\documentclass[10pt,a4paper]{article} \usepackage[T1]{fontenc} \usepackage[utf8]{inputenc} \usepackage{amsmath,amssymb,amsthm} \usepackage[margin=1in]{geometry} \usepackage[expansion=false]{microtype} \usepackage{longtable,booktabs} \usepackage[hidelinks]{hyperref} \hypersetup{pdftitle={A two-class Erdos--Rankin lower bound from finite estimates},pdfsubject={Exposition of three accepted mapped Lean conclusions}} \setlength{\emergencystretch}{2em} \allowdisplaybreaks[2] \newtheorem{theorem}{Theorem} \newcommand{\NN}{\mathbb N} \newcommand{\ZZ}{\mathbb Z} \newcommand{\RR}{\mathbb R} \newcommand{\CG}{\operatorname{Consec}} \title{A two-class Erd\H{o}s--Rankin lower bound\\from finite estimates} \author{} \date{LaTeX exposition edition, 9 October 2026} \begin{document} \maketitle \vspace{-1.8em} \begin{abstract} Let $P(y)=\prod_{p\le y}p$, with every prime at or below the real cutoff $y$ included, and let $\mathcal T_y$ be the set of integers $n$ for which $\gcd(n(n+2),P(y))=1$. We give a readable exposition of the accepted formal argument that the largest consecutive gap between members of $\mathcal T_y$ is at least a positive constant times \[ \frac{y(\log y)^3(\log\log\log y)^2}{(\log\log y)^4} \] for all sufficiently large real $y$. A three-band residue cover reduces the problem to three finite counting budgets. A finite Selberg sieve bounds the auxiliary survivors, a shifted finite Euler product bounds the smooth exceptions, and an elementary factorial argument supplies enough reserved primes to complete the cover. The sufficient estimates use the fixed parameter $A=12$ and require no additional prime-number-theorem or smooth-number-asymptotic hypothesis. The finite cyclic maximum is identified with an attained maximum over all integer starts, including negative starts and gaps crossing a period boundary. Seven stronger or more general obligations from the earlier manuscript are retained with their domains and remain outside the three accepted mapped conclusions. This edition presents the existing argument; it asserts neither novelty nor a new proof execution. \end{abstract} \paragraph{Edition and attribution.} This paper faithfully reuses and reorganizes the mathematical exposition in the accepted project manuscript accompanying \href{https://solveathome.org/projects/twin-primes/return/2597}{return \#2597}. Chris Benjaminsen supplied project direction. The current complete LaTeX draft was written by the actual selected \texttt{gpt-6-astra} AI assistant at high reasoning effort; Sol owns the subsequent integration, compilation, and fixes. The accepted source exposition itself disclosed preparation by \texttt{gpt-6.1-sol} at high reasoning effort. Historical contributors and adapted-source credits are preserved in Section~\ref{sec:credits}. No human endorsement, affiliation, sole authorship, or newly established mathematical result is implied by this edition. \section{The object and the three mapped conclusions}\label{sec:1} All logarithms are natural, and $\NN$ includes $0$. For a finite set $S$ of natural primes put \[ P_S=\prod_{p\in S}p,\qquad \mathcal T_S=\{n\in\ZZ:\gcd(n(n+2),P_S)=1\}. \] The gcd is nonnegative also for negative integer arguments. Define $\CG_S(s,d)$, for $s\in\ZZ$ and $d\in\NN$, to mean \[ d>0,\qquad s,s+d\in\mathcal T_S,\qquad \forall i\in\NN\ (00$ and $Y\in\RR$ such that, for every real $y\ge Y$, \begin{equation}\label{eq:main} cF(y)\le \bigl(G_2(P(y)):\RR\bigr).\tag{1} \end{equation} \end{theorem} The exact declaration is \path{PrimaryMain.manuscript_main_goal} in \path{PrimaryMain.lean}. Its target definition is \path{SieveParameters.ManuscriptMainGoal}. The same positive constant and threshold work for all subsequent real $y$; they are not selected anew for each cutoff. The construction gives $c=1/\operatorname{chooseB}(12)$. It proves the positivity of this coefficient and an eventual real threshold, without claiming an optimized numerical onset. \begin{theorem}[Attained integer maximum; mapped claim \texttt{integer\_maximum}] \label{claim:integer_maximum} For every $S:\operatorname{Finset}\NN$ such that every member of $S$ is prime, \[ \bigl(\exists s\in\ZZ\ \CG_S(s,G_2(P_S))\bigr) \quad\text{and}\quad \bigl(\forall s\in\ZZ\ \forall d\in\NN\ [\CG_S(s,d)\Longrightarrow d\le G_2(P_S)]\bigr). \] The