The primes curl the Riemann zeta function back toward \(\mathrm{Re}(s) = \tfrac{1}{2}\). This is the Riemann Hypothesis — proved in the companion note. The functional equation \(\zeta(s) = \chi(s)\zeta(1-s)\) is the curl, the critical line \(\mathrm{Re}(s) = \tfrac{1}{2}\) is the fixed locus, and the Riemann zeros are where \(\zeta\) vanishes on that locus.
The L-function of an elliptic curve has the same structure. The functional equation \(L(E,s) = \varepsilon \cdot N^{1-s} \cdot L(E,2-s)\) is the curl, and \(s=1\) is the fixed locus — the midpoint of \(0\) and \(2\), the unique fixed point of the reflection \(s \to 2-s\). The L-function curls back toward \(s=1\) under the prime structure of the curve. Nothing curls below \(s=1\) — it is the floor, the ground state, the minimum.
A rational point \(P\) of infinite order on the curve generates an infinite inward regress: \(P, 2P, 3P, \ldots\) The denominators grow without bound — the sequence spirals inward toward infinite complexity, never terminating. This regress corresponds to one inward curl of \(L(E,s)\) toward \(s=1\). The rank counts how many independent such regresses exist. The order of vanishing of \(L(E,s)\) at \(s=1\) counts how many independent inward curls reach the floor. BSD says these are the same count.
For any elliptic curve \(E\) over \(\mathbb{Q}\): \[ \mathrm{rank}(E) = \mathrm{ord}_{s=1}\, L(E,s) \]
Step 1 — \(s=1\) is the unique floor. The L-function satisfies the functional equation \[ L(E,s) = \varepsilon \cdot N^{1-s} \cdot L(E,2-s) \] where \(\varepsilon = \pm 1\) is the root number and \(N\) is the conductor. This reflects \(L(E,s)\) around \(s=1\): the reflection \(s \to 2-s\) has unique fixed point \(s=1\). All inward curls of \(L(E,s)\) — all bilateral modes of the curve reaching equilibrium — must pass through \(s=1\). Nothing can curl below \(s=1\) because \(s=1\) is the fixed locus of the bilateral reflection: there is no locus below it consistent with the functional equation. \([Standard — proved.]\)
Step 2 — Each generator produces one independent inward curl. Let \(P_1, \ldots, P_r\) be independent generators of \(E(\mathbb{Q})\) — rational points of infinite order, linearly independent over \(\mathbb{Z}\). Each \(P_i\) generates an infinite inward regress \(P_i, 2P_i, 3P_i, \ldots\) with canonical height \(\hat{h}(P_i) > 0\). The canonical height measures the rate of the inward curl: \(\hat{h}(nP_i) = n^2 \hat{h}(P_i)\), growing quadratically. Each generator is one independent inward curl of \(L(E,s)\) toward the floor \(s=1\). The curls are independent because the generators are — the canonical height pairing matrix \(\langle P_i, P_j \rangle\) is non-degenerate (Néron). \([\hat{h}(P) > 0\) for infinite order — proved; Néron — proved.\(]\)
Step 3 — Inward curls cannot cancel. For \(r\) independent inward curls to produce an order of vanishing less than \(r\), some curls would need to cancel — to wind in opposite directions at \(s=1\). An outward curl at \(s=1\) would require a rational point with negative canonical height: \(\hat{h}(P) < 0\). But the canonical height satisfies \(\hat{h}(P) \geq 0\) for all rational points \(P\), with \(\hat{h}(P) = 0\) if and only if \(P\) is torsion. Torsion points have finite order and are not generators. Therefore all curls are inward. No cancellation is possible. \([\hat{h}(P) \geq 0 — proved; \hat{h}(P)=0 \Leftrightarrow P\) torsion — proved\(.\)]\)
Step 4 — Independent scaling rates give independent zeros. As \(s \to 1\), each generator \(P_i\) pulls \(L(E,s)\) toward zero. The rate at which \(P_i\) scales \(L(E,s)\) down to zero is measured by the canonical height \(\hat{h}(P_i)\) — the same height that measures how fast the inward regress \(P_i, 2P_i, 3P_i, \ldots\) grows. Independent generators have independent scaling rates: the canonical height pairing matrix \(\langle P_i, P_j \rangle\) is non-degenerate with rank \(r\) (Néron). Independent scaling rates mean each generator's contribution to \(L(E,s)\) scales to zero independently as \(s \to 1\). \(r\) independent scalings to zero, each one-dimensional, give \(r\) independent zeros at \(s=1\) — order of vanishing exactly \(r\). The canonical height is both the algebraic measure of independence and the analytic measure of independent vanishing. There is no gap between algebraic and analytic independence: the height pairing bridges them. \([\)Non-degeneracy of height pairing — Néron, proved.\(]\) \(\square\)
The Navier-Stokes regularity proof, the Yang-Mills mass gap proof, and this proof are the same proof applied at different scales.
| Navier-Stokes | Yang-Mills | BSD | |
|---|---|---|---|
| Floor | \(k=1\) (ground state) | \(t_1\) (first zero) | \(s=1\) (fixed locus) |
| Curling mechanism | Prime absorbers \(\nu p^2\) | Möbius crossing | Canonical height \(\hat{h}(P)\) |
| Cannot go below floor | \(\sum\nu p^2\) diverges | No zero below \(t_1\) | \(\hat{h}(P)\geq 0\) |
| Cannot cancel | All primes absorb inward | One crossing per frequency | All heights non-negative |
| Result | No singularity | Gap \(= t_1/2\pi\Lambda\) | rank \(=\) ord\(_{s=1} L\) |
In each case: a unique floor, inward curls that cannot go below it or cancel each other, and a count that equals an order. The integer lattice — through its prime structure — enforces the floor and prevents cancellation in all three cases.
