# What is the limit of a sequence defined recursively as $x_1=2$, $x_{n+1}=1/(3-x_n)$ with $n \in \mathbb{N}$, and how do I prove it exists?

This is an exercise I've found online.

Find the limit of a sequence defined recursively as $x_1=2$, $x_{n+1}=\dfrac{1}{3-x_n}$ with $n\in \mathbb{N}$. Show that the limit exists before attempting to find it.

So far, I have shown that $\{x_{n}\}$ is bounded below by $0$ and above by $2$ since $\frac{1}{3-x_n}>0$ and $\frac{1}{3-x_n}\le 2$ for all $n$.

I'm stuck here because I'm not sure what to show next, and I don't know precisely how to show the limit exists.

-
Show that $(x_n)$ is a decreasing sequence and then appeal to the fact that a bounded monotone sequence is convergent. –  David Mitra Jan 25 '13 at 18:06

At first, you can show that $x_1>x_2$ easily.

If $x_{n-1}>a_n$ holds (and $0\le x_n \le 2$ for all $n$), we get \begin{align} x_{n-1}>x_n & \Longrightarrow & -x_{n-1}<-x_n \\ & \Longrightarrow & 3-x_{n-1}<3-x_n \\ & \Longrightarrow & \frac{1}{3-x_{n-1}}>\frac{1}{3-x_n} \\ & \Longrightarrow & x_n > x_{n+1}\\ \end{align}

so $x_n>x_{n+1}$. So this sequence convergent, by monotone convergence theorem.

-
Is this an inductive proof that $x_{n}>x_{n+1}$? –  000 Jan 25 '13 at 18:13
@Limitless Oh, it is my mistake. –  tetori Jan 25 '13 at 18:17
In the last line, how do you arrive at $x_n>x_{n+1}$? Isn't $\dfrac{1}{3-x_{n-1}}=x_{n}$? By definition, if $x_{n+1}=\dfrac{1}{3-x_{n}}$, then a substitution of $x=x'-1$ shows us that $x_{n}=\dfrac{1}{3-x_{n-1}}$. –  000 Jan 25 '13 at 18:26
@DavidMitra I do not follow his or her progression from line 3 to line 4. Is it correct? If so, why? –  000 Jan 25 '13 at 18:30
@Limitless Line 3 had a typo. It is now correct. –  David Mitra Jan 25 '13 at 18:31
For limit solve the equation $$l=\frac{1}{3-l} \implies l^2-3l+1=0$$ You get $l=\frac{3+\sqrt 5}{2},\frac{3-\sqrt 5}{2}$. Your limit will be $\frac{3-\sqrt 5}{2}$ since $\frac{3 +\sqrt 5}{2} >2$.
If a limit $L$ exists, it must satisfy $$x_{n+1} = \frac{1}{3-x_n} \Rightarrow L = \frac{1}{3-L} \Rightarrow L^2-3L+1=0$$
So that candidates for the limit are $\frac{3\pm \sqrt{5}}{2}$. Since you've proven the series is bounded above by 2, only one of these is possible.