2
$\begingroup$

Let $a_1=\frac\pi4, a_n = \cos(a_{n-1})$

Prove $\displaystyle\lim_{n\to\infty}a_n=\alpha$.

Where $\alpha$ is the solution for $\cos x=x$.

Hint: check that $(a_n)$ is a cauchy sequence and use Lagrange's theorem.

Well I tried to show that it's a cauchy sequence: $|a_m-a_n|<\epsilon , \ \forall m,n$ but I just don't see how it's done with a trig recursion sequence...

$\endgroup$
  • $\begingroup$ This is a recent duplicate which I'm failing to find. Edit: It's not a duplicate after all, but it is the same problem. See here. $\endgroup$ – Git Gud Jan 6 '14 at 9:15
  • $\begingroup$ you should have made the question a little better.. may be you write $\displaystyle\lim_{n\to\infty}a_n=\alpha$ and in the next line you write "where $\alpha$ is solution of $\cos x =x$" $\endgroup$ – user87543 Jan 6 '14 at 9:16
3
$\begingroup$

Note that $$ \cos(x):\left[\frac1{\sqrt2},\frac\pi4\right]\mapsto\left[\frac1{\sqrt2},\frac\pi4\right] $$ and on $\left[\frac1{\sqrt2},\frac\pi4\right]$, $$ \left|\frac{\mathrm{d}}{\mathrm{d}x}\cos(x)\right|=|\sin(x)|\le\frac1{\sqrt2}\lt1 $$ Thus, the Mean Value Theorem ensures that $\cos(x)$ is a contraction mapping on $\left[\frac1{\sqrt2},\cos\left(\frac1{\sqrt2}\right)\right]$. The Contraction Mapping Theorem says that $\cos(x)$ has a unique fixed point in $\left[\frac1{\sqrt2},\cos\left(\frac1{\sqrt2}\right)\right]$. Furthermore, for $n\ge3$, $$ \left|a_n-a_{n-1}\right|\le\left(\frac1{\sqrt2}\right)^{n-3}\left|a_3-a_2\right| $$ Thus, $a_n$ is a Cauchy sequence. The point to which it converges must be the unique fixed point where $\cos(x)=x$.

$\endgroup$
  • $\begingroup$ Without using the contraction mapping theorem, how can you conclude that: $$ \left|a_n-a_{n-1}\right|\le\left(\frac1{\sqrt2}\right)^{n-3}\left|a_3-a_2\right| $$ ? $\endgroup$ – GinKin Jan 6 '14 at 17:06
  • $\begingroup$ @GinKin: That is gotten using the Mean Value Theorem: since $|\sin(x)|\le\frac1{\sqrt2}$ on $\left[\frac1{\sqrt2},\frac\pi4\right]$, we have that, for some $\alpha_{n-1}$ between $a_{n-1}$ and $a_{n-2}$, $$ \begin{align} \left|a_n-a_{n-1}\right| &=\left|\cos(a_{n-1})-\cos(a_{n-2})\right|\\ &=\left|\sin(\alpha_{n-1})(a_{n-1}-a_{n-2})\right|\\ &\le\frac1{\sqrt2}\left|a_{n-1}-a_{n-2}\right| \end{align} $$ $\endgroup$ – robjohn Jan 6 '14 at 19:52
3
$\begingroup$

Hints: exists $\;c\;$ between $\;a_n\;,\;a_m\;$ s.t.

$$|a_n-a_m|=|\cos(a_{n-1})-\cos(a_{m-1})|\stackrel{\text{MVT}}= |(\cos c)'||a_{n-1}-a_{m-1}|=q|a_{n-1}-a_{m-1}|$$

since $\;|(\cos(c))'|=|\sin(c)|\;$ cannot be $\;1\;$ (why?) , and now

$$|a_n-a_{n-1}|\le q|a_{n-1}-a_{n-2}|\le\ldots\le q^{n-1}|a_1-a_0|$$

$\endgroup$
  • $\begingroup$ It cannot be 1 because $c\in[\frac\pi 4 , \frac1{\sqrt2}]$ but why it becomes a constant... Why is it a telescopic sequence ? $\endgroup$ – GinKin Jan 6 '14 at 14:51
  • $\begingroup$ What becomes constant? You know $\;\sin c=q<1\;$ for any possible $\;c\in\left[\frac\pi4\,,\,\frac1{\sqrt2}\right]\;$ , andthat's all that matters. $\endgroup$ – DonAntonio Jan 6 '14 at 14:59
  • $\begingroup$ @DonAntonio, why is $c \in \left[ {\pi \over 4},{1 \over \sqrt 2} \right]$? $\endgroup$ – AnnieOK May 31 '14 at 8:56
  • $\begingroup$ @DonAntonio, Why is the MVT make it $\le$? shouldn't it be an equality? $\endgroup$ – AnnieOK May 31 '14 at 9:33
  • 1
    $\begingroup$ @AnnieOK, this is prehistory! Anyway, I can't remember what I was thinking of when I answered this, but you're right: under that MVT thing there must be an equality sign though this doesn't matter that much. Now, $$\frac1{\sqrt2}\le\cos\frac\pi4=\frac1{\sqrt2}<\frac\pi4$$ , so iterating cosine on these values leave us all the time between $\;\frac1{\sqrt2}\;$ and $\;\frac\pi4\;$ (remember: on $\;[0,\pi/2]\;,\;\;\cos x$ is motone descending) $\endgroup$ – DonAntonio May 31 '14 at 10:01
1
$\begingroup$

Look over the Banach Fixed Point Theorem. Its practically the same deal.

Hint: Show with MVT that: $$ | \cos(a_{n}) - \cos(a_{n-1})| <= q^n|a_1 - a_0|$$ Where $q$ is a constant so $ 0 < q < 1 $. Use that in the Cauchy proof.

$\endgroup$
  • $\begingroup$ It's exactly the same proof as the BFPT's though I doubt it whether it'd be allowed to use that theorem in Calculus I. Yet the MVT allows to prove the sequence is Cauchy and that's almost all is needed. $\endgroup$ – DonAntonio Jan 6 '14 at 11:29
1
$\begingroup$

Over the interval $[0,1]$, the function $f(x) = \cos(x)$ is a strictly decreasing continuous function mapping $[0,1]$ into itself. So $f\circ f$ is strictly increasing there. If one pick $a_1 = \frac{\pi}{4} \in [0,1]$ and generate a sequence by iteration $a_n = f(a_{n-1})$, it is not hard to check

$$a_2 < a_3 < a_1$$

Since $f$ is strictly decreasing, $a_2 < a_3 < a_1 \implies a_3 > a_4 > a_2$ and hence

$$a_2 < a_4 < a_3 < a_1$$

Since $f \circ f$ is strictly increasing, this implies the even sub-sequence $a_{2n}$ is strictly increasing and the odd sub-sequence $a_{2n-1}$ is strictly decreasing. Since all $a_k$ belongs to $[0,1]$, the even and odd sub-sequences are both bounded monotonic sequences. As a result, following two limits exists:

$$a_{even} = \lim_{n\to\infty} a_{2n}\quad\text{ and }\quad a_{odd} = \lim_{n\to\infty} a_{2n-1}$$ Since $f\circ f$ is continuous, both $a_{even}$ and $a_{odd}$ are fixed points for $f \circ f$. If one make a plot of $f\circ f$ over the interval $[0,1]$, one will notice $f \circ f$ has a unique fixed point in $[0,1]$. That point is also the unique fixed point of $f$. This forces $a_{even} = a_{odd}$ and the sequence $a_n$ converges to the unique fixed point of $f$.

$\endgroup$

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.