Guccione, Guccione and Valqui list the possible shapes of a counterexample to the plane Jacobian conjecture of small degree; the first shape of maximal degree 125 is the family F_2 with (m,n)=(3,5), of degrees (75,125). We show that no standard (3,5)-pair of this shape exists. The proof combines three results of Guccione-Guccione-Valqui with the residue theorem on a generic fibre of P: on such a fibre the form y^2 Q dQ has no residue along the left side of the Newton polygon, while its residues along the right side add up to a quantity that cannot vanish. All statements are proved relative to the cited results; the exact computations are reproduced by an ancillary script. The argument has not yet been examined by specialists.
Statement
Throughout, K is an algebraically closed field of characteristic zero, L=K[x,y], and [P,Q]=PxQy−PyQx. We use the notation of [1] (Guccione–Guccione–Valqui): for a weight (ρ,σ), vρ,σ is the weighted degree, ℓρ,σ the leading form, and stρ,σ,enρ,σ its end points. Standard (m,n)-pairs, their regular corners and their starting triple are defined in [1 Definitions 4.3, 5.5, 6.2].
Standing data 1.1.(P,Q) is a standard (3,5)-pair in L with [P,Q]=J∈K×, degP=75, degQ=125, starting triple ((5,20),(1,0),(5,−1)), and (−1,1)<SuccP(1,0)<(−1,0) as in [1 Corollary 5.21(4)].
By [2 Theorem 2.20 and Sections 5–6], this is the shape (family F2, j=1) that a counterexample of degrees (75,125) produced by [1 Corollary 5.21] must have; see also [3] for the shapes of smaller degree.
Theorem 1.2. There is no pair (P,Q) as in Standing data 1.1.
Consequences of the standing data used below
Put t:=xy5. Adding constants and rescaling x, P, Q we may assume the following.
(D1) The support of P lies in {(i,j):5i−j≤15,i+j≤75}, that of Q in {5i−j≤25,i+j≤125}, and
ℓ5,−1(P)=(xh(t))3,ℓ5,−1(Q)=μ(xh(t))5,μ∈K×,
with h∈K[t] monic of degree 4, h(0)=0; thus st5,−1(P)=(3,0) and en5,−1(P)=(15,60)[1 Proposition 6.1, Theorem 7.6(2)]. We write u(t):=th(t).
(D2) Every root of h has multiplicity 1 or 2, and h has at least one double root. Indeed, write ℓ5,−1(P)=x3p(z) with z=x1/5y; a root of h of multiplicity k gives roots of p of multiplicity 3k. By [1 Proposition 5.16] some root has multiplicity ≥415, so k≥2 occurs. For any root with k≥2, [1 Proposition 5.18] produces a (3,5)-pair in L(5) with a regular corner at A(1)=(1+k/5,k). At such a corner, Case III several factors is excluded for k∈{3,4} by [1 Proposition 5.14(4)], because gcd(5+k,k)=1; Case III only translates [1 Proposition 5.17]; and Case I is excluded for k∈{3,4} by the end point conditions of [1 Proposition 5.13], which have no solution (ancillary script). The same shapes are recorded in the tables of [2 Section 5], where F2, with final corner (7/5,2), is the only family starting at (5,20) that contains (m,n)=(3,5).
Homogeneous brackets
Let N≥1 be an integer and Rω:=K[τ][ω,ω−1], with the bracket {f,g}N:=ωN(fτgω−fωgτ). For s∈Q>0 put νs(τaω−d):=d−sa, let νs(f) be the maximum over the monomials of f and ℓs(f) the sum of the terms realising it. An element with a single νs-value a has the form ω−aF(σ) with σ:=τω−s and F∈K[σ].
(b) νs({f,g}N)≤νs(f)+νs(g)+s+1−N, with equality if and only if {ℓsf,ℓsg}N=0.
(c) If {f,g}N=c∈K×, then Σs:=νs(f)+νs(g)+s+1−N≥0 for every s; if Σs>0 then aFG′=bF′G for the leading forms, and if Σs=0 then aFG′−bF′G=c.
Proof. (a) is a direct computation; (b) and (c) follow from (a) by comparing νs-values. □
The plane enters through (τ,W):=(t−θ,y) for θ∈K: since dx∧dy=y−5dt∧dy, one has [f,g]={f,g}5 with ω=W, and P=∑d≤15fd(t)W−d with f15=u3, Q=∑d≤25gd(t)W−d with g25=μu5.
