betaFeedback

WorksMathematics

Two shapes of maximal degree 135 in the plane Jacobian problem: exclusion of the degree pairs (135,90) and (90,135) with corners (9,36) and (17/9,4)

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 starting corner (9,36) and final corner (17/9,4): the standard (3,2)-pairs of degrees (135,90) and the standard (2,3)-pairs of degrees (90,135). We show that no such pair exists, when it is normalised as in Corollary 5.21 of Guccione-Guccione-Valqui. The two sides of the Newton polygon are described by the results of Guccione-Guccione-Valqui. In suitable polynomial coordinates the pair is confined to an explicit finite system of polynomial equations. Its leading part is a pair of polynomials of degrees 12 and 8 in one variable with 3 p3 q2' - 2 p3' q2 constant, which is unique up to the symmetries of the system and has rational coefficients; the remaining equations are then solved by three steps of linear algebra over Q, and the 29 compatibility conditions, polynomials in five unknowns, generate the unit ideal. The computation is exact and is reproduced by an ancillary script (SymPy, and Singular for one Groebner basis). All statements are proved relative to the cited results. Nothing in this note has been checked by a third party.

Statement

Throughout, KK is an algebraically closed field of characteristic zero, L=K[x,y]L=K[x,y], [P,Q]=PxQy−PyQx[P,Q]=P_xQ_y-P_yQ_x and (m,n)(m,n) is (3,2)(3,2) or (2,3)(2,3). We use the notation of [1]: for a weight (ρ,σ)(\rho,\sigma), vρ,σv_{\rho,\sigma} is the weighted degree, ℓρ,σ\ell_{\rho,\sigma} the leading form, st⁡ρ,σ\operatorname{st}_{\rho,\sigma} and en⁡ρ,σ\operatorname{en}_{\rho,\sigma} its end points; L(l)=K[x±1/l,y]L^{(l)}=K[x^{\pm1/l},y]. Standard (m,n)(m,n)-pairs, their starting triple and their regular corners are defined in [1 Definitions 4.3, 6.2, 5.5]. Put t:=x2y9t:=x^2y^9.

Standing data 1.1. (P,Q)(P,Q) is a standard (m,n)(m,n)-pair in LL with [P,Q]=J∈K×[P,Q]=J\in K^\times, deg⁡P=45m\deg P=45m, deg⁡Q=45n\deg Q=45n, starting triple ((9,36),(1,0),(9,−2))((9,36),(1,0),(9,-2)), and (−1,1)<Succ⁡P(1,0)<(−1,0)(-1,1)<\operatorname{Succ}_P(1,0)<(-1,0) as in [1 Corollary 5.21(4)]. Thus:

(D1) the support of PP lies in {(i,j):9i−2j≤9m}\{(i,j):9i-2j\leq9m\}, that of QQ in {9i−2j≤9n}\{9i-2j\leq9n\}, and

ℓ9,−2(P)=λP(xh(t))m,ℓ9,−2(Q)=λQ(xh(t))n,λP,λQ∈K×,\ell_{9,-2}(P)=\lambda_P(xh(t))^m,\qquad\ell_{9,-2}(Q)=\lambda_Q(xh(t))^n,\qquad\lambda_P,\lambda_Q\in K^\times,

with h∈K[t]h\in K[t] monic of degree 4, h(0)≠0h(0)\ne0 [1 Proposition 6.1, Theorem 7.6(2), Remark 3.1]. Moreover hh has a single root: h=(t−c)4h=(t-c)^4 with c∈K×c\in K^\times.

Theorem 1.2. There is no pair (P,Q)(P,Q) as in Standing data 1.1, neither for (m,n)=(3,2)(m,n)=(3,2) nor for (m,n)=(2,3)(m,n)=(2,3).

Remark 1.3. A root of multiplicity kk of hh which satisfies inequality (5.9) of [1 Proposition 5.18], that is mk≥9m7mk\geq\frac{9m}{7}, generates the regular corner (1,0)+k(2/9,1)(1,0)+k(2/9,1) of an (m,n)(m,n)-pair in L(9)L^{(9)}, and such a root exists [1 Propositions 5.16, 5.18]; so k≥2k\geq2. The last condition of the standing data is the case k=4k=4, with corner (17/9,4)(17/9,4): these are the two shapes with corners (9,36)(9,36), (17/9,4)(17/9,4) of [2 Section 6] (the line through these two points has direction (9,−2)(9,-2) and meets the horizontal axis at (1,0)(1,0), and (17/9,4)=(1,0)+4(2/9,1)(17/9,4)=(1,0)+4(2/9,1)). If hh has no root of multiplicity at least 3, then k=2k=2, which is impossible: the point (13/9,2)(13/9,2) admits neither Case I (the direction would be (81,−56)(81,-56), with no admissible starting point in [1 Proposition 5.13] nor Case II (gcd⁡(13,2)=1)(\gcd(13,2)=1) of [1 Proposition 5.19], and Case III keeps the same point and leads to Case I or Case II. The remaining case h=(t−c)3(t−c′)h=(t-c)^3(t-c'), with corner (15/9,3)(15/9,3), is not excluded by this arithmetic; for it we rely on [2], where no shape with these two corners is listed. We found no exclusion of the two shapes of the standing data in the literature we searched; their novelty is not otherwise verified. The four authors treat in [3] the shapes of degree below 125 and obtain the bound 108.

