---

1. Introduction

1.1 Background

Paper XXV reduced H' ($D(N) \to 1$) to pred gap control via dyadic telescoping. Papers XXVII–XXVIII established the upper reflection criterion, universal scale attenuation, and a three-step lower reflection framework.

The present paper performs three tasks: (1) a conceptual audit distinguishing $D(N)$ from $P(\text{pred gap} > 0)$; (2) a scale correction of Paper XXVIII's threshold choice via the discovery $\sigma(H) \propto 1/\ln N$; (3) reporting the discrete concentration of pred gap and its algebraic explanation.

This paper does not claim to prove H'.

1.2 Contributions

(A) Conceptual distinction between $D(N)$ and $P(\text{pred gap} > 0)$ (§2).

(B) Scale correction $\sigma(H) \propto 1/\ln N$ as a coordinate artifact (§3).

(C) Discrete concentration of pred gap on $\{+2, +3\}$ (§4).

(D) Dyadic differencing annihilates sub-leading terms (§5), with quantitative verification.

(E) Prime-composite decomposition of pred gap (§6).

1.3 Notation

Series conventions and Paper XXVI–XXVIII notation apply. New notation: $h_{\text{ref}} = h_0(10^7) = 3.7918$ (finite-scale empirical anchor); $h_\infty = \lim h_0(N)$ (if the limit exists; true asymptotic constant, currently unknown). $R_{\text{ref}}(n) = \rho_E(n) - h_{\text{ref}} \cdot \ln n$. $\text{pred gap}(m) = \rho_E(2m) - \rho_E(m)$. $D(N) = |\{n \leq N : \rho_E(n) > \rho_E(n-1)\}| / N$.

---

2. Conceptual Audit: D(N) vs P(pred gap > 0)

2.1 Two distinct quantities

D(N) (the target of H'): pointwise positive-jump density.

$$D(N) = \frac{|\{n \leq N : \rho_E(n) > \rho_E(n-1)\}|}{N}$$

Every prime $n$ contributes $\delta(n) = +1$. Every composite $n$ contributes $\delta(n) = 1 - j(n)$; when $j = 0$, $\delta = +1$. Numerically $D(N) \approx 0.574$ ($N = 10^7$). H' requires $D(N) \to 1$. The significant gap between 0.574 and 1 at finite scales reflects the ongoing competition between additive paths ($j = 0$, $\delta = +1$) and jump paths ($j \geq 1$, $\delta \leq 0$). Paper XXV's dyadic telescoping shows that if the macroscopic condition $E[\text{gap}] > \tau_k$ holds, the Cesàro averaging mechanism ultimately forces the zero-jump density to vanish in the limit. The current value 0.574 is a finite-scale transient, not an asymptotic obstruction.

P(pred gap > 0): dyadic gap positivity rate.

$$P(\text{gap} > 0) = \frac{|\{m \leq N/2 : \rho_E(2m) > \rho_E(m)\}|}{N/2}$$

Numerically $P(\text{gap} > 0) \approx 0.997$ ($N = 10^9$).

These are different quantities. $D(N)$ is a microscopic (single-step) density; $P(\text{gap} > 0)$ is a macroscopic (dyadic-interval) density. The latter being near 1 does not imply the former is near 1.

2.2 P(gap ≤ 0) ≈ 0.3% does not threaten H'

$P(\text{gap} \leq 0) \approx 0.29\%$ ($N = 10^9$), mildly increasing but saturating. This does not threaten H': $D(N) \to 1$ requires pointwise $\delta(n) = +1$ density approaching 1, not every dyadic interval having positive growth. Isolated $m$ with $\rho_E(2m) \leq \rho_E(m)$ are microscopic fluctuations — a compatible explanation is unusually smooth configurations causing composites in $[m, 2m]$ to cancel all prime $+1$ contributions. These are expected rare events.

2.3 The correct Route B target

Paper XXV's reduction targets $E[\text{gap}] > \tau_k$ for a positive threshold $\tau_k$, not $P(\text{gap} > 0) \to 1$. Numerically $E[\text{gap}] \approx 2.52$ ($N = 10^9$), stably positive and far from 0. The gap distribution is highly concentrated (96%+ on $\{+2, +3\}$) with strictly bounded variance — placing Route B in a very favorable statistical regime where mean-threshold criteria are plausibly effective.

