# Zero Genesis Game 23｜Critical Shell Stationarity、Compact Translation Hull 與 Covariance Completeness

**Series:** SECV / Zero Genesis Experimental Playground  
**Experiment:** Zero Genesis Game 23  
**Version:** v0.1  
**Author:** Neo.K  
**AI Collaboration:** Aletheia (GPT-5.6 Sol)  
**Status:** Exploratory / Log-Time Almost Periodicity and Stationarity Classification  

**Important:** This document does **not** claim a proof of RH. Wintner's RH-conditional limiting-distribution theorem for the normalized prime-number-theorem error and the modern $B^2$ almost-periodic framework of Akbary–Ng–Shahabi are classical. The purpose here is to combine those results with the zero-preserving localized shell operator from Game 21 and derive an exact stationarity/nonstationarity classification for the shell process.

---

## 0. Motivation

Game 22 reduced the local arithmetic innovation problem to one scalar degree of freedom:

$$
u_k=d_{k+1}-d_k,
$$

the difference of normalized prime densities in adjacent multiplicative shells.

The local congruence spectrum was shown to be asymptotically rank one.

Thus the remaining issue is no longer modular shape.

It is log-scale coherence.

Game 23 studies the critically normalized shell process

$$
\boxed{
Z_A(t)
=
e^{-t/2}
\mathfrak W_A(e^t),
}
$$

where

$$
\boxed{
\mathfrak W_A(x)
=
[\psi(Ax)-\psi(x)]
-
A[\psi(x)-\psi(x/A)],
\qquad
A>1.
}
$$

The question is:

> Is RH exactly the condition that this shell process is stationary/almost periodic in logarithmic time?

The answer is yes in several precise senses.

---

# 1. Base critical prime residual

Define

$$
\boxed{
R(t)
=
e^{-t/2}
[
\psi(e^t)-e^t
].
}
$$

Let

$$
\boxed{
\delta=\log A.
}
$$

Game 21 established the exact relation

$$
\boxed{
Z_A(t)
=
\sqrt A
[
R(t+\delta)+R(t-\delta)
]
-
(A+1)R(t).
}
$$

Define the positive shell operator

$$
\boxed{
\mathcal L_A
=
(A+1)I
-
\sqrt A
(
T_\delta+T_{-\delta}
).
}
$$

Then

$$
\boxed{
Z_A
=
-\mathcal L_A R.
}
$$

---

# 2. Critical frequency multiplier

On a log-frequency mode

$$
e^{i\omega t},
$$

the operator has eigenvalue

$$
\boxed{
P_A(\omega)
=
A+1
-
2\sqrt A
\cos(
\omega\log A
).
}
$$

Equivalently,

$$
P_A(\omega)
=
P_A
\left(
\frac12+i\omega
\right)
$$

in the notation of Game 21.

The bounds are

$$
\boxed{
(\sqrt A-1)^2
\le
P_A(\omega)
\le
(\sqrt A+1)^2.
}
$$

Thus the critical shell operator has a uniform spectral gap on the entire real frequency axis.

---

# 3. Exact inverse on the log lattice

The Green kernel of $\mathcal L_A$ is

$$
\boxed{
G_A(k)
=
\frac{
A^{-|k|/2}
}{
A-1
}.
}
$$

Hence, modulo the finite initial boundary region on the half-line,

$$
\boxed{
R(t)
=
-
\frac1{A-1}
\sum_{k\in\mathbb Z}
A^{-|k|/2}
Z_A(t+k\delta).
}
$$

The coefficients are absolutely summable:

$$
\boxed{
\sum_{k\in\mathbb Z}
A^{-|k|/2}
<
\infty.
}
$$

So the shell transform is invertible on all translation-invariant mean-square spaces considered below.

---

# 4. Besicovitch $B^2$ almost periodicity

For a locally square-integrable function $f$ on $[0,\infty)$, define the Besicovitch seminorm

$$
\boxed{
\|f\|_{B^2}^2
=
\limsup_{T\to\infty}
\frac1T
\int_0^T
|f(t)|^2dt.
}
$$

A function is $B^2$ almost periodic if it is approximable in this seminorm by trigonometric polynomials.