finite prime family $S$ may be empty. \end{theorem} This is \path{IntegerMainBinding.integer_maximum_binding} in \path{IntegerMainBinding.lean}. Its explicit Lean parameters are \path{S : Finset Nat} and \path{hS : KKFiniteCoreDraft.PrimeFamily S}. The latter means $\forall p\in S,\ p\text{ is prime}$. It is a defining domain condition and is retained in this statement. \begin{theorem}[Literal integer bound; mapped claim \texttt{integer\_main\_eq1}] \label{claim:integer_main_eq1} There exist $c\in\RR$, $c>0$, and $Y\in\RR$ such that for every real $y\ge Y$ there is $d\in\NN$ satisfying all three conditions \[ \exists s\in\ZZ\ \CG_{S_y}(s,d),\qquad \forall s\in\ZZ\ \forall e\in\NN\ [\CG_{S_y}(s,e)\Longrightarrow e\le d],\qquad cF(y)\le(d:\RR). \] \end{theorem} This is \path{IntegerMainBinding.manuscript_integer_main_goal} in \path{IntegerMainBinding.lean}. It follows by taking $d=G_2(P(y))$, using the proved prime-family property of $S_y$, and applying Theorems~\ref{claim:main_eq1} and~\ref{claim:integer_maximum}. In the accepted inventory, \path{assumptions: []} means that the final claims have no additional analytic input or custom axiom. It does not erase explicit parameters, domain conditions, or the hypotheses of intermediate lemmas. \paragraph{Why the maximum is an integer maximum.} The polynomial $n(n+2)$ has the same residue at congruent integer arguments modulo $P_S$. Hence survivor membership is periodic even at negative arguments. The residue $P_S-1$ survives, since its two factors are congruent to $-1$ and $1$. Every interval of $P_S$ consecutive positions contains a survivor; every consecutive distance is therefore at most $P_S$. There is a next survivor after any given survivor, so $\mathcal D_S$ is nonempty and its finite supremum is attained. For an arbitrary integer start $s$, reduce it to its nonnegative remainder modulo $P_S$. For each offset, adding it before or after this normalization preserves the survivor predicate. An integer consecutive gap consequently has the same distance as a gap starting in $\{0,\ldots,P_S-1\}$. Conversely, every finite natural start is an integer start. This proves both assertions of Theorem~\ref{claim:integer_maximum}; it also explains why omitting the crossing gap would change the object. For $S=\varnothing$, $P_S=1$, every integer survives, and the attained largest distance is $1$. \paragraph{Plan of the mathematical argument.} A covering phase turns a block of $m$ excluded positions into a survivor gap of distance at least $m+1$. The early prime choices almost cover that block. The uncovered indices split into auxiliary sieve survivors, a short initial interval, and two smooth-number sets. The three estimates are arranged to fit within a reserve of distinct primes, one for each remaining index. Sections~\ref{sec:2}--\ref{sec:8} give the finite details and the threshold assembly. They are mathematical explanations of the accepted argument, not additional independently mapped or separately reviewed claims of this edition. Section~\ref{sec:9} summarizes the additional formalization scope and links the exact preserved original obligation ledger. \section{Covering an interval and producing a gap}\label{sec:2} At each prime $p\in S$, choose one residue $a_p$. Its admissible pair is \begin{equation*} \{a_p,a_p-2\}\pmod p. \end{equation*} A cover of $1,\ldots,m$ means that for each index $i$ some prime satisfies $i\equiv a_p\pmod p$ or $i+2\equiv a_p\pmod p$. The Chinese remainder theorem supplies a common phase $s\equiv-a_p\pmod p$. At that phase these alternatives become $p\mid s+i$ or $p\mid s+i+2$. Every covered index therefore gives a nonsurvivor $s+i$. For every $m\in\mathbb N$, it follows that \begin{equation} \exists(a_p)\text{ covering }1,\ldots,m \quad\Longleftrightarrow\quad m+1\le G_2(P_S). \tag{2} \end{equation} For the forward implication, a covered block need not begin immediately after a survivor. One may translate it by a period and choose the greatest preceding survivor; its next survivor lies beyond the block. Their distance is at least $m+1$. For the reverse implication, take the first $m$ interior positions of an attained gap of distance at least $m+1$, and choose the negative of its left endpoint at every prime. This proves (2) with the required endpoints. The case $m=0$ is included: the cover is vacuous, while the attained positive gap gives $1\le G_2(P_S)$. For the empty prime family, $P_S=1$ and $G_2(P_S)=1$, so the only coverable length is $0$. No gap of distance zero is introduced. The completion step is finite. Suppose an early set of primes has left $R$ indices uncovered and a disjoint reserved set has at least $R$ primes. An injection assigns a different reserved prime $p_i$ to each uncovered $i$. Set $a_{p_i}=i\pmod {p_i}$, retaining the old choices at early primes. Every remaining index is now covered. The resulting cover uses the union of the early and reserved prime families, and (2) gives a gap of distance at least $m+1$. No prescribed location of this gap is asserted. The relevant proofs are the CRT and phase bridges in \path{Arithmetic} and \path{Bridges}, the gap arguments in \path{Normalization} and \path{Gaps}, and the injection and completion in \path{Reserved} and \path{ReserveBudget}. \section{The parameters and the two different residue systems}\label{sec:3} Write \begin{equation*} L=\log y,\qquad L_2=\log L,\qquad L_3=\log L_2. \end{equation*} The construction uses \begin{equation} z_0=L^{12},\qquad z_1=\exp\!\left(\frac{LL_3}{12L_2}\right),\qquad m=\left\lfloor\frac yB\frac{L^3L_3^2}{L_2^4}\right\rfloor. \tag{3} \end{equation} The fixed positive constant $B$ will be given explicitly in Section 5. On the proved sufficiently large domain all displayed logarithmic denominators are positive, the mass inside the floor is nonnegative, and \begin{equation} 3X$. Thus the actual weights satisfy \begin{equation} \sum_{d\mid t}w_t=\frac{Xg(d)}d+r_d, \qquad |r_d|\le g(d). \tag{8} \end{equation} The sifted weight sum is exactly $S(X,\Omega)$. These identities are established for the counting weights, not supplied as unspecified sieve hypotheses. Put $\nu(p)=g(p)/p$, extending multiplicatively on squarefree divisors, and $h(d)=\prod_{p\mid d}\nu(p)/(1-\nu(p))$. At a real level $H>1$, the Selberg denominator is \begin{equation*} J(H)=\sum_{\substack{d\mid P_{\mathrm{aux}}\\d^2\le H}}h(d). \end{equation*} The explicit coefficients are \begin{equation*} \lambda_d=\nu(d)^{-1}J(H)^{-1}\mu(d) \sum_{\substack{k\mid P_{\mathrm{aux}}\\d\mid k,\ k^2\le H}}h(k). \end{equation*} They have $\lambda_1=1$, vanish outside $d^2\le H$, and their quadratic main term is exactly $1/J(H)$, by finite M\"obius inversion and diagonalization. Expanding the square against (8) gives the bound by $X/J(H)$ plus the actual remainder sum. The construction uses the squared-level convention above throughout. The remainder is controlled coarsely but explicitly. One has $|\lambda_d|\le d$ on supported divisors. A squarefree $n$ has $3^{\omega(n)}$ ordered divisor pairs with least common multiple $n$. Their aggregate coefficient therefore has absolute value at most $n^2 3^{\omega(n)}$, while $g(n)\le4^{\omega(n)}$. Only $n\le H$ contributes. The factors at $2$ and $3$ give \begin{equation*} 12^{\omega(n)}\le4n^2, \end{equation*} so the entire remainder is at most $4H^5$. For a prime subset $Q$, the subset expansion of $\prod_{p\in Q}(1+h(p))$ has logarithmic first moment $\sum_{p\in Q}\nu(p)\log p$. If this is at most $\log H/4$, at least half the mass has $d^2\le H$. Consequently $J(H)\ge\tfrac12\prod_{p\in Q}(1-\nu(p))^{-1}$. A finite prime-log estimate provides \begin{equation} \sum_{p\le N}\frac{\log p}{p}\le\log4\,(2+\log N),\qquad N\ge2. \tag{9} \end{equation} It follows from the elementary primorial bound $\theta(x)\le x\log4$ and finite summation, and does not use the prime number theorem. Choose $k=512\log4$, $N=\lfloor\exp(L/k)\rfloor$, and $H=\exp(L/16)$. These satisfy the required moment condition on primes up to $N$. Passing from that restricted Euler product to the full auxiliary product costs a bounded factor. The proof bounds its logarithm by \begin{equation*} E=20\log4\,k+80\log4. \end{equation*} Also $4H^5\le y/(12L)$ on the stated domain. With $C_s=2e^E$, the resulting finite estimate is \begin{equation} S(X,\Omega)\le C_sX V(\sqrt y)+\frac y{12L}, \qquad V(\sqrt y)=\prod_{p\le\sqrt y}(1-g(p)/p). \tag{10} \end{equation} The product is bounded by elementary finite estimates as well. Define $C=4\exp(10\log4)$. The ordinary inverse product obeys \begin{equation*} \prod_{p\le N}(1-1/p)^{-1}\le C\log