Abstract
The Cremona conjecture, also called Jacobi problem, claims that a polynomial morphism \({{\mathbb C}^n \longrightarrow {\mathbb C}^n}\) is invertible as a polynomial morphism if its Jacobian is constant and not zero. In this paper, we show that the conjecture is true for \(n = 2\). The starting point of our proof is an important result of Shreeram Abhyankar. Then we use a computation in rigid geometry to achieve the result.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction
A polynomial map \((f,g):{\mathbb C}^{\,2} \longrightarrow {\mathbb C}^{\,2}\) is given by two polynomials f and g in two variables X and Y with complex scalars. We write f and g as sums of their homogenous components
where \(f_\mu \) respectively \(g_\nu \) are linear combinations of the terms of total degree \(\mu \) respectively \(\nu \). The forms \(f_m\) respectively \(g_n\) of highest degree are called the leading forms.
It was shown by Abhyankar that, for a given counterexample (f, g) to the Jacobian conjecture in dimension 2, one can assume that, after a suitable transformation of variables, the leading forms of f and g have the following shape
cf. [1, Theorem 8.7] or [5, Corollary 10.2.22]. In this paper, we will show that this assumption leads to a contradiction. Thus the Jacobian conjecture is true in dimension 2.
2 Division algorithm
In this section, let \({\mathbb K}\) be an algebraically closed field of characteristic 0. The \({\mathbb K}\)-algebra \(L:={\mathbb K}[X,Y]_{XY}\) consists of all Laurent polynomials in two variables. It carries a canonical graduation of type \({\mathbb Z}\) given by the total degree function. Let \(H_n\) be the subspace of all homogenous Laurent polynomials of degree n including the zero polynomial. For \(f=\sum _{\nu \in {\mathbb Z}}f_{\nu }\in L\) and \(f\ne 0\), we set
On L, we have a filtration \((L_{n};\,n\in {\mathbb N})\) where
The completion with respect to this filtration is denoted by
cf. [2, Chap. 3]. It consists of all series
cf. [5, Prop. 10.2.8]. The degree, the multiplication, and the filtration on A are declared as on L. The \({\mathbb K}\)-algebra A represents the formal functions on a neighborhood of the twice punctured projective line at infinity which behave like meromorphic functions there. The algebra A has similar properties as the algebra \(R^{\sim }\) defined in [5, Prop. 10.2.8]. In this section, we consider these functions without conditions of convergence; in Section 2, we will focus on that by means of rigid geometry.
Lemma 1.1
An element \(g\in A\) is a unit in A if and only if g is of the form
where \(c\in {\mathbb K}^{\times }\,,\, n_i\in {\mathbb Z}\,,\, \deg (v)<0\). Such a representation is unique. Such a unit g admits a k-th root for \(0\ne k\in {\mathbb Z}\) if and only if k divides both numbers \(n_1\) and \(n_2\).
Proof
The proof can be left to the reader. For example, we have for the inverse
For \(c=1\) and \(k\in {\mathbb N}\), the k-th root is given by
if k divides \(n_1\) and \(n_2\). We consider this as the canonical k-th root of g. \(\square \)
Corollary 1.2
Let \(g=X^{n_1}Y^{n_2}\cdot (1+v)\in A\) with \(\deg (v)<0\) and \(r\in {\mathbb Q}\) be such that \(r\cdot n_1\) and \(r\cdot n_2\) belong to \({\mathbb Z}\), then
is well defined.
In the following, we denote by \(\partial /\partial X\) respectively \(\partial / \partial Y\) the partial derivatives of Laurent series. Obviously they give rise to \({\mathbb K}\)-derivations on the \({\mathbb K}\)-algebra A. They satisfy the usual rules for \({\mathbb K}\)-derivations. Since the field \({\mathbb K}\) has characteristic 0, we have \(\ker (\partial /\partial X\,,\,\partial / \partial Y)={\mathbb K}\).
