\documentclass[11pt]{amsart} \usepackage[a4paper,margin=2.6cm]{geometry} \usepackage{amsmath,amssymb,amsthm,enumitem} \usepackage[colorlinks=true,linkcolor=blue,citecolor=blue,urlcolor=blue]{hyperref} \newtheorem{theorem}{Theorem}[section] \newtheorem{lemma}[theorem]{Lemma} \newtheorem{proposition}[theorem]{Proposition} \newtheorem{corollary}[theorem]{Corollary} \theoremstyle{definition} \newtheorem{remark}[theorem]{Remark} \newtheorem{hypothesis}[theorem]{Standing data} \newcommand{\st}{\operatorname{st}} \newcommand{\en}{\operatorname{en}} \begin{document} \title[The degree pair $(84,126)$ in the plane Jacobian problem]{The degree pair $(84,126)$ in the plane Jacobian problem:\\ exclusion of the shape with final corner $(19/7,5)$} \author{David Niedbala Giraudin} \address{Independent researcher, France} \thanks{ORCID: 0009-0009-1526-1178} \date{20 November 2026} \begin{abstract} In the list of Guccione, Guccione, Horruitiner and Valqui of the possible shapes of a counterexample to the plane Jacobian conjecture with degrees at most $150$, two shapes have maximal degree $126$. One of them consists of the standard $(2,3)$-pairs of degrees $(84,126)$ with starting corner $(7,35)$ and final corner $(19/7,5)$, normalised as in Corollary 5.21 of Guccione--Guccione--Valqui. We show that no such pair exists. The two sides of the Newton polygon are described by the results of Guccione--Guccione--Valqui; in suitable polynomial coordinates they confine the pair to an explicit finite system of polynomial equations, which is solved order by order and has no solution. All statements are proved relative to the cited results; the final step is an exact computation, reproduced by an ancillary script (SymPy, and Singular for one ideal membership). The other shape of maximal degree $126$ is the subject of a companion note. The argument has not yet been examined by specialists. \end{abstract} \maketitle \section{Statement} Throughout, $K$ is an algebraically closed field of characteristic zero, $L=K[x,y]$, $[P,Q]=P_xQ_y-P_yQ_x$ and $(m,n)=(2,3)$. We use the notation of \cite{GGV1}: for a weight $(\rho,\sigma)$, $v_{\rho,\sigma}$ is the weighted degree, $\ell_{\rho,\sigma}$ the leading form, $\st_{\rho,\sigma}$ and $\en_{\rho,\sigma}$ its end points; $L^{(l)}=K[x^{\pm1/l},y]$. \emph{Standard $(m,n)$-pairs}, their \emph{starting triple} and their \emph{regular corners} are defined in \cite[Definitions 4.3, 6.2, 5.5]{GGV1}. Put $t:=xy^7$. \begin{hypothesis}\label{data} $(P,Q)$ is a standard $(2,3)$-pair in $L$ with $[P,Q]=J\in K^\times$, $\deg P=84$, $\deg Q=126$, starting triple $\bigl((7,35),(2,0),(7,-1)\bigr)$, and $(-1,1)<\operatorname{Succ}_P(1,0)<(-1,0)$ as in \cite[Corollary 5.21(4)]{GGV1}. Thus, after rescaling $P$ and $Q$: \begin{enumerate}[label=(D\arabic*),leftmargin=3em] \item the support of $P$ lies in $\{(i,j):7i-j\le 28\}$, that of $Q$ in $\{7i-j\le 42\}$, and \[ \ell_{7,-1}(P)=\lambda_P\bigl(x^2h(t)\bigr)^2,\qquad \ell_{7,-1}(Q)=\lambda_Q\bigl(x^2h(t)\bigr)^3, \qquad \lambda_P,\lambda_Q\in K^\times, \] with $h\in K[t]$ monic of degree $5$, $h(0)\ne0$ \cite[Proposition 6.1, Theorem 7.6(2)]{GGV1}. \end{enumerate} Moreover $h$ has a single root: $h=(t-c)^5$ with $c\in