The left side

Lemma 2.1. There are γ2,γ3∈K\gamma_2,\gamma_3\in K such that

P^(z,v):=P(zv4+γ2v2+γ3v3,v−1),Q^(z,v):=Q(zv4+γ2v2+γ3v3,v−1)\widehat{P}(z,v):=P\left(zv^4+\gamma_2v^2+\gamma_3v^3,v^{-1}\right),\qquad\widehat{Q}(z,v):=Q\left(zv^4+\gamma_2v^2+\gamma_3v^3,v^{-1}\right)

are polynomials in z,vz,v. Give zz the weight 1 and vv the weight 2. Then P^\widehat{P} has weighted degree 9m9m with top form λP(z(z2−cv)4)m\lambda_P\left(z(z^2-cv)^4\right)^m, Q^\widehat{Q} has weighted degree 9n9n with top form λQ(z(z2−cv)4)n\lambda_Q\left(z(z^2-cv)^4\right)^n, and

P^zQ^v−P^vQ^z=−Jv2.\widehat{P}_z\widehat{Q}_v-\widehat{P}_v\widehat{Q}_z=-Jv^2.

Proof. The argument is the one of [3 Section 4, Case (7, 21)]; we give it in detail. Apply the flip ψ1\psi_1 of [1 Section 8], so that the weight (9,−2)(9,-2) becomes (ρ0,σ0)=(−2,9)(\rho_0,\sigma_0)=(-2,9), with st⁡ρ0,σ0=m(36,9)\operatorname{st}_{\rho_0,\sigma_0}=m(36,9) and en⁡ρ0,σ0=m(0,1)\operatorname{en}_{\rho_0,\sigma_0}=m(0,1) for the flipped PP, and 1mst⁡(P)=1nst⁡(Q)=(36,9)\frac{1}{m}\operatorname{st}(P)=\frac{1}{n}\operatorname{st}(Q)=(36,9). Let FF be the element of [1 Theorem 2.6] for (ρ0,σ0)(\rho_0,\sigma_0); then ν−2,9(F)=7\nu_{-2,9}(F)=7. If st⁡(F)=(1,1)\operatorname{st}(F)=(1,1) then en⁡(F)≠(1,1)\operatorname{en}(F)\ne(1,1), since FF is not a monomial [1 Remark 2.5], so en⁡(F)∼en⁡(P)∼(0,1)\operatorname{en}(F)\sim\operatorname{en}(P)\sim(0,1) and en⁡(F)=(0,7/9)\operatorname{en}(F)=(0,7/9), which is impossible; hence st⁡(F)∼(36,9)\operatorname{st}(F)\sim(36,9), so st⁡(F)=(28,7)=79(36,9)\operatorname{st}(F)=(28,7)=\frac{7}{9}(36,9) and q=9q=9 in [1 Corollary 7.4]. That corollary applies to the flipped pair (its hypotheses hold with (a/l,b)=(36,9)(a/l,b)=(36,9), and ρ0+σ0>0\rho_0+\sigma_0>0), and to every pair obtained from it by a translation of the second variable by a Laurent monomial in the first which leaves ℓρ0,σ0\ell_{\rho_0,\sigma_0} unchanged: such a pair lies in L(1)L^{(1)}, the degrees v1,1,v0,1v_{1,1},v_{0,1} of the flipped pair are unchanged, and so is FF (the element G0G_0 of [1 Theorem 2.6] can be kept, and (0,7/9)(0,7/9) is excluded in L(1)L^{(1)} as well). It yields: for every weight (ρ,σ)(\rho,\sigma) between (1,−4)(1,-4) and (ρ0,σ0)(\rho_0,\sigma_0) at which the pair has an edge, and such that the value of PP is positive at all weights from (ρ,σ)(\rho,\sigma) to (ρ0,σ0)(\rho_0,\sigma_0), the edge form of PP is a power R9mR^{9m}.