Definition 1.3
A couple (f, g) of elements of A is called a Jacobian couple if its Jacobian
is constant and not 0.
As for polynomials in two variables, we also have the notion of a leading form for a Laurent series in A.
Proposition 1.4
Consider a Jacobian couple (f, g) as introduced above, where \(m:=\deg f\) and \(n:=\deg g\). Assume that the leading form of g has the shape \(g_n=X^{n_1}Y^{n_2}\).
-
(a)
Then we always have \(m+n\ge 2\).
-
(b)
If \(m+n>2\), then \(f_m^{n}\cdot g_n^{-m}\) is constant.
Proof
(a) The homogenous components of the Jacobian of degree \(m+n>2\) vanish and for \(m+n=2\) it is given by
Since that the Jacobian is constant and the degree of a constant is 0, we see that \(m+n\ge 2\).
(b) If \(m+n>2\), then the expression (1) is zero. We compute
For homogenous polynomials, we have Euler’s differential equation
Then the term in parentheses of equation (2) is equal to
This vanishes due to (1) since \(m+n>2\). Thus we see that the left hand term of equation (2) is equal to 0. Analogously, one shows
So the total differential of \(f_m^{n}\cdot g_n^{-m}\) vanishes. Thus, the function \(f_m^{n}\cdot g_n^{-m}\) is constant. \(\square \)
Corollary 1.5
Let (f, g) be a Jacobian couple as in 1.4 satisfying \(g_n=X^{n_1}Y^{n_2}\) with integers \(n_1>0\,,\,n_2>0\). If \(m+n>2\), then we have \(f_m=c\cdot X^{m_1}Y^{m_2}\) with a constant \(c\in {\mathbb K}^{\times }\). Moreover it holds
Therefore the following expression is well defined
In particular, we have \(\left( g^{m/n} \right) ^n=g^{m}\). Furthermore \(|m_1-m_2|\ne |n_1-n_2|\) if \(|m|\ne |n|\).
Proof
The first assertion follows from 1.4(b). For the second assertion, we use
So we see \(|m_1-m_2|\ne |n_1-n_2|\) if \(|m|\ne |n|\). The formula for \(g^{m/n}\) follows from 1.2. \(\square \)
Now we turn to the division algorithm.
Proposition 1.6
Let (f, g) be as in 1.5. Then there exists a rational number \(r\in {\mathbb Q}\) such that \(r\cdot n_1\in {\mathbb Z}\) and \(r\cdot n_2\in {\mathbb Z}\) are integers, and a constant \(c\in {\mathbb K}^{\times }\) with \(\deg (f-c\cdot g^{r})<m\). The couple (c, r) is uniquely determined; actually we have \(r=m/n\) and \(f_m=c\cdot X^{m_1}Y^{m_2}\).
If, in addition, \(\deg (f-c\cdot g^{r})=2-n\), then \(n_1\ne n_2\) and the leading form of \(h:=f-c\cdot g^{r}\) is given by
where \(c_{1-n_1,1-n_2}\ne 0\). Furthermore there is at most one index (i, j) with \((i,j)\ne (1-n_1,1-n_2)\) with \(c_{i,j}\ne 0\). For this index, we have
Proof
The first assertion follows from 1.5.
