9

How can I find the following limit without the use of any series or expansion? $$ \lim_{x \to 0} \frac{x^2}{x+\sin (\frac 1 x)} $$ Thanks for help.

mnsh
  • 5,875

6 Answers6

12

As bob.sacramento suggests, let $y=1/x$ and then $$ \begin{align} \lim_{x\to0}\frac{x^2}{x+\sin(1/x)} &=\lim_{y\to\infty}\frac1{y+y^2\sin(y)} \end{align} $$ There are solutions near every non-zero integer multiple of $\pi$ for $y+y^2\sin(y)=0$. This can be easily verified since $y+y^2\sin(y)\gt0$ when $y=(2k+1/2)\pi$ and $y+y^2\sin(y)\lt0$ when $y=(2k-1/2)\pi$ for $k\ne0$.

For $y=k\pi$, $y+y^2\sin(y)=y$.

Thus, as $y\to\infty$, $\dfrac1{y+y^2\sin(y)}$ gets arbitrarily big and small. Thus, the limit does not exist.


Graphical Support

Here is a plot of $y+y^2\sin(y)$. It is easy to see why its reciprocal has no limit as $y\to\pm\infty$.

$\hspace{3cm}$enter image description here

robjohn
  • 345,667
11

Let $y=1/x$ and take the limit as $y\rightarrow \infty$.

5
  • Note that $x-1 \leq x+ \sin\frac{1}{x} \leq x+1$ and this says that $x^{2}$ increases much much faster than $x+\sin\frac{1}{x}$. So where should the limit go?
  • Very good hint. +1 – DonAntonio May 02 '13 at 17:07
  • for some strange reason, W|A gives the weird result +1 – S L May 02 '13 at 17:08
  • @DonAntonio Well, Thank you! – Seshadri May 02 '13 at 17:08
  • Good hint, except for the word "increases," since $x\to 0$. – Thomas Andrews May 02 '13 at 17:10
  • @ThomasAndrews what word you sugges then? grows ? – Seshadri May 02 '13 at 17:11
  • It doesn't grow faster as $x\to 0$, either. – Thomas Andrews May 02 '13 at 17:13
  • 5
    Actually, I'm not sure this hint leads in the right direction. There are values of $x$ arbitrarily close to zero where $x+\sin 1/x=0$. So I think there might not be a limit (even if we exclude those points.) – Thomas Andrews May 02 '13 at 17:14
  • I wouldn't believe WA too much, but in this case I get that normal calculation time exceeded... – DonAntonio May 02 '13 at 17:15
  • if not Limit[f(x), x->0] works then plotting gives the general idea. I always use it to verify answer. – S L May 02 '13 at 17:22
  • I think I can see what the problem may be in spite of what seemed a good hint...yet it doesn't work (still my +1 shall remain). Apparently Seshadri hinted to the squeeze theorem: $$\frac{x^2}{x+1}\le\frac{x^2}{x+\sin1/x}\le\frac{x^2}{x-1}$$...and the problem is the above inequalities are false for $,0<x<1,$ , since the leftmost side is positive yet the rightmost one is negative! – DonAntonio May 02 '13 at 17:23
  • "Obviously"? This word seems to mean for you something different than for me. Anyway, I was trying in my last comment to address the possible flaw in the given hint in this answer. – DonAntonio May 02 '13 at 17:25
  • And BTW, @Harold: could you please point one point $,0<x<0.1,$ for which $,f(x)>2,$ , say? – DonAntonio May 02 '13 at 17:27
  • @DonAntonio oh, I have misread the question. You are right. The limit seems now to be 0. – gukoff May 02 '13 at 17:31
  • Thanks for clearing that out, @Harold. We all must be careful when using the words "obvious, trivial, simple" , etc. – DonAntonio May 02 '13 at 17:33
  • 1
    Just use Alpha, or some other tool, to find a solution of $\sin(1/x)=-x$ to $20$ decimal places. It is geometrically clear there are such. How perturb the $x$ a very tiny bit. – André Nicolas May 02 '13 at 17:36
  • @DonAntonio Oh, I need to sleep, it seems like I was right and just misread it again. The problem is $\lim\limits_{x\rightarrow\infty}{\frac{1}{x(1+x\cdot\sin(x))}}$, where the fraction can be so big or small how you want. The function can be incredibly big, because $x(1+x\cdot\sin(x))$ can be incredibly small. You asked, why is it obvious to me. It is because $1+x\cdot\sin(x)$ equals to $0$ with period smaller than $2\pi$. So you can change $x$ a bit ($\Delta x < 2\pi$) and make $1+x\cdot\sin(x)$ close enough to $0$ to make $x(1+x\cdot\sin(x))$ so small as you want. – gukoff May 02 '13 at 18:12
  • That wasn't in fact my question to you, @Harold...and not even whether you were right or wrong: for something to be obvious it must be, imo, that it is something either completely basic and elementary in the theory we're dealing with or else something that follows at once and almost without thinking deep from some data or previous work. I don't think neither of these conditions is fulfilled in this case, whether the limit exists or not. – DonAntonio May 02 '13 at 21:47
3

In the comments, several folks indicate that WolframAlpha produces strange results and/or should not be trusted too much. Of course, no software should be trusted too much but, in this particular case, I think that Alpha's response is perfectly reasonable. Here's what I see when I enter limit x^2/(x+sin(1/x)) as x->0:

enter image description here

Now as Don pointed out, Alpha reports that standard computation time was exceeded. Nonetheless, the pod of primary interest, namely the "Limit" pod, did complete and simply says "no result found...". This seems perfectly reasonable, although, "does not exist" or "undefined" could be even better.

