\documentclass[11pt,a4paper]{article} \usepackage[utf8]{inputenc} \usepackage{amsmath,amssymb,amsthm} \usepackage[margin=2.7cm]{geometry} \usepackage{booktabs} \usepackage[hidelinks]{hyperref} \frenchspacing \theoremstyle{plain} \newtheorem{theorem}{Theorem}[section] \newtheorem{proposition}[theorem]{Proposition} \newtheorem{lemma}[theorem]{Lemma} \newtheorem{corollary}[theorem]{Corollary} \theoremstyle{definition} \newtheorem{definition}[theorem]{Definition} \newtheorem{remark}[theorem]{Remark} \newcommand{\eps}{\varepsilon} \newcommand{\Nst}{N_{*}} \title{Sums of three composite triangular numbers} \author{David Niedbala Giraudin\\[2pt] \small ORCID 0009-0009-1526-1178} \date{18 November 2026} \begin{document} \maketitle \begin{abstract} A triangular number $T_k=k(k+1)/2$ is composite for every $k\ge3$, the only non-composite ones being $0$, $1$ and $3$. We ask which integers are sums of three composite triangular numbers, a restricted form of Gauss's Eureka theorem; equivalently, when $8n+3$ is a sum of three squares none of which is $1$, $9$ or $25$. We prove that every integer $n>194$ is a sum of three composite triangular numbers, with at most three possible exceptions, all larger than $8.4\cdot10^{9}$. Such exceptions can occur only if Tatuzawa's form of Siegel's theorem, with $\eps=1/20$, has an exceptional discriminant $-\Nst$, and then only for $8n+3\in\{\Nst,9\Nst,25\Nst\}$. Up to $8.4\cdot10^{9}$ the integers that are not such sums are exactly $41$ explicit values, the largest being $194$. Under the Generalized Riemann Hypothesis these $41$ values are the complete list. The proof combines an elementary lemma, a union bound for the representations that use a forbidden square, the class number formula for sums of three squares, Tatuzawa's theorem, and explicit bounds for a divisor-type function. The analytic range begins at $8n+3=3.4561\cdot10^{10}$ and overlaps the computed range by a factor $1.944$. The computed range is certified by two independent algorithms, supplied as ancillary files together with a verifier for every constant. \end{abstract} \section{Introduction} Gauss's Eureka theorem states that every non-negative integer is a sum of three triangular numbers $T_k=k(k+1)/2$, $k\ge0$; it is equivalent to the representation of every $N\equiv3\pmod 8$ as a sum of three squares. Since $T_k=\frac{k}{2}(k+1)$ or $k\frac{k+1}{2}$ is a product of two integers larger than $1$ as soon as $k\ge3$, the only triangular numbers that are not composite are $T_0=0$, $T_1=1$ and $T_2=3$. It is natural to ask how much of Gauss's theorem survives when these three are removed. \begin{definition} Let $\mathcal E$ be the set of integers $n\ge0$ that cannot be written $n=T_a+T_b+T_c$ with $a,b,c\ge3$. \end{definition} \begin{theorem}\label{thm:main} \begin{enumerate} \item[(i)] The elements of $\mathcal E$ not exceeding $8.4\cdot10^{9}$ are exactly \begin{gather*} 0,1,\dots,17,\quad 19,20,21,23,24,25,28,29,32,34,38,\\ 39,43,47,50,54,60,68,69,73,95,113,194, \end{gather*} forty-one values, the largest being $194$. \item[(ii)] Every element of $\mathcal E$ larger than $8.4\cdot10^{9}$ is of the form $(f^2\Nst-3)/8$ with $f\in\{1,3,5\}$, where $-\Nst$ is the exceptional