---

3. σ(H) ∝ 1/ln N: Coordinate Correction

3.1 Data

Numerical Observation 1.

N	h₀	σ(H)	σ · ln N
10⁴	3.714	0.1006	0.927
10⁵	3.752	0.0824	0.948
10⁶	3.776	0.0681	0.941
10⁷	3.792	0.0576	0.929
10⁸	3.802	0.0498	0.917
10⁹	3.809	0.0440	0.912

$\sigma \cdot \ln N$ ranges from 0.93 to 0.91, extremely stable across six orders of magnitude.

3.2 Coordinate interpretation

$H(n) = \rho_E(n) / \ln n$. The relation $\sigma(H) \cdot \ln N \approx 0.93$ directly supports: centered at the local mean $h_0(N)$, $\rho_E$ fluctuates at $O(1)$ scale. This is a purely arithmetic consequence of dividing by $\ln n$, not a dynamical collapse.

The natural fluctuation variable is the locally centered $\rho_E$ (i.e., $\rho_E(n) - h_0(N) \cdot \ln n$), not $H(n)$. The fixed-anchor residual $R_{\text{ref}}(n) = \rho_E(n) - h_{\text{ref}} \cdot \ln n$ introduced in §5 is a related but distinct object: the former has std $\approx 0.93$, the latter std $\approx 0.75$ (the difference arising from the gap between $h_{\text{ref}}$ and $h_0(N)$).

3.3 Correction to Paper XXVIII

Paper XXVIII's three-step lower reflection used fixed $\varepsilon$ in $H$ coordinates, causing $\alpha$ to decline. The coordinate correction identifies:

• In $H$ coordinates: the natural scale is $\varepsilon(N) \propto \sigma(N) \propto 1/\ln N$.
• Equivalently in $\rho_E$ coordinates: the natural scale is $O(1)$ constant truncation.

This is a restatement of the Paper XXVIII Step 1 bottleneck, not its elimination. The rescaled formulation is more natural, but formal proof still requires nondegeneracy theory for the height distribution.

3.4 Occupancy concentration

Numerical Observation 2.

N	fraction in h₀ ± 0.02	fraction with |H−h₀| > 0.10
10⁴	17.6%	32.1%
10⁶	24.5%	13.6%
10⁸	31.9%	4.6%
10⁹	35.5%	2.4%

The distribution is concentrating — two-sided reflection controls an increasing fraction of the mass.

---

4. Discrete Concentration of Pred Gap

4.1 Gap distribution (N = 10⁹)

Numerical Observation 3.

Decade	P(g≤0)	P(g=1)	P(g=2)	P(g=3)	E[g]
[10², 10³)	0.0022	0.018	0.397	0.583	2.561
[10⁴, 10⁵)	0.0021	0.029	0.401	0.568	2.535
[10⁶, 10⁷)	0.0027	0.032	0.405	0.561	2.524
[10⁷, 10⁸)	0.0028	0.033	0.407	0.558	2.519
[10⁸, 10⁹)	0.0029	0.033	0.408	0.556	2.517

The gap distribution is extremely stable: 96%+ concentrated on $\{+2, +3\}$. No gap $> 3$ is observed. The ratio gap=2 to gap=3 slowly shifts (0.68 → 0.72 at tested scales).

4.2 Mild decrease of E[gap]

$E[\text{gap}]$ decreases from 2.561 ($[10^2, 10^3)$) to 2.517 ($[10^8, 10^9)$), but the rate of decrease is diminishing ($\Delta = -0.026, -0.011, -0.005, -0.002$ per decade). $E[\text{gap}]$ remains well above 0 at all tested scales.

Simultaneously $h_0 \cdot \ln 2$ increases from 2.57 to 2.64 — opposite direction. The negative correction term $E[R(2m)-R(m)]$ has not yet decayed to negligible at current scales. §5 provides a quantitative explanation.