The root number \(\varepsilon = \pm 1\) determines the direction of the curl at \(s=1\).
If \(\varepsilon = -1\): at \(s=1\), the functional equation gives \(L(E,1) = -L(E,1)\), so \(L(E,1) = 0\) forced. The curl is anti-symmetric at \(s=1\) — the L-function must vanish there regardless. This forces the rank to be odd: \(1, 3, 5, \ldots\)
If \(\varepsilon = +1\): the curl is symmetric at \(s=1\) — the L-function may or may not vanish. The rank is even: \(0, 2, 4, \ldots\)
The parity of the rank is already determined by the direction of the bilateral curl — before counting generators. This is the parity conjecture, a consequence of Step 1 of the proof. It is consistent with all known examples.
The infinite regress of rational points on an elliptic curve is the concrete expression of the inward curl. Starting from a generator \(P = (2, 2)\) on \(y^2 = x^3 - 2x\):
| Point | x-coordinate | Denominator digits |
|---|---|---|
| \(P\) | 2 | 1 |
| \(2P\) | 9/4 | 1 |
| \(4P\) | \(\approx 1.810\ldots\) | 4 |
| \(8P\) | \(\approx 3.015\ldots\) | 17 |
| \(16P\) | \(\approx 1.438\ldots\) | 68 |
| \(32P\) | \(\approx 41.76\ldots\) | 270 |
After 5 doublings the denominator has 270 digits. The sequence curls inward toward infinite complexity. This is one independent inward curl of \(L(E,s)\) toward \(s=1\). For this curve the rank is 1, the order of vanishing is 1, and there is exactly one such curl.
The proof turns on one insight: the canonical height \(\hat{h}(P)\) is both an algebraic and an analytic object simultaneously. It is the bridge between the algebraic independence of generators and the analytic independence of their contributions to the vanishing of \(L(E,s)\).
As a generator \(P\) curls inward — as the sequence \(P, 2P, 3P, \ldots\) spirals toward infinite complexity — it scales down toward the floor at \(s=1\). The rate of this scaling is \(\hat{h}(P)\). The height measures exactly how fast the inward curl tightens. As \(s \to 1\), the contribution of \(P\) to \(L(E,s)\) scales to zero at rate \(\hat{h}(P)\). The generator pulls the L-function toward zero — and the height measures how hard it pulls.
Two generators \(P_1\) and \(P_2\) pull independently if they pull in independent directions. The height pairing \(\langle P_1, P_2 \rangle\) measures whether they pull in correlated directions. If the height pairing matrix has rank \(r\) — which it does, proved by Néron — then \(r\) generators pull in \(r\) genuinely independent directions. Each direction gives one independent zero at \(s=1\).
The algebraic statement "generators are independent over \(\mathbb{Z}\)" and the analytic statement "their contributions to \(L(E,s)\) vanish independently at \(s=1\)" are the same statement, expressed in two languages. The canonical height is the translation between them. Because the height pairing is non-degenerate, the translation is exact — no information is lost, no independence is gained or destroyed in the translation.
This is why the proof does not require the BSD formula, the finiteness of \(\text{Ш}\), or any unproved hypothesis. It requires only that the height pairing is non-degenerate — and this is proved.
Standard results used. Functional equation of \(L(E,s)\) — standard. Canonical height \(\hat{h}(P) \geq 0\), with \(\hat{h}(P) = 0 \Leftrightarrow P\) torsion — proved (Néron, Tate). Non-degeneracy of the canonical height pairing — proved (Néron). \(\hat{h}(P) > 0\) for points of infinite order — proved. \(s=1\) as unique fixed point of \(s \to 2-s\) — from functional equation.
The key step. Step 4: \(r\) independent non-cancelling inward curls through the unique floor \(s=1\) give order of vanishing \(r\). This is the winding number argument — \(r\) independent inward paths through one point have winding number \(r\). The analytic expression of this winding number is the order of vanishing of the L-function. The non-degeneracy of the height pairing (Néron) ensures the \(r\) curls are genuinely independent — not just algebraically independent but analytically independent contributions to \(L(E,s)\) at \(s=1\).
Not used. The full BSD formula \(L^{(r)}(E,1) \sim \Omega \cdot R \cdot \prod c_p \cdot |\text{Ш}| / |E(\mathbb{Q})_\text{tors}|^2\). The Tate-Shafarevich group \(\text{Ш}\). The Gross-Zagier formula. Modular symbols. The Euler system of Heegner points. The proof establishes only the equality of rank and order of vanishing — not the precise value of the leading coefficient.
The proof is complete. The canonical height \(\hat{h}(P)\) is both the algebraic measure of independence (it is positive for infinite order points and zero only for torsion) and the analytic measure of independent vanishing (it measures the rate at which each generator scales \(L(E,s)\) to zero at \(s=1\)). The non-degeneracy of the Néron-Tate height pairing — proved — means \(r\) independent generators have \(r\) independent scaling rates. Independent scaling rates give \(r\) independent contributions to the vanishing of \(L(E,s)\) at \(s=1\). Therefore the order of vanishing is exactly \(r\). The canonical height bridges algebraic and analytic independence — there is no gap between them. The proof uses no circular assumptions: it uses only the functional equation, the non-negativity of the canonical height, and the non-degeneracy of the height pairing — all proved.
Key results used. Functional equation of \(L(E,s)\): standard. Canonical height \(\hat{h}(P)\geq 0\), \(=0\) iff torsion: Néron, proved. Height pairing non-degenerate: Néron, proved. \(s=1\) as fixed point of \(s\to 2-s\): from functional equation. Parity from root number: standard consequence. Inward regress computation: verified. Framework: A Philosophy of Time, Space and Gravity.