K^\times$. \end{hypothesis} \begin{theorem}\label{main} There is no pair $(P,Q)$ as in Standing data~\ref{data}. \end{theorem} \begin{remark}\label{roots} A root of multiplicity $k$ of $h$ which satisfies inequality (5.9) of \cite[Proposition 5.18]{GGV1}, that is $2k\ge 14/3$, generates the regular corner $(2,0)+k(1/7,1)$ of a $(2,3)$-pair in $L^{(7)}$, and such a root exists \cite[Propositions 5.16, 5.18]{GGV1}; so $k\ge3$. The last condition of the standing data is the case $k=5$, with corner $(19/7,5)$; it is the only shape with these degrees in the list of \cite[Section 6]{GGV4}. If $h$ has no root of multiplicity at least $4$, then $k=3$, which is impossible by the arithmetic of Lemma~\ref{corner} below: the corner $(17/7,3)$ admits neither Case I nor Case II. The remaining case $h=(t-c)^4(t-c')$, with corner $(18/7,4)$, is not excluded by that arithmetic alone; for it we rely on \cite{GGV4}. The first arXiv version of \cite{GGHV3} (August 2017) listed the shape of the standing data among those discarded, assuming that two expressions $I_M$ and $I_m$ for the intersection number of a Jacobian pair, due to Y.~Xu, coincide. The published version \cite{GGHV3} establishes the formula $I_M$ but only an inequality for $I_m$, and no longer contains that list. To our knowledge the shape has not been excluded elsewhere. \end{remark} \section{The left side} \begin{lemma}\label{left} There is $g=g_2v^2+g_3v^3+g_4v^4\in K[v]$ such that \[ \hat P(z,v):=P\bigl(zv^5+g(v),\,v^{-1}\bigr),\qquad \hat Q(z,v):=Q\bigl(zv^5+g(v),\,v^{-1}\bigr) \] are polynomials in $z,v$. Give $z$ the weight $2$ and $v$ the weight $1$. Then $\hat P$ has weighted degree $28$ with top form $\lambda_P z^4(z-cv^2)^{10}$, $\hat Q$ has weighted degree $42$ with top form $\lambda_Q z^6(z-cv^2)^{15}$, and \[ \hat P_z\hat Q_v-\hat P_v\hat Q_z=-J\,v^3 . \] \end{lemma} \begin{proof} Apply the flip $\psi_1$ of \cite[Section 8]{GGV1}, so that the weight $(7,-1)$ becomes $(\rho_0,\sigma_0)=(-1,7)$, with $\st_{\rho_0,\sigma_0}=(70,14)$ and $\en_{\rho_0,\sigma_0}=(0,4)$ for the flipped $P$, and $\frac1m\st=\frac1n\st=(35,7)$. Let $F$ be the element of \cite[Theorem 2.6]{GGV1} for $(\rho_0,\sigma_0)$; then $v_{-1,7}(F)=6$. As in the proof of \cite[Proposition 8.1]{GGV1}: if $\st(F)=(1,1)$ then $\en(F)\ne(1,1)$, since $F$ is not a monomial \cite[Remark 2.5]{GGV1}, so $\en(F)\sim\en(P)\sim(0,1)$ and $\en(F)=(0,6/7)$, which is impossible; hence $\st(F)\sim(35,7)$, so $\st(F)=(15,3)=\frac37(35,7)$ and $q=7$ in \cite[Corollary 7.4]{GGV1}. That corollary applies to the flipped pair and to every pair obtained from it by a translation of the second variable by a Laurent monomial in the first (the form $\ell_{\rho_0,\sigma_0}$, the degrees $v_{1,1}$, $v_{0,1}$ and $F$ are unchanged). It yields: for every weight between $(1,-5)$ and $(\rho_0,\sigma_0)$ at which the pair has an edge, the edge form of $P$ is a power $R^{14}$. In the original variables: for a weight $(-s,1)$ with $s<5$ carrying an edge with end point $(14,70)$, the form is $(xy^5-\gamma y^{5-s})^{14}$ up to a constant, so $s\in\mathbb Z$, and $s\ge2$ by the condition on $\operatorname{Succ}_P(1,0)$. Here $(14,70)=\en_{7,-1}(P)$ is the starting point of the edge at $\operatorname{Succ}_P(1,0)$, because $\deg P=84=v_{1,1}(14,70)$ (compare \cite[Section 8]{GGV1}); and the statement applies at every stage of the process below, so that no edge with a non-integer $s<5$ appears after a translation. The translation $x\mapsto x+\gamma y^{-s}$ removes this edge and leaves the forms of smaller $s$ unchanged \cite[Proposition 3.9]{GGV1}. After at most three such translations ($s=2,3,4$) no edge of weight $(-s,1)$, $s<5$, has end point $(14,70)$. Since the weight $(-5,1)$ lies within a half-turn of $(7,-1)$, the point $(14,70)$ is then still leading for $(-5,1)$ \cite[Remark 2.10, Proposition 3.7]{GGV1}; hence every monomial $x^iy^j$ of the translated $P$ satisfies $j\le 5i$, that is, $x^iy^j=z^iv^{5i-j}$ is a monomial in $z=xy^5$, $v=y^{-1}$. The same holds for $Q$. For every weight $(-s,1)$ with $s<5$ the values of $P$ and $Q$ are positive, since $(14,70)$ and $(21,105)$ remain in the supports of the translated polynomials and $v_{-s,1}(14,70)=14(5-s)$, $v_{-s,1}(21,105)=21(5-s)$. At a weight where $P$ has an edge, \cite[Proposition 7.3]{GGV1} with $T_0=Q$ (its hypothesis on the brackets holds, because $[Q,P]$ is a constant) gives $[\ell(Q),\ell(P)]=0$ and proportional values, so $\ell(Q)\sim R^{21}$ and the same translation applies. At the remaining weights the leading forms commute by \cite[Proposition 1.13]{GGV1}, so $\ell(Q)$ is proportional to a power of the monomial $\ell(P)$ \cite[Proposition 2.1(2)]{GGV1}, that is, $\ell(Q)$ is the monomial at $(21,105)$. Hence $j\le5i$ on the support of $Q$. A translation $x\mapsto x+\gamma y^{-s}$ with $s<7$ replaces $x^iy^j$ by monomials $x^{i-r}y^{j-sr}$ of $(7,-1)$-weight $7i-j-(7-s)r$, so (D1) holds for the translated pair with the same leading forms. Since $x^iy^j=z^iv^{5i-j}$ has $7i-j=2i+(5i-j)$, the bound $7i-j\le28$ is the weighted degree bound, and for $x=zv^5$, $y=v^{-1}$ one has $x^4(xy^7-c)^{10}=z^4(z-cv^2)^{10}$, which gives the top form; likewise for $Q$. The Jacobian follows from $dx\wedge dy=-v^3\,dz\wedge dv$. \end{proof} \section{The right side} Write $X:=x^{1/7}$, so that $L^{(7)}=K[X^{\pm1},y]$. \begin{lemma}\label{corner} There are $\lambda\in K^\times$ with $\lambda^7=c$ and $\lambda_2,\lambda_3\in K$ such that, with \[ Y:=\eta+\lambda X^{-1}+\lambda_2X^{-2}+\lambda_3X^{-3},\qquad P_1(X,\eta):=P\bigl(X^7,Y\bigr),\qquad Q_1(X,\eta):=Q\bigl(X^7,Y\bigr), \] and with the weight $W(X^i\eta^j):=3i-11j$, every monomial of $P_1$ has $W\le4$ and every monomial of $Q_1$ has $W\le6$. The monomials of $P_1$ with $W=4$ lie between $X^5\eta$ and $X^{38}\eta^{10}$, both with non-zero coefficient; those of $Q_1$ with $W=6$ lie between $X^2$ and $X^{57}\eta^{15}$, both with non-zero coefficient. \end{lemma} \begin{proof} In $\tilde z:=x^{1/7}y$ one has $\ell_{7,-1}(P)=\lambda_Px^4(\tilde z^7-c)^{10}$, so every root $\lambda$ of $\tilde z^7=c$ has multiplicity $10\ge\frac{m}{l}\frac{a\rho+bl\sigma}{\rho+\sigma}=\frac{14}{3}$ (with $l=1$, $(a,b)=(7,35)$, $(\rho,\sigma)=(7,-1)$). By \cite[Proposition 5.18]{GGV1} the automorphism $\varphi(y)=y+\lambda x^{-1/7}$ gives a $(2,3)$-pair $(\varphi P,\varphi Q)$ in $L^{(7)}$ with a regular corner at $A_1:=(2,0)+5(1/7,1)=(19/7,5)$, of direction $(\rho',\sigma')<(7,-1)$. We follow this corner as in