter to Mathlib’s zeta rather than to prose; Zlog_conj, that $\zeta'/\zeta$ is conjugation-symmetric; gCoef_odd_in_t and cCoef_even_in_t, the parities of $g$ and $c$ in $t$, which are what bridge $s=\sigma+it$ to $1-s=(1-\sigma)-it$; digamma_conj, that $\psi(\bar s)=\overline{\psi(s)}$; one_sub_conj, that $1-s=\bar s$ holds exactly on $\sigma=\tfrac{1}{2}$; and chiLog_real_on_critical_line, the reality of $\chi'/\chi$ on the critical line — the two digamma arguments are conjugates there, so their sum is real. That last one is why $\sigma=\tfrac{1}{2}$ is distinguished, and it is a theorem rather than a numerical observation.
Admitted: Result 12.1 alone, as reflection_law, which carries sorryAx. It is stated with four hypotheses — $s$ and $1-s$ off the poles of $\Gamma_{\mathbb{R}}$, and $\zeta(s),\zeta(1-s)\neq 0$ — because without them the identity would be a claim about junk values at the zeros, and its truth would depend on where those zeros are. Numerically confirmed is not proved, and the file says which is which.
In the dm³ system of Ch 10, the unit circle $r=1$ played a double role: (1) it was the unique fixed set of the Hopf normal form symmetry $r \mapsto r$ (trivially), and (2) it was a globally attracting helix under the dynamics. The key theorem was that $r=1$ is not just an invariant set but an attractor: every outer-basin orbit converges to it.
For the arithmetic system:
The contact-geometric framework makes this analogy precise: RH is the statement that the critical line is an attracting fixed set of the arithmetic contact flow. The dm³ system is the smooth prototype; the arithmetic system is the prime-number instantiation.
When we curve the contact manifold by the functional equation symmetry, the flat $\mathbb{R}^3_{(U,V,t)}$ becomes a quotient-like space where $\sigma$ and $1-\sigma$ are identified via $\Phi$. The fundamental domain is the half-space $\sigma \geq \tfrac{1}{2}$, with the boundary $\sigma = \tfrac{1}{2}$ being the critical wall.
The contact form $\alpha_{\mathrm{arith}}$ on this curved manifold inherits a boundary condition from the functional equation: at $\sigma = \tfrac{1}{2}$, the form must satisfy $\Phi^*\alpha = f\alpha$, which constrains $g(\tfrac{1}{2}, t)$ in terms of itself via the functional equation. This is the arithmetic analogue of the condition $r(1-r^2) = 0$ at $r=1$ in the dm³ system — the attractor condition made explicit.
In the GTCT operator chain, K is the threshold: the minimum input required to fire, the gate between compression and fold. In the arithmetic context, the K-threshold is the critical value of $\sigma$ at which the contact structure changes qualitatively.
Specifically: for $\sigma > 1$, the Dirichlet series converges absolutely and $\alpha_{\mathrm{arith}}$ is a well-defined smooth contact form. As $\sigma$ decreases through $1$ (the pole of $\zeta(s)$), the form acquires its first singularity. As $\sigma$ approaches $\tfrac{1}{2}$ from above, the form approaches the critical wall where the functional equation imposes its constraint. The K-threshold is $\sigma^* = \tfrac{1}{2}$.
| dm³ system (Ch 10) | Arithmetic system (Ch 11–12) |
| Attractor $r = 1$ | Critical line $\sigma = \tfrac{1}{2}$ |
| Hopf symmetry $r \mapsto 2-r$ | Functional eq. $s \mapsto 1-s$ |
| Inner basin $r^* \approx 0.776$ | Strip boundary $\sigma \in (0,1)$ |
| Decay rate $\mu \to -2$ | Positivity: $\partial_t g > 0$ on $\ker\alpha$ (?) |
What do the contact planes $\ker\alpha_{\mathrm{arith}}$ look like at $\sigma = \tfrac{1}{2}$? At a zero $s_0 = \tfrac{1}{2} + it_0$, the coefficient $g(\tfrac{1}{2}, t_0)$ diverges (simple pole of $\zeta'/\zeta$). The contact plane rotates infinitely rapidly — in differential-geometric language, the contact structure has a fold singularity at each zero.
This is the arithmetic F-fold of the GTCT chain: the fold event that makes K-crossings permanent. Each zero of $\zeta(s)$ is a fold in the arithmetic contact manifold. The Riemann Hypothesis says these folds occur exclusively on the critical wall $\sigma = \tfrac{1}{2}$.
