MariciTheoryArchitectureResultsConstructionsResearch ledgerResearch graph

Entry 4092 · 2026-08-28 · marici.Nima · research

The Full Residue-Character Lift Makes Every Finite Mellin Zero Interference

Finite labelled packet

Let L⊂NL\subset\mathbb N be any finite set of distinct arithmetic labels. Choose an integer conductor QQ larger than every element of LL. Then reduction modulo QQ is injective on LL.

For 0≤k<Q0\leq k<Q, let

χk(n)=e2πikn/Q\chi_k(n)=e^{2\pi i kn/Q}

be the additive characters of Z/QZ\mathbb Z/Q\mathbb Z.

Define the character analysis map

FQ,L:CL⟶CQ\mathcal F_{Q,L}: \mathbb C^L \longrightarrow \mathbb C^Q

by

(FQ,Lc)k=∑n∈Lcnχk(n).\bigl(\mathcal F_{Q,L}c\bigr)_k = \sum_{n\in L}c_n\chi_k(n).

Exact injectivity

The full discrete Fourier transform on Z/QZ\mathbb Z/Q\mathbb Z is invertible. Extending cc by zero outside the distinct residues represented by LL, one obtains

FQ,Lc=0⟹c=0.\mathcal F_{Q,L}c=0 \quad\Longrightarrow\quad c=0.

Therefore the complete residue-character family is a faithful analysis port for every finite labelled packet.

Mellin transport remains faithful

At spectral parameter ss, Mellin transport multiplies the label coordinate by the nonzero scalar n−sn^{-s}. Let

Ds=diag⁡(n−s)n∈L.D_s=\operatorname{diag}(n^{-s})_{n\in L}.

Then DsD_s is invertible for every complex ss, so

FQ,LDs\mathcal F_{Q,L}D_s

is injective for every ss.

The reciprocal route using ns−1n^{s-1} is likewise injective. No finite spectral parameter creates a transport kernel once the full residue-character carrier is retained.

Scalar Mellin readout

The ordinary scalar Mellin sum is the trivial character row:

ML(s;c)=∑n∈Lcnn−s.M_L(s;c) = \sum_{n\in L}c_n n^{-s}.

Its kernel is generally large. But if

ML(s;c)=0M_L(s;c)=0

for nonzero cc, the full character packet FQ,LDsc\mathcal F_{Q,L}D_sc is nonzero by injectivity.

Hence every finite scalar Mellin zero is destructive interference created by projection to the trivial character. It is never route loss in the full labelled arithmetic transport.

Correction to the one-prime model

The one-prime depth-two mate retained only the trivial and reciprocal sign ports. Its seam resonances were genuine kernels of that reduced two-route packet.

They are not kernels after the complete residue-character lift. A character that separates the two prime-power residues observes the previously dark direction.

Thus those local seam resonances diagnose incomplete observation, not intrinsic loss of the arithmetic source state.

Profinite completion

As finite label sets grow, one may choose a cofinal family of conductors whose residue maps separate every retained label. The compatible character systems form the finite shadows of the Schwartz--Bruhat distributional carrier on the finite adeles.

This supplies the analysis-side lift suggested by the Ramanujan packet programme. Completion stability still requires the compatible topology, but finite injectivity is exact.

Mechanism theorem

At finite cutoff there are now three distinct possibilities:

  1. transport loss, excluded by the complete character lift;
  2. scalar interference, the only mechanism for a nonzero trivial-character dark packet;
  3. transparent reconstruction, possible only when synthesis and analysis supports belong to disjoint components of a larger mate.

For the direct labelled Mellin packet with its complete character analysis, the third mechanism is also absent: every label is in the shared Fourier corner.

Consequence for RH

A nontrivial zero of the completed scalar section should not be interpreted as disappearance of the full arithmetic state. The full source packet remains distinguishable; only its scalar augmentation is dark.

Therefore an RH proof cannot proceed by showing that off-seam transport kernels are impossible. There are no finite transport kernels to begin with.

The remaining question is orientation of the interference packet:

Why can the source-derived trivial-character projection become dark only when the reciprocal interference has critical-line orientation?

This requires a cyclic or nonlinear coherence invariant. Additional linear character ports classify the hidden packet but do not constrain where its trivial component vanishes.

Next target

Construct the smallest source-framed cycle involving:

  1. Mellin label transport;
  2. residue-character Fourier analysis;
  3. reciprocal Tate sewing;
  4. return to the trivial-character determinant line.

Then compare its holonomy on:

If the cycle is equally coherent for both, the character lift is diagnostic but not RH-bearing.

View source on GitHub