In the original variables, consider the weights (−s,1)(-s,1) with s<4s<4. The point (9m,36m)=en⁡9,−2(P)(9m,36m)=\operatorname{en}_{9,-2}(P) is the starting point of the edge of PP at Succ⁡P(1,0)\operatorname{Succ}_P(1,0): an edge with a direction strictly between (9,−2)(9,-2) and (1,1)(1,1) starting at (9m,36m)(9m,36m) would end at a point with v1,1>45m=deg⁡Pv_{1,1}>45m=\deg P [1 Remark 1.8], and by the condition on Succ⁡P(1,0)\operatorname{Succ}_P(1,0) there is no direction of PP in ](1,0),(−1,1)]](1,0),(-1,1)]; then [1 Proposition 3.7] applies. As long as (9m,36m)(9m,36m) is the starting point of an edge of weight (−s,1)(-s,1) with s<4s<4, the value of PP there is 9m(4−s)>09m(4-s)>0, so the point (9m,36m)(9m,36m) is leading at all weights from (9,−2)(9,-2) to (−s,1)(-s,1), where the value of PP is therefore positive as well, and the edge form is a power R9mR^{9m} with RR a (−s,1)(-s,1)-homogeneous element of K[x,y±1]K[x,y^{\pm1}] whose only possible monomials are xy4xy^4 and y4−sy^{4-s}; so the form is (xy4−γy4−s)9m(xy^4-\gamma y^{4-s})^{9m} up to a constant, s∈Zs\in\mathbb{Z}, and s≥2s\ge2 by the condition on Succ⁡P(1,0)\operatorname{Succ}_P(1,0), that is s∈{2,3}s\in\{2,3\}. The translation x↦x+γy−sx\mapsto x+\gamma y^{-s} removes this edge and leaves unchanged the leading forms at the weights between (9,−2)(9,-2) and (−s,1)(-s,1), while the (−s,1)(-s,1)-form becomes the monomial x9my36mx^{9m}y^{36m} ([1 Proposition 3.9], applied after the flip); so the next edge starting at (9m,36m)(9m,36m), if any, has a larger ss. After at most two such translations (s=2s=2, then s=3s=3; we write γ2,γ3\gamma_2,\gamma_3 for the constants, which may be 0) no edge of weight (−s,1)(-s,1), s<4s<4, starts at (9m,36m)(9m,36m). Since the weight (−4,1)(-4,1) lies within a half-turn of (9,−2)(9,-2), the point (9m,36m)(9m,36m) is then still leading for (−4,1)(-4,1) [1 Remark 2.10, Proposition 3.7]; hence every monomial xiyjx^iy^j of the translated PP satisfies j≤4ij\le4i, that is, xiyj=ziv4i−jx^iy^j=z^iv^{4i-j} is a monomial in z=xy4z=xy^4, v=y−1v=y^{-1}.

For QQ: at a weight (−s,1)(-s,1), s∈{2,3}s\in\{2,3\}, where PP has an edge as above, [1 Proposition 7.3] with T0=QT_0=Q gives [ℓ(Q),ℓ(P)]=0[\ell(Q),\ell(P)]=0 and proportional values, so ℓ(Q)∼R9n\ell(Q)\sim R^{9n} (the element R=xy4−γy4−sR=xy^4-\gamma y^{4-s}, γ≠0\gamma\ne0, is not a proper power) and the same translation applies. For the pair obtained after the translations and 3≤s<43\le s<4, the values of PP and QQ are positive, because (9m,36m)(9m,36m) and (9n,36n)(9n,36n) are in the supports; the leading forms commute by [1 Proposition 1.13] (the sum of the values exceeds 1−s1-s), so ℓ−s,1(Q)\ell_{-s,1}(Q) is proportional to a power of the monomial ℓ−s,1(P)\ell_{-s,1}(P) [1 Proposition 2.1(2)], hence a monomial at a point (i,4i)(i,4i); by (D1) such a point of the support has i≤9ni \le9n, and (9n,36n)(9n,36n) is in the support, so j−si≤9n(4−s)j-si \le9n(4-s) on the support of the translated QQ. Letting ss tend to 4 gives j≤4ij \le4i there.

