One time I was bored and played around a bit with integrals and wolfram alpha and tested the following integral:


Note: The result at my laptop shows: $ \int_0^1 \lceil { x\sin({1 \over x})} \rceil \,dx = 1 - \frac{\log(4)}{2\pi} $

I was a bit surprised seeing this and wondered, if this result makes any sense and if yes, if there were an explanation, where $\frac{\log(4)}{2\pi}$ exactly comes from. The numbers look too neat to be random, so there might be a way to derive those values directly without using the help of a calculator etc.

As always: Thanks in advance for any constructive answer/comment.

  • 1
    $\begingroup$ I think the multiplicative factor of $x$ is redundant (sorry you'll have to paste the link): wolframalpha.com/inpu/?i=integral_0^1+ceil%28sin%28%2F%29%29 $\endgroup$ – Colm Bhandal Jul 29 '15 at 12:24
  • $\begingroup$ yeah, the factor is redundant, however somehow I never got an precise value for the integral.. (a friend of mine also got weird result for this integral and other values (which CAN'T be correct, guess wolfram alpha is a bit buggy in this regard) $\endgroup$ – Imago Jul 29 '15 at 12:27
  • $\begingroup$ It would be interesting to try the multiplicative factor for different intervals. In general I don't think it would be redundant. $\endgroup$ – Colm Bhandal Jul 29 '15 at 12:33
  • $\begingroup$ The one I don't understand, why is this is question so well received? Am I overlooking something? $\endgroup$ – Imago Jul 29 '15 at 17:40
  • 2
    $\begingroup$ @Imago First you try something that puzzles you and does not fit your understanding of the universe, then you start to think why, get intuition for a hidden law of the nature and search for a possible explanation - this is very scientific. Curiosity is a positive thing. $\endgroup$ – A.Γ. Jul 29 '15 at 18:02

The ceiling function gives one iff $\sin(1/x)> 0$, otherwise zero. So the integral is just a sum of interval lengths where the sin is positive, i.e. $$ \left(1-\frac{1}{\pi}\right)+\left(\frac{1}{2\pi}-\frac{1}{3\pi}\right)+\left(\frac{1}{4\pi}-\frac{1}{5\pi}\right)+\ldots=1-\frac{1}{\pi}\left(1-\frac12+\frac13-\frac14+\ldots\right)=\\ =1-\frac{1}{\pi}\ln(1+1)=1-\frac{\ln(2)}{\pi}. $$

| cite | improve this answer | |
  • $\begingroup$ It looks like your inequality should be strict. $\;$ $\endgroup$ – user57159 Jul 29 '15 at 16:48
  • $\begingroup$ @RickyDemer Yes, thanks. $\endgroup$ – A.Γ. Jul 29 '15 at 17:08

Note that $x \sin(1/x) \in [-1, 1]$ when $x \in [0,1]$ so your integral is the volume of the set on which $x \sin(1/x) > 0$. This set is

$$\left(\frac{1}{\pi}, 1\right) \cup \bigcup_{k=1}^{\infty}\left(\frac{1}{\pi (2k+1)}, \frac{1}{2\pi k}\right)$$ and its volume is $$1 - \frac{1}{\pi} + \sum_{k=1}^{\infty}\left(\frac{1}{2\pi k} - \frac{1}{\pi (2k+1)}\right)=1+\sum_{k=1}^{\infty} \frac{(-1)^k}{\pi k} = 1 - \frac{\log(2)}{\pi}.$$

| cite | improve this answer | |
  • $\begingroup$ Wait... $x\sin(x) \in [0,1]$ since both x and sin are in $[0,1]$ on $x\in [0,1]$, maybe you mean $x\sin\left(\frac{1}{x}\right) \in [-1,1]$, or am I confused? $\endgroup$ – mathreadler Jul 29 '15 at 17:38
  • $\begingroup$ @mathreadler. Thanks for pointing that out. That was a persistent typo. Fixed. $\endgroup$ – WimC Jul 29 '15 at 21:48

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.