N\quad(N\ge2), \end{equation*} and the harmonic sum gives the complementary upper bound for the ordinary product. Splitting a prime band therefore bounds its product by $C\log N_0/\log N_1$. At primes at least $3$, the basic two-class factor is at most the square of the ordinary factor; middle-band primes supply two additional powers. The exceptional prime $2$ costs at most a factor $2$. Converting real cutoffs to floors, with the proved logarithmic floor bounds, gives \begin{equation} V(\sqrt y)\le128C^2 \frac{(\log z_0/\log z_1)^2}{L^2}. \tag{11} \end{equation} Set $K=C_s\,128C^2$. Substitution of (3), using the natural floor bound, yields \begin{equation*} S(m,\Omega)\le\left(\frac{K12^4}{B}+\frac1{12}\right)\frac yL. \end{equation*} The actual choice is \begin{equation*} B=1+6K12^4, \end{equation*} which is positive, at least $1$, and gives $S(m,\Omega)\le y/(4L)$. These constants are deliberately generous; they are fixed expressions, not unspecified $O(1)$ factors. The proof establishes a sufficient one-sided estimate. It does not assert the original two-sided product asymptotic. \section{The smooth-number budget at \texorpdfstring{$A=12$}{A=12}}\label{sec:6} For any $\alpha>0$, $X\ge1$, and $Z\ge2$, the finite Rankin inequality is \begin{equation} \Psi(X,Z)\le X^\alpha \prod_{p\le Z}(1-p^{-\alpha})^{-1}. \tag{12} \end{equation} Indeed, each smooth $1\le n\le X$ has $1\le X^\alpha n^{-\alpha}$. Its weighted finite sum embeds into the sum over positive integers factored entirely over the finite set of primes $p\le Z$. The latter is a product of convergent geometric series because each $p^{-\alpha}<1$. This is local convergence for a finite prime family; the all-natural series need not converge when $\alpha\le1$. The identification with the actual smooth count uses \path{Nat.smoothNumbersUpTo X (floorNat(Z) + 1)}, preserving the inclusive condition $p\le Z$ despite that library's strict bound convention. Let $U=\log X/\log Z\ge1$, suppose $\log Z\ge1$ and $U\le\sqrt Z$, and set \begin{equation*} \delta=\frac{\log U}{2\log Z},\qquad\alpha=1-\delta. \end{equation*} Then $0\le\delta\le1/4$, $\alpha>0$, and $p^\delta\le\sqrt U$ for every prime $p\le Z$. The shifted local ratio satisfies \begin{equation} \frac{1-1/p}{1-p^{\delta-1}} =1+\frac{p^\delta-1}{p-p^\delta} \le\exp\!\left(\frac{4(p^\delta-1)}p\right). \tag{13} \end{equation} For this bound, $p^\delta\le\sqrt p\le3p/4$ makes the denominator positive and at least $p/4$; then $1+t\le e^t$ applies. The inequality $e^x-1\le xe^x$ gives \begin{equation*} p^\delta-1\le\delta\log p\sqrt U. \end{equation*} Summing (13)'s exponent with (9), and using $2+\log Z\le3\log Z$, bounds the total shift cost by $6\log4\sqrt U\log U$. Combining the product with the ordinary inverse bound and \begin{equation*} X^{1-\delta}=X\exp(-U\log U/2) \end{equation*} proves \begin{equation} \Psi(X,Z)\le CX\log Z \exp\!\left(-\tfrac12U\log U+6\log4\sqrt U\log U\right). \tag{14} \end{equation} Now use the actual $X=m+2,Z=z_1,U=u$. The proved parameter estimates give $X\le2yL^3$, $\log Z\le L$, and eventually $1\le u\le\sqrt Z$, $Z\ge e$. They also give the quantified limits \begin{equation*} \frac{\log X}{L}\longrightarrow1,\qquad \frac{\log u}{L_3}\longrightarrow1,\qquad \frac{u\log u}{L_2}\longrightarrow12, \end{equation*} and $u\longrightarrow\infty$. For clarity, these limits can be read directly from quantitative identities. Put $r=\log X/L$. The floor and mass estimates give $1\le r\le1+7/\sqrt L\le2$, and \begin{equation*} u=r\frac{12L_2}{L_3},\qquad \log u=L_3+\log12-\log L_3+\log r. \end{equation*} The bound $\log t\le2\sqrt t$ makes both $r-1$ and $(\log u)/L_3-1$ arbitrarily small at explicit sufficiently large thresholds. Their product gives $(u\log u)/L_2\to12$. Moreover $u\ge\sqrt{L_2}$ proves its growth. The same parameter estimates give $u\le48L_2$ and $\log Z\ge\sqrt L/48$; at the chosen larger threshold, a quadratic exponential bound implies $48L_2\le\sqrt Z$. Thus the range in (14) is proved for these exact parameters. These are parameter limits, not estimates for the number of smooth integers. Take the eventual error tolerance $1/2$, so $u\log u\ge(23/2)L_2$, and enlarge the threshold until $\sqrt u\ge288\log4$. Then the positive shift term in (14) is at most $u\log u/48$. Its total