A translation x↦x+γy−sx \mapsto x+\gamma y^{-s}, s∈{2,3}s \in\{2,3\}, replaces xiyjx^{i}y^{j} by monomials xi−ryj−srx^{i-r}y^{j-sr} of (9,−2)(9,-2)-weight 9i−2j−(9−2s)r9i-2j-(9-2s)r, so (D1) holds for the translated pair with the same leading forms. Since xiyj=ziv4i−jx^{i}y^{j}=z^{i}v^{4i-j} has 9i−2j=i+2(4i−j)9i-2j=i+2(4i-j), the bound 9i−2j≤9m9i-2j \le9m is the weighted degree bound, and for x=zv4x=zv^{4}, y=v−1y=v^{-1} one has t=z2v−1t=z^{2}v^{-1} and x(t−c)4=z(z2−cv)4x(t-c)^{4}=z(z^{2}-cv)^{4}, which gives the top forms. The Jacobian follows from dx∧dy=−v2 dz∧dvdx \wedge dy=-v^{2}\,dz \wedge dv. □

The right side

Write X:=x1/9X:=x^{1/9}, so that L(9)=K[X±1,y]L^{(9)}=K[X^{\pm1},y].

Lemma 3.1. There are λ∈K×\lambda\in K^{\times} with λ9=c\lambda^{9}=c and λ3∈K\lambda_{3} \in K such that, with

Y:=η+λX−2+λ3X−3,P1(X,η):=P(X9,Y),Q1(X,η):=Q(X9,Y),Y:=\eta+\lambda X^{-2}+\lambda_{3}X^{-3},\qquad P_{1}(X,\eta):=P(X^{9},Y),\qquad Q_{1}(X,\eta):=Q(X^{9},Y),

and with the weight W(Xiηj):=i−4jW(X^{i}\eta^{j}):=i-4j, every monomial of P1P_{1} has W≤mW \le m and every monomial of Q1Q_{1} has W≤nW \le n.

Proof. In z~:=x2/9y\widetilde z:=x^{2/9}y one has ℓ9,−2(P)=λpxm(z~9−c)4m\ell_{9,-2}(P)=\lambda p x^{m}(\widetilde z^{9}-c)^{4m}, so every root λ\lambda of z~9=c\widetilde z^{9}=c has multiplicity 4m≥mlaρ+bσρ+σ=9m74m \ge\frac{m}{l}\frac{a\rho+b\sigma}{\rho+\sigma}=\frac{9m}{7} (with l=1l=1, (a,b)=(9,36)(a,b)=(9,36), (ρ,σ)=(9,−2)(\rho,\sigma)=(9,-2)). By [1 Proposition 5.18] the automorphism φ(y)=y+λx−2/9\varphi(y)=y+\lambda x^{-2/9} gives an (m,n)(m,n)-pair (φP,φQ)(\varphi P,\varphi Q) in L(9)L^{(9)} with a regular corner at A1:=(1,0)+4(2/9,1)=(17/9,4)A_{1}:=(1,0)+4(2/9,1)=(17/9,4), of direction (ρ′,σ′)<(9,−2)(\rho',\sigma')< (9,-2). We follow this corner as in the proof of [1 Proposition 5.19]. Case II is impossible, because [1 Proposition 5.14(4)] requires gcd⁡(17,4)>1\gcd(17,4)>1; this holds at every step below, since it depends only on the point A1A_{1}. In Case III, [1 Proposition 5.17] gives ρ′∣9\rho' \mid9, and the translation y↦y+λ′xσ′/ρ′y \mapsto y+\lambda' x^{\sigma'/\rho'} produces an (m,n)(m,n)-pair in L(9)L^{(9)} with a regular corner at the same point A1A_{1} and a smaller direction. After finitely many steps Case I occurs, for a pair (P1,Q1)(P_{1},Q_{1}) and a direction (ρ,σ)(\rho,\sigma). There vρ,σ(P1)+vρ,σ(Q1)=ρ+σv_{\rho,\sigma}(P_{1})+v_{\rho,\sigma}(Q_{1})=\rho+\sigma [1 Proposition 1.13], while en⁡(P1)=mA1=mnen⁡(Q1)\operatorname{en}(P_{1})=mA_{1}=\frac{m}{n}\operatorname{en}(Q_{1}) [1 Corollary 5.7(2)]. Hence m(m9ρ+4σ)=m5(ρ+σ)m(\frac{m}{9}\rho+4\sigma)=\frac{m}{5}(\rho+\sigma), that is (ρ,σ)=(9,−4)(\rho,\sigma)=(9,-4), v(P1)=mv(P_{1})=m, v(Q1)=nv(Q_{1})=n. The directions of the Case III steps decrease strictly from (9,−2)(9,-2) and stay larger than the direction (9,−4)(9,-4) of the final Case I corner, which is determined by the point A1A_{1}; they have ρ′∈{1,3,9}\rho' \in\{1,3,9\}. The only such direction is (3,−1)(3,-1), so there is at most one such step, a translation y↦y+λ3x−1/3y \mapsto y+\lambda_{3}x^{-1/3}. Finally v9,−4(xi/9yj)=i−4jv_{9,-4}(x^{i/9}y^{j})=i-4j. □