discriminant allowed by Lemma~\ref{lem:tat} with $\eps=1/20$, if there is one. In particular $\mathcal E$ has at most $44$ elements, and every $n>194$ is a sum of three composite triangular numbers with at most three possible exceptions. \end{enumerate} \end{theorem} The exceptional discriminant may well not exist; if it exists, it need not be $\equiv5\pmod 8$, in which case part (ii) gives no exception at all. What the theorem cannot do is decide this, which is the familiar limitation of every effective form of Siegel's theorem. Under the Generalized Riemann Hypothesis the question disappears. \begin{corollary}\label{cor:grh} Assume the Generalized Riemann Hypothesis, in the form used in~\cite{KLO} and~\cite{LLS}. Then $\mathcal E$ consists exactly of the $41$ values listed in Theorem~\ref{thm:main}; every integer $n>194$ is a sum of three composite triangular numbers. \end{corollary} The closest result in the literature concerns sums of three \emph{nonzero} triangular numbers, that is, the exclusion of $T_0$ alone. Kim, Lee and Oh~\cite{KLO} proved, under the Generalized Riemann Hypothesis, that every positive integer except $1,2,4,6,11,20,29$ is such a sum. Excluding $T_0$, $T_1$ and $T_2$ amounts to excluding the three odd squares $1$, $9$, $25$ instead of the single square $1$. Their result enters our proof of Corollary~\ref{cor:grh}, through Proposition~\ref{prop:B}. We are not aware of an earlier treatment of the present question. The recent historical survey of Aygin and Williams~\cite{AW} is the natural place where one would be recorded; its abstract concerns formulas for the number of representations, and we have not been able to consult the full text. \paragraph{Plan.} Section~\ref{sec:red} states the problem in terms of sums of three squares and treats every $8n+3$ divisible by the square of an odd $q\ge7$. Section~\ref{sec:count} shows that the remaining cases follow from the inequality $H(N)>3D\bigl((N-1)/2\bigr)$ between a class number and a divisor-type function. Section~\ref{sec:D} bounds that function, and Section~\ref{sec:tat} recalls Tatuzawa's lower bound for the class number. Section~\ref{sec:proof} closes the chain of ranges, Section~\ref{sec:grh} proves Corollary~\ref{cor:grh}, and Section~\ref{sec:comp} describes the computation and the ancillary files. Section~\ref{sec:status} states precisely what rests on what. \section{Reformulation and the elementary range}\label{sec:red} \begin{lemma}\label{lem:dico} Put $N=8n+3$. Then $n=T_a+T_b+T_c$ with $a,b,c\ge0$ if and only if $N=x^2+y^2+z^2$ with $x,y,z$ odd, the correspondence being $x=\pm(2a+1)$, and similarly for $y,z$. Every representation of $N$ as a sum of three squares has $x,y,z$ odd. Consequently $n\notin\mathcal E$ if and only if $N=x^2+y^2+z^2$ with $|x|,|y|,|z|\ge7$, that is, with none of $x^2,y^2,z^2$ equal to $1$, $9$ or $25$. \end{lemma} \begin{proof} The identity $8T_k+1=(2k+1)^2$ gives $8(T_a+T_b+T_c)+3=(2a+1)^2+(2b+1)^2+(2c+1)^2$. Squares are $0$, $1$ or $4$ modulo $8$, and three of them sum to $3$ modulo $8$ only as $1+1+1$, so $x,y,z$ are odd. An odd $x$ is $\pm(2a+1)$ with $a\ge0$, and $a\ge3$ exactly when $|x|\ge7$; the odd squares below $49$ are $1$, $9$ and $25$. \end{proof} \begin{proposition}\label{prop:A} If $q^2\mid 8n+3$ for some odd $q\ge7$, then $n\notin\mathcal E$. \end{proposition} \begin{proof} Every odd square is $\equiv1\pmod 8$, so $M=(8n+3)/q^2\equiv3\pmod 8$. By Gauss's theorem $M=u^2+v^2+w^2$, and by Lemma~\ref{lem:dico} $u,v,w$ are odd, hence non-zero. Then $8n+3=(qu)^2+(qv)^2+(qw)^2$ with $|qu|,|qv|,|qw|\ge q\ge7$. \end{proof} \begin{proposition}\label{prop:B} Let $q\in\{3,5\}$ and suppose $8n+3=q^2(8m+3)$. If $m$ is a sum of three nonzero triangular numbers, then $n\notin\mathcal E$. \end{proposition} \begin{proof} If $m=T_a+T_b+T_c$ with $a,b,c\ge1$, then $8m+3=u^2+v^2+w^2$ with $u=2a+1\ge3$, and similarly for $v,w$. Hence $8n+3=(qu)^2+(qv)^2+(qw)^2$ with $qu\ge3q\ge9$. \end{proof} Write $N=8n+3=f^2N_0$ with $N_0$ squarefree and $f\ge1$; $f$ is odd. If $f\notin\{1,3,5\}$, then $f\ge7$ and Proposition~\ref{prop:A} applies with $q=f$. Since $N_0\equiv N\equiv3\pmod 8$, the number $-N_0\equiv5\pmod 8$ is a fundamental discriminant. \emph{From now on $f\in\{1,3,5\}$.} \section{The counting inequality}\label{sec:count} Let $t_3(n)$ be the number of ordered triples $(a,b,c)$ with $a,b,c\ge0$ and $T_a+T_b+T_c=n$, and $\rho_2(m)$ the number of ordered pairs $(b,c)$ with $b,c\ge0$ and $T_b+T_c=m$ (with $\rho_2(m)=0$ for $m<0$). \begin{lemma}\label{lem:r3} Let $N=8n+3=f^2N_0$ as above, with $N_0>3$. Then \[ t_3(n)=\tfrac18\,r_3(N)=3H(N),\qquad H(N)=\sum_{d\mid f}h(-d^2N_0), \] where $r_3$ counts representations as a sum of three squares and $h(-d^2N_0)$ is the class number of primitive positive definite binary quadratic forms of discriminant $-d^2N_0$. In particular \[ H(N)=h(-N_0)\cdot\begin{cases} 1 & \text{if } f=1,\\ 4-\chi(3) & \text{if } f=3,\\ 6-\chi(5) & \text{if } f=5, \end{cases} \] where $\chi$ is the Kronecker character of $-N_0$; hence $H(N)\ge f\,h(-N_0)$. \end{lemma} \begin{proof} By Lemma~\ref{lem:dico} each triple $(a,b,c)$ corresponds to exactly $8$ representations of $N$, one for each choice of signs, so $t_3(n)=r_3(N)/8$. For $N\equiv3\pmod 8$, Gauss's theorem on primitive representations~\cite[art.~291--292]{Gauss}, see also~\cite{Gro}, gives $24\,h(-N)$ primitive representations of $N>3$; summing over the squares $e^2\mid N$, that is over $N/e^2=d^2N_0$ with $d\mid f$, gives $r_3(N)=24H(N)$. The class number of the order of conductor $d$ is $h(-d^2N_0)=h(-N_0)\,d\prod_{p\mid d}\bigl(1-\chi(p)/p\bigr)$, the unit index being $1$ since $N_0>3$~\cite[Section~7]{Cox}; for $d=3$ and $d=5$ this is $h(-N_0)(3-\chi(3))$ and $h(-N_0)(5-\chi(5))$. \end{proof} \begin{lemma}\label{lem:rho} For $m\ge0$, $\rho_2(m)=d_1(4m+1)-d_3(4m+1)$, where $d_i(M)$ is the number of divisors of $M$ congruent to $i$ modulo $4$. Consequently $\rho_2(m)\le\prod_{p\mid 4m+1,\ p\equiv1(4)}(e_p+1)$, where $p^{e_p}\parallel 4m+1$. \end{lemma} \begin{proof} $T_b+T_c=m$ is $(2b+1)^2+(2c+1)^2=8m+2$. Every representation of $8m+2$ as a sum of two squares has both terms odd, so the pairs $(b,c)$ are the representations with both terms positive, that is $r_2(8m+2)/4$. By Jacobi's formula~\cite[Thm.