\documentclass[11pt]{amsart} \usepackage[T1]{fontenc} \usepackage{amsmath,amssymb,amsthm,mathtools} \usepackage{microtype} \usepackage{xcolor} \usepackage{booktabs} \usepackage{enumitem} \usepackage[a4paper,top=25mm,bottom=25mm,left=27mm,right=27mm]{geometry} \usepackage[colorlinks=true,linkcolor=blue!45!black,citecolor=blue!45!black,urlcolor=blue!55!black,bookmarks=true,bookmarksnumbered=true]{hyperref} \numberwithin{equation}{section} \newtheorem{theorem}{Theorem}[section] \newtheorem{lemma}[theorem]{Lemma} \newtheorem{proposition}[theorem]{Proposition} \newtheorem{corollary}[theorem]{Corollary} \theoremstyle{definition} \newtheorem{definition}[theorem]{Definition} \newtheorem{remark}[theorem]{Remark} \DeclareMathOperator{\rad}{rad} \DeclareMathOperator{\Supp}{Supp} \newcommand{\N}{\mathbb{N}} \newcommand{\Z}{\mathbb{Z}} \newcommand{\eps}{\varepsilon} \newcommand{\dd}{\mathrm{d}} \newcommand{\radout}[2]{\operatorname{rad}_{#1^{c}}(#2)} \newcommand{\fr}{\mathfrak} \title[Balanced radical triples and Pell discriminants]{Power-saving bounds for balanced radical triples\\via small Pell discriminants} \author{Chunping Zeng} \date{19 November 2026 (revised; first version 18 November 2026)} \subjclass[2020]{Primary 11D45; Secondary 11D09, 11N36, 11D75} \keywords{radical, abc exceptional set, Pell equation, squarefree kernel, dyadic decomposition} \hypersetup{ pdftitle={Power-saving bounds for balanced radical triples via small Pell discriminants}, pdfauthor={Chunping Zeng}, pdfsubject={Arithmetic counting via small fundamental solutions of Pell equations}, pdfkeywords={radical, Pell equation, abc exceptional set, squarefree kernel} } \begin{document} \begin{abstract} For $0<\lambda\leq 1$, let $\mathcal B_\lambda(N)$ be the set of integers $n\leq N$ for which $\rad(n(n+1)(2n+1))<(2n+1)^\lambda$. The elementary radical-box estimate gives $|\mathcal B_\lambda(N)|\ll N^{\lambda/3+\eps}$. We prove a strict power-saving refinement of this exponent. There is an absolute constant $C>0$ such that $|\mathcal B_\lambda(N)|\ll_{\lambda,\eps}N^{\lambda/3-C\lambda^3+\eps}$. The argument uses Diao's uniform bound for the number of small fundamental solutions to Pell equations, together with a squarefree-kernel transfer for the three unit-difference equations, a fixed-support radical-box lemma, and a three-branch dyadic optimization. We also formulate the resulting estimate for high-quality values and for Chinese-remainder orbits, and quantify the variable-right-hand-side obstruction to a direct extension to arbitrary abc triples. No assertion of the abc or Szpiro conjecture is used, and no pointwise abc inequality is claimed. The only non-elementary external input is Diao's Theorem 1.3 in arXiv:2610.28001v1, whose statement and preprint status are recorded explicitly. \end{abstract} \maketitle \tableofcontents % ===== BEGIN sections/01_intro.tex ===== \section{Introduction}\label{sec:intro} For a positive integer $m$, write $\rad(m)=\prod_{p\mid m}p$ for its radical. The three integers \begin{equation}\label{eq:three-coordinates} m_1(n)=n,\qquad m_2(n)=n+1,\qquad m_3(n)=2n+1 \end{equation} are pairwise coprime and satisfy the three unit-difference relations \begin{equation}\label{eq:unit-differences} m_2-m_1=1,\qquad m_3-2m_1=1,\qquad m_3-2m_2=-1. \end{equation} These