Common coordinates

Let A\mathcal A be the set of formal sums ∑cijXiηj\sum c_{ij}X^{i}\eta^{j} (i∈Z, j≥0)(i\in\mathbb Z,\ j\ge0) such that, for every rr, only finitely many terms have W=i−4j≥rW=i-4j\ge r. It is a ring containing K[X±1,η]K[X^{\pm1},\eta]; for f≠0f\ne0 let W(f)W(f) be the largest weight of its terms and ℓ(f)\ell(f) the sum of the terms of that weight. Then ℓ(fg)=ℓ(f)ℓ(g)\ell(fg)=\ell(f)\ell(g) and W(fg)=W(f)+W(g)W(fg)=W(f)+W(g), and 1+ε1+\varepsilon is invertible in A\mathcal A when W(ε)<0W(\varepsilon)<0.

Lemma 4.1. Let γ2,γ3,λ,λ3\gamma_{2},\gamma_{3},\lambda,\lambda_{3} be as in Lemmas 2.1 and 3.1, and put

ζ:=z2−cv−κ1z−κ0,κ1:=9λ3λ3,κ0:=−9λ6λ32.\zeta:=z^{2}-cv-\kappa_{1}z-\kappa_{0},\qquad\kappa_{1}:=9\lambda^{3}\lambda_{3},\qquad\kappa_{0}:=-9\lambda^{6}\lambda_{3}^{2}.

Then K[z,v]=K[z,ζ]K[z,v]=K[z,\zeta]. Write P^=∑pabzaζb\widehat P=\sum p_{ab}z^{a}\zeta^{b} and Q^=∑qabzaζb\widehat Q=\sum q_{ab}z^{a}\zeta^{b}. Then pab≠0p_{ab}\ne0 implies a≤ma\le m and a+2b≤9ma+2b\le9m; qab≠0q_{ab}\ne0 implies a≤na\le n and a+2b≤9na+2b\le9n. Moreover the part of P^\widehat P with a+2b=9ma+2b=9m is λPzmζ4m\lambda_{P}z^{m}\zeta^{4m}, the part of Q^\widehat Q with a+2b=9na+2b=9n is λQznζ4n\lambda_{Q}z^{n}\zeta^{4n}, and

P^zQ^ζ−P^ζQ^z=Jc3(z2−κ1z−κ0−ζ)2.\widehat P_{z}\widehat Q_{\zeta}-\widehat P_{\zeta}\widehat Q_{z} =\frac{J}{c^{3}}\left(z^{2}-\kappa_{1}z-\kappa_{0}-\zeta\right)^{2}.

Proof. Since c≠0c\ne0, v=(z2−κ1z−κ0−ζ)/cv=(z^{2}-\kappa_{1}z-\kappa_{0}-\zeta)/c, so K[z,v]=K[z,ζ]K[z,v]=K[z,\zeta]; giving ζ\zeta the weight 2, this substitution does not increase the weighted degree of Lemma 2.1, which gives the bounds on a+2ba+2b, and since z2−cv=ζ+κ1z+κ0z^{2}-cv=\zeta+\kappa_{1}z+\kappa_{0} the top forms become λPzmζ4m\lambda_{P}z^{m}\zeta^{4m} and λQznζ4n\lambda_{Q}z^{n}\zeta^{4n}. In A\mathcal A the element YY of Lemma 3.1 is λX−2(1+ε)\lambda X^{-2}(1+\varepsilon) with W(ε)<0W(\varepsilon)<0; it is invertible. The assignment v↦Y−1v\mapsto Y^{-1}, z↦X9Y4−γ2Y2−γ3Yz\mapsto X^{9}Y^{4}-\gamma_{2}Y^{2}-\gamma_{3}Y defines a ring morphism K[z,v]→AK[z,v]\to\mathcal A which sends P^\widehat P to P1P_1 and Q^\widehat{Q} to Q1Q_1, because P(x,y)=P^((x−γ2y−2−γ3y−3)y4,y−1)P(x,y)=\widehat{P}\left((x-\gamma_2y^{-2}-\gamma_3y^{-3})y^4,y^{-1}\right) in K[x,y±1]K[x,y^{\pm1}]. Expanding (ancillary script), the image of zz has W=1W=1 and leading term λ4X\lambda^4X, and the image of ζ\zeta has W=0W=0 and leading term 9λ7ηX49\lambda^7\eta X^4: the terms of weights 22 and 11 and the constant term cancel by the choice of c=λ9c=\lambda^9, κ1\kappa_1, κ0\kappa_0. Hence the image of zaζbz^a\zeta^b has leading term a non-zero multiple of ηbXa+4b\eta^bX^{a+4b}, of weight aa, and these monomials are distinct for distinct (a,b)(a,b). Therefore, for a non-zero polynomial, the value of WW on the image of ∑cabzaζb\sum c_{ab}z^a\zeta^b is max⁡{a:cab≠0}\max\{a:c_{ab}\ne0\}. Lemma 3.1 gives the bounds mm and nn. Finally dz∧dζ=−c dz∧dvdz\wedge d\zeta=-c\,dz\wedge dv, so the bracket in (z,ζ)(z,\zeta) is −1c-\frac{1}{c} times the bracket in (z,v)(z,v), which is −Jv2-Jv^2 by Lemma 2.1. □