Akbary–Ng–Shahabi use precisely this framework to prove limiting distributions for classical prime-number-theory error terms.

---

# 5. Wintner's RH-conditional prime process

Wintner proved, assuming RH, that

$$
\boxed{
R(t)
=
e^{-t/2}
[
\psi(e^t)-e^t
]
}
$$

possesses a limiting distribution.

The modern proof framework shows, under RH,

$$
\boxed{
R\in B^2_{\rm ap}.
}
$$

The truncated explicit formula is

$$
\boxed{
R(t)
=
\Re
\sum_{\substack{
\rho=1/2+i\gamma\\
0<\gamma\le X
}}
\frac{
-2e^{i\gamma t}
}{
\rho
}
+
\text{mean-square small error},
}
$$

with zeros counted with multiplicity.

---

# 6. RH implies shell $B^2$ almost periodicity

Because $Z_A$ is a finite linear combination of shifts of $R$,

$$
R\in B^2_{\rm ap}
\Longrightarrow
Z_A\in B^2_{\rm ap}.
$$

Therefore:

$$
\boxed{
\mathrm{RH}
\Longrightarrow
Z_A\in B^2_{\rm ap}.
}
$$

---

# 7. Shell $B^2$ almost periodicity implies base $B^2$ almost periodicity

The inverse Green representation is an absolutely summable linear combination of translates.

Therefore:

$$
\boxed{
Z_A\in B^2_{\rm ap}
\Longrightarrow
R\in B^2_{\rm ap}.
}
$$

So:

$$
\boxed{
Z_A\in B^2_{\rm ap}
\iff
R\in B^2_{\rm ap}.
}
$$

---

# 8. Finite mean square gives right-half-plane Laplace holomorphy

Suppose

$$
\boxed{
\limsup_{T\to\infty}
\frac1T
\int_0^T
|Z_A(t)|^2dt
<
\infty.
}
$$

Then for large $T$,

$$
\int_0^T
|Z_A(t)|^2dt
=
O(T).
$$

For every:

$$
a>0,
$$

split the Laplace integral into unit intervals and use Cauchy–Schwarz.

This gives absolute convergence of

$$
\boxed{
\int_0^\infty
Z_A(t)e^{-zt}dt
}
$$

for:

$$
\Re z>0.
$$

The resulting function is holomorphic in that half-plane.

---

# 9. Mellin transform of the shell process

Let

$$
s=z+\frac12.
$$

The full multiplicative Mellin transform of the shell wavelet is

$$
\boxed{
J_A(s)
\frac{
-\zeta'(s)
}{
s\zeta(s)
},
}
$$

where

$$
\boxed{
J_A(s)
=
A^s+A^{1-s}-(A+1).
}
$$

The one-sided transform beginning at $x=1$ differs only by a compact-range Mellin term, which is entire in $s$.

Therefore every nontrivial pole in the critical strip is unchanged.

---

# 10. Shell multiplier is zero-free in the open critical strip

Game 21 established

$$
\boxed{
J_A(s)\neq0
\qquad
0<\Re s<1.
}
$$

Its zeros lie only on the two boundary lattices:

$$
\Re s=0
$$

or

$$
\Re s=1.
$$

Therefore the shell filter cannot cancel a nontrivial zeta zero.

---

# 11. Mean-square boundedness implies RH

Suppose there exists

$$
\rho
=
\frac12+a+i\gamma,
\qquad
a>0.
$$

Then the shell Laplace transform has a pole at

$$
\boxed{
z
=
a+i\gamma
}
$$

inside:

$$
\Re z>0.
$$

This contradicts the holomorphy forced by finite Cesàro mean square.

Hence no zero has real part greater than $1/2$.

The functional equation then excludes real part below $1/2$.

Therefore:

$$
\boxed{
\limsup_{T\to\infty}
\frac1T
\int_0^T
|Z_A(t)|^2dt
<
\infty
\Longrightarrow
\mathrm{RH}.
}
$$

---

# 12. Mean-Square Stationarity Criterion

Combining the two directions:

$$
\boxed{
\mathrm{RH}
\iff
\limsup_{T\to\infty}
\frac1T
\int_0^T
|Z_A(t)|^2dt
<
\infty.
}
$$

This is the simplest stationarity criterion of Game 23.