relations make their joint radical a rigid one-dimensional Diophantine object, distinct from the full three-variable abc exceptional set. Background on the abc problem and its radical formulation can be found in Granville--Tucker \cite{GranvilleTucker2002}. For $0<\lambda\leq 1$ and $N\geq 2$, define \begin{equation}\label{eq:Blambda} \mathcal B_\lambda(N)= \left\{1\leq n\leq N:\ R_n<(2n+1)^\lambda\right\}, \qquad R_n=\rad(n(n+1)(2n+1)). \end{equation} Put $r_i(n)=\rad(m_i(n))$. Pairwise coprimality gives $R_n=r_1r_2r_3$. On $\mathcal B_\lambda(N)$, at least one $r_i$ is at most $N^{\lambda/3+o(1)}$. Since a coordinate determines $n$, the standard radical-box bound (Corollary \ref{cor:radical-box}) yields the elementary estimate \begin{equation}\label{eq:elementary-bound} |\mathcal B_\lambda(N)|\ll_{\lambda,\eps}N^{\lambda/3+\eps}. \end{equation} The purpose of this paper is to give all steps of a strict improvement of the exponent $\lambda/3$. The main new mechanism is a transfer from squarefree kernels to Pell discriminants. Write each coordinate uniquely as \begin{equation}\label{eq:kernel-intro} m_1=ax^2,\qquad m_2=by^2,\qquad m_3=cz^2, \end{equation} where $a,b,c$ are the squarefree products of the primes occurring to odd valuation in their respective coordinates. Each pair in \eqref{eq:unit-differences} then produces a solution of an ordinary Pell equation. For example, \begin{equation}\label{eq:first-pell-intro} (2n+1)^2-ab(2xy)^2=1. \end{equation} Thus, if $ab$ is small compared with $N$, the corresponding Pell discriminant has an unusually small fundamental solution. A uniform power-saving bound for such discriminants, proved by Diao \cite[Theorem 1.3]{Diao2026}, supplies the saving exactly in the dyadic range where the two elementary covers balance. There are three technical points that must not be suppressed. First, for $m=ax^2$ with $a$ squarefree, the correct radical formula is \[ \rad(m)=a\radout{a}{x}, \] where $\radout{a}{x}$ retains only prime divisors of $x$ coprime to $a$; in general it is not $a\rad(x)$. We prove a fixed-support box estimate that controls the primes shared by $a$ and $x$. Second, a fixed Pell discriminant can contribute more than one $n$, but its positive solutions grow exponentially, so the fibre size is only $O(\log N)$. Third, the square-part cover uses two-sided relations $x_i^2\asymp N/U_i$. We therefore prove the correlation estimates first on $N/20$ such that, for every $0<\lambda\leq1$ and every $\eps>0$, \begin{equation}\label{eq:main-bound} \boxed{|\mathcal B_\lambda(N)| \ll_{\lambda,\eps}N^{\lambda/3-C\lambda^3+\eps}.} \end{equation} More precisely, if $c_D$ is the constant in Diao's Theorem \ref{thm:diao}, then one may take $C=c_D/2160$ after harmlessly reducing $c_D$ so that $c_D\leq1$. \end{theorem} The improvement is small but strict. Its cubic shape is not an arbitrary choice: in the critical range the Pell discriminant has size $N^s$ with $s\asymp\lambda$, while the fundamental solution is at most $N$, so the relevant Diao parameter is $\theta\asymp s^{-1}$. The saving $c_D\theta^{-2}$ therefore becomes $O(s^3)$ on the $N$-exponent scale. For $\eta>0$, define the quality \begin{equation}\label{eq:quality-intro} q(n)=\frac{\log(2n+1)}{\log R_n}. \end{equation} Then $q(n)>1+\eta$ is equivalent to $R_n<(2n+1)^{1/(1+\eta)}$. Theorem \ref{thm:main} consequently gives \begin{equation}\label{eq:quality-bound-intro} \boxed{ \#\{n\leq N:q(n)>1+\eta\} \ll_{\eta,\eps} N^{\frac{1}{3(1+\eta)}-\frac{C}{(1+\eta)^3}+\eps}.