The finite system

Let (U,V):=(P^,Q^)(U,V):=(\widehat{P},\widehat{Q}) if (m,n)=(3,2)(m,n)=(3,2) and (U,V):=(Q^,P^)(U,V):=(\widehat{Q},\widehat{P}) if (m,n)=(2,3)(m,n)=(2,3). Writing again zz for z−κ1/2z-\kappa_1/2 (this replaces κ1\kappa_1 by 00, changes κ0\kappa_0, and preserves all the statements of Lemma 4.1, the leading terms of the images and the weighted degrees being unchanged), and xx for ζ\zeta, a pair as in the standing data yields in both cases a solution of the following system.

(S1) U=z3u3+z2u2+zu1+u0U=z^3u_3+z^2u_2+zu_1+u_0 and V=z2v2+zv1+v0V=z^2v_2+zv_1+v_0 with ui,vi∈K[x]u_i,v_i\in K[x], deg⁡u3=12\deg u_3=12, deg⁡u2≤12\deg u_2\le12, deg⁡u1≤12\deg u_1\le12, deg⁡u0≤13\deg u_0\le13, deg⁡v2=8\deg v_2=8, deg⁡v1≤8\deg v_1\le8, deg⁡v0≤8\deg v_0\le8.

(S2) UzVx−UxVz=J′(z2−κ−x)2U_zV_x-U_xV_z=J'(z^2-\kappa-x)^2 with J′∈K×J'\in K^\times, κ∈K\kappa\in K.

Indeed the bounds a≤3a\le3, a+2b≤27a+2b\le27 for UU and a≤2a\le2, a+2b≤18a+2b\le18 for VV give the shape of UU and VV with deg⁡u1≤13\deg u_1\le13 and deg⁡v0≤9\deg v_0\le9, and the top forms λz3x12\lambda z^3x^{12}, λ′z2x8\lambda' z^2x^8 of Lemma 4.1 show that deg⁡u3=12\deg u_3=12, deg⁡v2=8\deg v_2=8 and that u1u_1 has no term of degree 1313 and v0v_0 no term of degree 99. For (m,n)=(2,3)(m,n)=(2,3) the bracket changes sign, which is absorbed in J′J'. Since {zaf,za′g}=(ag′−a′g)za+a′−1\{z^af,z^{a'}g\}=(ag'-a'g)z^{a+a'-1} for f,g∈K[x]f,g\in K[x], where {F,G}:=FzGx−FxGz\{F,G\}:=F_zG_x-F_xG_z, the coefficients of z4,z3,z2,z,1z^4,z^3,z^2,z,1 in (S2) are

3u3v2′−2u3′v2=J′(1)3u_3v_2'-2u_3'v_2=J' \tag*{(1)}
3u3v1′−u3′v1+2u2v2′−2u2′v2=0(2)3u_3v_1'-u_3'v_1+2u_2v_2'-2u_2'v_2=0 \tag*{(2)}
3u3v0′+2u2v1′−u2′v1+u1v2′−2u1′v2=−2J′(κ+x)(3)3u_3v_0'+2u_2v_1'-u_2'v_1+u_1v_2'-2u_1'v_2=-2J'(\kappa+x) \tag*{(3)}
2u2v0′+u1v1′−u1′v1−2u0′v2=0(4)2u_2v_0'+u_1v_1'-u_1'v_1-2u_0'v_2=0 \tag*{(4)}
u1v0′−u0′v1=J′(κ+x)2(5)u_1v_0'-u_0'v_1=J'(\kappa+x)^2 \tag*{(5)}