exponent is therefore at most \begin{equation*} -\frac{23}{48}u\log u\le-\frac{529}{96}L_2\le-\frac{11}{2}L_2. \end{equation*} The factor $X\log Z\le2yL^4$ and $\exp(-11L_2/2)=L^{-11/2}$ give \begin{equation*} \Psi(m+2,z_1)\le\frac{2Cy}{L\sqrt L}. \end{equation*} Finally enlarge the threshold until $\sqrt L\ge96C$. It follows that \begin{equation} 2\Psi(m+2,z_1)\le\frac y{24L}. \tag{15} \end{equation} Every range and growth condition used here is discharged in \path{SmoothBudget.eventual_smooth_and_uncovered_budget}. Together with (6), (7), and the sieve quarter-budget, it proves the actual bound $R\le y/(3L)$. The loss of a factor roughly $1/2$ in the decay exponent is acceptable at $A=12$; this argument does not establish the full original smooth-number asymptotic or its asserted consequence for every fixed $A>4$. \section{Enough primes in the reserved interval}\label{sec:7} We next prove the needed literal supply \begin{equation} \#\{p\text{ prime}:y/2\le p\le y\}\ge\frac y{3\log y}. \tag{16} \end{equation} This is derived by a factorial argument. To avoid confusing smooth counts with prime-power counts, write $\psi(N)=\sum_{n\le N}\Lambda(n)$ and $\theta(N)=\sum_{p\le N}\log p$, where $\Lambda(p^j)=\log p$ for prime $p$ and $j\ge1$, and $\Lambda$ vanishes off positive prime powers. Put $f(x)=x\log x-x$, with $f(0)=0$. Logarithmic ratio inequalities show, for positive integers $n$, \begin{equation*} \log n\le f(n+1)-f(n)\le\log(n+1). \end{equation*} Induction gives, for $N\ge1$, \begin{equation*} N\log N-N+1\le\log(N!) \le N\log N-N+1+\log N. \end{equation*} For $F(x)=\log(\lfloor x\rfloor!)$, this implies the uniform estimate \begin{equation} |F(x)-f(x)|\le1+\log(x+1)\quad(x\ge0). \tag{17} \end{equation} For $x\ge1$, let $n=\lfloor x\rfloor$. Logarithmic ratio inequalities give $(x-n)\log n\le f(x)-f(n)\le(x-n)\log x$. Since $0\le x-n<1$, this difference lies between $0$ and $\log x$; combining it with the factorial bounds gives (17). The range $0\le x<1$ is handled separately, where the factorial is $0!=1$ and $-1\le f(x)\le0$. Thus applying (17) at $x/30$ does not require $x\ge30$. Define the signed floor wheel and the factorial combination \begin{equation*} w(n)=n-\lfloor n/2\rfloor-\lfloor n/3\rfloor -\lfloor n/5\rfloor+\lfloor n/30\rfloor, \end{equation*} \begin{equation*} T(x)=F(x)-F(x/2)-F(x/3)-F(x/5)+F(x/30), \qquad a=\frac7{15}\log2+\frac3{10}\log3+\frac16\log5. \end{equation*} Adding $30$ changes the five terms in the wheel by $30,15,10,6,1$, whose signed combination is zero. Checking the thirty remainders therefore proves that the wheel takes only $0$ and $1$; it equals $1$ for $1\le n<6$. These are signed integer identities; truncated natural subtraction is not used. The exact combination of the five $f$-terms is $ax$, so (17) gives \begin{equation} |T(x)-ax|\le5(1+\log(x+1)). \tag{18} \end{equation} The divisor identity $\sum_{d\mid n}\Lambda(d)=\log n$, followed by finite summation, gives \begin{equation*} \log(N!)=\sum_{n\le N}\Lambda(n)\lfloor N/n\rfloor. \end{equation*} Substituting it in the factorial combination yields the exact identity \begin{equation*} T(N)=\sum_{n\le N}\Lambda(n)w(\lfloor N/n\rfloor). \end{equation*} Nonnegativity and the wheel properties now imply \begin{equation} 0\le T(N)\le\psi(N),\qquad \psi(N)-\psi(\lfloor N/6\rfloor)\le T(N). \tag{19} \end{equation} With $L_N=1+\log(N+1)$, strong induction proves \begin{equation} aN-5L_N\le\psi(N) \le\frac65aN+5L_N^2. \tag{20} \end{equation} For the upper bound at $N\ge6$, take $M=\lfloor N/6\rfloor$. The induction hypothesis and (19) bound $\psi(N)$ by $aN+5L_N+(6/5)aM+5L_M^2$. Since $M\le N/6$, the linear terms are at most $(6/5)aN$. Also $N+1\ge3(M+1)$ and $\log3\ge1$, so $L_M+1\le L_N$. Hence $L_N^2-L_M^2\ge L_N$, absorbing the new error. For $N<6$, the elementary bound $\psi(N)\le N\log(N+1)$ handles the base case. The higher-prime-power tail satisfies, for $N\ge2$, \begin{equation} 0\le\psi(N)-\theta(N)\le2\sqrt N(\log N)^2. \tag{21} \end{equation} Indeed, every nonprime prime power $n\le N$ is $p^j$ with $p\le\lfloor\sqrt N\rfloor$ and $2\le j\le\lfloor\log N/\log2\rfloor$. This finite set lies in the image of those base-exponent pairs. Its size is at most the number of pairs, without needing an injectivity assertion. Each summand is at most $\log N$, and $\log2\ge2/3$ gives the stated loose coefficient $2$. The constants needed for a positive dyadic margin are proved rationally. The inequalities $(24/23)^{16}\le2$, $(12/11)^{12}\le3$, and $(8/7)^{12}\le5$, together with $\log r\ge1-r^{-1}$, give \begin{equation*} \log2\ge2/3,\qquad\log3\ge1,\qquad\log5\ge3/2, \qquad a\ge31/36. \end{equation*} Also $a\le2$ follows from $\log t\le t-1$ at $2,3,5$. For real $y\ge4$, use $N=\lfloor y\rfloor$, $M=\lfloor y/2\rfloor$. Combining (20) and (21), with $N\ge y-1$ and $M\le y/2$, proves \begin{equation*} \theta(y)-\theta(y/2)\ge\frac{2a}{5}y-\mathcal E(y), \end{equation*} \begin{equation*} \mathcal E(y)=a+5L_++5L_+^2+2\sqrt y(\log y)^2, \qquad L_+=1+\log(y+1). \end{equation*} There is a concrete error threshold. If $y\ge300000^4$ and $\log y\ge1$, put $s=\sqrt{\sqrt y}\ge300000$. Then $y=s^4$, $\sqrt y=s^2$, and $\log y=4\log s\le8\sqrt s$. Therefore $(\log y)^2\le64s$, $L_+\le10s$, and $L_+^2\le576s$. It follows that \begin{equation*} \mathcal E(y)\le128s^3+2932s\le3060s^3\le y/90. \end{equation*} The last inequality uses $90\cdot3060=275400\le s$. Since $2a/5\ge31/90$, the difference of theta sums is at least $y/3$. This difference sums $\log p$ over the strict interval $y/20$, with the fixed $B$ of Section 5. On the positive large domain the floor mass in (3) equals $c\cdot yL^3L_3^2/L_2^4$. The elementary floor inequality gives \begin{equation*} c\frac{yL^3L_3^2}{L_2^4}4$ (Eqs.~23--24), beyond the sufficient $A=12$ route proved here; \href{https://solveathome.org/projects/twin-primes/return/2601}{follow-up \#2601}. \item\label{obligation:O4} \textbf{O4:} the dyadic prime-count asymptotic (Eq.~10), stronger than the finite reserve bound used here; \href{https://solveathome.org/projects/twin-primes/return/2602}{follow-up \#2602}. \item\label{obligation:O5} \textbf{O5:} the external one-class comparison and the general quoted sieve (Eqs.~5--7, 11), beyond the sufficient finite sieve bound; \href{https://solveathome.org/projects/twin-primes/return/2603}{follow-up \#2603}. \item\label{obligation:O6} \textbf{O6:} the original harmonic-density expansion and two-sided Euler-product comparisons (Eqs.~20--21); \href{https://solveathome.org/projects/twin-primes/return/2604}{follow-up \#2604}. \item\label{obligation:O7} \textbf{O7:} the separate short construction and its priced constant (Eq.~28); \href{https://solveathome.org/projects/twin-primes/return/2605}{follow-up \#2605}. \end{itemize} Each obligation remains open outside the three mapped conclusions. Closing one would extend the formal record; it would not repair a missing hypothesis of the main theorem proved in this paper. \section{Exact source correspondence and assurance}\label{sec:10} The three theorem labels in Section~\ref{sec:1} provide distinct locators for the three accepted claim IDs. Their correspondence is: \begin{center}\small \begin{tabular}{ll} \toprule Claim ID & Exact accepted declaration\\ \midrule \texttt{main\_eq1} & \path{PrimaryMain.manuscript_main_goal}\\ \texttt{integer\_maximum} & \path{IntegerMainBinding.integer_maximum_binding}\\ \texttt{integer\_main\_eq1} & \path{IntegerMainBinding.manuscript_integer_main_goal}\\ \bottomrule \end{tabular} \end{center} The accepted \path{proof-to-exposition-map.json} records the frozen source correspondence for the mathematical sections. In particular, the finite core definitions in \path{FiniteCoreTargets.lean} are specifications; the target definitions themselves are not asserted as proofs. Their actual proofs are supplied by the arithmetic, bridge, normalization, and gap modules, and the final three declarations above supply the accepted endpoints. The dependencies of the exposition are worth distinguishing. The covering step belongs to \path{Arithmetic}, \path{Bridges}, \path{Normalization}, \path{Gaps}, and \path{Reserved}. Actual local residue counts and incidence moments belong to \path{ManuscriptResidues}, \path{ResidueMoment}, \path{CRTMoments}, \path{IncidenceMoments}, and \path{DivisorMoments}. The finite sieve passes through \path{ActualBoundingSieve}, \path{SelbergFinite}, \path{SelbergError}, \path{SelbergEuler}, \path{SieveCoarse}, \path{SieveInterface}, \path{EulerRatio}, and \path{SieveParameters}. The ordinary-product estimates in \path{PrimeLog} and \path{MertensBand}, and their real-cutoff versions in \path{RealBand}, feed the explicit choice in \path{ManuscriptParameters}. The smooth budget uses \path{SmoothRankin}, \path{ShiftedRankin}, \path{SmoothParameters}, \path{SmoothLogLimit}, and \path{SmoothBudget}. The independent reserve-supply estimate uses \path{FactorialLogBounds}, \path{FactorialWheel}, \path{PrimeFactorial}, \path{PrimePsiBounds}, \path{PrimePowerTail}, and \path{PrimeReserveBudget}. Finally, \path{EarlyCover} and \path{ReserveBudget} assemble the cover before \path{PrimaryMain} and \path{IntegerMainBinding} yield the mapped claims. These source correspondences do not enlarge the inventory of accepted claims to the whole prose paper. \subsection{Public evidence and the edition boundary} The accepted source is \href{https://solveathome.org/projects/twin-primes/return/2597}{return \#2597}. Its recorded execution is \href{https://solveathome.org/projects/twin-primes/return/2598}{return \#2598}, receipt \#31, under the named \texttt{lean-kernel-v1} assurance profile. The exact statements were addressed by \href{https://solveathome.org/projects/twin-primes/review/695}{statement review \#695}; \href{https://solveathome.org/projects/twin-primes/review/696}{independent review \#696} accepted the three mapped conclusions. The latter records a \texttt{claude-opus-5-5} review at high effort, with spot inspection of the retained evidence and no Lean rerun. It expressly excludes O1--O7, the additional $y\log y$ consequence, and the expository prose of the intermediate sections from its mapped coverage. These are records of the accepted source and proof package. The present LaTeX edition has different bytes and has not thereby inherited an independent fidelity review of its own. Its review status belongs to its ordinary submission record. No claim that this edition has already passed that review is made here. For reproducible identification, the exact source manuscript has 41,786 bytes and SHA-256 \begin{quote}\small \nolinkurl{6c287c180b331318a19b3fb0a0e571147e01d785c4efa4341cf3bba5bd1c8e80}. \end{quote} The accepted package fingerprint is \begin{quote}\small \nolinkurl{91107e1d12018bd0b37bcf47471394133c5c1d3fc793900b990727acedc986e6}, \end{quote} and the statement binding is \begin{quote}\small \nolinkurl{cf2f5db08226d9151d245d8e103296d61d99de317df7e3af05eab0af6deda232}. \end{quote} These identifiers refer to the accepted inputs; they are not a hash or a certification of this new LaTeX file. The receipt reports a pass, matching target statements and proof bodies, standard-axiom auditing, and replay of 23,756 reachable declarations in an empty Lean kernel environment. Four negative controls---a mistaken statement, a mistyped proof, \texttt{sorryAx}, and a custom axiom---were rejected. The axiom policy permits only \texttt{propext}, \texttt{Classical.choice}, and \texttt{Quot.sound}; the three target reports list those standard axioms. There is no extra mathematical premise named Input S, M, H, or P in the three final claims. This evidence has a precise trust boundary. The run used the released Lean 4.35.0-rc3 compiler and its kernel, together with retained build objects and their checked custody. No fresh clean source rebuild of the solution and dependencies was performed, and the compiler was not built from source during that check. The proof replay used one Lean kernel; there was no independent second kernel. The stronger comparator/export validation, primitive-export stage, and Nanoda check were not run. Retained exports were supplementary and unchecked under this profile. The execution was author-run; the independent review inspected its evidence rather than performing a new execution. No Lean build, replay, or proof check was run for this LaTeX drafting task. None of these records establishes a full-paper proof badge or the broader analytic obligations. \section{Mathematical lineage, contributions, and licenses}\label{sec:credits} The construction follows the three-band architecture of Kalmynin and Konyagin~\cite{KK} and the existing fixed-offset adaptation in the project record. Their polynomial-root extremal problem differs from the fixed-separation pair $n,n+2$ here. Their theorem is not transferred to this object merely by a change of notation. The present