Simultaneous approximation measures for functions satisfying generalized functional equations of Mahler type (Q1349474)

Mahler's method in transcendence theory began with questions he asked about the arithmetic properties of the function \(f(z)= \sum z^{2^n}\), which satisfies the typical functional equation \(f(z)= f(z^2)+ z\). Mahler considered generalisations to several complex variables with far-reaching consequences. In the case of one variable, interesting examples arise when the transformation \(z\to z^2\) is replaced by an algebraic relation. Thus, suppose \(f:U \to \mathbb{C}\) is holomorphic in a neighbourhood \(U\) of \(\omega\) in \(\mathbb{C}\) and transcendental over \(\mathbb{C} [z]\) and its power series coefficients about \(\omega\) are algebraic. Suppose \(T(z)\) is algebraic of degree \(n\) over \(\mathbb{Q} (z)\) with minimal polynomial \(Q(z,y)\), meromorphic in \(U\) and has \(T\omega= \omega\) and \(T(U) \subset U\). Suppose further that \[ (a)\quad (\text{ord}_\omega T)^3 >(mn \deg_z Q)^2, \qquad \text{and} (b)\quad \text{ord}_\omega T>mn \deg_yP. \] Suppose, finally, that \(P(z,f(z), f(Tz)) =0\) in \(U\) with a polynomial \(P\) in \(\mathbb{Q} [z,y,u]\) and \(\deg_u (P) = m\geq 1\). \textit{P. G. Becker} [J. Reine Angew. Math. 440, 111-128 (1993; Zbl 0770.11037)] proved transcendence theorems for such functions, but these appear to be ineffective. The author establishes an effective simultaneous approximation measure under the above hypotheses. Suppose \(\alpha\) in \(U\) satisfies \(T^k\alpha \to\omega\) as \(k\to \infty\), \(T^k \alpha \neq \omega, \infty\) and \(P(T^k \alpha, f(T^k\alpha), u) \neq 0\). Then, for any algebraic numbers \(\xi_1, \xi_2\) with \(\deg \xi_i \leq D_i\) and \(H(\xi_i) \leq H_i\), \[ \max \biggl\{|\alpha-\xi_1 |, \bigl|f(\alpha) -\xi_2 \bigr|\biggr\} \geq \exp \left[- c(D_1+ \log H_1)^2 D^2_2 \bigl( D_2 (D_1 +\log H_1) \bigr)^{\mu_1} +{\log H_2^{\mu_2} \over D_2} \right] \] with explicit \(\mu_1, \mu_2\) depending on \(mn\), \(\deg_z Q\), \(\text{ord}_\omega T\), \(\deg_yP\) and a constant \(c\) depending only on \(f\) and \(\alpha\). The key to the proof is a zero estimate. Take the above circumstances, with hypothesis (b) but without hypothesis (a). Then for each non-zero polynomial \(R\) in \(\mathbb{C} [z,y]\) with \(\deg_z R\leq N\), \(\deg_y R\leq M\), we have \[ \text{ord}_0 R \bigl(z, f(z)\bigr) \leq c_1 M(N+m) \] with an effectively computable constant \(c_1\) depending only on \(f\). The theorem is then proved by constructing an auxiliary polynomial \(R(z,y)\) such that \(R(z,f(z))\) vanishes to high order at the origin, iterating the functional relation to form polynomials \(R_k(\alpha,f(\alpha))\) from \(R(T^k\alpha,f(T^k\alpha))\) with very small values and approximating these by algebraic numbers \(R_k(\xi_1,\xi_2)\). For large enough \(k\) and as a result of hypothesis (a), the analytic upper bounds and the algebraic lower bounds are incompatible. Because of hypothesis (b), which is needed in the proof of the zero estimate, the theorem does not include all the examples treated by Becker. It would be of interest and may be possible to do without (b).