The left side
Lemma 3.1. There is g∈vK[v] with degg≤3 and v2∣g such that P(z,v):=P(zv4+g(v),v−1), Q(z,v):=Q(zv4+g(v),v−1) are polynomials in z,v. Moreover P has total degree 15, top form H3 with H(z,v):=v5u(z/v), P(z,0) has degree 15 in z, and
PzQv−PvQz=−Jv2.
Proof. Apply the flip ψ1 of [1 Section 8], so that the weight (5,−1) becomes (ρ0,σ0)=(−1,5), with stρ0,σ0=(60,15) and enρ0,σ0=(0,3) for the flipped P. Let F be the element of [1 Theorem 2.6] for (ρ0,σ0). As in the proof of [1 Proposition 8.1]: if st(F)=(1,1) then en(F)=(1,1), since F is not a monomial [1 Remark 2.5], so en(F)∼en(P)=(0,3) and ν−1,5(F)=4 would give en(F)=(0,4/5), which is impossible; hence st(F)∼(60,15), so st(F)=(16,4)=34st(P) and q=5 in [1 Corollary 7.4]. 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 ℓρ0,σ0, the degrees ν1,1, ν0,1 and F are unchanged). It yields: for every weight between (1,−4) and (ρ0,σ0) at which the pair has an edge, the edge form of P is a 15th power R15. In the original variables: for a weight (−s,1) with s<4 carrying an edge with end point (15,60), the form is (xy4−γy4−s)15, so s∈Z, and s≥2 by the condition on SuccP(1,0). The translation x↦x+γy−s removes this edge and leaves the forms of smaller s unchanged [1 Proposition 3.9]. After at most two such translations (s=2,3) no edge of weight (−s,1), s<4, has end point (15,60). Since the weight (−4,1) lies within a half-turn of (5,−1), the point (15,60) is then still leading for (−4,1)[1 Remark 2.10, Proposition 3.7]; hence every monomial xiyj of the translated P satisfies j≤4i, that is, xiyj=ziv4i−j is a monomial in z=xy4, v=y−1. The same holds for Q: at a weight where P has an edge, [1 Proposition 7.3] with T0=Q gives [ℓ(Q),ℓ(P)]=0 and proportional values, so ℓ(Q)∼R25 and the same translation applies; at the remaining weights (−s,1), s<4, the values of P and Q are positive, so the leading forms commute [1 Proposition 1.13] and ℓ(Q) is the monomial at (25,100); hence j≤4i on the support of Q by continuity. The degree statements follow from 5i−j≤15 in (D1), and the Jacobian from dx∧dy=−v2dz∧dv. □
Places at infinity of a fibre
Fix a generic p∈K and let Γp:={P=p}⊂Az,v2, with Γp the normalisation of its projective closure. Since H(1,0)=1, every place π of Γp at infinity satisfies v→∞ and z/v→θ with u(θ)=0; then y=v−1→0 and t=xy5→θ. So π is a branch, at (τ,W)=(0,0), of P−p regarded in K[τ][W,W−1], τ=t−θ.
Lemma 4.1. Let P,Q∈Rω with {P,Q}N=c∈K×, let (N,D) be one of (5,15), (5,12), (5,9), (9,27), (9,21) and let 0≤s∗<D/3. Assume that for all s in some interval (s∗,s∗+ε) one has ℓs(P)=A′τ3ω−D and ℓs(Q)=B′τ5ω−5D/3 with A′B′=0. Then D=3N (so the hypotheses cannot hold for the other three pairs), and there is κ∈K[ω] with ordωκ>s∗ such that, for τ=κ(ω)+σωD/3, both P and Q are polynomials in σ,ω. Every branch τ(ω) of P=p with ordωτ>s∗ has σ bounded; in particular Q is regular along it.
