2
$\begingroup$

When $a_1 = 1$, then $a_n$ converges to $\frac{1+\sqrt5}{2}$, But if we let $a_1 > 1$, does the conclusion still hold? Here's what I tried:

First, we can easily conclude $1<a_n<2$,let $f(x) = 1+ x/(1+x),\; (x>1)$, then $f'(x) = 1/(1+x)^2$. So $f(x)$ is strictly increasing. So $$ a_{n+1}> a_n\Leftrightarrow f(a_n) > f(a_{n-1})\Leftrightarrow a_n >a_{n-1} $$ Finally we can get $a_{n+1}> a_n\Leftrightarrow a_2 >a_1$.

Then $$ a_2 - a_1 = 1+a_1/(1+a_1) - a_1 = \dfrac{1+a_1 - a_1 ^2}{1+a_1} $$ The condition is $a_1> 1$, let $g(x)=1+x-x^2\; (x>1)$, the graph of $g(x)$: enter image description here

Here is the problem : must $a_2 - a_1$ be greater than 0, equals to 0, or less than 0 ?

So is the conclusion incorrect when a is greater than 1?

$\endgroup$
2
  • 1
    $\begingroup$ $a_2 > 1$ right? So just forget $a_1$ and rename $a_n$ as $b_{n-1}$ $\endgroup$
    – Tony Pizza
    Commented Oct 14, 2023 at 5:10
  • $\begingroup$ For $a_n > 0$ we have $1 < a_{n+1}< 2$ $\endgroup$
    – Cesareo
    Commented Oct 14, 2023 at 8:00

3 Answers 3

6
$\begingroup$

For a number $a$ we have $$a_{n+1}-a={(2-a)(a_n-a)-(a^2-a-1)\over 1+a_n}$$ If $a={1+\sqrt{5}\over 2}$ we get $$a_{n+1}-a=(2-a){a_n-a\over 1+a_n}$$ Hence $$|a_{n+1}-a|\le(2-a)|a_n-a|$$ Thus $$|a_{n+1}-a|\le (2-a)^n|a_1-a|$$ As $0<2-a={3-\sqrt{5}\over 2}<1$ we get $a_n\to a$ and the speed of convergence is exponential. Hence the series $\sum (a_n-a)$ is convergent.

Remark The above is a general method based on the contraction principle: if $a_{n+1}=f(a_n),$ $f(a)=a$ and $|f'(x)|\le c<1$ then $a_n\to a.$

$\endgroup$
1
  • $\begingroup$ Thanks a lot, your method uses contraction mapping to get the existence of fixed point and avoids the discussion of monotonicity, I totally get it! $\endgroup$
    – Zhiwei
    Commented Oct 14, 2023 at 13:47
1
$\begingroup$

The only thing you need to know is that the limit exists, and lies in $[1,2]$. Once you know that, letting $L=\lim_{n\to\infty} a_n$, you have $L=1+\frac{L}{L+1}$, giving you the only solution in $[1,2]$ as $L=\frac{1+\sqrt{5}}{2}$.

To show the limit exists, simply observe that every term other than the first is in $[1,2]$, so $(a_n)$ is bounded, and since you have shown $(a_n)$ is monotone (i.e., your same argument shows that if $a_1>a_2$ we get $a_n>a_{n+1}$ for all $n$, so in fact $(a_n)$ is either strictly increasing, strictly decreasing, or constant), we see that it must converge.

$\endgroup$
1
  • $\begingroup$ Ah, so I could have just done both $a_{n} \geq a_{n+1}$ and $a_{n} < a_{n+1}$ and the result gives those similar inequalities which shows monotonicity and with the series being bounded; $(a_n)$ converges for $a_{1} > 1$. $\endgroup$
    – Derek Luna
    Commented Nov 12, 2023 at 19:43
0
$\begingroup$

For boundedness, $a_{1}>1 \implies 0 \leq \frac{a_n}{1+a_n} < 1 \implies 1 \leq 1+ \frac{a_n}{1+a_n} < 2$.

To show montonocity for $a_{1} > 1$,

Start from $a_{n} \geq a_{n+1}$ and $a_{n} < a_{n+1}$ where the base cases are $a_{1} \geq a_{2}$ and $a_{1} < a_{2}$, respectively.

The result gives inequalities that imply the montonicity of $(a_{n})$, and with the fact that $(a_n)$ is bounded by $[1,2]$, $(a_n)$ converges for all $a_{1} > 1.$

$\endgroup$
8
  • $\begingroup$ @MW It is still a massive assumption that all terms of $a_{n}$ are positive, no? $\endgroup$
    – Derek Luna
    Commented Oct 14, 2023 at 6:44
  • $\begingroup$ You are free to write your proofs and responses however you like, and for completeness sake I will do the same for mine. Thanks. $\endgroup$
    – Derek Luna
    Commented Oct 14, 2023 at 6:48
  • $\begingroup$ Fair enough, didn't mean to make this a testy exchange, I'll delete these comments in a sec. $\endgroup$
    – M W
    Commented Oct 14, 2023 at 6:49
  • $\begingroup$ However, it might be worth pointing out that it will not necessarily be true that $$\dfrac{a_{n+1}}{1+a_{n+1}} \geq \dfrac{a_{n}}{1+a_{n}}.$$ We will sometimes have the reverse inequality for all $n$, which is still fine. $\endgroup$
    – M W
    Commented Oct 14, 2023 at 7:06
  • $\begingroup$ @MW Are you referring to my part about not assuming $a_{n}$ is positive? I concede it is clear, but it can still be shown too before the next part using basic facts like you said (and/or induction), so I don't think one must bother with the reverse inequality. $\endgroup$
    – Derek Luna
    Commented Oct 14, 2023 at 7:23

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .