35
$\begingroup$

If I want to prove that the function$\ f(x)=\sin(x^2)$ is not uniformly continuous on the domain $\mathbb{R}$, I need to show that:

$\exists\epsilon>0$ $\forall\delta>0$ $\exists{x,y}\in\mathbb{R}\ : |x-y|<\delta$ and $|\sin(x^2) - \sin(y^2)|\geq\epsilon$.

So let's take $\epsilon = 1$. Then I want $|\sin(x^2)-\sin(y^2)|\ge1$. That's the case if $\sin(x^2)$=0 and $\sin(y^2)=\pm1$. Thus $x^2=n\pi$ and $y^2=n\pi + \frac{1}{2}\pi$. Now I'm stuck on expressing x and y, which I want to express in $\delta$, to ensure that $|x-y|<\delta$.

Thanks in advance for any help.

$\endgroup$
47
$\begingroup$

You have chosen $x^2=n\pi$ and $y^2 = n\pi+\frac{\pi}{2}$, so you can take $x=\sqrt{n \pi}$ and $y=\sqrt{n \pi + \frac{\pi}{2}}$.

Then,

$|x-y|=\sqrt{n \pi + \frac{\pi}{2}}-\sqrt{n \pi}=\frac{n\pi + \frac{\pi}{2}-n\pi}{\sqrt{n \pi + \frac{\pi}{2}}+\sqrt{n \pi}}=\frac{\frac{\pi}{2}}{\sqrt{n \pi + \frac{\pi}{2}}+\sqrt{n \pi}}<\frac{2}{2\sqrt{n \pi}}<\frac{1}{\sqrt{n}}$

If $n > \frac{1}{\delta^2}$ then $|x-y|<\delta$ but $|f(x)-f(y)|\geq 1$. The values $x,y$ are very close but $f(x)$ and $f(y)$ are far apart. Intuitively, for $\epsilon = 1$ there is no $\delta$ that allows you to know $f(x)$ within precision $\epsilon$ if you know $x$ within precision $\delta$. The oscillations in $\sin(x^2)$ get faster and faster, on arbitrarily small intervals the function changes its value from 0 to 1.

Note that if you change the function to $\sin x$ the proof will fail, because taking $x=n\pi$ and $y=n\pi+\frac{\pi}{2}$ does not make $|x-y|<\delta$ for small $\delta$ (indeed, $y-x$ is constant). In fact, $\sin x$ is uniformly continuous.

$\endgroup$
  • 2
    $\begingroup$ +1 nice answer, Is $f(x) = sinx^2$ is uniformly continous on (0,1) ? $\endgroup$ – user525416 May 12 '18 at 9:58
  • $\begingroup$ Thank you! That's a very nice proof :) $\endgroup$ – Hendrra Nov 3 '18 at 21:55
  • $\begingroup$ How to know that it is less than $1/√(nπ)$ and it is less than $1/√n$ $\endgroup$ – Pranita Gupta Jan 30 '19 at 16:07
  • 1
    $\begingroup$ @PranitaGupta If you are asking how I got the penultimate inequality, I used $\sqrt{n \pi + \pi/2} + \sqrt{n \pi} > 2\sqrt{n \pi}$ and $\pi/2 < 2$. $\endgroup$ – sdcvvc Feb 4 '19 at 10:27
1
$\begingroup$

Hint: If $x^2=n\pi$ and $\delta>0$, note that $(x+\delta)^2>x^2+2\delta x$. To make this $>n\pi +\frac12\pi$, it suffices to have $2\delta x > \frac12 \pi$. This gives you a condition on $n$.

$\endgroup$
  • $\begingroup$ Why would I want $(x+\delta)^2 \gt n\pi+\frac{1}{2}\pi$ ? I'm still not feeling it. $\endgroup$ – Jeroen Oct 22 '12 at 22:27
  • $\begingroup$ +1 nice answer, Is $f(x) = sinx^2$ is uniformly continous on (0,1) ? $\endgroup$ – user525416 May 12 '18 at 9:59
-7
$\begingroup$

if f(x)=sin(x^2) then derivative of f(x) is not bounded on real line, but it is bounded on any bounded subset of R, so f(x) is not uniformly continuous on the domain R but uniformly continuous on any bounded subset of R.

$\endgroup$
  • 1
    $\begingroup$ This even gives a reason that $f$ is Lipschitz on each bounded subset of $\mathbb R$. For uniform continuity on bounded intervals, it suffices that the function is continuous. This doesn't answer the question. $\endgroup$ – Jonas Meyer Sep 17 '13 at 13:41

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.