# Different ways of computing $\sqrt{1+\sqrt{1+\sqrt{1+\cdots}}}$ [duplicate]

Possible Duplicate:
On the sequence $x_{n+1} = \sqrt{c+x_n}$

I am wondering how many different solutions one can get to the following question:

Calculate $\sqrt{1+\sqrt{1+\sqrt{1+\cdots}}}$

-

## marked as duplicate by J. M., Jyrki Lahtonen♦, Byron Schmuland, Pedro Tamaroff♦, Grumpy ParsnipJun 9 '12 at 18:02

If $$\phi = \sqrt{1+\sqrt{1+...}}$$ then $$\phi^2= 1+\sqrt{1+\sqrt{1+...}}=1+\phi$$ hence $$\phi^2-\phi-1=0$$ $$\phi=\frac{1\pm\sqrt{5}}{2}$$

EDIT: the sequence $\phi_n=\sqrt{1+\phi_{n-1}}$ is apparently increasing. $\phi_1<2$. Assume $\phi_{k-1}<2$, $k>2$. Then $1+\phi_{k-1}<3\implies \phi_k=\sqrt{1+\phi_{k-1}}<\sqrt{3}<2$, hence the sequence is bounded.

-
the second answer obviously arises when squaring – Valentin Jun 9 '12 at 16:54
It isn't a "wrong answer"...just isn't completely rigorous. It shows that IF a limit exists then it must be the golden ratio (agreed on the presence of two values though). – fretty Jun 9 '12 at 16:56
You got $\phi^2 - \phi + 1 = 0$ as your quadratic when it should be $\phi^2 -\phi - 1 = 0$. Note that this makes the 1 positive in the expression for $\phi$ – Carl Jun 9 '12 at 17:05
@Carl, thanks, corrected. Also added a note on convergence. This is not supposed to be a discussion forum, but i have to add: if asked for my "favourite" proof of say $\sum\frac{1}{n^2}=\frac{\pi^2}{6}$ i would quote Euler's argument expanding sine in a series and a product and comparing coefficients. Although that is not exactly rigorous. Cheers – Valentin Jun 9 '12 at 17:10

Consider the sequence $$x_n = \sqrt{1 + \sqrt{1 + \sqrt{1 + \cdots \text{ n times}}}}$$ i.e. $$x_{n+1} = \sqrt{1+x_n}.$$ For instance, $x_1 = \sqrt{1}$, $x_2 = \sqrt{1+\sqrt{1}}$, $x_3 = \sqrt{1+\sqrt{1+\sqrt{1}}}$.

We will prove that the sequence is monotonically increasing and is bounded above by $\phi$ where $\phi = \dfrac{1+\sqrt{5}}{2}$ satisfies $\phi^2 = 1 + \phi$ and $\phi > 0$.

Boundedness

First we will prove that the sequence is bounded above by $\phi$. The proof follows from induction.

For $n=1$, $x_n = 1 = \dfrac{1+1}{2} < \dfrac{1+\sqrt{5}}{2} = \phi$. Hence, the claim is true for $n=1$.

Now assume it is true for some $k \in \mathbb{N}$.

$x_k < \phi \implies 1+x_k < 1 + \phi = \phi^2$. (Since $\phi = \dfrac{1+\sqrt{5}}{2}$ satisfies $\phi^2 = 1 + \phi$).

Hence, $x_{k+1} = \sqrt{1+x_k} < \phi$ and thereby the sequence is bounded above.

By a similar argument, we also have that $x_n$ is bounded below by $1$. Hence, we now have that $$1 \leq x_n < \phi.$$

Monotonically increasing

We will now prove that the sequence is monotonically increasing i.e. $x_n < x_{n+1}$.

Consider $x_{n+1}^2 - x_n^2$. We have $x_{n+1}^2 - x_n^2 = (1+x_n) - x_n^2 = \dfrac54 - \left( x_n - \dfrac12\right)^2$.

The quadratic equation $f(x) = 1 + x - x^2$ is positive whenever $x \in \left( \dfrac{1-\sqrt{5}}{2}, \dfrac{1+\sqrt{5}}{2}\right)$.

Since we proved earlier that $x_n \in \left[1,\dfrac{1+\sqrt{5}}{2}\right)$, we have that $x_{n+1}^2 - x_n^2 = 1 + x_n - x_n^2 > 0$.

Since $x_n$'s are positive, we have that $$x_{n+1} > x_n.$$

Hence, now combining both we get that $$1 \leq x_n < x_{n+1} < \phi$$ Hence, by monotone sequence theorem, (or equivalently by completeness of $\mathbb{R}$), the sequence $\{x_n\}_{n=1}^{\infty}$ converges.

Once we know that the sequence converges, we can then make use of the following limit rules.

1. If $y_n \rightarrow L$, then $y_{n+1} \rightarrow L$.
2. If $y_n \rightarrow L$, then $1+y_n \rightarrow 1+L$.
3. If $z_n \rightarrow L$, and $z_n > 0$, then $L \geq 0$.
4. If $z_n \rightarrow L$, and $z_n > 0$, then $\sqrt{z_n} \rightarrow \sqrt{L}$.

Let us now assume that $x_n \rightarrow L$, then making use of the above limit rules, we get that $$L = \sqrt{1+L}$$ Since $L \geq 0$ (because $x_n > 0)$, solving the quadratic we get that $$L = \dfrac{1+\sqrt{5}}2$$

-
I'm working on a similar, yet slightly more abstract sequence, and was wondering when you were proving the boundedness why did you have to select $\phi = \dfrac{1+\sqrt{5}}{2}$? Or could you have chosen it to be something else as well? – user66807 Mar 17 '13 at 5:45
@user66807 The choice of $\phi = \dfrac{1+\sqrt5}2$ is because $\phi$ satisfies $\phi = \sqrt{1+\phi}$. – user17762 Mar 17 '13 at 6:55

Assuming convergence, here is one:
$$S=\sqrt{1+\sqrt{1+...}}=\sqrt{1+S}\implies S^2=1+S \implies S=\frac{1+\sqrt{5}}{2}$$

-