The condition says that the critically normalized localized shell process has bounded long-time average power.

---

# 13. $B^2$ Almost-Periodic Criterion

Since RH implies $B^2$ almost periodicity, and $B^2$ almost periodicity implies finite mean square:

$$
\boxed{
\mathrm{RH}
\iff
Z_A
\text{ is }B^2\text{-almost periodic}.
}
$$

Thus RH is equivalent to deterministic mean-square almost periodicity of the local shell innovation process.

---

# 14. Fixed-window local mean square

Define the base log-time local energy:

$$
\boxed{
\mathcal I_R(T,L)
=
\int_T^{T+L}
|R(t)|^2dt.
}
$$

For:

$$
L=\log2,
$$

set:

$$
X=e^T.
$$

Then:

$$
\boxed{
\mathcal I_R(T,\log2)
=
\int_X^{2X}
\frac{
|\psi(x)-x|^2
}{
x^2
}dx.
}
$$

---

# 15. RH gives uniform fixed-window control

Brent–Platt–Trudgian prove under RH:

$$
\boxed{
\limsup_{X\to\infty}
\frac1{X^2}
\int_X^{2X}
|\psi(x)-x|^2dx
\le
0.8603.
}
$$

Therefore:

$$
\boxed{
\sup_{T\gg1}
\int_T^{T+\log2}
|R(t)|^2dt
<
\infty.
}
$$

Any other fixed window length can be covered by finitely many dyadic log windows.

Thus for every fixed:

$$
L>0,
$$

$$
\boxed{
\sup_{T\gg1}
\int_T^{T+L}
|R(t)|^2dt
<
\infty.
}
$$

---

# 16. Shell fixed-window control

Because $Z_A$ is a finite combination of fixed shifts of $R$:

$$
\boxed{
\mathrm{RH}
\Longrightarrow
\sup_{T\gg1}
\int_T^{T+L}
|Z_A(t)|^2dt
<
\infty
}
$$

for every fixed:

$$
L>0.
$$

---

# 17. Fixed-window control implies RH

Conversely, if for one fixed:

$$
L>0
$$

we have:

$$
\boxed{
\sup_{T\gg1}
\int_T^{T+L}
|Z_A(t)|^2dt
<
\infty,
}
$$

then covering:

$$
[0,T]
$$

by $O(T/L)$ intervals gives finite Cesàro mean-square growth.

The Mean-Square Stationarity Criterion then implies RH.

Thus:

$$
\boxed{
\mathrm{RH}
\iff
Z_A
\text{ is eventually Stepanov-}L^2\text{ bounded}.
}
$$

Here "Stepanov bounded" means uniform $L^2$ control over translates of one fixed finite window; it is not the stronger assertion of Stepanov almost periodicity.

---

# 18. Stationarity Ladder Collapse

For this arithmetic signal, the following properties are equivalent:

$$
\boxed{
\mathrm{RH}
}
$$

$$
\Longleftrightarrow
$$

$$
\boxed{
Z_A\in B^2_{\rm ap}
}
$$

$$
\Longleftrightarrow
$$

$$
\boxed{
\limsup_{T\to\infty}
\frac1T
\int_0^T
|Z_A(t)|^2dt
<
\infty
}
$$

$$
\Longleftrightarrow
$$

$$
\boxed{
\sup_{T\gg1}
\int_T^{T+L}
|Z_A(t)|^2dt
<
\infty.
}
$$

These conditions are not equivalent for arbitrary functions.

Their equivalence here is caused by the prime explicit formula, the zero-free shell multiplier, and RH-conditional prime mean-square theory.

We call this:

$$
\boxed{
\textbf{Stationarity Ladder Collapse}.
}
$$

---

# 19. Fourier coefficients under RH

Group zeros by positive ordinate.

