I'm looking this recursive formula to approximate $\sqrt{2}$
\begin{align*} x_1 &= 2 \\ x_{n+1} &= \frac{1}{2}\left(x_n + \frac{2}{x_n}\right) \end{align*}
I noticed an easier way to approximate by hand would be to think of each $x_k$ as a rational number $x_k = \frac{a_k}{b_k}$ which when substituted into the formula above produces
\begin{align*} x_1 &= \frac{a_1}{b_1} =\frac{2}{1} \\ x_{n+1} &= \frac{a_{n+1}}{b_{n+1}} = \frac{a_n^2 + 2b_n^2}{2a_nb_n} \end{align*}
But what's interesting is that
\begin{align*} x_1 = &\frac{2}{1} \\ &x_2 = \frac{3}{2} \\ x_3 = \frac{9+8}{12} \\ x_4 &= \frac{289 + 288}{408} \\ x_5 = \frac{332929 + 332928}{470832} \end{align*}
So it appears that after $n = 2$, that $a_{n+1} = 2{a_n}^2 -1$. If this were true for all $n > 2$ then I could save a lot of time computationally, and that's what I want to prove. From there on out the numerator would only be dependent on the previous numerator. But I'm having a hard time proving that. I've tried induction but it seems like everything comes back to depending on a $b,$-term.