4

I have posted previously on a problem in a similar vein here: Limit evaluation: very tough question, cannot use L'hopitals rule

I believe this problem is very similar, but it has stumped me.

$$\lim_{x \to 0}\frac{1-\frac12 x^2 - \cos\left(\frac{x}{1-x^2}\right)}{x^4}$$

Really appreciate it if someone has some insight on this.This comes out to be indeterminate if one plugs in zero. Following the idea from the link above, I tried to recognize this as derivative evaluated at zero of a function, BUT I could not find the function, because I tried to make this all over x, so that means the function I would create would generate a rational type with x^3 on the bottom. I guess I should also try to look at some trig limit identities as well.

Hope someone out there can see how to navigate this problem.

P

Palu
  • 851
  • You should use Taylor formula to solve these kind of limits. – Emanuele Paolini Jan 03 '14 at 05:44
  • Is there a typo in your problem? Perhaps $$\lim_{x \to 0}\frac{1-\frac12x^2 - \cos(\frac{x}{1-x^2})}{x^4} ?$$ – Stephen Montgomery-Smith Jan 03 '14 at 05:54
  • Yes, you are correct, I missed the 1/2 in front of x^2 – Palu Jan 03 '14 at 05:56
  • Hi,I am using the taylor cosine series expansion. – Palu Jan 03 '14 at 05:57
  • @user99279: You should edit your question rather than leaving the correction in the comments. – user21820 Jan 03 '14 at 05:58
  • Stephen beat me to the edit, there was a conflict, we edited it at the same time. Thanks Stephen. – Palu Jan 03 '14 at 06:04
  • Ok, I did a two terms taylor cosine expansion, and I got the x^4 to go away and was able to take the limit and got -2. – Palu Jan 03 '14 at 06:06
  • BUT, since finite talyor expansions are just approximations of the true function, how do you know how far to expand a function, as in this case being the cosine function! – Palu Jan 03 '14 at 06:08
  • I'll explain in my answer. – user21820 Jan 03 '14 at 06:09
  • You need to use 3 terms of the cosine expansion. – Stephen Montgomery-Smith Jan 03 '14 at 06:13
  • Hi there Stephen. Thanks. But I am still left not understanding how one knows how far to expand. AND why doing a finite expansion is sufficient for an infinite function in a limit based question. – Palu Jan 03 '14 at 06:17
  • Use the big-Oh notation like in the answer Mohamed provided. – Stephen Montgomery-Smith Jan 03 '14 at 06:20
  • With the big-Oh notation, if you ended up with $\frac{O(x^4)}{x^4}$, then you know you haven't done enough terms. – Stephen Montgomery-Smith Jan 03 '14 at 06:22
  • Hi, I have done Big-Oh notation back in my university days in a course on data structures. – Palu Jan 03 '14 at 06:22
  • BUT, one needs to be able to explain this to students who are doing calculus in their first year, Calc2 course. Big-Oh notation and meaning is not taught in first year Calculus courses as far as I am aware. – Palu Jan 03 '14 at 06:23
  • See my answer. For many simple problems like this you can convert the Big-O notation into elementary inequalities. This preserves the intuitiveness of the Big-O notation, and is yet rigorous and easy to follow. – user21820 Jan 03 '14 at 06:25
  • Big-O notation can actually be made rigorous but the current convention in its notation does not allow that, because it is not clearly specified what the hidden constants depend on. I try to avoid that problem by explicitly saying "as $x \to 0$" when necessary – user21820 Jan 03 '14 at 06:27
  • Hi user21820, Based on you saying that O(x^5) goes to zero as x goes to zero now makes sense. And all other terms further down will also be zero, because they will all be divisible by x^4, leaving power polynomials with integer coefficients. – Palu Jan 03 '14 at 06:28
  • Yup you got the basic idea.. As long as you faithfully keep track of all the error terms it can never go wrong, and even if the limit doesn't exist it will still give you a correct and meaningful bound. – user21820 Jan 03 '14 at 06:31
  • Thanks so much for all your help. I consider this solved now. – Palu Jan 03 '14 at 06:32