4.3 Pred gap can be negative

Full scan ($N = 10^7$): min pred gap $= -2$ at $m = 1{,}396{,}679$. Negative and zero gaps occur at all scales; $P(\text{gap} \leq 0) \approx 0.29\%$ mildly increasing but saturating. As established in §2.2, this does not threaten H'.

---

5. Dyadic Differencing Annihilates Sub-Leading Terms

5.1 Behavior of R(n)

$R_{\text{ref}}(n) = \rho_E(n) - h_{\text{ref}} \cdot \ln n$ ($h_{\text{ref}} = h_0(10^7) = 3.7918$, a finite-scale empirical anchor).

Numerical Observation 4. $R$ is systematically negative ($E[R] \approx -3$), with $\text{std}(R) \approx 0.75$ stable at tested scales. The negative offset admits multiple compatible explanations: a constant correction $c \approx -3$, a slowly varying sub-leading term $c_1 \cdot \ln\ln n$, or finite-scale bias in the choice of $h_{\text{ref}}$. Current data cannot distinguish these.

5.2 The ln ln n hypothesis

Linear regression $R(n) = a + b \cdot \ln(\ln n)$ yields $b \approx -1.18$, explaining approximately 57% of the sample variance. This suggests $R(n)$ may contain a $c_1 \cdot \ln\ln n$ sub-leading term.

5.3 The algebraic annihilation mechanism

Dyadic differencing naturally annihilates $\ln\ln n$-scale slow divergence. Even if $R(n)$ contains slowly diverging sub-leading terms, pred gap $= h_{\text{ref}} \ln 2 + (R(2m) - R(m))$ remains stable, because differencing smooths out the divergence.

5.4 Quantitative closure

Substituting $c_1 = -1.18$ into the difference formula. At $m \approx 10^8$, $\ln m \approx 18.4$:

$$\Delta R \approx -1.18 \times \frac{0.693}{18.4} \approx -0.044$$

Using the early-scale mean $E[\text{gap}] \approx 2.56$ ($[10^2, 10^3)$) plus this correction: $2.56 - 0.044 = 2.516$.

5.5 Asymptotic behavior of E[gap]

The data are compatible with "$E[\text{gap}]$ remains positive and approaches a positive constant." If the dominant sub-leading term of $R$ is indeed of $\ln\ln n$ type, the dyadic correction decays as $1/\ln m$, and $E[\text{gap}]$ would slowly rise. With $h_0(10^9) = 3.809$ still rising, the asymptotic value $h_\infty$ may lie in $3.83$–$3.85$, corresponding to $E[\text{gap}] \to 2.65$–$2.67$. However, this is heuristic extrapolation, not a theorem established in this paper.

---

6. Prime-Composite Decomposition

6.1 Decomposition

$$\text{pred gap}(m) = \sum_{n=m+1}^{2m} \delta(n) = \underbrace{\pi(2m)-\pi(m)}_{\text{prime contribution}} + \underbrace{\sum_{\text{composite}} \delta(n)}_{\text{composite contribution}}$$

Numerical Observation 5.

m	gap	prime contrib.	composite contrib.	E[δ|comp]
10⁴	+3	+1,033	−1,030	−0.115
10⁵	+2	+8,392	−8,390	−0.092
10⁶	+3	+70,435	−70,432	−0.076

Pred gap is a fine balance between two $O(m/\ln m)$ quantities, with $O(1)$ net surplus. The absolute value of composite negative drift decreases at the tested points (0.115 → 0.076); if this trend continues (more data needed), the composite drag weakens and primes increasingly dominate the net balance.

6.2 Connection to Paper XXVI

Paper XXVI established $\rho_E(N) = \pi(N) + S_{\text{comp}}(N)$. The decomposition of §6.1 verifies this structure at the dyadic level: pred gap $= [\pi(2m)-\pi(m)] + [S_{\text{comp}}(2m)-S_{\text{comp}}(m)]$. The first term $\approx m/\ln m$ (PNT); the second numerically behaves like $-(m/\ln m) + O(1)$ at tested points. The $O(1)$ net surplus is the pred gap.