As several folks also pointed out, an examination of a graph makes sense as well:

enter image description here

It now becomes crystal clear what's going on. For every $\varepsilon>0$ there are $x$ values with $0<|x|<\varepsilon$ and $x+\sin(1/x) = 0$ so that the function is simply not even defined on any entire punctured neighborhood of the origin.

Also, while it is certainly sufficient to show that the function is arbitrarily large and small in any neighborhood of the origin, it's not necessary to show this; really the problem is much simpler than that. Again, in order for the limit definition to be satisfied a function must be everywhere defined in some punctured neighborhood of the origin, i.e. some set of the form

$$\{x: 0<|x|<\varepsilon\}$$

and this function isn't - done. As a similar example, you might consider $f(x)=x\sin(1/x)/\sin(1/x)$. This function is, in fact, bounded and the same as $y=x$ except at an infinite sequence of points that converges to the origin where it is undefined. A graphical representation that somewhat captures this is the following:

enter image description here

Mark McClure
  • 30,510
2

According to WA and to this site , the limit seems to exist and be equal to zero, yet in any neighbourhood of zero there is one point (well, in fact infinitely many) where the function's denominator vanishes and, thus, the limit cannot exist in fact. We can try to show this as follows: define

$$f(x)=x+\sin\frac1x$$

We now choose

$$(1)\;\;w_n:=\frac2{(4n+1)\pi}\;,\;n\in\Bbb Z\implies f(w_n)=\frac2{(4n+1)\pi}+1>0$$

$$(2)\;\;z_n:=\frac2{(4n+3)\pi}\;,\;n\in\Bbb Z\implies f(z_n)=\frac2{(4n+3)\pi}-1<0$$

Since clearly $\;w_n\,,\,z_n\xrightarrow[n\to\infty]{}0\;$ and since $\;f\;$ is continuous in the positive reals, we can see that in any neighborhood of zero $\,f\,$ changes sign infinite times, and this means our original function isn't defined in infinite points (using the MVT for continuous functions) within any "small" right neighbourhood of zero. The same can be shown on the negative side of zero.

DonAntonio
  • 211,718
  • 17
  • 136
  • 287
  • What limit "seems to exist and be equal to zero"? The original?? – Mark McClure May 03 '13 at 00:02
  • Yes @MarkMcClure. In the links you can check it. It is pretty misleading. – DonAntonio May 03 '13 at 00:04
  • Again - what is misleading. I'm honestly not sure what you're referring to. – Mark McClure May 03 '13 at 00:13
  • The graph of the function $,\frac{x^2}{x+\sin 1/x},$ is pretty misldeading, as it looks as if the limit exists, whereas, as pointed out, it doesn't... – DonAntonio May 03 '13 at 00:25
  • Oh the graph. Hmm... It certainly doesn't look that way to me. Of course, I plotted just one graph from -0.1 to 0.1. In the course of numerical experimentation, one would naturally check the interval from -0.01 to 0.01 and see that the sequence of asymptotes continues. It's just a short step from there to examine the graph of both sides of $sin(1/x)=-x$ from which its geometrically clear that there are infinitely many solutions. – Mark McClure May 03 '13 at 00:31
  • Well, it certainly looked that way to me...and this is a good lesson for me, and perhaps others, that even in the tiny-tiny neighbourhoods of points one has to be way more careful when dealing with infinitesimal stuff. – DonAntonio May 03 '13 at 00:34
1

note that this function is odd symmetric with respect to $x$. So if we can show that $\lim_{x\to 0^+}\frac{x^2}{x+\sin{1/x}}=0$, then $\lim_{x\to 0^-}\frac{x^2}{x+\sin{1/x}}=0$ as well. We get down to prove this.

Using squeeze theorem, we arrive at: $\lim_{x\to 0^+}\frac{x^2}{x-1}<\lim_{x\to 0^+}\frac{x^2}{x+\sin1/x}<\lim_{x\to 0^+}\frac{x^2}{x-0}$ and, $\lim_{x\to 0^+}\frac{x^2}{x+1}<\lim_{x\to 0^+}\frac{x^2}{x+\sin1/x}<\lim_{x\to 0^+}\frac{x^2}{x+0}$ and $\lim_{x\to 0^+}\frac{x^2}{x+1}=0$ ,also $\lim_{x\to 0^+}\frac{x^2}{x-1}=0$ ,and also $\lim_{x\to 0^+}\frac{x^2}{x+0}=0$, so

$\lim_{x\to 0^+}\frac{x^2}{x+\sin{1/x}}=0$, and using the symmetric property mentioned above, we get

$\lim_{x\to 0}\frac{x^2}{x+\sin{1/x}}=0$