0

I have a question and hope for some help, because I not so sure somehow..

QUESTION:

We have a sequence defined by $a_0= 1$ and $a_{n+1}=\sqrt{2+a_n}$ for $n\in \mathbb{N}$. Show with induction that, for all $n\in \mathbb{N}$, $0\leq a_n\leq 2$ and $a_n\leq a_{n+1}$. Decide if the sequence is convergent and show its limit.

SOLUTION:

So, at first I start with induction, do I need to start with $0$ or $1$? of course both if them pass for it. by $n=0$ we have $0\leq a_0=1 \leq 2$ and for $n=1$ we have $0\leq a_2=\sqrt{3} \leq 2$ .

Then for $n+1$ we have $0\leq a_n \leq 2\implies2\leq2+a_n\leq4\implies\sqrt2\leq\sqrt{2+a_n}\leq2\implies\sqrt2\leq a_{n+1}\leq2)$ right? So the first part is done, I guess.

Next part is $a_{n+1}-a_n=\sqrt{2+a_n}-a_n=\frac{(\sqrt{2+a_n}-a_n){(\sqrt{2+a_n}+a_n)}}{{(\sqrt{2+a_n}+a_n)}}=\frac{\left(\sqrt{2+a_n}\right)^2-a_n^2}{\sqrt{2+a_n}+a_n}$

and ${a_{n+1}-a_n}=\frac{(2+a_n)-a_n^2}{\sqrt{2+a_n}+a_n}=\frac{(1+a_n)(2-a_n)}{\sqrt{2+a_n}+a_n}>0$ everything is positive because of $\sqrt2\leq a_n\leq2$ and $a_{n+1}-a_n>0 \implies a_{n+1}>a_n$.

And the last part would be convergence: Our sequence is monotonic increasing and bounded so there is a limit. $a=\sqrt{2+a}\implies a^2=2+a\implies a^2-a-2=0\implies (a-2)(a+1)=0\implies a=2$ Another root $a=-1$ is wrong because for all $a_n\ge\sqrt2$.

Is my solution okay? Thanks in advance : )

Semiclassical
  • 15,842
Ann
  • 29
  • 1
    It suffices that the initial conditon is satisfied for $n=0.$ There is no need to calculate $a_1.$ The monotonicity and $a_n<2$ can be proved simultaneously from the formula $2-a_{n+1}={2-a_n\over 2+\sqrt{2+a_n}}$ – Ryszard Szwarc Jul 12 '23 at 17:28
  • 1
    https://math.stackexchange.com/tags/solution-verification/info Please point precisely where you have some doubts. This site is not a proof-checker. – Anne Bauval Jul 12 '23 at 17:35
  • By sheer coincidence, there happens to be a closed-form solution to this particular recurrence relation: $a_n = 2 \cos \left(\frac{\pi}{3 \cdot 2^n}\right)$. (Though it's definitely not necessary to know that to solve this problem.) – Daniel Schepler Jul 12 '23 at 17:42
  • @RyszardSzwarc thank you, good point! – Ann Jul 12 '23 at 17:46
  • The monotonicity and $a_n<2$ and the convergence can be proved simultaneously form Ryszard's formula (or from the simpler one $4-a_{n+1}^2=\frac{4-a_n^2}{2+a_n}$). – Anne Bauval Jul 12 '23 at 17:49
  • @AnneBauval why can we use $4-a_{n+1}^2=\frac{4-a_n^2}{2+a_n}$ , we can just take anything bigger then $2$ also? – Ann Jul 12 '23 at 17:54
  • By induction, we derive from this formula that $4-a_n^2$ has the same sign as $4-a_0^2.$ (Not sure this is what you are asking.) – Anne Bauval Jul 12 '23 at 17:55

2 Answers2

1

$a_0 = 1, a_1 = \sqrt{3}.$

So, $~a_1 > 0.$

Inductively assume that $~a_n > 0.$
This implies that $~(2 + a_n) > 0 \implies a_{n+1} = \sqrt{2 + a_n} > 0.$