~278]{HW}, $r_2(8m+2)=4\bigl(d_1(8m+2)-d_3(8m+2)\bigr)$, and the odd divisors of $8m+2$ are those of $4m+1$. The last assertion is the usual evaluation of $d_1-d_3$ as a multiplicative function: it vanishes unless every prime $\equiv3\pmod 4$ occurs to an even power, and then equals the product. \end{proof} \begin{proposition}\label{prop:count} If $H(N)>\rho_2(n)+\rho_2(n-1)+\rho_2(n-3)$, then $n\notin\mathcal E$. \end{proposition} \begin{proof} For $i\in\{1,2,3\}$ and $j\in\{0,1,2\}$, let $A(i,j)$ be the set of triples counted by $t_3(n)$ whose $i$-th entry is $j$. Then $|A(i,j)|=\rho_2(n-T_j)$, and $T_0,T_1,T_2=0,1,3$. A triple outside $\bigcup A(i,j)$ has all entries $\ge3$. Such a triple exists as soon as \[ t_3(n)=3H(N)\ >\ \sum_{i,j}|A(i,j)|=3\bigl[\rho_2(n)+\rho_2(n-1)+\rho_2(n-3)\bigr]. \qedhere \] \end{proof} \begin{definition} For $Y\ge1$ let \[ D(Y)=\max_{1\le M\le Y}\ \prod_{p\mid M,\ p\equiv1\,(4)}(e_p+1),\qquad p^{e_p}\parallel M . \] \end{definition} The three integers $4n+1$, $4n-3$, $4n-11$ attached by Lemma~\ref{lem:rho} to $n$, $n-1$, $n-3$ are at most $4n+1=(N-1)/2$. Since $D$ is non-decreasing, Proposition~\ref{prop:count} gives: \begin{corollary}\label{cor:cond} If $H(N)>3\,D\bigl((N-1)/2\bigr)$, then $n\notin\mathcal E$. \end{corollary} \begin{remark} The union bound in Proposition~\ref{prop:count} is crude: a triple with two forbidden entries is counted twice. Inclusion--exclusion would improve the constants of Section~\ref{sec:proof}, but the chain closes without it, and the cruder form is easier to check. \end{remark} \section{The function \texorpdfstring{$D$}{D}}\label{sec:D} Let $5=q_110^{13}$, and are enumerated exhaustively by the ancillary verifier. The records that matter are in Table~\ref{tab:D}; the enumeration is checked against a brute-force scan of all $M\le10^{6}$. \begin{table}[ht] \centering \begin{tabular}{rrl} \toprule $D$ & smallest $M$ & factorisation\\ \midrule 96 & $243\,061\,325$ & $5^2\cdot13\cdot17\cdot29\cdot37\cdot41$\\ 128 & $1\,215\,306\,625$ & $5^3\cdot13\cdot17\cdot29\cdot37\cdot41$\\ 144 & $3\,159\,797\,225$ & $5^2\cdot13^2\cdot17\cdot29\cdot37\cdot41$\\ 160 & $6\,076\,533\,125$ & $5^4\cdot13\cdot17\cdot29\cdot37\cdot41$\\ 192 & $12\,882\,250\,225$ & $5^2\cdot13\cdot17\cdot29\cdot37\cdot41\cdot53$\\ 216 & $53\,716\,552\,825$ & $5^2\cdot13^2\cdot17^2\cdot29\cdot37\cdot41$\\ \bottomrule \end{tabular} \caption{Records of $D$: $D(Y)$ first reaches the value in the first column at $Y=M$.}\label{tab:D} \end{table} In particular \begin{equation}\label{eq:D192} D(Y)\le192\qquad\text{for } Y<53\,716\,552\,825 . \end{equation} \begin{lemma}\label{lem:uniform} For $t>0$ put $g(t)=\prod_{p\equiv1(4),\ p^{t}<2}\ \max_{e\ge0}(e+1)p^{-et}$. Then $D(Y)\le g(t)\,Y^{t}$ for all $Y\ge1$. Moreover $g(0.2)<2.18407$, the maxima being attained at $e=2,1,1,1$ for $p=5,13,17,29$. \end{lemma} \begin{proof} For $M\le Y$, $v(M)=\prod_p (e_p+1)p^{-e_pt}\cdot\prod_p p^{e_pt}\le\prod_p (e_p+1)p^{-e_pt}\cdot M^{t}$, the products running over $p\equiv1\pmod 4$. If $p^t\ge2$ then $(e+1)p^{-et}\le(e+1)2^{-e}\le1$. The numerical value is computed by the verifier. \end{proof} \section{Tatuzawa's theorem}\label{sec:tat} \begin{lemma}[Tatuzawa]\label{lem:tat} Let $0<\eps<1/2$. There is at most one fundamental discriminant $D<0$ with $|D|\ge\max\bigl(e^{1/\eps},e^{11.2}\bigr)$ such