---

7. Discussion

7.1 Core contributions

This paper does not advance analytic proofs; it establishes the correct coordinate system and target quantities for the series:

(1) Conceptual audit: $D(N)$ and $P(\text{gap} > 0)$ are distinct. H' targets $D(N)$. The behavior of $P(\text{gap} \leq 0)$ does not directly constrain $D(N)$.

(2) Coordinate correction: $\sigma(H) \propto 1/\ln N$ is a coordinate artifact. The natural fluctuation variable is $R(n)$, with $O(1)$ fluctuations. Paper XXVIII's threshold should use $O(1)$ truncation in $\rho_E$ coordinates.

(3) Dyadic annihilation of sub-leading terms: $R(n)$ may have $\ln\ln n$-scale slow divergence, but dyadic differencing naturally smooths it. This is the algebraic core of pred gap stability.

(4) Discrete concentration: Pred gap is almost exclusively $+2$ or $+3$, with bounded variance.

7.2 The critical-gap picture

$E[\text{gap}] \approx 2.52$ lies between gap $= 2$ and gap $= 3$ (closer to 2.5). The system operates near criticality — $c_{\text{main}} \cdot \ln 2$ slightly exceeds 2.5, while the negative correction ~0.11 pulls the mean to 2.52. As the correction decays ($\propto 1/\ln m$), $E[\text{gap}]$ will slowly rise toward 2.6+.

7.3 Open problems

(1) Analytic proof of $\sigma \propto 1/\ln N$.

(2) Asymptotic limit of $E[\text{gap}]$ — does it equal $c_{\text{main}} \cdot \ln 2$?

(3) Precise sub-leading structure of $R(n)$ — constant $+ \ln\ln n$, or more complex?

(4) Route B proof of $D(N) \to 1$ — requires analytic confirmation of $E[\text{gap}] > \tau_k$.

(5) Asymptotic behavior of $P(\text{gap} = 0)$ — tends to 0, to a small positive constant, or other? Does not affect H' but has independent number-theoretic interest.

7.4 Acknowledgments

The thermodynamics thread (Claude) identified the critical conceptual error $D(N) \neq P(\text{pred gap} > 0)$, averting an incorrect pursuit of $P(\text{gap} \leq 0) \to 0$. Gemini contributed the coordinate correction ($\sigma$ collapse as artifact), the $\ln\ln n$ hypothesis, and the quantitative closure of $E[\text{gap}]$ decline (theoretical 2.516 vs measured 2.521). ChatGPT contributed systematic tightening of claim boundaries and unification of target hierarchy. Grok ensured consistency with the preceding 28 papers. Claude (mathematics thread) wrote all numerical computation scripts and drafted working notes v1–v3. The final text was completed independently by the author; all mathematical judgments are the author's responsibility.

---

8. Data Sources and Reproducibility

Script	Measurement	Section
p29_block123.py	Occupancy + moving threshold α + pred gap statistics	§3, §4
p29_block45.py	R(n) sequence + R(2m)−R(m) distribution + global extremes	§5, §4.3
p29_block67.py	gap≤0 full scan + gap distribution by decade + ln ln n regression	§4, §5.2
p29_block7_only.py	gap distribution by decade (lightweight, N=10⁸/10⁹)	§4.1

All scripts pass sanity checks: $\rho_E(10^7) = 58$, $\rho_E(10^8) = 66$, $\rho_E(10^9) = 74$.

---

References

1. ZFCρ Papers I–XXVIII. H. Qin. Paper XXVI DOI: 10.5281/zenodo.19059834. Paper XXVII DOI: 10.5281/zenodo.19062684. Paper XXVIII DOI: 10.5281/zenodo.19078654.
2. J. Arias de Reyna. Complexity of natural numbers and arithmetic compact coding. Preprint.
3. K. Cordwell, S. Epstein, A. Hemmady, S. J. Miller, E. Steiner (2018). On the number of 1's needed to represent n. J. Number Theory, 189:17–34.
4. H. Altman (2014). Integer complexity, addition chains, and well-ordering. PhD thesis, Rutgers University.
5. S. P. Meyn, R. L. Tweedie (1993). Markov chains and stochastic stability. Springer-Verlag, London.