$~a_1 < 2.$

Inductively assume that $~a_n < 2.$
This implies that $~(2 + a_n) < 4 \implies a_{n+1} = \sqrt{2 + a_n} < 2.$


$a_1 > a_0 > 0.$

Inductively assume that $a_{n+1} > a_n > 0.$
This implies that $~(2 + a_{n+1}) > (2 + a_n) > 0 \implies $
$\sqrt{2 + a_{n+1}} > \sqrt{2 + {a_n}} > 0 \implies $ $a_{n+2} > a_{n+1} > 0.$


Therefore, the sequence is strictly positive, strictly increasing, and bounded above by $~2.$

Therefore, the sequence must be convergent.

Let $~L~$ denote the limit of this sequence.

So, $a_n$ goes to $~L,~$ as does $~a_{n+1} = \sqrt{2 + a_n}.$

Therefore,

$$L = \sqrt{2 + L} \implies L^2 = L + 2 \implies (L-2)(L+1) = 0.$$

Since the limit, $~L,~$ must be positive, you must have that $~L = 2.$


Addendum

My answer begs the following question.

Given that the sequence must be convergent, with some unique limit $~L,~$ and given that this implies that $~a_n \to L,~$ and $~a_{n+1} = \sqrt{2 + a_n} \to L,~$ can you then assert that $~L = \sqrt{2 + L}?$

In my opinion, it depends on whether the instructor/class has presented on point worked examples or a Real Analysis theorem, that justifies this step.

Presumably, you know that the function $~f(x) = \sqrt{x}~$ is a continuous function, and that the composition of two continuous functions is a continuous function. Therefore, you know that the function $~g(x) = \sqrt{2 + x}~$ is a continous function.

This implies that as $~x \to L,~$ that $~g(x) \to g(L).~$

So, as $~a_n~$ goes to $~L,~ g(a_n)~$ goes to $~g(L).$ Further, $~g(a_n) = a_{n+1},~$ which (also) goes to $~L.~$