Let:

$$
\rho_\gamma
=
\frac12+i\gamma,
$$

with multiplicity:

$$
m_\gamma.
$$

The base $B^2$ coefficient is:

$$
\boxed{
r_\gamma
=
-\frac{
2m_\gamma
}{
\rho_\gamma
}.
}
$$

The shell operator multiplies frequency $\gamma$ by:

$$
J_A(\rho_\gamma)
=
-P_A(\gamma),
$$

where:

$$
\boxed{
P_A(\gamma)
=
A+1
-
2\sqrt A
\cos(
\gamma\log A
).
}
$$

Hence:

$$
\boxed{
Z_A(t)
=
\Re
\sum_{\gamma>0}
\frac{
2m_\gamma P_A(\gamma)
}{
\rho_\gamma
}
e^{i\gamma t}
}
$$

in the $B^2$ sense.

---

# 20. Parseval mean square

Besicovitch Parseval gives:

$$
\boxed{
\lim_{T\to\infty}
\frac1T
\int_0^T
|Z_A(t)|^2dt
=
2
\sum_{\gamma>0}
\frac{
m_\gamma^2
P_A(\gamma)^2
}{
1/4+\gamma^2
}.
}
$$

The series converges because:

$$
P_A(\gamma)
$$

is uniformly bounded and:

$$
\sum_{\gamma>0}
\frac{
m_\gamma^2
}{
1+\gamma^2
}
<\infty
$$

for the zeta-zero divisor.

---

# 21. Shell variance is equivalent to Wintner variance

The uniform gain bounds give:

$$
\boxed{
(\sqrt A-1)^4
\le
P_A(\gamma)^2
\le
(\sqrt A+1)^4.
}
$$

Therefore the shell mean-square is bounded above and below by fixed multiples of the base prime-error $B^2$ mean-square.

The shell filter changes conditioning, not spectral completeness.

---

# 22. Autocorrelation

Define the Besicovitch autocorrelation:

$$
\boxed{
C_A(\tau)
=
\lim_{T\to\infty}
\frac1T
\int_0^T
Z_A(t+\tau)
Z_A(t)
dt.
}
$$

Under RH:

$$
\boxed{
C_A(\tau)
=
2
\sum_{\gamma>0}
\frac{
m_\gamma^2
P_A(\gamma)^2
}{
1/4+\gamma^2
}
\cos(
\gamma\tau
).
}
$$

This is a positive-definite almost-periodic covariance function.

---

# 23. Pure-point stationarity spectrum

The corresponding symmetric spectral measure is:

$$
\boxed{
\Sigma_A
=
\sum_{\gamma>0}
\frac{
m_\gamma^2
P_A(\gamma)^2
}{
1/4+\gamma^2
}
\left(
\delta_\gamma
+
\delta_{-\gamma}
\right).
}
$$

Thus the stationary shell spectrum is pure point.

Its frequencies are exactly the ordinates of the nontrivial zeta zeros.

---

# 24. Covariance Completeness

The support of:

$$
\Sigma_A
$$

recovers every positive zero ordinate:

$$
\gamma.
$$

Since:

$$
P_A(\gamma)>0
$$

is explicitly known, the atom weight determines:

$$
m_\gamma^2.
$$

Hence multiplicity is also recoverable.

We call this:

$$
\boxed{
\textbf{Covariance Completeness}.
}
$$

> Under RH, the second-order stationary covariance of the localized prime shell process is spectrally complete for the nontrivial zeta zero divisor.

No zero-height information is lost by passing to covariance.

---

# 25. Limiting distribution

Every $B^2$ almost periodic function has a limiting distribution in the Besicovitch/Wintner framework.

Therefore under RH:

$$
\boxed{
Z_A(t)
}
$$

possesses a limiting distribution:

$$
\boxed{
\mu_A.
}
$$

Its mean is zero and its second moment is:

$$
\boxed{
\int_{\mathbb R}
u^2\,d\mu_A(u)
=
2
\sum_{\gamma>0}
\frac{
m_\gamma^2
P_A(\gamma)^2
}{
1/4+\gamma^2
}.
}
$$

No linear-independence conjecture is needed for existence.

---

# 26. Conditional torus model under LI

If one additionally assumes the usual linear-independence conjecture for the positive zero ordinates over $\mathbb Q$, then the phase flow:

$$
t
\mapsto
(
e^{i\gamma t}
)_\gamma
$$

equidistributes on the corresponding infinite torus in the finite-dimensional projection sense.

The limiting distribution is then the pushforward of Haar measure by the Fourier series.