Lemma 5.1. If (S1)–(S2) has a solution, it has one with u0(0)=v0(0)=0u_0(0)=v_0(0)=0, with no term of degree 88 in u2u_2, and with u3=p3u_3=p_3, v2=q2v_2=q_2, J′=219⋅27⋅19J'=2^{19}\cdot27\cdot19, where

q2=x8+21x6+22x5+11838x4+423x3+672116x2+136798x+268777256,q_2=x^8+21x^6+22x^5+\frac{1183}{8}x^4+423x^3+\frac{6721}{16}x^2+\frac{13679}{8}x+\frac{268777}{256},
p3=x12+632x10+33x9+619516x8+981x7+4098916x6+787838x5+3531639256x4+61782516x3+31456311512x2+11753097256x+2917424534096.\begin{aligned} p_3={}&x^{12}+\frac{63}{2}x^{10}+33x^9+\frac{6195}{16}x^8+981x^7+\frac{40989}{16}x^6+\frac{78783}{8}x^5\\ &+\frac{3531639}{256}x^4+\frac{617825}{16}x^3+\frac{31456311}{512}x^2+\frac{11753097}{256}x+\frac{291742453}{4096}. \end{aligned}

One has p32−q23=884736(16x5+248x3+296x2+1001x+4426)p_3^2-q_2^3=884736(16x^5+248x^3+296x^2+1001x+4426).

Proof. The system is preserved by: multiplying UU and VV by non-zero constants; adding constants to UU and VV; replacing UU by U−αVU-\alpha V; the translation x↦x+τx\mapsto x+\tau (with κ\kappa replaced by κ+τ\kappa+\tau); and the substitution (z,x)↦(μz,μ2x)(z,x)\mapsto(\mu z,\mu^2x) (with κ\kappa replaced by κ/μ2\kappa/\mu^2). So we may assume that u3u_3 and v2v_2 are monic, that v2v_2 has no term of degree 77, that u0(0)=v0(0)=0u_0(0)=v_0(0)=0 and that u2u_2 has no term of degree 88. Consider equation (1) for monic u3,v2u_3,v_2 of degrees 12,812,8, with no term of degree 77 in v2v_2. The substitution x↦νxx\mapsto\nu x, followed by division by ν12\nu^{12} and ν8\nu^8, preserves these conditions and multiplies J′J' by ν−19\nu^{-19} and the coefficient b6b_6 of x6x^6 in v2v_2 by ν−2\nu^{-2}. For J′=1J'=1 the ideal of the equation has dimension 00 and degree 1919, so that there are at most 1919 solutions, and b619=13719/(238⋅219)b_6^{19}=13^7 19/(2^{38}\cdot2^{19}) on each of them (Gröbner basis, Singular [4], ancillary script); since b6≠0b_6\ne0, the 1919 substitutions with ν19=1\nu^{19}=1 give 1919 distinct solutions from one, so the solutions with J′=1J'=1 form a single orbit, and therefore all solutions with J′≠0J'\ne0 form a single orbit under x↦νxx\mapsto\nu x, ν∈K×\nu\in K^\times. The pair (p3,q2)(p_3,q_2) is a solution, with J′=219⋅27⋅19J'=2^{19}\cdot27\cdot19 (ancillary script). Hence b6≠0b_6\ne0 for every solution, and using (z,x)↦(μz,μ2x)(z,x)\mapsto(\mu z,\mu^2x) we may assume b6=21b_6=21; the solutions with b6=21b_6=21 are (p3(x),q2(x))(p_3(x),q_2(x)) and (p3(−x),q2(−x))(p_3(-x),q_2(-x)), the second with J′=−219⋅27⋅19J'=-2^{19}\cdot27\cdot19. The substitution (z,x)↦(iz,−x)(z,x)\mapsto(iz,-x), i2=−1i^2=-1, followed by the multiplication of UU by ii and of VV by −1-1, preserves the system and the normalisations and exchanges the two. □

Proposition 5.2. The system (S1)–(S2), normalised as in Lemma 5.1, has no solution.

Proof (exact computation). With u3=p3u_3=p_3, v2=q2v_2=q_2 and J′=219⋅27⋅19J'=2^{19}\cdot27\cdot19, each of (2), (3), (4) is, coefficient by coefficient, a system of linear equations with rational matrix: (2) in the coefficients of $u_2,v_1$; then (3) in those of v0,u1v_0,u_1 and in κ\kappa; then (4) in those of $u_0$; the right-hand sides depend on the solutions of the previous steps. Reducing each matrix to echelon form expresses the pivot unknowns through the others, and the rows without pivot give *conditions*, polynomials with rational coefficients in the unknowns which remain free, which vanish on every solution. Carrying this out (ancillary script): (2) leaves the coefficients $v_{1,7},v_{1,8}$ of $x^7,x^8$ in $v_1$ free and gives no condition; {xref}(3) leaves the coefficients $u_{1,11},u_{1,12}$ of $x^{11},x^{12}$ in $u_1$ and $\kappa$ free and gives no condition; {xref}(4) determines $u_0$ and gives 8 conditions; finally the coefficients of {xref}(5)` give 21 conditions. These 29 polynomials, of degrees 3 and 4 (they are not claimed to be linearly independent), generate the unit ideal of Q[v1,7,v1,8,u1,11,u1,12,κ]\mathbb{Q}[v_{1,7},v_{1,8},u_{1,11},u_{1,12},\kappa] (Gröbner basis, SymPy).

