Exceptional isomorphisms between complements of affine plane curves (Q2280533)

Let \(k\) be an arbitrary field and denote by \(\bar{k}\) its algebraic closure. The Complement Problem asks whether two geometrically irreducible hypersurfaces \(E,F\) in the affine space \(\mathbb{A}^n\) are isomorphic if their complements in \(\mathbb{A}^n\) are isomorphic. Equivalently the problem can be formulated in commutative algebra as follows: given two polynomials \(P,Q\in k[x_1,\cdots,x_n]\) which are irreducible in \(\bar{k}[x_1,\cdots,x_n]\), and an isomorphism of \(k\)-algebras \(k[x_1,\cdots,x_n,\frac{1}{P}]\cong k[x_1,\cdots,x_n,\frac{1}{Q}]\), is it true that the \(k\)-algebras \(k[x_1,\cdots,x_n]/(P)\) and \(k[x_1,\cdots,x_n]/(Q)\) are isomorphic? This very natural problem was unsolved until Pierre-Marie Poloni recently constructed counterexamples for \(n\geq 3\) in [\textit{P.-M. Poloni}, Comment. Math. Helv. 94, No. 3, 439--444 (2019; Zbl 1441.14202)]. The paper under review gives a negative answer to the Complement Problem for \(n=2\). Note that the counterexamples in dimension two give rise to new counterexamples in higher dimensions different from those of Pierre-Marie Poloni. As a corollary the authors give also a negative answer to the Complement Problem in the holomorphic category, i.e.\ two non biholomorphic algebraic hypersurfaces in \(\mathbb{C}^n\) may have biholomorphic complements. Before constructing counterexamples for the Complement Problem, the authors establish various fundamental results concerning the complement of a geometrically irreducible curve in the affine plane \(\mathbb{A}^2\). Let \(C\) and \(D\) be two geometrically irreducible curves in \(\mathbb{A}^2\) such that there is an isomorphism \(\varphi:\mathbb{A}^2\backslash C\rightarrow \mathbb{A}^2\backslash D\) that does not extend to an automorphism of \(\mathbb{A}^2\). Then \(C\) is isomorphic to \(\mathbb{A}^1\backslash A\) and \(D\) is isomorphic to \(\mathbb{A}^1\backslash B\) where \(A\) and \(B\) are both finite sets of \(k\)-points of cardinality \(d\). If \(d=0\) then both \(C\) and \(D\) are images of a line under automorphisms of \(\mathbb{A}^2\). If \(d1\) then \(\varphi\) is uniquely determined up to an automorphism of \(\mathbb{A}^2\). When \(d=2\), \(C\) and \(D\) are always isomorphic. Thus if the pair \((C,D)\) gives a counterexample to the Complement Problem then \(d2\). The case where \(d=1\) is further investigated. For any \(n1\), the authors construct explicitly curves \(C_1,\cdots,C_n\) in \(\mathbb{A}^2\), all isomorphic to \(\mathbb{A}^1\backslash\{0\}\), such that \(\mathbb{A}^2\backslash C_i\) are all isomorphic but \(C_i\) is not the image of \(C_j\) under an automorphism of \(\mathbb{A}^2\) for any \(i\neq j\). The paper is well written. We refer the reader to the clear introduction for more complete and precise statements of the main results.