3

How to compute the limit, $$\lim_{x\rightarrow 0}\frac{3x^2-3x\sin x}{x^2+x\cos\frac{1}{x}}$$

memerson
  • 923

4 Answers4

5

Rewrite $$\lim_{x\to 0}\frac{3x^2-3x\sin x}{x^2+x\cos\frac{1}{x}}=\lim_{x\to 0}\frac{3x^2-3x\sin x}{x(x+\cos\frac{1}{x})},$$ we will deduce the nature of $x+\cos\frac1x$ near $0$.
There are monotone decreasing sequence $a_n>0$ with $\lim_{n\to\infty}a_n=0$ such that $\cos\frac 1{a_n}=(-1)^n$.
Let $a_n\ll 1$, we have $a_{n+1}+\cos \frac 1{a_{n+1}}$ have a different sign from $a_{n}+\cos \frac 1{a_{n}}$. Since $(x+\cos\frac{1}{x})$ is continuous, there is a zero point between $a_{n+1}$ and $a_n$.
Hence there are infinite undefined points near $0$. Therefore, the limit does not exist.
Notice that this answer is only available with definition $\lim_{x\to x_0}f(x)=A\Leftrightarrow\forall \varepsilon>0\exists\delta>0\ s.t.\ 0<|x-x_0|<\delta\implies|f(x)-A|<\varepsilon$

Kemono Chen
  • 8,629
  • Some authors prefer to take limit only via points of domain. – Paramanand Singh Nov 10 '18 at 04:16
  • The problem is far deeper than the vanishing of denominator in each deleted neighborhood of $0$. See my answer for details. – Paramanand Singh Nov 10 '18 at 07:07
  • Thanks, both of you! – Alexander Lau Nov 10 '18 at 08:04
  • I agree with Paramanand, that's not a good argument, in an more advanced perspective, to discuss the existence of the limit. We always exclude from the domain the points where the expression/function is not defined. – user Nov 10 '18 at 08:25
  • Please revise the answer accordinlgy to the comments given, otherwise I'll be forced to downvote that because it is completely uncorrect. – user Nov 10 '18 at 10:03
  • Let me clarify that my downvote is not of course a personal issue but I think that it could be very misleading for not skilled users which could assume your argument as a good way to dsicuss and solve the limit. I'll eliminate the downvote as you'll revise that accordingly. – user Nov 10 '18 at 10:34
  • @gimusi I corrected it. – Kemono Chen Nov 10 '18 at 11:52
  • @KemonoChen Sorry but I think that it is not a correct way to prove that. You could also mantain your initial claim and way to prove it. I see now you are an high school student and often at that level the definition of limit is given accordingly to your definition. Maybe you could simply specify that noticing that you solution is valid only according to this not general definition and not to the more general one discussed in other answers here. Thanks, I remove the downvote as promised and convert into a (+1). Bye – user Nov 10 '18 at 11:58
  • National difference is sometimes annoying. I asked some university students, they all use the ancient epsilon-delta language ($\lim_{x\to\infty}\sin x/\sin x$ does not exist). – Kemono Chen Nov 10 '18 at 12:11
  • @KemonoChen I understand you point and I think you have solved that properly accordingly to your current knowledge. But according to the most general definition of limit, we can go beyond that limitation and discuss it in a deeper and interesting way. Refer to the discussion HERE about that. In some answer you can also find some good reference for that (e.g. Rudin). – user Nov 10 '18 at 12:43
  • @KemonoChen What I mean is that according to your basic definition of course the limit doesn't exist but in taht way the problem becomes trivail and very boring. By the more general definition, that is assuming $x\in(-\delta,\delta)\cap D$, the problem becomes more challenging and we need a more sophisticated way to show that the limit doesn't exist. – user Nov 10 '18 at 12:47
  • @KemonoChen Your current answer is still wrong. You can rollback to the previous one explaining the motivation for your conclusion (that is simplified definition of limit). – user Nov 10 '18 at 12:49
  • @KemonoChen I think that now it is fine! – user Nov 10 '18 at 13:13
2

This is a tricky problem. The first issue as pointed out in other answers is that the denominator vanishes in every deleted neighborhood of $0$ so that in every deleted neighborhood of $0$ there are points where the function is not defined. The usual definition of limit requires that the function be defined in some deleted neighborhood of the point under consideration and hence based on this definition the limit does not exist. In fact one should say that it does not make sense to talk of the limit of this function based on this definition as the prerequisites of the definition are not fulfilled.

However there is a relaxed definition of limit which considers only points of domain of definition of the function:

Let $D $ be a non-empty subset of $\mathbb {R} $ and let $a\in\mathbb{R} $ be an accumulation point of $D$. Let $f:D\to \mathbb{R}$ be a function. A number $L$ is said to be the limit of $f$ at $a$ if for any given $\epsilon >0$ there exists a corresponding $\delta >0$ such that $|f(x) - L|<\epsilon $ whenever $x\in(a-\delta, a+\delta) \cap D, x\neq a$.

The limit in question does not exist even under this relaxed definition of limit but it is difficult to show that.

Let $g(x) =x^4-x$ and note that $g(x) <0$ if $0<x<1$ and $g(x) \to 0$ as $x\to 0^{+}$. Further as $x\to 0^{+}$ the function $\cos(1/x)$ oscillates between $1$ and $-1$ and therefore by intermediate value theorem $\cos(1/x)=g(x)$ or $x+\cos (1/x)=x^4$ for infinitely many values of $x$ as $x\to 0^{+}$.