Formally, its characteristic function is:

$$
\boxed{
\widehat\mu_A(\xi)
=
\prod_{\gamma>0}
J_0
\left(
\frac{
2m_\gamma P_A(\gamma)
}{
|\rho_\gamma|}
\xi
\right),
}
$$

where:

$$
J_0
$$

is the Bessel function.

This LI assumption is not part of the RH equivalence.

---

# 27. Translation hull

Define the $B^2$ translation hull:

$$
\boxed{
\mathcal H_A
=
\overline{
\{
Z_A(\cdot+\tau):
\tau\in\mathbb R
\}
}^{\,B^2}.
}
$$

Standard Besicovitch almost-periodic theory gives relative compactness of the family of translates.

Therefore under RH:

$$
\boxed{
\mathcal H_A
\text{ is compact}.
}
$$

---

# 28. Compact Translation-Hull Criterion

If the translation hull is compact in the $B^2$ topology, then in particular its orbit is norm-bounded.

Thus the Cesàro mean-square of $Z_A$ is finite, and RH follows.

Therefore:

$$
\boxed{
\mathrm{RH}
\iff
\mathcal H_A
\text{ is compact in }B^2.
}
$$

We call this:

$$
\boxed{
\textbf{Compact Shell-Translation Hull Criterion}.
}
$$

---

# 29. Phase-flow geometry under RH

Under RH, translation acts on every Fourier mode by:

$$
\boxed{
e^{i\gamma t}
\mapsto
e^{i\gamma\tau}
e^{i\gamma t}.
}
$$

So translation changes only phase.

Amplitude is unchanged.

The hull is the closure of the one-parameter orbit:

$$
\boxed{
\tau
\mapsto
(
e^{i\gamma\tau}
)_\gamma
}
$$

in the weighted phase torus determined by the square-summable shell coefficients.

Thus the RH shell dynamics is compact and quasiperiodic.

---

# 30. Rational relations among zero ordinates

If the zero ordinates satisfy rational relations, the phase orbit lies in a proper compact subgroup of the full torus.

If they are rationally independent, finite-dimensional projections are dense in the full corresponding torus.

Therefore the topology of the shell hull encodes arithmetic relations among zero ordinates.

The existence of the compact hull itself, however, requires only RH.

---

# 31. Off-line zero as hyperbolic translation mode

Suppose:

$$
\rho
=
\frac12+a+i\gamma,
\qquad
a>0.
$$

Its shell contribution has the form:

$$
\boxed{
C_{\rho,A}
e^{at}
e^{i\gamma t}.
}
$$

Under translation by:

$$
\tau,
$$

the mode is multiplied by:

$$
\boxed{
e^{a\tau}
e^{i\gamma\tau}.
}
$$

Its modulus grows as:

$$
e^{a\tau}.
$$

Therefore the translation orbit escapes every bounded $B^2$ set.

---

# 32. Compact phase flow versus hyperbolic escape

This gives a geometric dichotomy:

### RH

$$
\boxed{
\text{pure phase rotation}
}
$$

with compact orbit closure.

### not RH

$$
\boxed{
\text{phase rotation}
+
\text{radial exponential escape}.
}
$$

Thus the horizontal displacement of a zero is exactly a dynamical radial exponent.

---

# 33. Koopman Unitarity Criterion

The translation operator:

$$
(U_\tau f)(t)=f(t+\tau)
$$

acts on a formal zero mode by:

$$
\boxed{
U_\tau:
e^{(\rho-1/2)t}
\mapsto
e^{(\rho-1/2)\tau}
e^{(\rho-1/2)t}.
}
$$

Therefore:

$$
\boxed{
\mathrm{RH}
}
$$

is equivalent to every arithmetic shell translation mode having unit-modulus multiplier.

We call this:

$$
\boxed{
\textbf{Koopman Unitarity Criterion}.
}
$$

This is a dynamical-language restatement of the spectral abscissa condition.

---

# 34. Phase–Radial Separation

Write:

$$
\boxed{
\rho-\frac12
=
a+i\gamma.
}
$$

Then:

- $\gamma$ controls phase rotation;
- $a$ controls radial expansion/contraction.

