Consider the recursively defined sequence $a_n = 1$

$$a_{n+1} = 3 - \frac{1}{a_n} \text{ for n > 1}$$

Is the sequence convergent?

This is my attempt:

First, we prove that the sequence is positive, and monotonically increasing, using induction.

$\textbf{Base Case:}$ $a_1 = 1$ and $a_2 = 2$. $a_1 < a_2$ and $a_1,a_2 > 0$.

$\textbf{Inductive Hypothesis:}$ $a_n < a_{n+1}$ where $a_n, a_{n+1}$.

$\textbf{Inductive Step:}$ We prove that $a_n < a_{n+1} \Rightarrow a_{n+1} < a_{n+2}$.

By our induction hypothesis:

$$a_n < a_{n+1}$$

$$-a_n > -a_{n+1}$$

$$-\frac{1}{a_n} < -\frac{1}{a_{n+1}}$$ (True by our IH since $a_n, a_{n+1} > 0$ and thus, $-a_n, -a_{n+1}$ share the same sign).

$$3-\frac{1}{a_n} < 3-\frac{1}{a_{n+1}}$$

$$a_{n+1} < a_{n+2}$$.

Note that $a_{n+1} > 0$ by our IH, so $a_{n+2} > 0$.

Now we prove that the sequence is bounded.

Observe that $a_n$ is monotonically increasing, which means that $ - \frac{1}{a_n}$ is monotonically increasing as well and it is upped bounded by $0$. Thus, $3-\frac{1}{a_n}$ is upper bounded by $3$.

We have a sequence that is monotone and bounded. Hence, by the Monotone Convergence Theorem, This sequence converges.

I was wondering if this method is correct.

  • $\begingroup$ FWIW, since neither of the answers actually says this: yes, your method is entirely correct (modulo the assertion that $a_n\gt 0$ for all $n$, which is straightforward to prove from what you've already got). $\endgroup$ – Steven Stadnicki Feb 7 '18 at 20:14
  • $\begingroup$ @StevenStadnicki that's not accurate, Michael points to the fact that $a_n>0$ is implied, but should be proved first. I presented a version where this is obvious. $\endgroup$ – rtybase Feb 10 '18 at 13:40

A shorter approach, look at the function $f(x)=3-\frac{1}{x}$ because this function "generates" the sequence, i.e. $f(a_n)=a_{n+1}$. This function is ascending, because $f'(x)=\frac{1}{x^2}$ and $\frac{1}{3}<a_1<a_2 \Rightarrow 0< \color{red}{f(a_1)\leq f(a_2)} \Rightarrow \frac{1}{3}<a_1<\color{red}{a_2 \leq a_3}$ and by induction $\frac{1}{3}< a_{n} \leq a_{n+1}$. So, the sequence is ascending and positive.

Given $a_n > \frac{1}{3} > 0, \forall n$, then $\frac{1}{a_n}>0 \Rightarrow a_{n+1}=3-\frac{1}{a_n} < 3 \Rightarrow 0<a_{n+1} < 3, \forall n$. So the sequence in bounded.


In your proof of $a_{n+1}>a_n$ you used that $a_n>0$, but it is not proven.

I like the following reasoning. $$a_{n+1}-\frac{3-\sqrt5}{2}=3-\frac{3-\sqrt5}{2}-\frac{1}{a_n}=\frac{3+\sqrt{5}}{2}-\frac{1}{a_n}=\frac{1}{\frac{3-\sqrt5}{2}}-\frac{1}{a_n}>0$$ by induction because $a_1=1>\frac{3-\sqrt5}{2}.$ $$a_{n+1}-\frac{3+\sqrt5}{2}=3-\frac{3+\sqrt5}{2}-\frac{1}{a_n}=\frac{3-\sqrt{5}}{2}-\frac{1}{a_n}=\frac{1}{\frac{3+\sqrt5}{2}}-\frac{1}{a_n}<0$$ by induction because $a_1=1<\frac{3+\sqrt5}{2}.$

Thus, for all natural $n$ we got: $$\frac{3-\sqrt5}{2}<a_n<\frac{3+\sqrt5}{2}.$$ In another hand, $$a_{n+1}-a_n=3-a_n-\frac{1}{a_n}=\frac{\left(a_n-\frac{3-\sqrt5}{2}\right)\left(\frac{3+\sqrt5}{2}-a_n\right)}{a_n}>0,$$ which says that $a$ converges.

  • $\begingroup$ If $a_n=\frac32$, then it doesn't converge by your inequality. $\endgroup$ – Takahiro Waki Feb 10 '18 at 14:39
  • $\begingroup$ I proved that it converges to $\frac{3+\sqrt5}{2}.$ $\endgroup$ – Michael Rozenberg Feb 10 '18 at 15:55
  • $\begingroup$ If you say there is such a proof in this answer, you are not a mathematical graduate student. Also this question can be proved by even high school math. $\endgroup$ – Takahiro Waki Feb 10 '18 at 16:06
  • $\begingroup$ Yes, of course. My proof is very easy and it's a not hard problem. $\endgroup$ – Michael Rozenberg Feb 10 '18 at 16:11

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.