Should monotonic sequence theorem be used? Thanks in advance.
-
1Yes, that's a good idea. – José Carlos Santos Nov 26 '17 at 14:54
-
If they just ask for the value of the limit under the assumption there's one, you just have to use the continuity of the function which defines the recurrence relation. – Bernard Nov 26 '17 at 14:54
-
Prove the monotony by induction. – Nov 26 '17 at 15:12
3 Answers
First, it's trivial by Mathematical Induction that $a_n \ge 0$.
Assuming the limit exists, you have
$$x=\frac{x}{2}+\frac{2}{x}$$ $$\implies x=\pm 2$$
Therefore, $$\lim_{n\to\infty}a_n=2$$
- 9,961
Note: For $x,$ real, positive:
$2/x +x/2 \ge 2\sqrt{(2/x)(x/2)} =2$. (AM-GM)
$a_{n+1} -a_n = -a_n/2 +2/a_n = \dfrac {-a_n^2 +4}{2a_n}.$
$a_n > 0$, $n \in \mathbb{Z+}$, and
$a_n \ge 2$ for $n \gt 1.$
Hence $a_{n+1} - a_n \le 0.$
$a_n$ is decreasing and has a lower bound $2$, implies convergent.
Can you find the limit?
Hint: L= L/2 + 2/L , why?
- 20,344
- 2
- 17
- 28
Consider the function
$F(x)=x^k-c$, where $k>1$ is a positive integer and $c>0$
$F$ is differentiable and $F'(x)=kx^{k-1}$ $\forall$ $x$ $\in$ $\mathbb{R}$
We can prove that the Newton-Raphson method converges to its positive soltution given a starting point $a_0>0$
Then, if we take the sequence defined by $a_1=1$ and $a_{n+1}=a_n-\frac{F(a_n)}{F'(a_n)}$ converges to $\sqrt[k]{c}$
Substituting $F$ and $F'$ with their polynomial expressions, we can obtain $a_{n+1}=a_{n}-\frac{a_{n}^k-c}{ka_{n}^{k-1}}=(1-\frac{1}{k})a_{n}+\frac{c}{ka_{n}^{k-1}}$
Now take $k=2$ and $c=4$ and we'll get our original sequence.
We can then conclude that the sequence is convergent to $\sqrt{4}=2$