Thus the zero coordinate splits dynamically as:

$$
\boxed{
\text{zero displacement}
=
\text{radial exponent}
+
i\,
\text{phase frequency}.
}
$$

RH is exactly the vanishing of every radial exponent.

---

# 35. Stationarity exponent

Define:

$$
\boxed{
\chi_A
=
\inf
\left\{
a\ge0:
\limsup_{T\to\infty}
\frac1T
\int_0^T
|
e^{-at}Z_A(t)
|^2dt
<
\infty
\right\}.
}
$$

This measures the exponential damping required to stationarize the shell process.

---

# 36. Stationarity exponent equals zero spectral abscissa

Let:

$$
\boxed{
\Theta_\zeta
=
\sup_\rho
\Re\rho.
}
$$

If:

$$
a<
\Theta_\zeta-\frac12,
$$

the damped Laplace transform still contains a pole in the right half-plane.

So bounded mean square is impossible.

Conversely, for every:

$$
a>
\Theta_\zeta-\frac12,
$$

the standard zero-free-half-plane / prime-error correspondence gives enough exponential damping to place the signal in a polynomially bounded and hence mean-square-bounded class.

Therefore:

$$
\boxed{
\chi_A
=
\Theta_\zeta-\frac12.
}
$$

The value is independent of $A$.

---

# 37. Unification with previous escape invariants

Game 07 defined the zero escape radius:

$$
\boxed{
\Omega_\xi
=
\sup_\rho
\left|
\Re\rho-\frac12
\right|.
}
$$

By functional symmetry:

$$
\boxed{
\Omega_\xi
=
\Theta_\zeta-\frac12.
}
$$

Hence:

$$
\boxed{
\chi_A
=
\Omega_\xi.
}
$$

So the same invariant has now appeared as:

- zero horizontal escape radius;
- prime residual excess growth exponent;
- right-half-plane Laplace pole abscissa;
- shell nonstationarity exponent;
- radial translation/Koopman exponent.

---

# 38. Stationarity Defect Principle

We define:

$$
\boxed{
\textbf{Stationarity Defect Principle}.
}
$$

> The distance of the rightmost zeta zero from the critical line is exactly the amount of exponential damping required to convert the localized critical shell process into a mean-square stationary class.

This is the dynamical meaning of the RH defect.

---

# 39. Minimal logical requirement

The full $B^2$ almost-periodic structure is stronger than logically necessary.

To prove RH from the shell side, it is enough to prove:

$$
\boxed{
\limsup_{T\to\infty}
\frac1T
\int_0^T
|Z_A(t)|^2dt
<
\infty.
}
$$

Or even the stronger-but-local fixed-window estimate:

$$
\boxed{
\sup_{T\gg1}
\int_T^{T+L}
|Z_A(t)|^2dt
<
\infty.
}
$$

So the minimal stationarity proof obligation is only bounded power, not explicit Fourier decomposition.

---

# 40. Why the almost-periodic structure is still useful

Although mean-square boundedness suffices logically, $B^2$ almost periodicity gives more:

1. a pure-point Fourier spectrum;
2. exact Parseval energy;
3. all lag autocorrelations;
4. a limiting distribution;
5. a compact translation hull;
6. a dynamical phase-flow representation.

Thus it describes **what RH-positive dynamics looks like**, not just how to exclude RH-negative dynamics.

---

# 41. Stationarity is not randomness

The shell process under RH is not stationary because it becomes random.

It is stationary-like because after critical normalization it is a deterministic superposition of bounded log-frequency phases.

Its spectral measure is pure point.

So the appropriate picture is:

$$
\boxed{
\text{deterministic quasiperiodic stationarity}
}
$$

rather than stochastic white-noise stationarity.

---

# 42. Relation to Game 22

Game 22 showed that local modular prediction collapses to a one-dimensional shell amplitude.

Game 23 shows that the remaining amplitude innovation, after critical normalization, has an exact dynamical classification:

### RH

$$
\boxed{
\text{compact almost-periodic phase dynamics}.
}
$$

### not RH

$$
\boxed{
\text{hyperbolically expanding phase dynamics}.
}
$$

So the remaining scalar coherence problem is now completely identified at the dynamical level.