} \end{equation} A Chinese-remainder orbit version is stated and proved in Section \ref{sec:crt}. \subsection*{Status of the input and scope} Theorem \ref{thm:main} uses no conjectural hypothesis. Its only non-elementary external ingredient is Diao's preprint theorem on small Pell solutions \cite{Diao2026}; we do not claim to reprove that theorem here. The new arguments are the support-adjusted radical covering, the uniform three-pair Pell transfer, the dyadic synthesis, and the passage to \eqref{eq:main-bound}. The result is a counting theorem for a special one-dimensional family. It neither proves a pointwise abc inequality for this family nor implies the full abc or Szpiro conjecture. \subsection*{Organization} Section \ref{sec:prelim} proves the fixed-support and radical-box estimates. Section \ref{sec:pell} derives all three Pell equations and controls their fibres. Section \ref{sec:dyadic} establishes the three-branch dyadic bound. Section \ref{sec:diag} proves the diagonal theorem and deduces Theorem \ref{thm:main}. Section \ref{sec:general-abc} analyzes the variable-square obstruction for arbitrary abc triples and records the resulting open problem. Section \ref{sec:crt} treats CRT orbits and discusses limitations. Appendix \ref{app:bookkeeping} collects the exponent bookkeeping and the linear-progression details used in the main proof. % ===== END sections/01_intro.tex ===== % ===== BEGIN sections/02_prelim.tex ===== \section{Fixed-support and radical-box estimates}\label{sec:prelim} Throughout, $N\geq2$ is the global height parameter. The notation $A\lesssim B$ is never used; $A\ll B$ has its usual meaning, and a subscript indicates permitted dependence of the implied constant. We write $f(q,X)$ for the number of integers $1\leq m\leq X$ all of whose prime factors divide $q$, and $\Psi(X,Y)$ for the number of $Y$-friable integers $m\leq X$. \subsection{A uniform fixed-support estimate} \begin{lemma}[Fixed-support box]\label{lem:fixed-support} Uniformly for integers $1\leq q,X\leq N$, \begin{equation}\label{eq:fixed-support} f(q,X)=N^{o(1)}. \end{equation} More precisely, for every $\eps>0$ there is $C_\eps$ such that $f(q,X)\leq C_\eps N^\eps$ in this range. \end{lemma} \begin{proof} We may replace $q$ by its radical without changing $f(q,X)$. Let $q_1<\cdots1$ and $U_{ij}>0$. Thus $D_{ij}$ is nonsquare. Finally, $\eta=T_{ij}+U_{ij}\sqrt{D_{ij}}$ is a Pell unit greater than $1$, so $\eps_{D_{ij}}\leq\eta$. Since $U_{ij}\sqrt{D_{ij}}=\sqrt{T_{ij}^2-1}\sqrt D\geq\sqrt2$, \[ k\log\eps_D\leq\log(2T_{ij})\ll\log N, \] and hence there are $O(\log N)$ possible exponents $k$. Finally, $T_{12}=2n+1$, $T_{13}=4n+1$, and $T_{23}=4n+3$ each determine $n$ uniquely. \end{proof} \subsection{Diao's uniform small-solution bound} For $\theta>0$ and $X\geq2$, define \begin{equation}\label{eq:Etheta} E_\theta(X)= \#\{00$ such that, for all sufficiently large $X$, \begin{equation}\label{eq:diao-bound} \boxed{E_\theta(X)\ll X^{1-c_D/\theta^2}} \end{equation} uniformly in the range \begin{equation}\label{eq:diao-range} 3\leq\theta\leq c_D\sqrt{\frac{\log X}{\log\log