Proof. Let s0 be the supremum of the s>s∗ with ℓs(P)=A′τ3ω−D. It is finite, since otherwise τ3∣P and {P,Q}N would vanish at τ=0. Suppose s0<D/3 and write ℓs0(P)=ω−aFP, ℓs0(Q)=ω−bFQ; then a=D−3s0>0, degFP=3, FP has at least two terms, b≥5D/3−5s0=35a and degFQ≤5 (a monomial of P of τ-degree >3, or of Q of τ-degree >5, is not leading for any s>s∗). By Lemma 2.1(c), either FPb is proportional to FQa, or Σs0=0 and aFPFQ′−bFP′FQ=c. In the first case degrees give b=35a, FP5∼FQ3, so FP=A′(σ−κ0)3 with κ0=0 and s0∈Z; the translation τ↦τ+κ0ωs0 preserves the bracket and the hypotheses, with s∗ replaced by s0. The second case requires a monomial pair (τ1,τ0) or (τ0,τ1) with integer exponents in (ℓs0P,ℓs0Q) together with Σs0=0 and b≥35a>0; for the five pairs (N,D) this has no solution s0<D/3 (ancillary script). Hence after finitely many translations s0≥D/3, so νD/3(P)=0 and P∈K[σ,ω] with P(σ,0) of degree 3. Let b:=νD/3(Q)≥0. If b>0, Lemma 2.1(c) gives FP′FQ=0 or −bFP′FQ=c, both impossible; so b=0 and Q∈K[σ,ω]. Now dP∧dQ=cω−Ndτ∧dω=cωD/3−Ndσ∧dω is regular, so D≥3N; among the five pairs this leaves only D=3N, the other three being contradictory. Finally, for every s with s∗<s<D/3 other than the finitely many slopes removed by the translations, ℓs(P−p) is a monomial, so no root τ(ω) of P−p has order s; at a removed slope s0 the roots of order s0 are those with σ→κ0, and the translation moves them to higher order. Hence every root with ordτ>s∗ has order ≥D/3 in the final coordinate, that is, σ is bounded. □
Corollary 4.2. At every place of Γp at infinity lying over a simple root θ of u (including θ=0), the function Q is regular.
Proof. For small s, ℓs(P)=u′(θ)3τ3W−15 and ℓs(Q)=μu′(θ)5τ5W−25; apply Lemma 4.1 with (N,D)=(5,15), s∗=0. □
Lemma 4.3. Let θ=c be a double root of h and u(c+τ)=Uτ2+O(τ3), U=0. Then exactly one of the following holds.
Q is regular at every place of Γp over c.
After a translation τ↦τ+κ1W+κ2W2, with σ=τW−12/5,
with BG=−J and G=5μB2/(9U). There are exactly two places of Γp over c: a place β1 with ordβ1(y)=5, at which Q has a pole of order 5 and yQ→−G/5; and a place β2 with ordβ2(y)=1, at which Q has a simple pole and yQ→G. In case (ii), G6=25J4μ4/(81U2), and κ1,κ2,B,G do not depend on p.
Proof. Work with f:=P−p and Q in K[τ][W,W−1], N=5. For small s, ℓs(f)=U3τ6W−15, ℓs(Q)=μU5τ10W−25. Let s0 be the first s at which ℓs(f) is not a monomial; as f has a nonzero constant term, s0≤5/2. Put a=νs0(f)=15−6s0 and b=νs0(Q)≥35a, so that Σs0≥36−15s0. Put a=νs0(f)=15−6s0 and b=νs0(Q)≥35a, so that Σs0≥36−15s0. Thus s0<12/5. Then Σs0>0, so FPb∼FQa, whence b=35a and FP=U3G3, FQ=μU5G5 with G monic quadratic, G=σ2. If G=(σ−κ)2, then s0∈{1,2} and we translate by κWs0 and start again. If G has two distinct roots and s0∈{1,2}, translating to either root gives the situation of Lemma 4.1 with N=5, D=15−3s0∈{12,9}, s∗=s0, which is contradictory. Otherwise G=σ2−κ with 2s0∈{1,3}; with W=ω2 one has {⋅,⋅}5=21{⋅,⋅}9, and translating to τ=κω2s0 gives the situation of Lemma 4.1 with N=9, D=30−6s0∈{27,21}, s∗=2s0.
::: For D=21 this is contradictory; for D=27 it gives (i), the six roots of f with τ→0 being those near ±κω, exchanged by ω↦−ω.
s0=12/5. The only lattice point on the edge is τW−3, so FP=Bσ+U3σ6, a=3/5. If b>1, proportionality FPb∼FQa is impossible by degrees; so b=1, Σs0=0 and 53FP′FQ−53FPFQ′=J with FQ=G+Eσ5+Dσ10, whose unique solution is E=3U3G/B, D=9U6G/(5B2), BG=−J. Comparing D with the coefficient μU5 of τ10W−25 gives G=5μB2/(9U), hence B3=−9JU/(5μ) and the value of G6. For s>12/5 the leading terms are BτW−3 and GW−1, with Σs=0; the next edge of f can only end at the constant term, at s=3, where Lemma 2.1(c) forces ℓ3(Q)=GW−1. Thus the roots τ(W) of f with τ→0 are five conjugate roots of order 12/5, forming one place β1 on which σ5→−B/U3 and WQ→G(1−3+59)=−G/5, and one root of order 3, the place β2, on which WQ→G.
