I'm tryig to use Newton–Raphson's method to compute the inverse of $a$. Given $f(x) = a - \frac{1}{x}, a \neq 0$ we are looking for $x^*$ such that $f(x^*) = 0$. Newton–Raphson's method gives the sequence $x_{n+1} = -ax_n^2 + 2x_n$. For example let $a = 2$, so $x_{n+1} = -2x_n^2 + 2x_n$. So want to find for which initial values $x_0$ the method converges. Playing around with an algorithm I implemented I found that $x_0$ must be in $(0, 1)$ for it to converge.
But how can I prove that? And in the more general case of $a$. Any help would be appreciated.
Have you heard of Kantorovich's theorem for Newton's method?
Here a expository Article: Kantorovich's s theorem on Newton's method.. See too other expository Article here.. See Ortega's paper here for a simple proof.
See personal home page of Prof. Ferreira for articles of Newton-Kantorovich method and variants in Banach spaces.