For the supplement, set \(m':=\deg (f-c\cdot g^{r})=2-n\). The Jacobian d of the couple (h, g) is equal to the Jacobian of (f, g). Then we have
For \((i,j)=(1-n_1,1-n_2)\), it follows that
Thus, we see \(n_1\ne n_2\) and \(c_{1-n_1,1-n_2}\ne 0\). For all the other indices, we have
If \(c_{i,j}\ne 0\), then
Moreover we know \(i+j=2-n\) and \(n=n_1+n_2\). This yields
and \(i\cdot n_2=(2-n-i)\cdot n_1\) and hence \(i\cdot n=i\cdot (n_2+n_1)=(2-n)\cdot n_1\) \(\square \)
Corollary 1.7
Keep the assumptions of 1.6. Then we have \(n_1\ne n_2\) and there exist a natural number \(s\in {\mathbb N}\), constants \(c_{\sigma }\in {\mathbb K}\), and rational numbers \(r_{\sigma }\in {\mathbb Q}\) satisfying
such that
belongs to A and the leading term of \((f-G)\) fulfills
Proof
Apply 1.6 inductively. Note that \(\deg (f-G)\) is always an integer and that \((f-G\,,\,g)\) is a Jacobian couple. Therefore the procedure stops after finitely many steps until we arrive at the situation \(\deg (f-G)=2-n\) since there is at each step at most one term which has to be cancelled. In the case \(\deg (f-G)=2-n\), we apply the additional claim of 1.6. Then we obtain for the leading form
where \(i/n_1=j/n_2=(2-n)/n\) as follows from 1.6. Then we subtract
which cancels the term \(c_{i,j}X^{i}Y^{j}\). Thus the assertion is proved. \(\square \)
3 Convergence of the division algorithm
In the following, we make use of some elementary results in rigid geometry; for a general reference, we cite [3] or [4]. We consider an algebraically closed field \({\mathbb K}\) which is complete with respect to a non-Archimedean valuation and which has residue characteristic 0. We assume that \({\mathbb K}\) contains the field K of characteristic 0 as a subfield, where K is the algebraically closed field over which the Jacobian problem is posed. Such a field can be constructed in the following way: Consider the field of fractions \(K'\) of K[[T]] and define \({\mathbb K}\) as the topological algebraic closure of \(K'\). The canonical valuation on K[[T]] extends to a valuation of \({\mathbb K}\). Note that we write valuations in the multiplicative way. So we obtain on \({\mathbb K}^{2}\) a canonical structure of rigid space in the sense of Tate. On each subset \(V\subset {\mathbb K}^{\,2}\), we have the spectral norm of functions f
In particular, we have the notion of an affinoid domain \(V\subset {\mathbb K}^{2}\); for example, bounded domains described by finitely many inequalities
with polynomials \(f_i,g_j\in {\mathbb K}[X,Y]\) are affinoid domains. Affinoid functions on such a domain are functions which can be uniformly approximated by rational functions without poles in V. Such functions are bounded and take their maximal absolute value in V. Thus the spectral norm \(|f|_V\) is always a non-negative real number which actually lies in the value group of \({\mathbb K}\). We are mainly interested in domains of the following shape
for values \(\varepsilon \le \rho \) belonging to the value group of \({\mathbb K}\). The affinoid functions on \(W_{\varepsilon ,\rho }\) are exactly the Laurent series which converge on \(W_{\varepsilon ,\rho }\). Of particular interest will be the following domains
These subsets are also affinoid and they are open subsets in the rigid analytic sense.
Lemma 2.1
Keep the above notations. Let \(\varepsilon \,,\,\rho \) be elements of the value group \(|{\mathbb K}^{\times }|\) with \(\rho \ge \varepsilon \).
-
(a)
If v is an affinoid function on \(U:=U_{\varepsilon ,\rho }\) with \(|v|_U<1\), then the series
$$\begin{aligned} h:=\sum _{\nu =0}^{\infty } {{r}\atopwithdelims (){\nu }} v^{\nu }\,, \end{aligned}$$for any \(r\in {\mathbb Q}\), converges uniformly on \(U_{\varepsilon ,\rho }\) and gives rise to an affinoid function there. In particular, \((1+v)^{r}\) is well-defined and affinoid on \(U_{\varepsilon ,\rho }\).