The given function in question can be written as $$3\cdot\frac {x-\sin x} {x+\cos (1/x)}$$ and the numerator behaves like $x^3/2$ as $x\to 0^{+}$. As discussed in previous paragraph there is a sequence of positive values of $x$ tending to $0$ for which denominator is equal to $x^4$ and thus if $x\to 0^{+}$ through such a sequence of values then the given function tends to $\infty$. On the other hand using similar strategy we can find a sequence of positive values of $x$ such that denominator equals $x^3$ so that the fraction tends to $1/2$. Thus the given function oscillates infinitely as $x\to 0^{+}$ and the desired limit does not exist.

The technique used above can be applied to prove that given any number $L$ there is a sequence $x_n$ such that $x_n\to 0$ and $f(x_n) \to L$ as $n\to\infty $. One can also have $L=\pm\infty$. The cases $L=1/2,L=\infty$ were discussed above.

  • Nice initial discussion about the issue for the existence. I've added another reference for that. For the strategy to show the not existence of the limit, I agree that the best strategy is to use the fact that we can find arbitrary sequences $x_n\to 0$ such that $cos\frac1{x_n}=f(x_n)$ for $|f(x_n)|\le 1$ and therefore we can obtain any real value for the limit. – user Nov 10 '18 at 08:47
  • @gimusi: once we encounter a continuous oscillating function there is no limit to its devilish behavior using intermediate value property. For example here if we have $x^{1000}/(x+\cos(1/x))$ we can find many $x$ with $x+\cos(1/x)=x^{1001}$ and then there is a sequence of values of $x$ for which function diverges. – Paramanand Singh Nov 10 '18 at 08:51
  • @ParamanandSingh That's the point, we know that such sequences exist by IVT. I hope the asker can clarify his doubts and that the other answers will be revised accordingly. – user Nov 10 '18 at 08:55
  • @gimusi : wonder if graphing tools can graph the function in question near $0$. – Paramanand Singh Nov 10 '18 at 08:58
  • 1
    @ParamanandSingh One can enjoy here zooming in GRAPH – user Nov 10 '18 at 09:01
  • @ParamanandSingh And also with that one to se the idea for the case $x_n$ such that $\cos\frac{1}{x_n}=f(x_n)$ GRAPH 2 – user Nov 10 '18 at 09:04
  • I'll leave deleted my comment as I see no way to solve this problem with the things the OP says can be used.. For example, how do we know the function's numerator behaves as $;\frac{x^3}2;$ ? If not by Taylor or something pretty close, how can deduce that? The rest of the argument afterwards is somehow messy but it should be clear, in principle, by the IVT for continuous functions for anyone who's already studied. Anyway, the fact about the numerator still bewilders me. – DonAntonio Nov 10 '18 at 11:52
  • Without Taylor we can consider $x_n\to 0$ such that $\cos \frac1{x_n}=2x_n$ and the limit reduces to $3-3\sin{x_n}/x_n \to 0$. Then we can consider for example $x_n\to 0$ such that $\cos \frac1{x_n}=-x_n+x_n^3$ and the limit reduces to $3\frac{x_n-\sin x_n}{x_n^3}=\frac12$ which can be solved also without Taylor (see THAT). – user Nov 10 '18 at 13:02
  • @DonAntonio I've edited my answer using that way just discussed here in order to avoid Taylor. – user Nov 10 '18 at 13:11
2

As noticed the limit doesn't exist indeed we can consider the sequence as $x_n \to 0$ such that

$$\cos\frac{1}{x_n}=2x_n \implies \frac{3x^2-3x\sin x}{x^2+x\cos\frac{1}{x}}=\frac{3x_n^2-3x_n\sin x_n}{3x_n^2}=1-\frac{\sin x_n}{x_n} \to 1-1=0$$

and the sequence $x_n \to 0$ such that

$$\cos\frac{1}{x_n}=-x_n+x_n^3 \implies \frac{3x^2-3x\sin x}{x^2+x\cos\frac{1}{x}}=3\frac{x_n^2-x_n\sin x_n}{x_n^4}=3\frac{x_n-\sin x_n}{x_n^3}\to \frac12$$

indeed as $t \to 0$ we have that $\frac{t-\sin t}{t^3} \to \frac16$ wich can be proved by l'Hopital, Taylor or by the method shown here: Are all limits solvable without L'Hôpital Rule or Series Expansion.

For the issue already discussed here in detail by Paramanand Singh about the not satisfactory way to show that the limit doesn't exist because not defined at "infinitely many points", for a useful discussion refer also to the related

user
  • 154,566
-1

For any $\delta > 0$ there exists an $0<x<\delta$ such that $x - \cos \frac {1}{x} = 0$

As $x-\cos \frac 1x$ is continuous, the are points close to these zeros such that $ x-\cos \frac 1x $ is arbitrarily close to zero, and $\frac {3x^2 - 3x\sin x}{x(x-\cos \frac 1x)}$ is arbitrarily large.

Doug M
  • 57,877
  • Some authors prefer to consider values of $x$ only in domain of function $f(x) $ while defining limit. The issue here is that the function oscillates infinitely even when $x\to 0$ via points in domain. – Paramanand Singh Nov 10 '18 at 07:09
  • That's not a good point to discuss the existence of the limit. Refer to Paramanand discussion here and to the link I've given as a reference. – user Nov 10 '18 at 08:36
  • Please revise the answer accordinlgy to the comments given, otherwise I'll be forced to downvote that because it is completely uncorrect. – user Nov 10 '18 at 10:03
  • Let me clarify that my downvote is not of course a personal issue but I think that it could be very misleading for not skilled users which could assume your argument as a good way to dsicuss and solve the limit. I'll eliminate the downvote as you'll revise that accordingly. – user Nov 10 '18 at 10:34