X}}. \end{equation} The implied constant is absolute. \end{theorem} \begin{remark}\label{rem:preprint-status} Theorem \ref{thm:diao} is quoted from arXiv:2610.28001v1, dated 20 October 2026. The present paper uses it as a cited preprint result and does not assert an independent verification or a peer-reviewed publication status. Diao's Section 8.1 has already passed from the squarefree Proposition 8.1 to general nonsquare discriminants by using the squarefree estimate with parameter $2\theta$ and putting $c=c_0/4$. Our $c_D$ denotes the resulting constant in the final Theorem 1.3, so no further halving or other reduction is required. Diao explicitly leaves $c$ unoptimized, and the auxiliary absolute constants in the later propositions are not assigned numerical values; accordingly, we claim an absolute positive constant but no numerical value for it. Reducing $c_D$ weakens \eqref{eq:diao-bound} and shrinks the upper endpoint in \eqref{eq:diao-range}; this is harmless in our application because only bounded values of $\theta$ are used. We therefore assume $c_D\leq1$ when convenient. \end{remark} The next lemma converts Theorem \ref{thm:diao} directly into a count of $n$ in a dyadic kernel-product box. \begin{lemma}[Small-discriminant branch]\label{lem:small-discriminant} Fix $1\leq i8N+6, \] so every discriminant in the box belongs to $E_\theta(CW)$ for an absolute $C>0$. The upper restriction in \eqref{eq:diao-range} is satisfied eventually because $\theta=2/s$ is bounded on the fixed compact set, whereas the permitted upper endpoint tends to infinity. Theorem \ref{thm:diao} gives \[ E_\theta(CW)\ll W^{1-c_D/\theta^2} =W^{1-(c_D/4)s^2}. \] Lemma \ref{lem:pell-fibres} multiplies this by only $O(\log N)$. Rewriting the result in terms of $N$ proves \eqref{eq:small-discriminant-count}; the dyadic constant $C$ contributes only $N^{o(1)}$. \end{proof} % ===== END sections/03_pell.tex ===== % ===== BEGIN sections/04_dyadic.tex ===== \section{The three-branch dyadic estimate}\label{sec:dyadic} For $1\leq i0$, \begin{equation}\label{eq:coefficient-cover} \mathcal C_{ij}(A,B;N;W)\ll_\eps W N^\eps. \end{equation} If in addition $W\leq N$, then \begin{equation}\label{eq:square-part-cover} \mathcal C_{ij}(A,B;N;W)\ll_\eps \frac{AB}{W}N^\eps. \end{equation} The implied constants are independent of $i,j$. \end{lemma} \begin{proof} The first bound is immediate from the Pell transfer. There are $O(W)$ possible discriminants $D_{ij}=\kappa_{ij}w_{ij}$ in the box, and each has $O(\log N)$ preimages by Lemma \ref{lem:pell-fibres}. For the second bound, dyadically localize the odd-valuation kernels: \[ u_i\in[U_i,2U_i),\qquad u_j\in[U_j,2U_j), \] where $(u_1,u_2,u_3)=(a,b,c)$ and $(x_1,x_2,x_3)=(x,y,z)$. Since $w_{ij}=u_iu_j\in[W,2W)$, we have $U_iU_j\asymp W$. There are $O((\log N)^2)$ such pairs $(U_i,U_j)$, which are absorbed by $N^\eps$ after replacing $\eps$. The exact radical formula \eqref{eq:three-radical-formulae} gives \[ \radout{u_i}{x_i}\leq\frac{A}{u_i}\ll\frac{A}{U_i}, \qquad \radout{u_j}{x_j}\ll\frac{B}{U_j}. \] Because $N/20 \] for $s<2/5$ and $c_0\leq1$, since then $3c_0s^2<12/25$. Therefore \begin{align} s-c_0s^3 &\leq \rho+4\tau-c_0(\rho+4\tau)^3\notag\\ &\leq \rho+4\tau-c_0\rho^3 =\rho+4\tau-10\tau =\rho-6\tau.\label{eq:diag-case-middle} \end{align} This is stronger than the required $\rho-\tau$. Combining \eqref{eq:diag-case-small}, \eqref{eq:diag-case-large}, and \eqref{eq:diag-case-middle}, and summing over the dyadic boxes, proves \eqref{eq:diagonal-bound}. \end{proof} \subsection{From two small radicals to the balanced family} The remaining step is a combinatorial amplification. If the smallest coordinate radical is already below the improved threshold, Corollary \ref{cor:radical-box} suffices. Otherwise, the radical budget forces the two smallest radicals into the diagonal window of Theorem \ref{thm:diagonal}. \begin{lemma}[Diagonal amplification]\label{lem:amplification} Suppose that for some $\vartheta,\tau>0$ and all $iN^{\rho-\delta}$. Dyadically localize $r_{(1)}$ and $r_{(2)}$, and write \begin{equation}\label{eq:a-b-exponents} r_{(1)}=N^{a+o(1)},\qquad r_{(2)}=N^{b+o(1)}. \end{equation} The $o(1)$ terms include both the dyadic localization and the $N$-dependence of the boxes. Because $r_{(1)}^3\leq r_1r_2r_3$, \begin{equation}\label{eq:a-upper} a\leq\rho+o(1), \end{equation} while the assumption on $r_{(1)}$ and the ordering give \begin{equation}\label{eq:a-b-lower} a\geq\rho-\delta+o(1),\qquad b\geq a+o(1). \end{equation} Moreover, $r_{(3)}\geq r_{(2)}$ implies $r_{(1)}r_{(2)}^2\leq N^{\lambda+o(1)}$, hence \begin{equation}\label{eq:a-plus-2b} a+2b\leq\lambda+o(1)=3\rho+o(1). \end{equation} Combining \eqref{eq:a-b-lower} and \eqref{eq:a-plus-2b} gives \begin{equation}\label{eq:b-upper} b\leq\frac{3\rho-a}{2}+o(1) \leq\rho+\frac{\delta}{2}+o(1). \end{equation} After shrinking the dyadic grid by a further $o(1)$ allowance, the two exponents lie in \[ [\rho-\delta,\rho+\delta/2]\subset[\rho-\tau,\rho+\tau], \] because $\delta<\tau$. Thus the pair of coordinates attaining $r_{(1)}$ and $r_{(2)}$ is covered by \eqref{eq:general-diag-assumption}. There are only three pairs and $O((\log N)^2)$ dyadic boxes, so this contribution is \[ \ll_{\lambda,\eps}N^{\rho-\vartheta+\eps} \leq N^{\rho-\delta+\eps} \] after replacing $\eps$ appropriately. Adding the small-$r_{(1)}$ contribution proves \eqref{eq:general-amplified-bound}. \end{proof} \subsection{Proof of the main theorem and the quality corollary} \begin{proof}[Proof of Theorem \ref{thm:main}] Theorem \ref{thm:diagonal} verifies the hypothesis of Lemma \ref{lem:amplification} with \[ \vartheta=\tau=\frac{c_0\rho^3}{10}. \] Choose $\delta=\tau/2=c_0\rho^3/20$. The dyadic conclusion is therefore \begin{equation}\label{eq:main-dyadic-bound} |\mathcal B_\lambda^{\mathrm{dy}}(M)| \ll_{\lambda,\eps}M^{\rho-c_0\rho^3/20+\eps}, \end{equation} uniformly in the scale $M$. Apart from a bounded residual interval, the intervals $(M/2,M]$ with $M=N/2^j$ and $j\geq0$ partition $[1,N]$. Applying \eqref{eq:main-dyadic-bound} on each interval gives \[ |\mathcal B_\lambda(N)| \ll_{\lambda,\eps}1+ \sum_{j\geq0}\left(\frac{N}{2^j}\right)^{\rho-c_0\rho^3/20+\eps}. \] The exponent $\gamma=\rho-c_0\rho^3/20$ is positive: indeed $c_0\leq1$ and $\rho\leq1/3$, so $c_0\rho^2/20<1$. Applying the dyadic estimate first with $\eps<\gamma/2$ makes the geometric series convergent with sum $\ll N^{\gamma+\eps}$; enlarging $\eps$ afterwards yields the usual formulation for every $\eps>0$. Substituting $\rho=\lambda/3$ gives the exponent \[ \frac{\lambda}{3}-\frac{c_0}{540}\lambda^3 =\frac{\lambda}{3}-\frac{c_D}{2160}\lambda^3, \] as claimed. \end{proof} \begin{corollary}[High-quality values]\label{cor:quality} For every $\eta>0$ and $\eps>0$, \begin{equation}\label{eq:quality-bound} \#\{n\leq N:q(n)>1+\eta\} \ll_{\eta,\eps} N^{\frac{1}{3(1+\eta)}-\frac{C}{(1+\eta)^3}+\eps}, \end{equation} where $C$ is the constant of Theorem \ref{thm:main}. \end{corollary} \begin{proof} Since $R_n\geq6$ for $n\geq1$, the quality $q(n)$ in \eqref{eq:quality-intro} is defined. The inequality $q(n)>1+\eta$ is equivalent to $R_n<(2n+1)^\lambda$ with $\lambda=(1+\eta)^{-1}$. Theorem \ref{thm:main} applies, and $\lambda^3=(1+\eta)^{-3}$. \end{proof} \begin{remark}\label{rem:baseline-recovery} Ignoring the negative term $C\lambda^3$ recovers \eqref{eq:elementary-bound}. The proof shows that the entire improvement comes from the middle dyadic range in Theorem \ref{thm:diagonal}; the two outer ranges are already covered by the elementary estimates. \end{remark} % ===== END sections/05_diag.tex ===== % ===== BEGIN sections/06_general_abc.tex ===== \section{Variable right-hand sides and the general abc obstruction}\label{sec:general-abc} This section explains why Theorem \ref{thm:main} does not extend to arbitrary abc triples by a change of parameters. The distinction is structural rather than numerical. Nothing in this section is used in the proof of the balanced-family theorem. \subsection{The variable-square identity} Let $a+b=c$ be positive integers with $a\neq b$, and write \[ a=\alpha x^2,\qquad b=\beta y^2 \] using the odd-valuation squarefree kernels from \eqref{eq:unique-square-kernel}. Then \begin{equation}\label{eq:general-square-identity} c^2-\alpha\beta(2xy)^2=(a-b)^2. \end{equation} Thus, with $D=\alpha\beta$, $t=c$, $u=2xy$, and $k=|a-b|$, \begin{equation}\label{eq:general-pell} t^2-Du^2=k^2. \end{equation} The balanced coordinates $(n,n+1,2n+1)$ force $k=1$ in the corresponding identity, and \eqref{eq:general-pell} becomes an ordinary Pell equation. In general, however, $k$ is an additional height variable. Factorization now gives \begin{equation}\label{eq:general-factorization} (t-k)(t+k)=Du^2,\qquad \gcd(t-k,t+k)\mid 2k. \end{equation} When $k$ is unbounded, this does not lead to Diao's bounded list $h\in\{\pm1,\pm2\}$ in the factorization (2.1) of \cite{Diao2026}. Theorem \ref{thm:diao} consequently cannot be quoted for \eqref{eq:general-pell} merely by replacing the right-hand side. \subsection{Generalized Pell fibres can be linear} The next proposition makes the loss of logarithmic fibres explicit. It also shows that an assertion of exact order $X\log^2X$ is unnecessary for the obstruction. \begin{proposition}[Linear generalized fibres]\label{prop:general-fibres} Fix a positive nonsquare integer $D$, and put \[ \mathcal S_D(X)= \#\{(t,u,k)\in\Z_{>0}^3:t,u,k\leq X,\ t^2-Du^2=k^2\}. \] Then, for fixed $D$, \begin{equation}\label{eq:general-fibre-order} X\ll_D \mathcal S_D(X)\leq X^{1+o(1)}. \end{equation} \end{proposition} \begin{proof} For the upper bound, note that $u\leq X/\sqrt D$. For each fixed $u$, the factorization \eqref{eq:general-factorization} determines $(t,k)$ from a same-parity ordered factorization of $Du^2$. The divisor bound gives $X^{o(1)}$ such factorizations, uniformly for $Du^2\leq X^2$. Summing over $u$ proves the