the proof of \cite[Proposition 5.19]{GGV1}. Case II is impossible, because \cite[Proposition 5.14(4)]{GGV1} requires $\gcd(19,5)>1$. In Case III, \cite[Proposition 5.17]{GGV1} gives $\rho'\mid7$, hence $\sigma'/\rho'=-j/7$ with $2\le j\le6$, and the translation $y\mapsto y+\lambda'x^{-j/7}$ produces a $(2,3)$-pair in $L^{(7)}$ with a regular corner at the same point $A_1$ and a smaller direction. After finitely many steps Case I occurs, for a pair $(P_1,Q_1)$ and a direction $(\rho,\sigma)$. There $[\ell(P_1),\ell(Q_1)]\ne0$, so $v_{\rho,\sigma}(P_1)+v_{\rho,\sigma}(Q_1)=\rho+\sigma$ \cite[Proposition 1.13]{GGV1}, while $\en(P_1)=2A_1=\frac23\en(Q_1)$ \cite[Corollary 5.7(2)]{GGV1}. Hence $2(\tfrac{19}{7}\rho+5\sigma)=\tfrac25(\rho+\sigma)$, that is $(\rho,\sigma)=(21,-11)$, $v(P_1)=4$, $v(Q_1)=6$. By \cite[Proposition 5.13]{GGV1} either $\frac12\st(P_1)=(a'/7,0)$ with $a'\in\mathbb Z$, which would give $6a'=4$; or $\{\st(P_1),\st(Q_1)\}=\{(k/7,0),(1-k/7,1)\}$ with $k\in\mathbb N$. If $\st(P_1)=(k/7,0)$ then $3k=4$; so $\st(Q_1)=(k/7,0)$ with $3k=6$, that is $\st(Q_1)=(2/7,0)$ and $\st(P_1)=(5/7,1)$. The directions of the Case III steps lie strictly between $(21,-11)$ and $(7,-1)$, so $j\in\{2,3\}$, each at most once. Finally $v_{21,-11}(x^{i/7}y^j)=3i-11j$, and $\en(Q_1)=(57/7,15)$. \end{proof} \section{Common coordinates} Let $\mathcal A$ be the set of formal sums $\sum c_{ij}X^i\eta^j$ ($i\in\mathbb Z$, $j\ge0$) such that, for every $r$, only finitely many terms have $W=3i-11j\ge r$. It is a ring containing $K[X^{\pm1},\eta]$; for $f\ne0$ let $W(f)$ be the largest weight of its terms and $\ell(f)$ the sum of the terms of that weight. Then $\ell(fg)=\ell(f)\ell(g)$ and $W(fg)=W(f)+W(g)$, and $1+\varepsilon$ is invertible in $\mathcal A$ when $W(\varepsilon)<0$. \begin{lemma}\label{coord} Let $g$, $\lambda,\lambda_2,\lambda_3$ be as in Lemmas~\ref{left} and~\ref{corner}, and put \[ \zeta:=z-c\,v^2-\kappa_1v-\kappa_2,\qquad \kappa_1:=7\lambda^5\lambda_2,\quad \kappa_2:=7\lambda^3(\lambda\lambda_3+2\lambda_2^2). \] Write $\hat P=\sum p_{ab}\zeta^av^b$ and $\hat Q=\sum q_{ab}\zeta^av^b$ in $K[z,v]=K[\zeta,v]$. Then $p_{ab}\ne0$ implies $3b-2a\le4$, and $q_{ab}\ne0$ implies $3b-2a\le6$. Moreover $p_{1,2}\ne0$ and $q_{0,2}\ne0$. \end{lemma} \begin{proof} In $\mathcal A$ the element $Y$ of Lemma~\ref{corner} is $\lambda X^{-1}(1+\varepsilon)$ with $W(\varepsilon)<0$; it is invertible. The assignment $v\mapsto Y^{-1}$, $z\mapsto X^7Y^5-g_2Y^3-g_3Y^2-g_4Y$ defines a ring morphism $K[z,v]\to\mathcal A$ which sends $\hat P$ to $P_1$ and $\hat Q$ to $Q_1$, because $P(x,y)=\hat P\bigl((x-g(y^{-1}))y^5,y^{-1}\bigr)$ in $K[x,y^{\pm1}]$. Expanding, the image of $v$ has $W=3$ and leading term $X/\lambda$, and the image of $\zeta$ has $W=-2$ and leading term $7\lambda^4\eta X^3$: the terms of weights $6$, $3$, $0$ cancel by the choice of $c=\lambda^7$, $\kappa_1$, $\kappa_2$, and no monomial has weight $-1$. Hence the image of $\zeta^av^b$ has leading term a non-zero multiple of $\eta^aX^{3a+b}$, of weight $3b-2a$, and these monomials are distinct for distinct $(a,b)$. Therefore, for a non-zero polynomial, $W\bigl(\sum c_{ab}\zeta^av^b\bigr)=\max\{3b-2a:c_{ab}\ne0\}$, and its leading part is the