PROFILE / MASK / COHERENT MEAN · PRINCIPAL ν · S ABOVE 2,3
One common profile.
Every surviving column.
A norm determines the actual coefficient and unit phase, while several ideals can share that norm. Inspect how one complex smooth profile aligns them all, and why a sixth-power factor still deletes raw columns.
Loading the hash-bound norm groups and exact raw masks…
THE ACTUAL NORM GROUPS · U² = D · D < N <2D
Read a profile at its arithmetic nodes.
Inspect this complete norm group
Multiplicity counts distinct ideal columns. Duplicate divisor norms are retained. Ideal squarefreeness permits nonsquarefree rational norms.
SAME PRIMITIVE CHARACTER · DIFFERENT RAW DELETIONS
A sixth power still leaves a mask.
π =−2−3ω, π̄ =1+3ω. The two prime squares meet one prime over7; ππ̄ meets both. The common WD is unchanged when v changes.
u =εβv6
χm(u) =χm(ε)1(m,v)=1
These conductor norms are the reviewed prose result, separate from the finite experiment. Squaring drops the 2-part of the primitive conductor. The original S-deletions at2 and3 and the actual v-prime deletions stay in the raw function.
At the selected norm
EXACT SELECTED-ROW ENERGIES · FINITE DATA
Keep the row boundary explicit.
Fixed-X probes, including X=64, have H=X⁶ and are separate from H=⌊D^(3/4)⌋. In these modest capacity-matched cases X≤4, so only v=1 survives the good-S cap. The asymptotic cDX proof requires D,X→∞.
SEPARATE EXACT q₅ FIXTURE · D =4000 · U =63
Zero variance can hide a nonzero mean.
Four actual orientations share the same correlation Z=3 after the displayed unit phase is separated. The selected h=1 A–A sum is positive. Removing that phase gives a wrong answer.
The four v generators are the complete norm 13/19 set in the source fixture. One prime over13 meets q and is deleted; three remain. This fixture is independent of the six D cases above.
| Orientation | Nm / Nn | θ(ε) phase | Signed coefficient basis(a,b) |
|---|
NEW / ONE FIXED COMMON W / ACTUAL A₆=1 ROWS
Keep the diagonal. Keep the mean.
Twenty nonprincipal rows give a positive Gram matrix. Keeping only coprime column pairs removes its diagonal and leaves a signed, indefinite slice. Explore the exact arithmetic behind its negative contribution.
One selected q, two column ideals. D=60000, U=244, H=3833; A₁=37, A₂=7, other ideal factors1 and literal unit1. This is a complete row block for that class, not the whole column family or h=1 sum.
Loading the saved forty raw character values…
EXACT BILINEAR PAIRING
Convention: Q(X,Y)=XᵀG conjugate(Y)/D. ω²+ω+1=0. Cross pairings can be complex; the full matrix is positive for X=Y. Floating gauge coordinates are display projections of exact j/√N values.
The exact mean and the tempting wrong bound
Γ=(−4−2ω, −2+2ω), mean −3. Centered values are (−1−2ω, 1+2ω), with variance average3 and sum6. For A–A, deleting the mean would demand |−6|²=36 ≤ (2)(6)=12. The missing mean term is −6. This is an invalid deletion, not a failure of Cauchy–Schwarz.
The full eigenvalues are 17±√21. The h=1 matrix has eigenvalues ±2√3. Its negative pairing is an off-diagonal contribution, not a negative total row energy.
All twenty rows. All forty raw values.
44 candidate pairs − 2 coprimality rejections − 22 capacity rejections =20. Six rows collide with q, giving12 zero orientation/row evaluations. A raw zero is different from phase0, which equals1.
| Row / Nu | χₘ · Nm97051 | χₙ · Nn96091 |
|---|
The same W is fixed as D varies. W(y)=s(8(y−1))s(8(2−y)), using the displayed smoothstep source. It is supported in[1,2] and equals1 on[9/8,15/8]; both97051/60000 and96091/60000 lie inside that plateau. Other column ideals remain active.
A fixed Hecke twist, not a new bound. Ψ has infinity type(1,0) and finite conductor norm12; it is not a finite-order character. On squarefree columns ρ=μΨ/√N=j/√N. Rotating coefficients and the kernel together preserves the exact pairing. Rotating coefficients alone changes the problem. The growing row conductor and every raw mask remain.
The actual signed all-q estimate, the remaining A₆=1 neighborhoods, positive determinant and GRH remain open. This finite selected-q calculation supplies no asymptotic power gain.
WHAT THE CONSTRUCTION DOES AND DOES NOT ESTABLISH
A varying profile is a different quantifier.
The exact norm-group constancy lets WD align the original AU coefficients, rather than replace them by arbitrary weights for individual ideals. The selected rows εβv6 have the same nonprincipal order6 primitive character, so their actual principal residue is 0. Raw collision zeros are retained.
A separate classical coprime-ideal count gives a cDH1/6 lower bound in this profile-varying class. It matches the DH1/6 term when 0<θ<2/3 and H=Dθ. At θ=3/4 its exponent is 9/8, below the current 7/6 upper envelope; it does not establish7/6 sharpness.
The profiles have ||WD||∞=1 but derivatives Ok(Dk). A bound uniform in this class with Mk(W)α(HD)εH for every ε>0 needs kα≥1−5θ/6; at θ=3/4 this is 3/8. That is a necessary cost, not a sufficient estimate.
The fixed-W target, remaining A₆=1 witness sectors, signed all-row cancellation and arithmetic determinant/zero-location bridge remain open. Positivity on the selected h=1 block does not imply positivity of h=1 over other row classes. No new upper-bound improvement, training, Lean theorem, quantum advantage or GRH proof is claimed.