Irrationality of power series for various number theoretic functions (Q2487843)

The authors study formal power series whose coefficients are taken to be a variety of number theoretic functions, such as the Euler, Möbius and divisor functions. They show that these power series are irrational over \(Z[X]\), and they obtain lower bounds on the precision of their rational approximations. To describe their results lets recall some definitions. For an integer \(n>1\), let \(n=p_1^{\alpha_1}\ldots p_k^{\alpha_k}\) be its prime factorization. -- The Euler function, which gives the cardinality of the group \((\mathbb Z/n\mathbb Z)^*\), is defined by \(\varphi(n) =\prod_{j:1}^kp_j^{\alpha_j-1}(p_j-1)\). -- The functions \(\omega(n)\) and \(\Omega(n)\) have their usual meanings: \(\omega(n)=k\) is the number of distinct prime factors of \(n\), and \(\Omega(n)=\alpha_1+\ldots+\alpha_k\) is the number of prime divisors of \(n\), counted with multiplicity. -- The function \(\tau(n)\) counts the number of positive divisors \(d\) of \(n\); one has \(\tau(n) =\prod_{j:1}^k(\alpha_j+1)\). -- The function \(\rho(n)=2^{\omega(n)}=2^k\) counts the number of squarefree positive divisors of \(n\). -- The function \(\sigma(n)\) gives the sum of the positive divisors \(d\) of \(n\); we have \(\sigma(n)= \prod_{j:1}^k {p_j^{\alpha_j+1}-1\over p_j-1}\). -- The Liouville function is given by \(\lambda(n)=(-1)^{\omega(n)}=(-1)^k\), while \(\mu(n)\) denotes the Möbius function; we recall that \(\mu(n)=\lambda(n)\) if \(n\) is squarefree, and \(\mu(n)=0\) otherwise. Following standard conventions, we also put \(\omega(1)=\Omega(1)=0\) and \(\varphi(1)=\tau(1)=\rho(1)=\sigma(1) =\lambda(1)=\mu(1)=1\). \noindent - For every positive integer \(n\), let \(p(n)\) denote the \(n-th\) prime number. The goal of this paper is to study the irrationality of the corresponding formal power series: \noindent \(\displaystyle{\mathcal F}(X)= \sum_{n:1}^{\infty}\varphi(n)X^n\), \(\displaystyle{\mathcal W_1}(X)= \sum_{n:1}^{\infty}\omega(n)X^n\), \(\displaystyle{\mathcal W_2}(X)= \sum_{n:1}^{\infty}\Omega(n)X^n\), \(\displaystyle{\mathcal T}(X)= \sum_{n:1}^{\infty}\tau(n)X^n\), \(\displaystyle{\mathcal P}(X)= \sum_{n:1}^{\infty}p(n)X^n\), \(\displaystyle{\mathcal R}(X)= \sum_{n:1}^{\infty}\rho(n)X^n\), \(\displaystyle{\mathcal S}(X)= \sum_{n:1}^{\infty}\sigma(n)X^n\), \(\displaystyle{\mathcal L}(X)= \sum_{n:1}^{\infty}\lambda(n)X^n\), \(\displaystyle{\mathcal M}(X)= \sum_{n:1}^{\infty}\mu(n)X^n\). For any positive integer \(d\) and any formal power series \({\mathcal U}\in Z[[X]]\), we define \(\Delta_d({\mathcal U})= \max\{ \text{ord} ({\mathcal U}(X)-{f(X)\over g(X)})\); \(f,g\in Z[X]\), \(\deg f,\deg g\leq d, g\not=0\}\). The main result of the paper is Theorem: For a positive integer \(d\), the following bounds hold: \(\Delta_d({\mathcal F})= \exp (O(d^{2\over 3}log^{2\over 3}d))\), \(\Delta_d({\mathcal W}_i)=O(d.3^{1.965d})\), \(i=1,2\), \(\Delta_d({\mathcal T})=O(d^2)\), \(\Delta_d({\mathcal P}) =O({d^29^d\over log^2d})\), \(\Delta_d({\mathcal R})=O(d.3^{{11d\over 4}})\), \(\Delta_d({\mathcal S}) = \exp(O(d^{2\over 3}\log^{{2\over 3}}d))\), \(\Delta_d({\mathcal L})=O(d.3^{{11d\over 4}})\), \(\Delta_d({\mathcal M})=O(d.3^{{11d\over 4}})\). I end by the following remark. On page 184, it is said that the field of formal Laurent series \(Q((X))\) is the field of fractions of \(Z[[X]]\). This is not true. Indeed the transcendence degree of \(Q((X))\) over the field of fractions of \(Z[[X]]\) is infinite. As a reference I indicate chapter 3 ``Corps des fractions de l'anneau des séries formelles'' of the reviewer's book [Les anneaux de séries formelles, Queen's Papers in Pure and Applied Mathematics 124. (Kingston: Queen's University) (2003; Zbl 1054.13010)].