exposition makes no novelty or literature-completeness claim. Bibliographic metadata for \cite{KK,HT} was checked for this edition; the historical sieve citations and source-access qualifications in O1--O7 are preserved rather than represented as newly audited literature. The historical contribution record distinguishes earlier drafting from this edition. The original disclosure records \href{https://solveathome.org/projects/twin-primes/return/1093}{return \#1093} under nielsegberts with \texttt{gpt-6-astra} (naming GitHub Copilot CLI), and \href{https://solveathome.org/projects/twin-primes/return/1730}{return \#1730} under Benjaminsen with \texttt{deepseek-v4-flash} (naming Freebuff). \href{https://solveathome.org/projects/twin-primes/return/158}{Return \#158} retains the historical OCR/source-custody and priced-constant discussion. The \href{https://solveathome.org/projects/twin-primes/docs/research/RESEARCH-CONTRIBUTIONS-2026-09-27.md}{project contribution record} and the accepted manuscript preserve those credits and the recorded Codex revisions. No contributor chronology or author dates have been reconstructed for this edition. The new Astra drafting disclosed above is a separate, actual drafting contribution, not a relabeling of the historical Astra return; Sol's subsequent integration remains a distinct role. The following adapted-source acknowledgments remain attached to the Lean files and are retained here for readers of the exposition: \begin{itemize} \item \path{SelbergFinite.lean} and \path{SelbergError.lean} adapt finite M\"obius/diagonalization and lambda-square/lcm-pair arguments from Arend Mellendijk's PNT+, revision\\ \nolinkurl{2667e414c38e5a5dc9aa1946f16f13001e5cd3ed}. They preserve the 2023 copyright, author credit, and Apache 2.0 notice. \item \path{PrimeLog.lean}, \path{MertensBand.lean}, and \path{SelbergEuler.lean} adapt \texttt{openai/math}, revision \nolinkurl{adc7f1241b42e322a6451854ab7e4b4c146bf78a}, under Apache 2.0. The headers identify the finite weighted prime-log, Mertens-factor/Rankin, finite Euler truncation, and denominator arguments and their local changes. \item \path{PrimeLog.lean} also credits Mathlib's \path{Chebyshev.lean}, copyright 2025 Alastair Irving, and names Alastair Irving, Terry Tao, and Ruben Van de Velde. The pinned Mathlib dependencies retain their contributors' notices and attribution. \end{itemize} Project prose is distributed under \href{https://creativecommons.org/licenses/by/4.0/}{CC BY 4.0}; this edition adapts the accepted exposition under those prose terms with attribution. The adapted Lean sources retain their separate Apache 2.0 notices and licenses. These acknowledgments do not grant new rights in external material or settle the distinct source-ledger repair described in \href{https://solveathome.org/projects/twin-primes/job/3880}{job \#3880}. The seven open obligations already have follow-up returns \#2599--\#2605 and jobs \#5411--\#5417; recording them here creates no new follow-up or closure claim. \begin{thebibliography}{9} \bibitem{KK} A.~B.~Kalmynin and S.~V.~Konyagin, \emph{A polynomial analogue of Jacobsthal function}, Izvestiya: Mathematics \textbf{88} (2) (2024), 225--235. \href{https://doi.org/10.4213/im9467e}{doi:10.4213/im9467e}. \href{https://arxiv.org/abs/2302.00459v2}{arXiv:2302.00459v2}. \bibitem{HT} A.~Hildebrand and G.~Tenenbaum, \emph{Integers without large prime factors}, Journal de Th\'eorie des Nombres de Bordeaux \textbf{5} (2) (1993), 411--484. \href{https://www.numdam.org/item/JTNB_1993__5_2_411_0/}{Primary journal record}. \bibitem{project} Solve@Home twin-primes project, \emph{A two-class Erd\H{o}s--Rankin lower bound from finite estimates}, accepted source manuscript accompanying \href{https://solveathome.org/projects/twin-primes/return/2597}{return \#2597}, with proof-to-exposition map, statement bundle, and frozen Lean sources; execution evidence in \href{https://solveathome.org/projects/twin-primes/return/2598}{return \#2598}, statement review \href{https://solveathome.org/projects/twin-primes/review/695}{\#695}, and independent review \href{https://solveathome.org/projects/twin-primes/review/696}{\#696}. This is the primary source for the argument and the historical quotations in this edition. \end{thebibliography} \end{document}