The reflection is superseded, and this is where that is recorded.
On 8 September 2026 Mathlib merged
logDeriv_riemannZeta_one_sub (PR #43252), in the new file
Mathlib/NumberTheory/LSeries/RiemannZetaLogDeriv.lean:
theorem logDeriv_riemannZeta_one_sub {s : ℂ} (hs : ∀ n : ℤ, s ≠ n) (hz : riemannZeta s ≠ 0) :
logDeriv riemannZeta s =
-logDeriv riemannZeta (1 - s) + log (2 * π) - digamma s + π / 2 * tan (π * s / 2)That is the same identity as this chapter's Zlog_add_Zlog_one_sub.
Applying the digamma duplication formula to log(2π) − ψ(s) and then the
reflection formula at (1+s)/2 — where
ψ((1−s)/2) − ψ((1+s)/2) = π·cot(π(1+s)/2) = −π·tan(πs/2) — carries one
right-hand side into the other exactly. The result stated here as new is not new.
It is in the library, under someone else's name, and that is the correct outcome. The convergence reading of the same event — one theorem reached twice in one week, by two routes, inside a library that keeps a version-controlled record of both — is written up as a multiple in Omega Point · In the Air, where it is also conceded that shared prerequisites make it constrained discovery rather than independent discovery.
What does not follow from it is the domain. The merged statement
assumes hs : ∀ n : ℤ, s ≠ n — every integer excluded — because its
right-hand side carries digamma s, whose poles are the whole of
{0, −1, −2, …}. The route through the archimedean factor carries
digamma (s/2) and digamma ((1−s)/2) instead. Halving the
argument halves the pole set: the singularities are exactly
s ∈ {0, −2, −4, …} and s ∈ {1, 3, 5, …}, which are exactly
the hypotheses hΓ and hΓ' — no more and no less.
Consequence. The Gammaℝ-route functional
equation is stated at every negative odd and every positive even
integer, where the ζ-direct form is not.
Verification status. That the hypotheses admit s = −1
and s = 2 while hs excludes them is machine-checked
— DomainCheck.lean, five theorems, no sorry, compiled clean
against Mathlib v4.32.0 on 11 September 2026. That the identity itself holds
at those points is not yet machine-checked; it requires instantiating
Zlog_add_Zlog_one_sub there, and until that is done the claim stands at
the grade of its hypotheses only.
One hypothesis comes out. Given hΓ and hΓ',
Gammaℝ is non-zero at both s and 1−s, so
Λ(s) = Λ(1−s) makes ζ(s) = 0 ⟺ ζ(1−s) = 0. Therefore
hζ' follows from hζ and is redundant. Three hypotheses,
same theorem. hΓ' is not redundant — it is the hypothesis doing
the work that widens the domain.
And the lemma underneath is absent from the library.
Mathlib/Analysis/SpecialFunctions/Gamma/Deligne.lean carries eighteen
declarations and no derivative lemma for Gammaℝ of any kind — only
differentiable_Gammaℝ_inv. There is no differentiableAt_Gammaℝ
and no logDeriv_Gammaℝ. Checked against the published documentation,
11 September 2026. Since Gammaℝ is the archimedean factor for every
completed L-function and not for ζ alone, that gap is the part of this chapter's
work that is still open ground.
| Claim | Status |
|---|---|
| Functional equation $\zeta(s) = \chi(s)\zeta(1-s)$ | ✓ Classical theorem |
| $s \mapsto 1-s$ has unique fixed line $\sigma = \tfrac{1}{2}$ | ✓ Trivially true |
| The critical line is an invariant set of the symmetry | ✓ Proved |
| Functional equation induces a contactomorphism of $(M,\alpha_{\mathrm{arith}})$ | ○ Conjecture 12.1 — open, well-posed |
| Critical line is an attractor for zeros of $\zeta(s)$ | ○ = Riemann Hypothesis |
Chapter 12 works globally: the functional equation is a symmetry of the whole function $\zeta(s)$. Chapter 13 goes local: the Euler product $\zeta(s) = \prod_p (1 - p^{-s})^{-1}$ decomposes the global function into local pieces, one for each prime $p$. The arithmetic contact manifold decomposes accordingly into p-adic local pieces — the adelic structure. Each local piece is a simpler contact form $\alpha_p$, and the global form is assembled from these by the adelic product. Chapter 13 builds this decomposition explicitly.