---

# 43. Representation fixed point

The route has now become:

$$
\boxed{
\text{prime shell density}
}
$$

$$
\Downarrow
$$

$$
\boxed{
\text{scalar shell innovation}
}
$$

$$
\Downarrow
$$

$$
\boxed{
Z_A(t)
}
$$

$$
\Downarrow
$$

$$
\boxed{
\text{translation dynamics}
}
$$

$$
\Downarrow
$$

$$
\boxed{
\text{compact phase hull}
\quad\text{or}\quad
\text{hyperbolic escape}.
}
$$

This is the current dynamical representation fixed point.

---

# 44. What Game 23 does not solve

Game 23 does not provide an arithmetic reason that the shell process must be $B^2$ almost periodic.

Proving bounded mean-square stationarity from prime arithmetic is still RH-hard.

The transformation has made the positive and negative dynamical pictures very clear, but it has not generated the missing cancellation law.

Thus the final problem remains:

> Why does the actual prime shell process belong to the compact phase-flow class rather than the hyperbolic escape class?

---

# 45. Game 24 candidate

The next step should use the newly available stationary spectral structure rather than invent another filter.

Three possible directions are meaningful.

## A. Covariance positivity and Toeplitz completion

The RH covariance

$$
C_A(\tau)
$$

is positive definite and has a pure-point spectral measure.

Construct finite Toeplitz covariance matrices:

$$
\boxed{
[
C_A((j-k)\Delta)
]_{j,k=0}^{N}.
}
$$

Ask whether prime-side finite-scale data admits a positive-definite completion whose failure would detect an off-line radial mode.

This reconnects to Games 02 and 10 but now with an explicitly local shell signal.

## B. Translation-hull compactness from arithmetic recurrence

Search for an arithmetic recurrence theorem strong enough to prove precompactness of shell translates without first proving the full prime-error bound.

A useful result would need to establish recurrence of the scalar shell innovation, not modular shape.

## C. Spectral-measure reconstruction from finite prime data

Use the covariance-complete shell process to define finite empirical spectral measures.

Ask whether positivity, tightness, and zero-free shell passband can force limiting mass onto the real frequency axis.

This would reinterpret RH as spectral tightness rather than pointwise zero localization.

---

# 46. Most promising next step

The most promising route is:

$$
\boxed{
\text{finite covariance Toeplitz positivity}
}
$$

because:

1. stationarity naturally produces positive-definite covariance matrices;
2. off-line modes correspond to nonunitary radial growth;
3. Toeplitz/Schur machinery from earlier Games is already available;
4. the shell signal has removed equilibrium and local congruence distractions;
5. covariance is spectrally complete under RH.

The question is whether this creates a genuinely new finite positivity obligation or merely returns to the old Toeplitz RH criteria.

That is exactly the kind of representation-cycle test the SECV program is designed to perform.

---

# 47. Monster status

After Game 23:

$$
\boxed{
\text{Shell Mean-Square RH Criterion: FOUND}
}
$$

$$
\boxed{
\text{Shell }B^2\text{-Almost-Periodic RH Criterion: FOUND}
}
$$

$$
\boxed{
\text{Stepanov Local-}L^2\text{ RH Criterion: FOUND}
}
$$

$$
\boxed{
\text{Stationarity Ladder Collapse: FOUND}
}
$$

$$
\boxed{
\text{Exact Shell Parseval Variance: FOUND}
}
$$

$$
\boxed{
\text{Pure-Point Shell Spectral Measure: FOUND}
}
$$

$$
\boxed{
\text{Covariance Completeness: FOUND}
}
$$

$$
\boxed{
\text{Compact Translation-Hull Criterion: FOUND}
}
$$

$$
\boxed{
\text{Koopman Unitarity Criterion: FOUND}
}
$$

$$
\boxed{
\text{Stationarity Exponent }=\Theta_\zeta-\tfrac12\text{: FOUND}
}
$$

