Given a real number $a>0$, $(x_n)$ is recursively defined by $$ \begin{cases} x_1=a\\ x_{n+1}=\frac{2x_n+1}{x_n+2} \end{cases} $$ Find a proof that $(x_n)$ is convergent and then find the limit.

Can anyone give me a hint on how to determine the limit of a sequence like that? I tried to write the first terms to see if I could find an equation that was not recursive for the $n$-th term, but I failed miserably. I dont really know how to find the answer to that without such equation.

Grateful for any help!

  • 2
    $\begingroup$ Once you determined that the sequence converges (and only after that), assume that the limit is $L\,$, then pass the recurrence relation to the limit and solve $L = (2L+1) / L+2\,$. $\endgroup$
    – dxiv
    Jul 28 '17 at 3:07
  • 1
    $\begingroup$ On the contrary, first you should solve $L = (2L + 1)/(L + 2)$ to find what the limit should be, then you should proceed to prove that this $L$ really is a limit $\endgroup$ Jul 28 '17 at 3:09
  • $\begingroup$ @BoltonBailey The point of my comment was that you can not find the limit of a sequence this way unless you first prove that the sequence does in fact converge. Take for example $x_{n+1}=-x_n\,$, using your approach the limit would come up as $0\,$, which of course it's not. $\endgroup$
    – dxiv
    Jul 28 '17 at 3:13

HINT: if the sequence converges in $\Bbb R$ then applying limits both sides of the definition of $x_{n+1}$ we get

$$\lim_{n\to\infty} x_{n+1}=\lim_{n\to\infty} \frac{2 x_n+1}{x_n+2}\implies L=\frac{2L+1}{L+2}\implies L^2=1\implies L=1$$

where we had discarded $L=-1$ because the sequence $(x_n)$ is positive so it cannot converges to $-1$.

Then suppose that the limit of the sequence is $1$ and divide by cases. If $0<a<1$ then $x_1<x_2<1$, so you can try to setup a proof by induction to show that $x_n<1$ for all $n\in\Bbb N$ and that the sequence $(x_n)$ is strictly increasing.

If $a>1$ then $x_1>x_2$ and you can try to setup a proof by induction to show that the sequence $(x_n)$ is strictly decreasing and positive.

If the above fails for some case then you can try to show that the sequence doesnt converges to $1$, and because $1$ is the unique possible candidate for the limit then the sequence doesnt converge for the studied case.

An alternative method is studying the fixed points of $f(x):=\frac{2x+1}{x+2}$ in $(0,\infty)$. In this case we can see that $f$ is a contraction in this region (because $|f'(x)|<1$ for $x\in(0,\infty)$), so it converges to a unique fixed point in this region.


Since $a_1>0$ the the recursive def'n of $a_{n+1}$ implies,by induction on $n,$ that $a_n>0$ for all $n.$

Let $a_n=1+d_n.$ We have $a_n>0$ so $ d_n>-1.$ By direct calculation we have $$d_{n+1}=a_{n+1}-1=[(2a_n+1)/(a_n+2)]-1=d_n/(3+d_n).$$

Now $d_n>-1\implies 3+d_n>2,$ so $$|d_{n+1}|=|d_n|/|3+d_n|\leq |d_n|/2.$$

The result $|d_{n+1}|\leq |d_n|/2$ implies that $d_n\to 0$ and hence $a_n\to 1.$

Note: The first sentence of the Answer by Masacroso shows you how to deduce that if $L=\lim_{n\to \infty}$ exists, then $L^2=1.$ And since $a_n>0$ for all $n,$ if $L$ exists then $L=1.$


Coincidence! I just added a similar answer to another question using matrices.

Note: This is to answer your question about a non-recursive formula for the $n^{th}$ term (and consequently to find the limit). Also note that this is a very general approach for solving such problems.

$$x_n = \frac{p_n}{q_n}$$


$$ p_{n+1} = 2p_n + q_n$$ $$ q_{n+1} = p_n + 2q_n$$

[You can consider $s_n = p_n + q_n$ and do something clever, but we won't go there as we want a general approach]


$$A^n \begin{bmatrix}a\\1\end{bmatrix} = \begin{bmatrix}p_n\\q_n\end{bmatrix}$$


$$A = \begin{bmatrix}2&1\\1&2\end{bmatrix}$$

This matrix is diagonalizable and gives the general form of $p_n$ and $q_n$ to be

$$ u3^n + v (-1)^n$$

($3$ and $-1$ are the eigenvalues of $A$)

I will leave it to you to compute the the formulae for $p_n$ and $q_n$, which will prove convergence of $x_n$.

btw, the limit is $1$, because the matrix $A$ is symmetric. $A^n$ will be symmetric with entries of the form $x3^n + y(-1)^n$, and so the $u$ for $p_n$ will be the same as the $u$ for $q_n$ and the limit is their ratio.


When you have a sequence $x_{n+1}=f(x_n)$ with f(x) an increasing function ($f'(x)>0$).

You can prove the fact that the sequence has a ceiling (or a floor) and is increasing (or decreasing) by induction :

using the property of increasing function f : $a < b \Rightarrow f(a) < f(b)$

example :

if $x_{n} < L \Rightarrow f(x_{n})=x_{n+1} < f(L)=L$ the sequence has a ceiling. Provided that $x_1<L$

if $x_{n} < x_{n+1} \Rightarrow f(x_{n})=x_{n+1} < f(x_{n+1})=x_{n+2}$ the sequence is going up. Provided that $x_1<x_2$

Then you use the theorem : an increasing sequence with a ceiling necessarily converges. (or the other way)


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.