that \[ L(1,\chi_D)\ \le\ 0.655\,\eps\,|D|^{-\eps}. \] \end{lemma} This is~\cite[Theorem~2]{Ta}, proved there from Lemma~9; it is restated by Elsenhans, Kl\"uners and Nicolae~\cite[Lemma~2]{EKN} and used by Hopkins and Stopple~\cite{HS}. We take $\eps=1/20$, so that $\max(e^{20},e^{11.2})=e^{20}=4.8517\cdot10^{8}$. If $-N_0\ne-\Nst$ is a fundamental discriminant with $N_0\ge e^{20}$, then, by the class number formula $h(-N_0)=\sqrt{N_0}\,L(1,\chi)/\pi$, \begin{equation}\label{eq:hlow} h(-N_0)\ >\ A\,N_0^{0.45},\qquad A=\frac{0.655}{20\pi}=0.01042464877\ldots \end{equation} Here and below, $-\Nst$ denotes the exceptional discriminant of Lemma~\ref{lem:tat} for $\eps=1/20$, if there is one. \begin{remark} The proof requires the form of Lemma~\ref{lem:tat}, with the constant $0.655\,\eps$, as given in~\cite[Theorem~2]{Ta}. \end{remark} \section{Proof of Theorem~\ref{thm:main}}\label{sec:proof} Let $N=8n+3=f^2N_0$ with $f\in\{1,3,5\}$, $N_0$ squarefree, and suppose $N_0\ne\Nst$. By Lemma~\ref{lem:r3} and~\eqref{eq:hlow}, as long as $N_0\ge e^{20}$, \[ H(N)\ \ge\ f\,h(-N_0)\ >\ f\cdot f^{-0.9}\,A\,N^{0.45}\ \ge\ A\,N^{0.45}, \] since $3^{0.1}=1.116\ldots$ and $5^{0.1}=1.174\ldots$. By Corollary~\ref{cor:cond} it is therefore enough to have \begin{equation}\label{eq:target} A\,N^{0.45}\ \ge\ 3\,D\bigl((N-1)/2\bigr). \end{equation} The condition $N_0\ge e^{20}$ holds throughout the analytic ranges below, which start at $N>3.4\cdot10^{10}$, so that $N_0\ge N/25>1.38\cdot10^{9}>e^{20}$. \paragraph{Range 1: $N>8.9635\cdot10^{10}$.} By Lemma~\ref{lem:uniform} with $t=0.2$, $3D((N-1)/2)\le3g(0.2)(N/2)^{0.2}$, so~\eqref{eq:target} holds as soon as \[ N^{0.25}\ \ge\ \frac{3\,g(0.2)}{A\cdot2^{0.2}},\qquad\text{that is}\qquad N\ \ge\ 8.9635\cdot10^{10}. \] \paragraph{Range 2: $3.4561\cdot10^{10}3.4561\cdot10^{10}$, that is every $n>4.3201\cdot10^{9}$. \paragraph{Range 3: $N\le6.72\cdot10^{10}$.} By direct computation (Section~\ref{sec:comp}), the elements of $\mathcal E$ up to $8.4\cdot10^{9}$, that is with $N\le8\cdot8.4\cdot10^{9}+3=67\,200\,000\,003$, are exactly the $41$ values of part~(i). \paragraph{Closure.} Ranges~2 and~3 overlap, by the factor \[ \frac{67\,200\,000\,003}{3.456051\cdot10^{10}}\ =\ 1.9444 . \] Hence every $n>8.4\cdot10^{9}$ with $N_0\ne\Nst$ lies outside $\mathcal E$. The remaining integers have $N\in\{\Nst,9\Nst,25\Nst\}$, which is part~(ii). Part~(i) is the computation. \qed \begin{remark} The factor $1.9444$ is the whole of the slack between the analytic and the computed ranges: a factor $2$ lost in~\eqref{eq:target} would separate them. Every constant entering the chain is recomputed in the ancillary verifier in $50$-digit decimal arithmetic, with $\pi$ replaced by an upper bound so that~\eqref{eq:hlow} is never overstated, and the closure of the chain is checked by assertions. \end{remark} \section{Proof of Corollary~\ref{cor:grh}}\label{sec:grh} By Theorem~\ref{thm:main} it suffices to show that, under GRH, no $n>8.4\cdot10^{9}$ with $8n+3\in\{\Nst,9\Nst,25\Nst\}$ lies in $\mathcal E$. If $8n+3=q^2\Nst$ with $q\in\{3,5\}$, then $\Nst\equiv3\pmod 8$, say $\Nst=8m+3$, and $m>(e^{20}-3)/8>29$. By the theorem of Kim, Lee and Oh~\cite[Cor.