upper bound. For the lower bound, consider positive integers $r,s$ in the ellipse \begin{equation}\label{eq:ellipse} Dr^2+s^2\leq X. \end{equation} Define \[ t=Dr^2+s^2,\qquad u=2rs,\qquad k=|Dr^2-s^2|. \] Then $t^2-Du^2=k^2$ and $t\leq X$. Moreover, $2rs\leq X/\sqrt D$ by $Dr^2+s^2\geq2\sqrt D\,rs$. The diagonal $Dr^2=s^2$ is empty because $D$ is nonsquare, and hence $k3.5\times10^9$. For a minimality certificate, the continued fraction is \[ \sqrt{61}=[7;\overline{1,4,3,1,2,2,1,3,4,1,14}], \] with period length $11$, and the convergent of index $21$ (the initial term having index $0$) is $1766319049/226153980$. The standard continued-fraction parametrization of Pell solutions therefore gives the fundamental solution above. Direct multiplication also verifies both \eqref{eq:D61-generalized} and $1766319049^2-61\cdot226153980^2=1$. Thus the small generalized solution in the fixed-$\kappa$ family does not place $D=61$ on the small-$\varepsilon_D$ branch to which Diao's theorem applies. Also, $(125+16\sqrt{61})/3$ is not an algebraic integer, so it is not itself a Pell unit. \end{remark} For the same reason, the fixed-$\kappa$ assertion \[ \#\{a\leq N:\rad(a(a+\kappa)(2a+\kappa))<(2a+\kappa)^\lambda\} \ll_\kappa N^{\lambda/3-C_\kappa\lambda^3+\eps} \] is \emph{not} a corollary of the argument in this paper. It remains an open construction with the methods currently available here. Nor do we assume that all solutions of \eqref{eq:fixed-kappa} form one Pell coset: in general, orbit multiplicities must be proved separately. \subsection{Revised status of G-B86d6-GenPell} The preceding counterexample also forces a correction to the proposed reduction. Adjoining the condition $\varepsilon_D\leq D^\theta$ to \eqref{eq:general-pell} does not capture the general family: Diao already counts the discriminants satisfying that condition, while Remark \ref{rem:kappa-counterexample} shows that the generalized solutions of interest need not satisfy it. A useful G-B86d6-GenPell statement must instead estimate generalized solutions themselves, with the fibre cardinality included. For example, the fibre-weighted object is \begin{equation}\label{eq:genpell-weighted} \mathcal G(W,X;\kappa)= \#\left\{(D,t,u):\begin{array}{l} W\leq D<2W,\ D\text{ nonsquare},\ t,u\leq X,\\ t^2-Du^2=\kappa^2 \end{array}\right\}. \end{equation} A power saving must be proved for \eqref{eq:genpell-weighted}, uniformly in the required $\kappa$-ranges and after comparison with the elementary dyadic covers; a discriminant-only saving multiplied by an $X^{o(1)}$ fibre is incompatible with Proposition \ref{prop:general-fibres}. Moreover, the exponent ledger for the full two-dimensional abc problem must be rebuilt rather than copied from the one-dimensional balanced family. \medskip \noindent\textbf{Status.} G-B86d6-GenPell is an open construction; no ready-made paper supplying the required uniform bound has been located. The balanced-family result G-B86d5-Diag remains proved modulo the cited Diao preprint theorem, but no fixed-$\kappa$ or full-abc extension is asserted. \subsection{The external general-abc benchmark} For orientation, let $N_\lambda(X)$ denote the number of coprime positive triples $a+b=c$ in $[1,X]^3$ with $\rad(abc) \begin{cases} 