A remark on Zagier's observation of the Mandelbrot set (Q888450)

Let \(d \geq 2\) be an integer and consider the set \(\mathcal{M}_d\) of complex numbers \(c\) such that the Julia set of the unicritical polynomial \(z \mapsto z^d + c\) is connected. The set \(\mathcal{M}_d\) is called the Mandelbrot set for \(d = 2\) and the Multibrot for \(d \geq 3\). Most of the properties of the bifurcations inside the family of unicritical polynomials can be used to study the bifurcations of general family of holomorphic functions. Douady and Hubbard showed that the Mandelbrot set is connected. More precisely, for any \(d \geq 2\), there exists a conformal isomorphism \(\Phi_d\) from the complement of \(\mathcal{M}_d\) in the Riemann sphere to the complement of the closed unit disk. This conformal isomorphism can be expressed into a Laurent series expansion: \[ \Phi_d (z) = z + \sum_{m = 0}^{\infty} b_{d, m} z^{-m}. \] This article is concerned with the arithmetic properties of the coefficients \(b_{d,m}\). \textit{I. Jungreis} [Duke Math. J. 52, 935--938 (1985; Zbl 0589.30023)] showed that those numbers are \(d\)-adic rational numbers. Moreover the coefficients \(b_{d,m}\) for which \((d-1)\) doesn't divide \((m+1)\) are equal to \(0\). Several authors studied the \(d\)-adic valuation of \(b_{d,m}\), in particular giving some lower bounds or identites. This includes \textit{J. H. Ewing} and \textit{G. Schober} [Numer. Math. 61, No. 1, 59--72 (1992; Zbl 0745.30026)], \textit{G. M. Levin} [Teor. Funkts. Funkts. Anal. Prilozh. 51, 94--106 (1989; Zbl 0698.30021)] in the case \(d=2\). Another notable work is from Yamashita's master thesis, whose result, as pointed out by themselves, is valid only when \(d\) is a prime number. An upper bound on the \(2\)-adic valuation on the denominators of the non zero coefficients \(b_{2, m}\) has been verified up to \(m = 1000\) [\textit{B. Bielefeld} et al., Adv. Appl. Math. 14, No. 1, 25--38 (1993; Zbl 0772.30025)], which is referred to as Zagier's observation in the article. Yamashita's result extends and prove this inequality for any prime degree \(d\). It has been independently proved by different authors (Levin, Lau and Schleicher, Yamashita) that if \((d-1) {\not |} (m + 1)\) then \(b_{d, m} = 0\). The main result of this article concerns the case \((d-1) | (m + 1)\). Let \(\nu_p\) denote the \(p\)-adic valuation. It states that if \(d\) is factored into its prime factors as \(d = p_1^{t_1} p_2^{t_2} \cdots p_s^{t_s}\), where \(p_i\) are the distinct primes numbers in the factorization of \(d\), then, for all \(i\), \[ - \nu_{p_i} \left(b_{d, m}\right) \leq \nu_{p_i} (a!) + t_i a, \] where \(a\) is the integer \(a = \frac{m + 1}{d - 1}\). Moreover equality holds precisely when \(m = d - 2\) or \(p_i {\not |} m\). This result encompasses all the aforementioned results. It is based on a formula from \textit{J. H. Ewing} and \textit{G. Schober} [Numer. Math. 61, No. 1, 59--72 (1992; Zbl 0745.30026)] giving the coefficients \(b_{d, m}\) in terms of a sum of products of (noninteger) binomial coefficients and on the properties of valuations.