Everything this chapter does with the functional equation has an exact counterpart one column over in Weil’s analogy, and in that column it is a theorem. Setting the two side by side is the fastest way to see both what the contact reading is aiming at and what it cannot inherit.
| $\zeta(s)$ over $\mathbb{Q}$ — this chapter | $Z_C(T)$ over $\mathbb{F}_q$ — Weil | |
|---|---|---|
| the object | $\zeta(s) = \prod_p (1-p^{-s})^{-1}$ | $Z_C(T) = \exp\!\big(\sum_{n\ge1} \#C(\mathbb{F}_{q^{n}})\,T^{n}/n\big)$ |
| continuation | meromorphic on $\mathbb{C}$, one simple pole at $s=1$ | rational: $P(T)/\big((1-T)(1-qT)\big)$, $\deg P = 2g$ — Dwork 1960, Grothendieck |
| functional equation | $\Lambda(s) = \Lambda(1-s)$, $\Lambda = \Gamma_{\mathbb{R}}\cdot\zeta$ | $Z_C\!\big(1/(qT)\big) = q^{1-g}T^{2-2g}\,Z_C(T)$ |
| the involution | $s \mapsto 1-s$ | $T \mapsto 1/(qT)$ |
| its fixed locus | $\sigma = \tfrac12$ | $|T| = q^{-1/2}$ |
| zeros on that locus | ○ open — the Riemann Hypothesis | ✓ proved — Hasse (genus 1), Weil 1940/48 (all genus) |
| what closes it | ○ nothing yet | ✓ intersection positivity on the surface $C \times C$ (Castelnuovo–Severi); in higher dimension, $\ell$-adic positivity (Deligne 1974) |
Read the fifth and sixth rows together. In the right-hand column, the sentence “the fixed locus of the functional-equation involution is exactly where the zeros are” is not a conjecture about an attractor. It is a proved statement, and it has been one since 1940. That sentence is precisely what Conjecture 12.1 is reaching for. The analogy is therefore doing real work: it tells you the target is the right shape.
Over $\mathbb{F}_q$ the Frobenius map is an honest endomorphism of an honest variety, and the curve $C$ supplies a surface $C \times C$ on which correspondences can be intersected. Weil’s proof runs on the Castelnuovo–Severi inequality — an intersection-theoretic positivity on that surface. Deligne’s 1974 proof for varieties of any dimension runs on a different positivity, inside $\ell$-adic cohomology. In both cases a geometric object supplies an inequality, and the inequality is the proof.
Over $\mathbb{Q}$ there is no such surface. No Frobenius, no ambient variety to intersect in, and so no source for the inequality. Weil himself gave the sharpest statement of the gap: the explicit formula of 1952 converts the Riemann Hypothesis into the positivity of a distribution — Weil’s criterion. That conversion is an exact equivalence and it is not a proof. It relocates the whole difficulty into one sentence and leaves it standing there.
Where this chapter actually stands. Conjecture 12.1, if proved, would establish that the critical line is an invariant set of a contactomorphism. It would not produce an inequality. The invariance is already known and is cheap — it is row five, left column, and it follows from the functional equation alone. The expensive object is the positivity, and the contact form $\alpha_{\mathrm{arith}}$ does not yet supply one. Stating that plainly is the difference between a programme and a claim.
RELATED-WORK.md that appears to pass the wall should be read again.
Attribution. The mathematics above is Weil’s, Hasse’s,
Dwork’s, Grothendieck’s and Deligne’s; nothing in this section is claimed
by this series. The author chapter is
Book 7 · André Weil, which carries the history, the
1940 Rosetta-stone letter, the record that Grothendieck’s own intended route — the
standard conjectures on algebraic cycles — is still open, and a verification
script, book7/ch-weil-verify.py, that establishes the genus-1 case by exhaustive
count: 66 curve/prime pairs with every $p^{2}$ point tested, 132 reciprocal roots at
$|\alpha| = \sqrt{p}$, and the $\mathbb{F}_{p^{2}}$ counts predicted from the roots and
confirmed by exhaustion over all $p^{4}$ pairs. The operator-algebra side is
Book 7 · Alain Connes. The screw-line comparison
is in Ch 11 · Honest Inventory.
sorryAx. A clean axiom report is not a reading of the statement: per R20, a theorem can assume its conclusion and still report clean. Follow the link before citing one as evidence.