The surviving monster is:

$$
\boxed{
\textbf{Arithmetic Origin of Critical Shell Stationarity}.
}
$$

---

# 48. Final diagram

$$
\boxed{
\begin{aligned}
\mathfrak W_A(x)
&\rightarrow
Z_A(t)
=
e^{-t/2}
\mathfrak W_A(e^t)
\\
Z_A
&=
-\mathcal L_A R
\\
\mathcal L_A
&\text{ invertible on real log frequencies}
\\
\mathrm{RH}
&\Longleftrightarrow
Z_A\in B^2_{\rm ap}
\\
&\Longleftrightarrow
\sup_T
\frac1T
\int_0^T
|Z_A|^2
<
\infty
\\
&\Longleftrightarrow
\text{compact translation hull}
\\
&\Longleftrightarrow
\text{unit-modulus shell modes}.
\end{aligned}
}
$$

Under RH:

$$
\boxed{
\Sigma_A
=
\sum_{\gamma>0}
\frac{
m_\gamma^2P_A(\gamma)^2
}{
1/4+\gamma^2
}
(
\delta_\gamma+\delta_{-\gamma}
).
}
$$

Off RH:

$$
\boxed{
\rho=\frac12+a+i\gamma
\quad\Rightarrow\quad
e^{at}e^{i\gamma t}
}
$$

and the translation hull escapes compactness.

---

# 49. Conclusion

Game 23 turns the scalar shell-amplitude problem into a precise log-time dynamical classification.

The localized prime shell process

$$
Z_A(t)
=
e^{-t/2}
\mathfrak W_A(e^t)
$$

is linked to the classical normalized PNT error by an invertible massive discrete shell operator.

Under RH, Wintner's explicit-formula theory and the modern Akbary–Ng–Shahabi framework place the normalized prime error in the Besicovitch $B^2$ almost-periodic class.

The shell operator preserves and reflects this property.

Conversely, finite shell mean-square power is already enough to force the shell Laplace transform to be holomorphic in the right half-plane.

Because the shell multiplier has no zeros in the open critical strip, any off-line zeta zero would survive as a right-half-plane pole.

Therefore:

$$
\boxed{
\mathrm{RH}
\iff
Z_A\in B^2_{\rm ap}
\iff
\limsup_{T\to\infty}
\frac1T
\int_0^T
|Z_A(t)|^2dt
<
\infty.
}
$$

Using the RH-conditional dyadic mean-square theorem of Brent–Platt–Trudgian, this is further equivalent to uniform boundedness of fixed-length log-time $L^2$ windows.

Thus the shell signal satisfies a Stationarity Ladder Collapse.

Under RH its Fourier spectrum is pure point, with frequencies exactly the zeta-zero ordinates.

The shell autocovariance is spectrally complete: it recovers zero ordinates and multiplicities.

The translation hull is compact and consists of phase rotations.

An off-line zero instead generates a radial exponential direction and destroys compactness.

The amount of exponential damping required to restore stationarity is exactly:

$$
\Theta_\zeta-\frac12,
$$

the same escape invariant found in earlier Games.

The RH problem can therefore be stated dynamically:

> Why does the actual localized prime-shell innovation generate a compact quasiperiodic phase flow rather than a hyperbolically expanding translation flow?

That is the surviving **Arithmetic Origin of Critical Shell Stationarity** problem.

---

# References

1. Wintner, A. (1935). *On the asymptotic distribution of the remainder term of the prime number theorem*. American Journal of Mathematics.
2. Akbary, A., Ng, N., & Shahabi, M. (2014). *Limiting distributions of the classical error terms of prime number theory*. Quarterly Journal of Mathematics 65(3), 743–780; arXiv:1306.1657.
3. Akbary–Ng–Shahabi, Theorem 1.14: Parseval-type mean-square identity for the relevant $B^2$ almost-periodic error terms.
4. Brent, R. P., Platt, D. J., & Trudgian, T. S. (2021). *The mean square of the error term in the prime number theorem*. arXiv:2008.06140.
5. Besicovitch, A. S. Classical theory of generalized / $B^2$ almost periodic functions and Parseval identities.
6. Standard characterization of $B^2$ almost periodic functions through trigonometric-polynomial closure and relative compactness of translates in the Besicovitch topology.
7. Classical explicit formula for $\psi(x)$ and the zero-free-half-plane / PNT-error correspondence.
8. Game 21 of this series for the zero-free localized shell multiplier and its exact discrete Green inverse.