4 Answers4

2

Let $t=\dfrac{x} {2(1-x^2)}$ to simplify typing. Note that $t\to 0$ and $t/x\to 1/2$ as $x\to 0$. The numerator can be rewritten as $$2\sin^2t-2\cdot\frac{x^2}{4}=2\left(\sin t-\frac {x} {2}\right)\left(\sin t+\frac{x}{2}\right)=2AB\text{ (say)} $$ Clearly we have $$\frac{B} {x} =\frac{1}{2}+\dfrac{\sin t}{t}\cdot\frac{t}{x}\to \frac {1}{2}+1\cdot\frac{1}{2}=1$$ as $x\to 0$. And $$\frac{A} {x^3}=\frac{\sin t-t}{t^3}\cdot\frac{t^3}{x^3}+\frac{1}{2}\cdot\frac{1}{1-x^2}\to-\frac{1}{6}\cdot\frac{1}{8}+\frac {1}{2}=\frac{23}{48}$$ Therefore the given expression tends to $23/24$ as $x\to 0$. In the above process we have used the standard limit $$\lim_{t\to 0}\frac{\sin t} {t} =1$$ and the limit $$\lim_{t\to 0}\frac {\sin t-t} {t^3}=-\frac{1}{6}$$ which is easily proved via a single application of L'Hospital's Rule or via Taylor series for $\sin t$.

2

$\frac 1{1-x^2}=1+x^2+O(x^4)$ then $\frac x{1-x^2}=x+x^3+O(x^5)$ then $\cos\left(\frac x{1-x^2}\right)=1-\frac{x^2}{2}-x^4+\frac{x^4}{24}+O(x^5)=1-\frac{x^2}{2}-\frac{23x^4}{24}+O(x^5)$ the limit is $\frac{23}{24}$.

Mohamed
  • 3,771
2

Like what Emanuele said, asymptotic expansions are useful for this kind of limits, and in fact better than L'Hopital (which fails miserably for some limits):

$\frac{x}{1-x^2} \in x + x^3 + O(x^5) \to 0$ as $x \to 0$

[We keep the error term so that at the end we know the error of the final approximation.]

$\cos( \frac{x}{1-x^2} ) \in 1 - \frac{1}{2} ( \frac{x}{1-x^2} )^2 + \frac{1}{24} ( \frac{x}{1-x^2} )^4 + O( ( \frac{x}{1-x^2} )^6 ) \\ \subset 1 - \frac{1}{2} (x+x^3+O(x^5))^2 + \frac{1}{24} (x+x^3+O(x^5))^4 + O(x^6) \text{ as } x \to 0 \\ \subset 1 - \frac{1}{2} x^2 - \frac{23}{24} x^4 + O(x^6) \text{ as } x \to 0$

[We can make the substitution into the Taylor expansion only because the input to $\cos$ tends to 0.]

$\frac{ 1 - \frac{1}{2} x^2 - \cos( \frac{x}{1-x^2} ) }{ x^4 } \in \frac{ \frac{23}{24} x^4 + O(x^6) }{ x^4 } = \frac{23}{24} + O(x^2) \to \frac{23}{24}$ as $x \to 0$

But you must make sure you understand the meaning of the Big-O notation and when and why they can be used. To make it more concrete, you can in many cases find explicit constants for bounds instead of using Big-O notation. For example:

$1 - \frac{1}{2} x^2 + \frac{1}{24} x^4 - \frac{1}{720} x^6 \le cos(x) \le 1 - \frac{1}{2} x^2 + \frac{1}{24} x^4$ [obtained by repeated differentiation and Mean-value theorem]

$x + x^3 \le \frac{x}{1-x^2} \le x + x^3 + 2 x^5$ for sufficiently small $x \ge 0$

$x + x^3 \ge \frac{x}{1-x^2} \ge x + x^3 + 2 x^5$ for sufficiently small $x \le 0$

user21820
  • 60,745
  • Can you please tell me where I can learn these methods of solving the limits. I have searched so much but couldn't find anywhere these things and many other like limit of a summation, Newton-Leibniz Formula, etc. – Knight wants Loong back Nov 18 '19 at 14:15
  • @adeshmishra: You can take a look at some other posts linked from my profile under "Asymptotic expansions". It's actually very simple once you fully grasp all the Landau notations (not just Big-O notation). – user21820 Nov 18 '19 at 17:10
0

Just expand $\cos\left(\frac{x}{1-x^2}\right)$ using series expansion and simplify a bit

user85503
  • 741
Aman
  • 19
  • Ignore the terms which have the numerator greater than x^4 because that will become zero anyway – Aman May 30 '15 at 07:21