The η-Ladder — Sacred Inputs, Free Outputs

There is a temptation that shadows any system built from a few revered numbers. The engine takes three of them, e, the golden ratio φ, and √3, and folds the world out of their company. So surely, the thought goes, those numbers should be everywhere. Surely if you measured the engine deeply enough you would keep finding e and φ staring back. It is the warm, dangerous instinct of the pattern-seeker, and the honest thing to do with it is to put it to a hard test. So I did, and the answer is the opposite of the romance, and better than it.

The engine has a fingerprint

For this engine the quantities at issue are the log-area cumulants: take the logarithm of the absolute Jacobian determinant and ask after its mean, variance, and higher cumulants. The exact calculation uses ε=0, μ=1, α≠0, and independent uniform phases. With Z = log|det J| − 2 log|α|, the normalized observable has the same law as log|cos θ| for uniform θ. This is the engine’s analytic phase model; its agreement with a deterministic orbit is a separate measurement.

So I listened. What came out was not a scatter of numbers, and it was certainly not the sacred constants. It was a ladder: one clean, ordered family, each rung an exact special-function value, each following the one below it by a strict rule.

The ladder

The first rung is −ln 2. The second is π²/12. The third is −3ζ(3)/2, carrying Apéry’s constant; the fourth, 7π⁴/120; and every rung above is a fresh value from the same family, the Dirichlet-η function, the alternating sibling of Riemann’s zeta. A single law generates the whole stack, and once you have it you can write down rung after rung exactly. For the first seven, two evaluations of the uniform-phase formulas agree to 28–41 decimal places. Those digits verify the analytic calculation, not an orbit measurement.

The lowest rung is worth a pause. −ln 2 is the normalized mean at μ=1. The related −2 ln 2 is the intercept in the general uniform-phase area law, 2 log|αμ| − 2 ln 2 + log(1+μ⁻²), for μ≠0. They belong to the same calculation, but they are not the same rung. To recover the raw log-area mean at μ=1, add 2 log|α| to −ln 2.

What is not there

The recorded finite census found no integer relation in the tested bases and search bounds for five invariants, and no tested ladder rung within two percent of its list of ten sacred constants. That is a useful result: these observables did not simply echo the selected inputs.

A finite relation search does not prove arithmetic independence or exclude every possible algebraic relation. Nor does this census cover every observable the engine can produce. A different full-period derivative-gain observable contains Catalan’s constant. The η-ladder is a coherent family with a precise scope, not a complete inventory of the machine’s voice.

Inputs and outputs

There is a natural way to read the wall, and it is the cleanest thing in the whole note. The sacred constants are the engine’s inputs: the dials you turn, the settings you feed it. The Dirichlet-η ladder is its output, what this specified phase average produces. At μ≠1 an additional μ-dependent term changes the cumulants; a non-uniform orbit measure changes them again. The engine does not derive its own sacred constants out of itself; it derives ln 2, π²/12, ζ(3), and the rest. You hand it φ; it hands you back the zeta function.

Said plainly: the engine speaks a different dialect than the one it was written in. Its chosen constants are the dials; its computed special functions are another voice. The gap between the two is not a wound in the theory, it is one of its quietly satisfying features, because it means the engine is generative rather than circular. A machine that only ever echoed its own inputs would be a mirror. This one is a voice.

Why this matters for the rest of the work

A nearby numerical value and an exact identity are different discoveries. Here the exact formulas carry the result: they tell us which constants occur and why. The finite comparison with the sacred pool adds a narrower observation about the tested entries. The reading of inputs and outputs can stand on that distinction without turning a relation-finder’s silence into a theorem.

Where the ladder bends

The ladder is exact under independent uniform phases at the stated slice. Actual engine orbits have their own invariant measures. The recorded large-α comparisons approximate this law, while the canonical operating point shows substantial deviations; those finite observations do not prove an asymptotic invariant-measure limit.

The “inputs versus outputs” sentence is a reading. The normalized cumulant formula is the theorem. The absence of a relation in the tested finite search is a computation. Keeping all three visible gives the interpretation a firm place to stand.

:: ARKHEIN ::

Lineage

Those who arrived at the same question before. This is not the first arrival.

1. 1740

   Leonhard Euler · De summis serierum reciprocarum

   Summed the reciprocal squares to π²/6 and opened the whole world of ζ-values, the family the engine’s voice turns out to belong to. The second rung of the ladder is his π²/12, halved straight from his most famous sum.

2. 1837

   Peter Gustav Lejeune Dirichlet

   The alternating series bearing his name supplies the η values in the cumulant formula. The “free” family has his name on it.

3. 1979

   Roger Apéry · Irrationalité de ζ(3)

   Proved ζ(3) irrational. Its transcendence remains unproved; the third cumulant here is −3ζ(3)/2, before normalization to a skewness.

4. 1997

   David Bailey & Jonathan Borwein · Experimental Mathematics

   Developed high-precision experimental mathematics and integer-relation methods. Such searches can find candidate identities; a finite unsuccessful search does not prove arithmetic independence.