0,&0<\lambda\leq9/11,\\[2pt] \dfrac{11\lambda-9}{120},&9/11<\lambda\leq21/23,\\[6pt] \dfrac{2\lambda}{3}-\dfrac35,&21/23<\lambda<1. \end{cases} \end{equation} The second and third expressions agree at $\lambda=21/23$, and the last threshold at $\lambda=1$ is $1/15$. These are comparison thresholds for the full two-dimensional problem only; they play no role in Theorem \ref{thm:main}. % ===== END sections/06_general_abc.tex ===== % ===== BEGIN sections/06_crt_conclusion.tex ===== \section{Chinese-remainder orbits and concluding remarks}\label{sec:crt} \subsection{CRT normalization} Let $(a_0,b_0,c_0)$ be a pairwise coprime source triple with $a_0$ even and $b_0,c_0$ odd. Put \begin{equation}\label{eq:CRT-source} M=a_0b_0c_0,\qquad D_0=\rad(M),\qquad S=\Supp(M). \end{equation} The subscripts distinguish the source data from the squarefree kernels $a,b,c$ used earlier. Consider the simultaneous congruences \begin{equation}\label{eq:CRT-congruences} n\equiv0\pmod{a_0},\qquad n\equiv-1\pmod{b_0},\qquad n\equiv-2^{-1}\pmod{c_0}. \end{equation} They are pairwise compatible because $a_0,b_0,c_0$ are pairwise coprime and $c_0$ is odd. Let $00$, and put \begin{equation}\label{eq:X-definition} X=2M(T+1). \end{equation} Then \begin{equation}\label{eq:CRT-bound} \boxed{ \#\{0\leq t\leq T:q_s(t)>1+\eta\} \ll_{\eta,\eps} X^{\frac{1}{3(1+\eta)}-\frac{C}{(1+\eta)^3}+\eps}.} \end{equation} The implied constant is independent of $(a_0,b_0,c_0)$ and $T$. \end{corollary} \begin{proof} Since $03$. The parameter $\theta$ is bounded above because $s>\rho/2$; hence Diao's growing upper restriction on $\theta$ is automatic for large $N$. With $\tau=c_0\rho^3/10$, the middle exponent is at most $\rho-6\tau$. The outer exponents are at most $\rho-\tau$ and $\rho-2\tau$, respectively. This verifies the uniform diagonal saving without leaving an uncovered dyadic interval. Finally, the amplification lemma is applied with $\delta=\tau/2$. Since \[ \tau=\frac{c_0\rho^3}{10},\qquad c_0=\frac{c_D}{4},\qquad \rho=\frac{\lambda}{3}, \] the final negative term is \[ \delta=\frac{c_0\rho^3}{20} =\frac{c_D}{2160}\lambda^3. \] This is the source of the explicit constant in Theorem \ref{thm:main}. % ===== END sections/07_appendix.tex ===== % ===== BEGIN sections/08_refs.tex ===== \begin{thebibliography}{99} \raggedright \bibitem{Diao2026} Yijie Diao, \emph{On the average of class numbers of indefinite binary quadratic forms}, preprint (20 October 2026). \url{https://arxiv.org/abs/2610.28001v1} \bibitem{GranvilleTucker2002} Andrew Granville and Thomas J. Tucker, \emph{It's as easy as $abc$}, Notices Amer. Math. Soc. \textbf{49} (2002), no. 10, 1224--1231. \bibitem{LebowitzLockard2025} Noah Lebowitz-Lockard, \emph{On pairs of consecutive sequences with the same radicals}, preprint (2025). \url{https://arxiv.org/abs/2507.09899} \bibitem{BBLT2026} Christian Bernert, Tim Browning, Jared Duker Lichtman, and Joni Ter\"av\"ainen, \emph{Bounds on the exceptional set in the $abc$ conjecture}, preprint (2026), arXiv:2410.12234v2. \url{https://arxiv.org/abs/2410.12234v2} \bibitem{Tenenbaum2015} G\'erald Tenenbaum, \emph{Introduction to Analytic and Probabilistic Number Theory}, third edition, Graduate Studies in Mathematics \textbf{163}, American Mathematical Society, Providence, RI, 2015. \end{thebibliography} % ===== END sections/08_refs.tex ===== \end{document}