# 12 — Arithmetic Realizability Bridge ## From the Unconditional Prime Side at $\sigma=1$ to the Arithmetic Requirements of $P_{70}/P_{80}/P_{90}/P_{95}/P_{99}$ **Date:** 2026-08-11 **Status:** source-grounded reconstruction + derived conditional schema **Core Question:** $$ \boxed{ \text{The extremal side already knows how much support is needed; what exactly must the arithmetic side provide to legitimately realize it?} } $$ --- # 0. Known Boundary: Why is Claude Stuck at $\sigma=1$? Claude Proposition 5.6 decomposes the quadratic mean of the prime polynomial $$ P_X $$ as: $$ \mathcal M[P_X,P_X] = D+O_1+O_2. $$ where the diagonal is: $$ D \sim \frac{TL^3}{6\pi}, $$ and is currently bounded using the Montgomery–Vaughan generalized Hilbert inequality: $$ \boxed{ O_1\ll L^2X. } $$ If: $$ X=(T/2\pi)^\sigma, $$ ignoring fixed constants: $$ \frac{|O_1|}{D} \ll \frac{X}{TL} \asymp \frac{T^{\sigma-1}}{\log T}. $$ Thus: ### $\sigma<1$ $$ \frac{|O_1|}{D}\to0. $$ ### $\sigma=1$ $$ \frac{|O_1|}{D} \ll \frac1{\log T} \to0. $$ ### Any Fixed $\sigma>1$ $$ \frac{|O_1|}{D} \gg \frac{T^{\sigma-1}}{\log T}, $$ the current absolute value bound fails. This is not merely a matter of the "proof technique not being elegant enough"; rather, the off-diagonal terms must begin to be **evaluated**, not just **bounded away**. --- # 1. What Prime-Pair Scale Does the Off-Diagonal Actually Probe? Claude's $O_1$ contains: $$ \frac{1}{\log n-\log m} $$ and: $$ (n/m)^{iT} $$ -like oscillations. For: $$ n\sim m\sim X, $$ let: $$ h=n-m. $$ When: $$ |h|\ll X $$ we have: $$ \log(n/m) = \log\left(1+\frac{h}{m}\right) \approx \frac{h}{X}. $$ The time interval length is $T$, so the near-diagonal region where the off-diagonal cannot sufficiently oscillatory-cancel roughly satisfies: $$ T|\log(n/m)| \lesssim1. $$ That is: $$ \boxed{ |h| \lesssim \frac XT. } $$ If: $$ X=T^\sigma, $$ then: $$ \boxed{ H_\sigma \asymp T^{\sigma-1}. } $$ Expressed in the prime scale $X$: $$ T=X^{1/\sigma}, $$ thus: $$ \boxed{ H_\sigma \asymp X^{1-1/\sigma}. } $$ **This section is a scale inference derived from the explicit $O_1$ formula in Claude Proposition 5.6; it is not a formula directly listed in Claude's original text.** --- # 2. Prime-Pair Shift Scales Corresponding to CPL Proportions From the v8 generalized one-delta reconstruction: | target | $\sigma$ | $H_\sigma$ in $T$ scale | $H_\sigma$ in prime scale $X$ | |---|---:|---:|---:| | $70\%$ | $1.042628$ | $T^{0.042628}$ | $X^{0.040885}$ | | $80\%$ | $1.257848$ | $T^{0.257848}$ | $X^{0.204991}$ | | $90\%$ | $1.701455$ | $T^{0.701455}$ | $X^{0.412268}$ | | $95\%$ | $2.260790$ | $T^{1.260790}$ | $X^{0.557677}$ | | $99\%$ | $4.187215$ | $T^{3.187215}$ | $X^{0.761178}$ | The physical meaning here is not "we need to know all $h$ up to this point." Rather: > When the Dirichlet polynomial grows to $X=T^\sigma$, the near-diagonal pair interactions naturally fall into this additive-shift scale; without sufficient averaged or shift-by-shift prime-pair information, $O_1$ cannot yield the required main term / cancellation. --- # 3. A Classical Strong Hardy–Littlewood Input Goldston's pair-correlation notes record the strong prime-pair hypothesis used by Montgomery: $$ \boxed{ \sum_{n\le N} \Lambda(n)\Lambda(n+k) = \mathfrak S(k)N + O_\varepsilon(N^{1/2+\varepsilon}), } $$ uniformly for: $$ 02, $$ $$ \sigma_{99}\approx4.187>2. $$ Thus: $$ \boxed{ P_{95},P_{99} } $$ have already stepped outside the standard region where "the shift-by-shift Hardy–Littlewood pair conjecture with square-root error is known to imply SPC." This is a more substantial **arithmetic regime change** than a mere $q=90\%\to95\%$. --- # 5. The Second Arithmetic Input: Prime Variance / Short Intervals The classical Goldston–Montgomery equivalence (contextualized under RH in the original literature) connects the Strong Pair Correlation with the second moment of primes in short intervals: $$ \int_1^X \left( \psi(x+h)-\psi(x)-h \right)^2dx \sim hX\log\frac Xh. $$ Goldston's notes specifically point out: > The square-root error of the shift-by-shift twin-prime conjecture only directly gives SPC up to $\alpha<2$; > but this weaker, averaged short-interval second-moment hypothesis can supply the full fixed support range for SPC. So for $P_{95},P_{99}$, the more natural arithmetic hypothesis is not to require an extremely strong pointwise Hardy–Littlewood error for every shift, but rather to require: $$ \boxed{ \text{a prime short-interval variance asymptotic over a sufficiently wide scale}. } $$ --- # 6. Defining Three Tiers of Arithmetic Bridge Hypotheses To avoid stating "HL is needed" too vaguely, CPL temporarily divides this into three tiers. ## ABH-1 — Direct Prime-Side Trace Hypothesis For a specified support $\sigma$ and optimized one-delta family, assume Claude's prime-side trace evaluation can be extended such that: $$ \frac{ (\operatorname{tr}\widetilde G)^2 }{ N\operatorname{tr}(\widetilde G^2) } \to c(\sigma), $$ where: $$ c(\sigma) = \frac{1}{2-q(\sigma)}. $$ This is the hypothesis **closest to the Claude proof pipeline**. It avoids pre-specifying which prime conjecture must be responsible. --- ## ABH-2 — Partial Strong Pair Correlation Assume: $$ F(\alpha,T)=1+o(1) $$ uniformly for: $$ 1\le|\alpha|\le\sigma. $$ Together with the known $|\alpha|\le1$ region, this provides the full pair data up to $\sigma$ required for the generalized one-delta certificate. This is in the language of zero/pair-statistics. --- ## ABH-3 — Arithmetic Realization Using prime-side sufficient conditions, for example: ### For $\sigma<2$ The strong Hardy–Littlewood prime-pair: $$ \sum_{n\le N}\Lambda(n)\Lambda(n+h) = \mathfrak S(h)N+O(N^{1/2+\varepsilon}) $$ classical framework. ### For Larger Fixed $\sigma$ Adopt a: $$ \int (\psi(x+h)-\psi(x)-h)^2dx $$ -type short-interval variance asymptotic, or directly assume the arithmetic input required for full SPC / PCC. --- # 7. Derived Conditional Theorem Schema Below is **a conditional schema reconstructed in this study based on Claude's linear-algebra pipeline + the v8 one-delta operator; it is not a verbatim theorem from the source paper.