So, you can conclude that

  • $~g(a_n) = a_{n+1}~$ goes to $~L.$

  • $~g(a_n) = \sqrt{2 + a_n}~$ goes to $~g(L) = \sqrt{2 + L}.$

Therefore, $~L = g(L) = \sqrt{2 + L}.$

user2661923
  • 35,619
  • 3
  • 17
  • 39
  • 1
    @AnneBauval The point of my response was to indicate to the OP (i.e. original poster) where they were over-complicating, and to show what the problem composer's intent was. From the OP's perspective, this is clearly better than referring the OP to a different article, where they have to dig out the alternative approaches. Further, the OP definitely showed work, so is justified in expecting a direct response. Given the over-complications, it is reasonable to expect the OP to compare their work with mine, and then ask direct questions, if any. – user2661923 Jul 12 '23 at 17:56
  • 1
    thank you! you are right, I overdid it I guess.. So it is acceptable if I just show it as you did, without going too deep? That's great! I was struggling over it because the question is actually not so complex but I was not really sure how to show it in a good way. Your comment helped a lot! – Ann Jul 12 '23 at 17:59
  • @Ann The question in your last comment reduces to asking whether any of my response (i.e. answer) is invalid (as opposed to inaccurate). If it is all valid, then the teacher has to accept it. – user2661923 Jul 12 '23 at 18:02
  • No, I did not mean that it is invalid in your case, but it often happens by us that we do not get full amount of points if the prof does not like the art of solution. That is the reason why I posted this question here, we had nothing similar in our course so I have no reference how this kind of exercise should be solved ( I mean style) (excuse me for my bad English). – Ann Jul 12 '23 at 18:07
  • @Ann Your teacher can not reject work that is both on point and valid. To the best of my knowledge, all of the work in the answer is valid. Therefore, it is irrelevant whether the teacher likes the response or not. – user2661923 Jul 12 '23 at 18:09
  • okay! Got it, thank you : ) – Ann Jul 12 '23 at 18:11
  • @Ann The one questionable area is my assertion that since $~a_{n}~$ and $~a_{n+1}~$ both converge to $~L,~$ that $~L = \sqrt{2 + L}.$ Prior to this problem, you should have been presented with worked examples and/or a Real Analysis Theorem that justifies this assertion. – user2661923 Jul 12 '23 at 18:12
  • You are right hmm, I think I should think about it. – Ann Jul 12 '23 at 18:27
  • I can just write that it is because limit of sub sequence of convergent sequence is same as limit of sequence. Basically this is the reason why we can assume it and define the answer. – Ann Jul 12 '23 at 18:36
  • Are you sure that $a_n>0$ requires a proof, especially by induction ? – Ryszard Szwarc Jul 12 '23 at 18:38
  • @RyszardSzwarc My answer used the conclusion that $a_n > 0$ (for all $n$) to justify the analysis that $$a_{n+1} > a_n \implies (2 + a_{n+1}) > (2 + a_n) \implies \sqrt{2 + a_{n+1}} > \sqrt{2 + a_n}.$$ For counterexample, $~-3 > -5, ~$ but $~\sqrt{-3 +2} < \sqrt{-5 + 2}.$ Using induction to show that the series is strictly positive seemed very easy. – user2661923 Jul 12 '23 at 18:58
  • @Ann See the Addendum that I have just added to my answer. – user2661923 Jul 12 '23 at 18:59
  • The formula $a_{n+1}=\sqrt{a_n+2}$ cannot produce negative values. Concerning your comment: the inequality $\sqrt{-3+2}<\sqrt{-5+2}$ does not make sens. – Ryszard Szwarc Jul 12 '23 at 19:03
  • @RyszardSzwarc Circular reasoning. You are assuming that $~a_n~$ is positive. Superficially, you have know way of knowing whether $~\sqrt{a_n + 2}~$ will be well defined, for all $~n.~$ It is only by demonstrating that no $~a_n~$ is less than $~-2,~$ can you assert that $~a_n~$ is well defined, for all $~n.~$ To an experienced Math student, I agree that the result can be construed as immediate. However, for the new student, the assertion that $~\sqrt{a_n + 2}~$ will always be well defined requires thought. ...see next comment – user2661923 Jul 12 '23 at 19:08
  • @RyszardSzwarc Further, the way that the problem is presented, the (new) student is expected to prove the conclusion that each $~a_n~$ is non-negative. What then is the intent of the problem composer? I would argue that the problem composer does not intend that since $a_0 > 0,~$ that no value $~\sqrt{a_n + 2},~$ can produce a negative number. This is true, but asking a new student to prove this seems iffy. I suspect that the approach that I took was intended by the problem composer, for the new student. – user2661923 Jul 12 '23 at 19:10
0

Prove: $a_0=1. 0\le a_n\le 2. \ a_n\le a_{n+1}$ when $a_{n+1}=\sqrt{2+a_n}$.

$\sqrt{2+1}=1.7>1.0$ check!

$\sqrt{2+\sqrt{2+u}}>\sqrt{2+u}$?

$2+\sqrt{2+u}>2+u \iff \sqrt{2+u}>u$

$a_{n+1}>a_n. $QED-I.

Let $u=2-\epsilon, 0<\epsilon<1$.

$ \sqrt{2+u}=\sqrt{4-\epsilon}\le 2. $ QED-II

Monotonic and bounded implies convergence.

Suppose $f(x)=x$ and $|f'(x)|<1$ $|f(x+\delta)-f(x)|=|f(x+\delta)-x|<\delta |f'(x)|<\delta$

So $f(x+\delta)$ is closer to $x$ than $x+\delta$ was.

$f(x)=\sqrt{2+x}\implies f'(x)=\frac{1}{2\sqrt{2+x}}<1$ for $0<x<2$.

This suggests the limit is a positive root of $x=\sqrt{2+x}$. So $x =2$.

TurlocTheRed
  • 5,683