-
(b)
Let \(g=g_n+\cdots +g_0\in {\mathbb K}[X,Y]\) be a polynomial with homogenous components \(g_{\nu }\) of degree \(\nu \). Assume \(g_n=X^{n_1}Y^{n_2}\) . Then there exists an \(\varepsilon \) in \(|{\mathbb K}^{\times }|\) such that \(|g_{\nu }(x,y)|<|g_{n}(x,y)|\) for all \((x,y)\in U_{\varepsilon ,\rho }\) and all \(\nu =0,\ldots ,n-1\) and \(\rho \ge \varepsilon \). Especially, for any \(r\in {\mathbb Q}\) with \(n_1\cdot r\in {\mathbb Z}\) and \(n_2\cdot r\in {\mathbb Z}\), the function \(g^{r}\) is well-defined and affinoid on \(U_{\varepsilon ,\rho }\) for all \(\rho \) with \(\rho \ge \varepsilon \).
Proof
(a) Since the residue field of \({\mathbb K}\) has characteristic 0, the absolute value \(|{{r}\atopwithdelims (){\nu }}|=1\) is equal to 1. Therefore the series converges on \(U_{\varepsilon ,\rho }\) for all \(\rho \ge \varepsilon \). (b) For all monomials \(X^{\nu _1}Y^{\nu _2}\) of \(g_{\nu }\) with \(\nu <n\), we have \(|x^{\nu _1}y^{\nu _2}|\le |x^{n_1}y^{n_2}|\) if \((x,y)\in U_{\varepsilon ,\rho }\) and \(\varepsilon >1\). If we now choose \(\varepsilon \ge |c_{\nu _1,\nu _2}|\) for all the coefficients \(c_{\nu _1,\nu _2}\) of \(g_{\nu }\) for all \(\nu =0,\ldots ,n-1\), then the assertion follows by (a).
For the last assertion, note that \(\left( X^{n_1}Y^{n_2}\right) ^{r}=X^{m_1}Y^{m_2}\), where \(n_1\cdot r=m_1\) and \(n_2\cdot r=m_2\) with \(m_1,m_2\in {\mathbb Z}\) . Then it follows from (a). \(\square \)
Proposition 2.2
Let (f, g) be a Jacobian couple of polynomials with homogenous decompositions
in \({\mathbb K}[X,Y]\) with \(n_1>0\,,\,n_2>0\), where \(m:=m_1+m_2\) and \(n:=n_1+n_2\).
If we apply the division algorithm of 1.6 and 1.7 to f and set \(v:=\sum _{\nu =0}^{n-1}g_{\nu }g_n^{-1}\), then there exists an \(\varepsilon \in |\mathbb {K}^{\times } |\) such that the formal series G defined in 1.7 converges on every affinoid domain \(U_{\varepsilon ,\rho }\) for all \(\rho \ge \varepsilon \) and gives rise to an affinoid function there.
After a possible enlarging of \(\varepsilon \), the function \((f-G)\) has the form
with \(e\in {\mathbb K}^{\times }\,,\,\deg (u)<0\,\), is affinoid on each \(U_{\varepsilon ,\rho }\), and satisfies \(\,|u|_{U_{\varepsilon ,\rho }}<1\).
Proof
The claim follows from Lemma 2.1. \(\square \)
In the following, we will compute the cardinality of the fibers of \((f-G\,,\,g)\) on \(U_{\varepsilon ,\rho }\).
Proposition 2.3
Let (f, g) be a Jacobian couple as in 2.2. Thus we have the map
Set \(k:=\gcd (n_1,n_2)\). Then, for any domain \(V:=U_{\varepsilon ',\rho '}\subset U_{\varepsilon ,\rho }\), the fibers of the morphism \(\left( f-G\,,\,g^{1/k} \right) |_V\) consist of exactly \(|n_1-n_2|/k\) points. The fibers of \(\left( f\,,\,g \right) |_V\) consist of exactly \(|n_1-n_2|\) points.
Proof
We abbreviate
Since \(|u|_V<1\) and \(|v|_V<1\), the map \(\Psi \) gives rise to a map
Due to the construction, all numbers \(k\cdot r_{\sigma }\) are integers by 1.6 since \(r_{\sigma }\cdot n_1\in {\mathbb Z}\) and \(r_{\sigma }\cdot n_2\in {\mathbb Z}\). Obviously, this map is injective. So, for every \((x_0,y_0)\in V\), the map \(\Psi \) induces a mapping
Next we will compute the cardinality of the fibers of \(\Psi _{(x_0,y_0)}\). After adjusting the radii \(|x_0|\) and \(|y_0|\) to 1 and the constant e to 1, we are concerned with a morphism of type \(\Phi :W\longrightarrow W\) with
sending \((x,y)\in W\) to \((x^{1-n_1}\cdot y^{1-n_2}\cdot (1+u(x,y))\,,\,x^{n_1/k}\cdot y^{n_2/k}\cdot (1+v(x,y))^{1/k}\). The degree of this map can be calculated via its reduction. The algebra of affinoid functions on W which are bounded by 1 is given by \({\mathbb K}^{\circ }\langle X,Y,X^{-1},Y^{-1}\rangle \), where \({\mathbb K}^{\circ }\) denotes the valuation ring of \({\mathbb K}\). Denote by \({\tilde{\mathbb K}}\) the residue field of the valued field \({\mathbb K}\) and by \({\tilde{\mathbb K}}^{\times }\) its multiplicative group. The reduction of W is given by the spectrum of the \({\tilde{\mathbb K}}\)-algebra \({\tilde{W}}= {\tilde{\mathbb K}}[{\tilde{x}}\,,\,{\tilde{x}}^{-1}\,,\,{\tilde{y}}\,,\,{\tilde{y}}^{-1}]\), where \({\tilde{x}}\) resp. \({\tilde{y}}\) is the reduction of x resp. y. Since \(|u|<1\) and \(|v|<1\), the map of the reductions coincides with the mapping
The degree of this map is \(|n_1-n_2|/k\) as claimed; cf. Lemma 2.4 below. This degree is the degree of \(\Phi \) since a finite generating system of the reduced module via \({\tilde{\Phi }}\) lifts to a generating system via \(\Phi \) due to the lemma of Nakayama [3, 1.2.4/6]. Linear independence is also preserved as one easily checks.
It remains to compute the cardinality of the fibers of \(\left( f\,,\,g \right) |_V\). Recall from 1.7 that G(x, y) is a function of \(g(x,y)^{1/k}\). Therefore, we have that the fiber of \((f-G,g^{1/k})|_V\) of a point \((x,y)\in V\) with image \((z_1,z_2):=(f-G,g^{1/k})(x,y)\) coincides with the fiber of \((f,g^{1/k})|_V)\) over the point \((z_1+c_1,z_2)\) where \(c_1:=G(x,y)\) depends only on \(z_2=g^{1/k}(x,y)\). Thus, we see that the cardinality of the fiber of \(\Psi \) coincides with that of \((f,g^{1/k})|_V\). Therefore the cardinality of the fiber of \((f,g)|_V\) is equal to \(|n_1-n_2|\) since \(\Phi \) is finite and étale. \(\square \)
Lemma 2.4
Let k be a field and let \(m_1,m_2\in {\mathbb Z}\) be non-zero and \(m_1\ne m_2\). Let x, y be variables and \(r\in {\mathbb Z}\). Then the extension of k-algebras
is finite flat of degree \(|m_1-m_2|\).
Proof
Obviously we have
Moreover, we have that the extension
is finite flat of degree \(|m_1-m_2|\). \(\square \)
4 The contradiction
Now we have all the preparations to deduce the main result of our article.
Proposition 3.1
There does not exist a Jacobian couple (f, g) of polynomials \(f,g\in {\mathbb C}[X,Y]\) with homogenous decompositions
where \(f_m= X^{m_1}Y^{m_2}\) and \(g_n= X^{n_1}Y^{n_2}\) with \(m_1m_2\ne 0\,,\,n_1n_2\ne 0\).