~3.11]{KLO}, valid under GRH, $m$ is a sum of three nonzero triangular numbers, and Proposition~\ref{prop:B} gives $n\notin\mathcal E$. If $8n+3=\Nst$, then $\Nst>6.72\cdot10^{10}$. Under GRH, Lamzouri, Li and Soundararajan~\cite[Thm.~1.5]{LLS} prove, for every primitive character $\chi$ modulo $q\ge10^{10}$, \[ \frac1{|L(1,\chi)|}\ \le\ \frac{12e^{\gamma}}{\pi^2}\Bigl(\log\log q-\log2+\frac12+ \frac1{\log\log q}+\frac{14\log\log q}{\log q}\Bigr). \] At $q=6.72\cdot10^{10}$ this gives $L(1,\chi)\ge0.0898$, while the exceptional inequality of Lemma~\ref{lem:tat} would require $L(1,\chi)\le0.655\cdot\frac1{20}\,q^{-1/20}=0.00942$; the ratio between the two bounds is at least $9.54$ for every $q\ge6.72\cdot10^{10}$, since the first decreases only like $1/\log\log q$ while the second decreases like $q^{-1/20}$. Hence $\Nst$ does not exist in this range. \qed \section{Computation}\label{sec:comp} The ancillary file \texttt{VERIFY\_range.py} recomputes $\mathcal E\cap[0,8.4\cdot10^{9}]$ by two algorithms that share no code except an exact test used on rare survivors. Both work on segments of $2^{26}$ consecutive integers. \emph{Method A.} On each segment, the set of sums $T_a+T_b$ with $3\le a\le b$ is built on a window reaching just below the segment, and $n$ is declared covered as soon as $n-T_k$ lies in it for one of the $64$ composite triangular numbers $T_3,\dots,T_{66}$. \emph{Method B.} The set of sums $T_a+T_b$ with $3\le a\le b$ is built once on $[0,4\cdot10^{6}]$, and $n$ is declared covered as soon as $n-T_c$ lies in it for some $c$. For $n\le4\cdot10^{6}$ this test is exhaustive. Every integer that is not declared covered is passed to an exact test, which tries every $c\ge3$ and decides whether $n-T_c$ is a sum of two composite triangular numbers; the exceptions are therefore certified by exhaustion. As a guard against a marking error, $24$ integers declared covered are drawn at random in each segment and re-tested exactly. Up to $8.4\cdot10^{9}$, method~A passes $368$ integers to the exact test and method~B passes $16\,341$; both return exactly the $41$ values of Theorem~\ref{thm:main}\,(i). The two runs together take about six and a half minutes on two cores. The file \texttt{VERIFY\_composite\_triangular.py} recomputes the exceptional set up to $2\cdot10^{7}$ by a third implementation, the records of $D$ (Table~\ref{tab:D}) by the enumeration of Lemma~\ref{lem:shape} checked against a brute-force scan up to $10^{6}$, the constants $A$, $g(t)$, the thresholds and the overlap of Section~\ref{sec:proof}, and the identities of Lemma~\ref{lem:r3} for $f=1$ ($11\le N<2000$), $f=3$ ($11\le N_0<400$) and $f=5$ ($11\le N_0<200$), with class numbers computed by counting reduced forms. The files \texttt{CHECK\_classnumbers.magma} and \texttt{CHECK\_exceptions.magma} repeat the class number identities and the exceptional set up to $2\cdot10^{5}$ under Magma, with Magma's own class number routine. Both Python files require only Python~3 and \texttt{numpy}, and take no input. As consistency checks, none of the $41$ exceptions has $8n+3$ divisible by $q^2$ with $q\ge7$ odd, as Proposition~\ref{prop:A} requires; and the six exceptions with $9\mid 8n+3$ or $25\mid8n+3$, namely $n=3,9,12,21,34,39$, give $m=0,0,1,2,1,4$ in Proposition~\ref{prop:B}, all of them on the list of Kim, Lee and Oh. \section{Status of the argument}\label{sec:status} \textbf{Proved here:} Lemmas~\ref{lem:dico}, \ref{lem:rho}, \ref{lem:shape}, \ref{lem:uniform}, Propositions~\ref{prop:A}, \ref{prop:B}, \ref{prop:count}, and the deduction of Theorem~\ref{thm:main}\,(ii) and Corollary~\ref{cor:grh} from the inputs below. \textbf{Verified by computation:} the $41$ exceptions up to $8.4\cdot10^{9}$; the records of $D$; the constants and the closure of the chain; the identities of Lemma~\ref{lem:r3} on the ranges stated. \textbf{Quoted:} Gauss's formula for $r_3$ and the class number of orders (Lemma~\ref{lem:r3}); Jacobi's formula for $r_2$; Tatuzawa's theorem (Lemma~\ref{lem:tat}), checked against the original~\cite[Theorem~2]{Ta}; and, for Corollary~\ref{cor:grh} only, \cite[Cor.~3.11]{KLO} and~\cite[Thm.~1.5]{LLS}, both conditional on GRH. \textbf{Open:} whether the exceptional discriminant can be excluded unconditionally, which would make the list of $41$ values complete without any reserve. \paragraph{Acknowledgements.} Technical assistance: Claude (Anthropic). \begin{thebibliography}{99} \bibitem{AW} Z.~S. Aygin and K.~S. Williams, Historical survey of sums of three squares and sums of three triangular numbers, \emph{Expo. Math.} \textbf{44} (2026), no.~3, article 125776. \href{https://doi.org/10.1016/j.exmath.2026.125776}{doi:10.1016/j.exmath.2026.125776}. \bibitem{Cox} D.~A. Cox, \emph{Primes of the Form $x^2+ny^2$: Fermat, Class Field Theory, and Complex Multiplication}, Wiley, New York, 1989. \bibitem{EKN} A.-S. Elsenhans, J.~Kl\"uners and F.~Nicolae, Imaginary quadratic number fields with class groups of small exponent, \emph{Acta Arith.} \textbf{193} (2020), 217--233. \href{https://doi.org/10.4064/aa180220-20-3}{doi:10.4064/aa180220-20-3}. \bibitem{Gauss} C.~F. Gauss, \emph{Disquisitiones Arithmeticae}, Leipzig, 1801. \bibitem{Gro} E.~Grosswald, \emph{Representations of Integers as Sums of Squares}, Springer, New York, 1985. \href{https://doi.org/10.1007/978-1-4613-8566-0}{doi:10.1007/978-1-4613-8566-0}. \bibitem{HW} G.~H. Hardy and E.~M. Wright, \emph{An Introduction to the Theory of Numbers}, Oxford University Press. \bibitem{HS} K.~Hopkins and J.~Stopple, Lower bounds for the principal genus of definite binary quadratic forms, \emph{Integers} \textbf{10} (2010). \href{https://doi.org/10.1515/integ.2010.021}{doi:10.1515/integ.2010.021}. \bibitem{KLO} D.~Kim, J.~Lee and B.-K. Oh, A sum of three nonzero triangular numbers, \emph{Int. J. Number Theory} \textbf{17} (2021), no.~10, 2279--2300. \href{https://doi.org/10.1142/S1793042121500883}{doi:10.1142/S1793042121500883}. \bibitem{LLS} Y.~Lamzouri, X.~Li and K.~Soundararajan, Conditional bounds for the least quadratic non-residue and related problems, \emph{Math. Comp.} \textbf{84} (2015), no.~295; Corrigendum, \emph{Math. Comp.} \textbf{86} (2017), 2551--2554. Corrected version: arXiv:1309.3595v3. \bibitem{Ta} T.~Tatuzawa, On a theorem of Siegel, \emph{Japan. J. Math.} \textbf{21} (1951), 163--178. \href{https://doi.org/10.4099/jjm1924.21.0_163}{doi:10.4099/jjm1924.21.0\_163}. \end{thebibliography} \end{document}