On special extrema of polynomials with applications to Diophantine problems (Q2280012)

\textit{Y. F. Bilu} and \textit{R. F. Tichy} proved in [Acta Arith. 95, No. 3, 261--288 (2000; Zbl 0958.11049)] a remarkable theorem that enabled the classical theorem of Siegel to work in an effective (in a sense) manner. Ever since, there have been a large number of research papers dealing with Diophantine equations of the form \(f(x)=g(y)\) for polynomials \(f,g\). One aspect which proved crucial is that of finding for a given complex polynomial \(f\), all the complex numbers \(\lambda\) for which \(f(x)- \lambda\) has \textit{many} multiple roots. This is crucial in proving finiteness or infiniteness of the number of solutions of the Diophantine equations of the form \(f(x)=g(y)\). In the paper under review, the author characterizes the set of polynomials \(f \in \mathbb{Q}[X]\) for a given \(t>0\) such that there are exactly \(t\) complex numbers \(\lambda_1, \cdots, \lambda_t\) with \(f(x) - \lambda_i\) of the form \(u g(x)^q h(x)^k\) with \(u \in \mathbb{C}\), \(g,h \in \mathbb{C}[X]\) and either deg \(g = 2, k=2q\) or \(\deg g \leq 1\), \(k \geq 2\) and \(q \geq 1\). The cases \(t \neq 2\) are known and fairly easy to deal with but the crucial case of \(t=2\) dealt with in this paper requires some deep tools including work of Avanzi and Zannier. The author applies the results to problems on Diophantine equations and proves what may be viewed as an effective, algebraic version of Bilu-Tichy's theorem in the special case when \(g(y)\) is of the form \(\beta y^k + c\). The above-mentioned characterization involves two families of polynomials denoted by \(\mathbb{P}_1\) and \(\mathbb{P}_2\). The first family is defined as: \noindent \(\mathbb{P}_1 := \bigg\{ u g(x)^q h(x)^k + v, u,v \in \mathbb{Q}, u \neq 0, g,h \in \mathbb{Q}[X]\), and \(~~~~~~~~~~\)either deg \(g = 2, k=2q\) or deg \(g \leq 1\), \(k \geq 2\) and \(q \geq 1 \bigg\}\). The second family involves the ubiquitous Dickson polynomials. To define the family, we recall Avanzi and Zannier's result. For a square-free, monic polynomial \(G \in \mathbb{Q}[X]\), consider the equation \[F(x)^2 - G(x)H(x)^2 = \gamma \hskip 12mm (\spadesuit)\] in polynomials \(F,H \in \mathbb{Q}[X]\), and \(\gamma \in \mathbb{Q}^{\ast}\). Avanzi and Zannier completely parametrized all the solutions of the above equation in terms of the elliptic curve \(y^2=G(x)\). They proved roughly that if \(F_1\) is monic, of minimal degree \(N\) satisfying the equation, then the solution \(F_d(x)\) monic of degree \(dN\) is given as the Dickson polynomial \(D_d\) of \(F_1(x)\) for some parameters. For a given \(G\) as above, and a positive integer \(n\), let \(f_{G,n}(x)\) denote the unique monic polynomial of degree \(n\) appearing as a solution in \((\spadesuit)\) -- if it exists. In this case, \(\gamma_{G,n}\) denotes the corresponding \(\gamma\) value in \((\spadesuit)\). Define: \(\mathbb{P}_2 := \{ uD_d(x+v,a) + w: d \in \mathbb{N}, a,u,v,w \in \mathbb{Q}, au \neq 0 \} \cup\) \(~~~~~~~~~\{uf_{G,n}(x)+v: n \in 2 \mathbb{N}, u,v \in \mathbb{Q}, u \neq 0, f_{G,\frac{n}{2}}(x)^2 - \gamma_{G,\frac{n}{2}} \neq 2f_{G,n}(x) \}\). The main point is that due to Mazur's theorem on torsion of elliptic curves over \(\mathbb{Q}\), the family \(\mathbb{P}_2\) contains linear transforms of polynomials of the form \(D_d(F(x),a)\) with \(\deg F \leq 12\) and \(\deg F \neq 11\). Before stating the main theorem, we recall one more definition. Given any polynomial \(f \in \mathbb{C}[X]\), and a complex number \(\lambda\), let \(f(X) - \lambda\) have \(r\) distinct roots with multiplicities \(\mu_1, \cdots, \mu_r\). The number \(\lambda\) is said to be \textit{special} for \(f\) if there is some \(k\geq 2\) for which \(k/(k,\mu_i)\)'s form either \(1,1, \cdots, 1,2,2\) or \(1,1, \cdots, 1,q\) for some \(q \geq 1\). The main result of the paper asserts the following: Theorem. For \(i \geq 1\), let \(S_i\) denote the set of polynomials \(f\) of degrees at least \(5\) for which precisely \(i\) complex numbers are special. Then, \[S_1 = \mathbb{P}_1 \setminus \mathbb{P}_2, S_2 = \mathbb{P}_2, S_i = \emptyset~~\forall~ i>2.\] Combining this result with well known results of Schinzel-Tijdeman and of Brindza etc., the author deduces finiteness theorems for solutions of Diophantine equations of the form \(f(x)= \beta y^k + c\) or of polynomial values in Pell-type equations and for solutions of \(u_n = f(x)\) where \(u_n\)'s satisfy a binary recurrence.