Proof
First of all we perform a field extension \({\mathbb C}\hookrightarrow {\mathbb K}\) as introduced in Section 2. It is clear that it suffices to show the assertion for the field \({\mathbb K}\). Assume that (f, g) is such a couple in the \({\mathbb K}\)-algebra A.
Assume first \(m=n\). If \(m+n=2\), then we would have \(m=m_1=1\) and \(n=n_2=1\) without loss of generality. Obviously, that case cannot occur as a counterexample. If \(m+n>2\), then we have \(m_1=n_1\) and \(m_2=n_2\) due to 1.5. So we can replace f by \(h:=f-g\). Due to 2.2, the leading form of the polynomial h also has the shape \(h_r=a\cdot X^{r_1}Y^{r_2}\) and \(r:=r_1+r_2<n\). Thus we can assume \(m\ne n\). Moreover, we have \(m_1\ne n_1\) and \(m_2\ne n_2\) and \(|m_1-m_2|\ne |n_1-n_2|\) due to 1.5. Thus we see that we can start just from the beginning with \(m\ne n\).
Now we apply Proposition 2.3. So there exists a function in A
as in 1.5 such that \(h:=(f-G)\) is of degree \((2-n)\) and h has a leading form of the shape
Note that all the monomials of h have negative degree. Furthermore, there exists a domain \(U:=U_{\varepsilon ,\rho }\) such that for any subdomain \(V:=U_{\varepsilon ',\rho '}\subset U\) the restriction \((h,g)|_V\) has fibers with cardinality \(n'=|n_1-n_2|\), which coincides with the degree of the map \((f,g)|_V\); cf. 2.3.
If we interchange f und g, then, after a possible shrinking of U, the degree of \((f,g)|_V\) would be \(m'=|m_1-m_2|\). This has to be equal to \(n'\), but we have \(m'\ne n'\) due to 1.5. Contradiction! \(\square \)
Summarizing the arguments we obtain the main result. Indeed, by Abhyankar’s result, a counterexample would give rise to a Jacobian couple of the given shape, which cannot exist due to Proposition 3.1.
Theorem 3.2
Let \((f,g): {\mathbb C}^{2}\rightarrow {\mathbb C}^{2}\) be a polynomial morphism. If the Jacobian of (f, g) is constant and unequal zero, then (f, g) is an isomorphism.
Change history
27 May 2023
A Correction to this paper has been published: https://doi.org/10.1007/s00013-023-01863-0
References
Abhyankar, S.S.: Some thoughts on the Jacobian conjecture. I. J. Algebra 319(2), 493–548 (2008)
Bourbaki, N.: Commutative Algebra. Chapters 1–7. Translated from the French. Reprint of the 1989 English translation. Elements of Mathematics (Berlin). Springer, Berlin (1998)
Bosch, S., Güntzer, U., Remmert, R.: Non-Archimedean Analysis. A Systematic Approach to Rigid Analytic Geometry. Grundlehren der mathematischen Wissenschaften, 261. Springer, Berlin (1984)
Bosch, S.: Lectures on Formal and Rigid Geometry. Lecture Notes in Mathematics, 2105. Springer, Cham (2014)
van den Essen, A.: Polynomial Automorphisms and the Jacobian Conjecture. Progress in Mathematics, 190. Birkhäuser, Basel (2000)
Acknowledgements
My thanks go to Werner Lütkebohmert, who critically read several versions of the manuscript and made several useful suggestions. Moreover, I am grateful to the referee for his attentive consideration of the manuscript and his valuable suggestions.
Funding
Open Access funding enabled and organized by Projekt DEAL.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Bartenwerfer, W. The Cremona problem in dimension 2. Arch. Math. 119, 53–62 (2022). https://doi.org/10.1007/s00013-022-01733-1
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00013-022-01733-1