** If for some fixed: $$ \sigma>1 $$ one can legitimately extend the prime-side second-trace evaluation of Claude Proposition 5.6 / Theorem 5.8 to the optimized support-$\sigma$ one-delta test family, while keeping all localization / taper errors: $$ o(N(T,2T)), $$ such that the corresponding trace ratio is: $$ c(\sigma) = \frac1{2-q(\sigma)}, $$ then the same zero-side inertia + rank–trace mechanism should yield: $$ \boxed{ \liminf_{T\to\infty} \frac{ N_0^s(T,2T) }{ N(T,2T) } \ge q(\sigma). } $$ Thus: ### If the Arithmetic Side Can Realize $\sigma=1.04263$ $$ P_{70} $$ enters conditional closure. ### If It Can Realize $\sigma=1.25785$ $$ P_{80}. $$ ### If It Can Realize $\sigma=1.70146$ $$ P_{90}. $$ This breaks the problem down very cleanly: $$ \boxed{ \text{Proportion threshold} \rightarrow \text{extremal support} \rightarrow \text{arithmetic realization}. } $$ --- # 8. How Can the Minimal Arithmetic Problem for $P_{70}$ Be Posed Now? We no longer ask: > "Can we prove $70\%$?" Instead, we ask: $$ \boxed{ \text{Can we push the prime-side second trace from }X\le T \text{ to } X\le T^{1.04263} \text{?} } $$ Converted to an additive shift: $$ \boxed{ h \lesssim T^{0.04263}. } $$ Or in the prime scale: $$ \boxed{ h \lesssim X^{0.04089}. } $$ This scale is very small. But the real difficulty is not that the exponent is small, but rather: $$ \boxed{ \text{it has already stepped outside the diagonal-only mean-value regime.} } $$ Therefore, even just going from: $$ 1 \to 1.0001 $$ is already a qualitative change. --- # 9. A Noteworthy Historical Comparison Goldston's notes state that Montgomery's strong Hardy–Littlewood hypothesis allows the program to be carried out up to: $$ x\le T\le x^{2-\varepsilon}, $$ which is: $$ \alpha<2. $$ So from a historical perspective: $$ P_{90} $$ requires: $$ \sigma\approx1.70 $$ which still falls within the range that Montgomery's original prime-pair heuristic was expected to handle. The first CPL node that truly begins to demand a new arithmetic regime is instead: $$ \boxed{ P_{95} } $$ because: $$ \sigma_{95}\approx2.26. $$ This point is highly non-intuitive in the "proportion ladder," but is visible at a glance in the "support ladder." --- # 10. What is Most Worth Doing in the Next Round? There are currently three routes. ## Route A — $P_{70}$ Weighted Prime-Pair Hypothesis Starting from the actual: $$ O_1 $$ formula in Claude Proposition 5.6, instead of assuming the full Hardy–Littlewood, we only extract **a weighted prime-pair asymptotic sufficient to make the optimized $\sigma=1.04263$ test hold**. This might be much weaker than the standard pointwise HL. ## Route B — Average-$h$ Hypothesis Because $O_1$ itself has smooth weights on $n,m$, we might not need a prime-pair asymptotic for: $$ \forall h $$ We can investigate: $$ \sum_h W(h/H) \left[ \sum_n\Lambda(n)\Lambda(n+h)-\mathfrak S(h)X \right] $$ requiring only: $$ o(\text{main}). $$ This would be closer to the modern "average Hardy–Littlewood" or short-interval variance. ## Route C — Unconditional Partial Escape Search for whether there exist currently proven results that, although insufficient to yield: $$ F(\alpha)\sim1 $$ can provide a one-sided bound for: $$ 1<\alpha<1+\delta $$ and test whether we can push $67.25\%$ up a bit. This might not reach $70\%$, but if we can unconditionally break through: $$ 68.185\% $$ that would be a result on an entirely different level. --- # 11. Conclusion of This Round We have advanced from: $$ \text{"$\sigma>1$ requires prime pairs"} $$ to: $$ \boxed{ \text{Target proportion} \rightarrow \sigma_q \rightarrow X=T^{\sigma_q} \rightarrow h\sim X/T \rightarrow \text{required arithmetic correlation scale}. } $$ This makes the true QCI of $P_{70}$ operational: $$ \boxed{ X:T\to T^{1.04263} } $$ is the first wall.