image of the corresponding terms. Lemma~\ref{corner} gives $W(\hat P)=4$, $W(\hat Q)=6$; the monomials $X^5\eta$ and $X^2$ correspond to $\zeta v^2$ and $v^2$. \end{proof} \section{The finite system} By Lemmas~\ref{left} and~\ref{coord}, and since the substitution $z=\zeta+cv^2+\kappa_1v+\kappa_2$ preserves the weighted degree ($\zeta$ of weight $2$) and turns the top form of $\hat P$ into $\lambda_P\zeta^{10}(\zeta+cv^2)^4$, a pair as in the standing data yields a solution of the following system. \begin{enumerate}[label=(S\arabic*),leftmargin=3em] \item $\hat P=\sum p_{ab}\zeta^av^b$ with $a,b\ge0$, $2a+b\le28$, $3b-2a\le4$; $\hat Q=\sum q_{ab}\zeta^av^b$ with $a,b\ge0$, $2a+b\le42$, $3b-2a\le6$; and $p_{1,2}\,q_{0,2}\ne0$. \item $\hat P_\zeta\hat Q_v-\hat P_v\hat Q_\zeta=-J\,v^3$ with $J\in K^\times$. \item The part of $\hat P$ on the line $2a+b=28$ is $\lambda_P\,\zeta^{10}(\zeta+c\,v^2)^4$, with $\lambda_P,c\in K^\times$. \end{enumerate} Call the terms with $3b-2a=4-k$ the \emph{layer $k$} of $\hat P$ and those with $3b-2a=6-k$ the layer $k$ of $\hat Q$, and write $\hat P=\sum_{k\ge0}P_k$, $\hat Q=\sum_{k\ge0}Q_k$. The bracket of a term of layer $i$ of $\hat P$ with a term of layer $j$ of $\hat Q$ is a monomial $\zeta^av^b$ with $3b-2a=9-(i+j)$, and $v^3$ has $3b-2a=9$. So (S2) is equivalent to the \emph{equations of order $k$}: \begin{equation}\label{order} \sum_{i+j=k}\{P_i,Q_j\}=\begin{cases}-J\,v^3,&k=0,\\ 0,&k\ge1,\end{cases} \qquad \{f,g\}:=f_\zeta g_v-f_vg_\zeta . \end{equation} \begin{lemma}\label{normal} If \textup{(S1)--(S3)} has a solution, it has one with $c=1$, $\lambda_P=\frac1{12}$, $J=-2$, \begin{align*} P_0&=\zeta v^2\bigl(1+\tfrac54s+\tfrac12s^2+\tfrac1{12}s^3\bigr),\qquad s:=\zeta^3v^2,\\ Q_0&=v^2\bigl(1+5s+\tfrac{16}{3}s^2+\tfrac83s^3+\tfrac23s^4+\tfrac2{27}s^5\bigr), \end{align*} and $p_{0,0}=q_{0,0}=q_{10,8}=0$. \end{lemma} \begin{proof} The system is preserved by $(\hat P,\hat Q)\mapsto(\gamma\hat P(\alpha\zeta,\beta v), \delta\hat Q(\alpha\zeta,\beta v))$, which multiplies $p_{1,2}$ by $\gamma\alpha\beta^2$, $q_{0,2}$ by $\delta\beta^2$, $c$ by $\beta^2/\alpha$ and $\lambda_P$ by $\gamma\alpha^{14}$. So we may assume $p_{1,2}=q_{0,2}=c=1$ and $\lambda_P=\frac1{12}$; the remaining freedom is $\alpha^{12}=1$, $\beta^2=\alpha$, $\gamma=\alpha^{-2}$, which multiplies $s$ by $\alpha^4$, a cube root of unity. By (S1) the layers of order $0$ are $P_0=\zeta v^2(1+a_4s+a_7s^2+a_{10}s^3)$ and $Q_0=v^2(1+h_1s+\dots+h_5s^5)$, and $a_{10}=\lambda_Pc^4=\frac1{12}$ by (S3). The equation of order $0$ then has exactly three solutions, with $J=-2$, permuted by $s\mapsto\omega s$, $\omega^3=1$; one of them is the pair displayed (ancillary script). Finally replace $\hat Q$ by $\hat Q-12q_{10,8}\hat P-\mathrm{const}$ and $\hat P$ by $\hat P-\mathrm{const}$: this preserves (S1)--(S3) and the layers of order $0$, since $\hat P$ has no term in layers $0$ and $1$ of $\hat Q$, and $p_{10,8}=\frac1{12}$. \end{proof} \begin{proposition}\label{nosol} The system \textup{(S1)--(S3)}, normalised as in Lemma~\ref{normal}, has no solution. \end{proposition} \begin{proof}[Proof (exact computation)] For $k\ge1$ the equation of order $k$ in \eqref{order} is $\{P_k,Q_0\}+\{P_0,Q_k\}=-\sum_{0