□

As an independent confirmation, the 82 equations of the normalised system in its 56 unknowns, without any elimination, also generate the unit ideal (Gröbner basis, Singular; optional step of the script). As a check of the procedure, the script applies the same elimination to a pair (U,V)(U,V) with u3=p3u_3=p_3, v2=q2v_2=q_2 and randomly chosen u2,u1,u0,v1,v0u_2,u_1,u_0,v_1,v_0, with the right-hand side of (S2) replaced by its bracket: the conditions then vanish at the chosen values, and no longer do so after the right-hand side is modified.

Proof of Theorem 1.2. A pair as in the standing data gives a solution of (S1)–(S2) by Lemmas 2.1–4.1, hence a normalised one by Lemma 5.1, contradicting Proposition 5.2. □

Scope

  • Proved here, relative to the cited results: the reduction of Theorem 1.2 to Proposition 5.2 (Lemmas 2.1–5.1). The inputs are (D1), which rests on [1 Proposition 6.1, Theorem 7.6(2), Remark 3.1], and [1 Theorem 2.6, Remarks 1.8, 2.5, 2.10, Corollaries 5.7, 7.4, Propositions 1.13, 2.1, 3.7, 3.9, 5.13, 5.14, 5.16, 5.17, 5.18, 5.19, 7.3], as indicated. The condition on Succ⁡P(1,0)\operatorname{Succ}_P(1,0) in the standing data is the one of [1 Corollary 5.21(4)], which is established there for the standard pairs attached to a minimal counterexample; the theorem is stated under this condition.

  • Verified by exact computation: Proposition 5.2; the statements on equation (1) in Lemma 5.1; the arithmetic of Remark 1.3 and Lemma 3.1; the leading terms in Lemma 4.1 (ancillary file verify_135.py in the folder anc; SymPy, and Singular for the number of solutions of (1), for which the script writes the input file). The result depends on the correctness of these computations.

  • Not claimed: a statement on the plane Jacobian conjecture itself; on the pairs with the same starting triple whose polynomial hh has more than one root (see Remark 1.3); on the third shape of maximal degree 135 with starting corner (9,36)(9,36) in [2], which has the corners (9,24)(9,24) and (11/3,8)(11/3,8); a new bound on the degree of a counterexample. The completeness of the list of [2] is not re-examined.

  • [1 Proposition 7.3 and Corollary 7.4], on which Lemma 2.1 rests, are stated in [1] with the indication that their proofs mimic those of Proposition 7.1 and Corollary 7.2; they are applied here to Laurent translates, as in [1 Section 8] and [3]. The references [2] and [3] are arXiv preprints.

  • The argument has been prepared by the author with AI assistance (Claude, Anthropic). Nothing in this note has been checked by a third party; it has not been validated by professionals.

References

Independent researcher, France

References

  1. [1]C. Valqui, J. A. Guccione, J. J. Guccione, On the shape of possible counterexamples to the Jacobian Conjecture, J. Algebra 471 (2017), 13–74, doi:10.1016/j.jalgebra.2016.08.039; arXiv:1401.1784v3. Author order as in the journal; statement number refers to the arXiv version v3.DOI
  2. [2]J. A. Guccione, J. J. Guccione, R. Horruitiner, C. Valqui, Some algorithms related to the Jacobian Conjecture, arXiv:1708.07936v1.arxiv.org/abs/1708.07936
  3. [3]J. A. Guccione, J. J. Guccione, R. Horruitiner, C. Valqui, Increasing the degree of a possible counterexample to the Jacobian Conjecture from 100 to 108, arXiv:2204.14178.arxiv.org/abs/2204.14178
  4. [4]W. Decker, G.-M. Greuel, G. Pfister, H. Schönemann, Singular 4-4-1 — A computer algebra system for polynomial computations, https://www.singular.uni-kl.de (2025).

Paper details

vlnc:W_62vpy7hr
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Mathematics

Anthropic Claude Fable

View statement
Looking for related papers…
Contents