12/5≤s0≤5/2. For s0<5/2 there is no lattice point on such an edge. For s0=5/2 one has a=0 and Σs0=b−23≥0, and Lemma 2.1(c) gives FP′FQ=0 or b=3/2 and −23FP′FQ=J, both impossible since degFP=6.
Two double roots
Lemma 5.1. If h=(t−c1)2(t−c2)2 with c1=c2, then c12−3c1c2+c22=0. In particular c12+c22=0.
Proof. In K(t)((y)) let C:=uy−5(1+⋯) be the cube root of P, and recall [a(t)yα,b(t)yβ]=(βa′b−αab′)yα+β+4. From 3C2[C,Q]=J we get [C,Q]=3u2Jy10(1+⋯). If R=Q−∑λiCi has leading term b(t)yβ with β<11, then βu′b+5ub′=0, so b=λu−β/5; as u has a simple zero at 0 and b∈K(t), 5∣β and the term is removed by subtracting λC−β/5. Hence there are λ5,…,λ−2∈K with Q−∑i=−25λiCi=φ(t)y11+⋯, φ∈K(t), and
5uφ′+11u′φ=3u2J.
A rational solution has poles only at the roots of u, of order 3k−1 at a root of multiplicity k≤2, and is O(t−11) at infinity. So φ=(a0+a1t)t−2(t−c1)−5(t−c2)−5, and eliminating a0,a1 from the resulting polynomial identity gives c12−3c1c2+c22=0 (ancillary script). If moreover c12=−c22, then c1c2=0, a contradiction.
Proof of Theorem 1.2
On Γp consider the meromorphic forms
ω1:=v−1dQ,ω2:=v−2QdQ.
Finite points. They are regular where v=0. The points of Γp with v=0 are the (z0,0) with P(z0,0)=p; for generic p, Pz(z0,0)=0, so v is a local parameter and, by Lemma 3.1, dQ=−Jv2Pz−1dv on Γp. Hence ω1,ω2 are regular at these points.
Places at infinity. There y=v−1 vanishes, so ωj=yjQj−1dQ is regular wherever Q is regular. By Corollary 4.2 and Lemma 4.3, Q has poles only at the places β1,β2 over the double roots c of h that are in case (ii); there, writing ω2=(yQ)2dQ/Q,
Conclusion. By the residue theorem, ∑cGc2=0, the sum being over the double roots in case (ii). If there is no such root, Q has no pole on Γp and is constant on each component, whereas dQ has no zero on the affine curve {P=p} because [P,Q]=J=0. If there is exactly one, Gc=−J/B=0, a contradiction. By (D2) the only remaining possibility is h=(t−c1)2(t−c2)2 with both roots in case (ii) and Gc12=−Gc22. Then Gc16=−Gc26, and Lemma 4.3 gives U12=−U22, where Uk=ck(c1−c2)2; thus c12+c22=0, contradicting Lemma 5.1.
Scope
Proved here, relative to the cited results:Theorem 1.2. The inputs are (D1), (D2) and [1 Theorem 2.6, Corollary 7.4, Propositions 5.13–5.18], as indicated.
Verified by exact computation: the bracket formula, the solution of the terminal bracket equation and the two residue sums, the formula for G6, the arithmetic exclusions in (D2) and in Lemmas 4.1 and 4.3, and the elimination in Lemma 5.1 (ancillary file verify-residue.py, in the folder anc; it uses SymPy).
Not claimed: a statement on the plane Jacobian conjecture itself, or on other shapes of degree ≥125. Whether a counterexample of degrees (75, 125) is excluded unconditionally depends on the reduction to standard pairs in [1, 2].
[1 Proposition 7.3 and Corollary 7.4], on which Lemma 3.1 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].
The argument has been checked only by the author with AI assistance (Claude, Anthropic); it has not been validated by professionals.
Independent researcher, France
References
[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. Statement numbers refer to the arXiv version 3.DOI
[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]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