9

$$\lim_{x\to0} \frac{x-\sin x}{x-\tan x}=?$$

I tried using $\lim\limits_{x\to0} \frac{\sin x}x=1$.

But it doesn't work :/

PinkyWay
  • 4,565
IndyZa
  • 509
  • 1
    I just don't know why people say solve *** without L'Hopital. – Shobhit Sep 29 '13 at 11:33
  • 5
    Either the instructor doesn't want them to use l'H, or else they haven't yet studied it... – DonAntonio Sep 29 '13 at 11:34
  • the limit formula you have used is incorrect. its equal to 1 – Aman Mittal Sep 29 '13 at 11:35
  • @Shobhit I can't use it if its in my upcoming test because we didn't learn that in class yet thats why. – IndyZa Sep 29 '13 at 11:37
  • @AmanMitall Thanks I correct it already – IndyZa Sep 29 '13 at 11:40
  • @DavidMitra: Far more natural, in my opinion, to divide numerator and denominator both by $x$. This is a procedure that one will apply in numerous settings, with limits both at $0$ and at $\infty$. – Ted Shifrin Sep 29 '13 at 11:46
  • @TedShifrin That will serve no purpose to avoid L'Hospital. – OR. Sep 29 '13 at 11:48
  • 1
    @TedShifrin I tried divide it with x all the way but the result is 0/0 – IndyZa Sep 29 '13 at 11:50
  • 1
    Ah right, my carelessness and apologies. So @Sami's solution is the best, although at first I was assuming the OP didn't get know about Taylor polynomials. When I've taught my class Taylor polynomials, I want them to realize how much easier they would be to apply in problems like this than a triple application of L'Hôpital's rule. – Ted Shifrin Sep 29 '13 at 11:51
  • 6
    No, Sami's solution is not the best either. Using Taylor is just the same as using L'Hospital. – OR. Sep 29 '13 at 11:53
  • 8
    @ABC: I disagree with you ardently. An experienced student can do the Taylor polynomial calculation in his head here; I challenge most people to do the product rule (and often chain rule) three or more times in succession in both numerator and denominator and not make any mistakes. The Taylor polynomial is both operationally and conceptually more appropriate and more indicative of what is actually going on. – Ted Shifrin Sep 29 '13 at 12:25
  • The question is not about being able to do it in your head. Iterating L'Hospital and Taylor are both (third order in this problem) approximations. Conceptually is the same thing. The idea here, if you want to avoid completely using derivatives, is to disguise the Taylor formula by replacing the derivatives by their definitions. All derivatives of all orders of the trigonometric functions follow from the given limit $\sin(x)/x\rightarrow 1$, as $x\rightarrow0$. A big ugly limit is going to appear and you can show it is equal to the one it is asked. – OR. Sep 29 '13 at 12:33
  • I have been asked this problem before. It is terribly ugly to completely do it using only the limit of $\sin(x)/x$, but it is possible. On the other hand, it is not really a useful exercise. – OR. Sep 29 '13 at 12:36
  • @ABC, oh, can you, and rigorously? Let's see it. You get all sorts of iterated limits which are hard to handle rigorously without some serious analysis. Or perhaps I don't understand what you're saying. – Ted Shifrin Sep 29 '13 at 12:37
  • Not just me. Everybody can. As I said, it is just about disguising what you do with Taylor. – OR. Sep 29 '13 at 12:38
  • I have the feeling that you can actually do it without specifically invoking $\frac{\sin x}x$, through some clever substitutions, but I'm still testing my idea. – Glen O Sep 29 '13 at 12:52
  • As it turns out, I was right. – Glen O Sep 29 '13 at 13:51
  • @GlenO You are assuming existence of the limit in your solution. Existence is necessarily going to come from somewhere. That somewhere is the definition of $\sin(x)$, which is where the limit of $\sin(x)/x$ is coming from. In principle you are right too (although your partial solution is not the reason). You can use geometry to get an application of the sandwich lemma and get the existence of some limit, equivalent to knowing $\sin(x)/x$, but not necessarily the same. – OR. Sep 29 '13 at 16:31
  • @ABC: The same type of assumption must be made in the solution provided by experimentX. And as with that solution, only the limit of $\sin(x)/x$ existing is necessary to prove that the limits seen in my solution exist (note that one can replace the $x\sin^2(x)$ denominators with $x^3$ by considering that limit, at which point they simplify down to the limits found in experimentX's solution). – Glen O Sep 30 '13 at 01:09

5 Answers5

13

In the beginning of this answer, it is shown that $$ \begin{align} \frac{\color{#C00000}{\sin(2x)-2\sin(x)}}{\color{#00A000}{\tan(2x)-2\tan(x)}} &=\underbrace{\color{#C00000}{2\sin(x)(\cos(x)-1)}\vphantom{\frac{\tan^2(x)}{\tan^2(x)}}}\underbrace{\frac{\color{#00A000}{1-\tan^2(x)}}{\color{#00A000}{2\tan^3(x)}}}\\ &=\hphantom{\sin}\frac{-2\sin^3(x)}{\cos(x)+1}\hphantom{\sin}\frac{\cos(x)\cos(2x)}{2\sin^3(x)}\\ &=-\frac{\cos(x)\cos(2x)}{\cos(x)+1}\tag{1} \end{align} $$ Therefore, $$ \lim_{x\to0}\,\frac{\sin(x)-2\sin(x/2)}{\tan(x)-2\tan(x/2)}=-\frac12\tag{2} $$ Thus, given an $\epsilon\gt0$, we can find a $\delta\gt0$ so that if $|x|\le\delta$ $$ \left|\,\frac{\sin(x)-2\sin(x/2)}{\tan(x)-2\tan(x/2)}+\frac12\,\right|\le\epsilon\tag{3} $$ Because $\,\displaystyle\lim_{x\to0}\frac{\sin(x)}{x}=\lim_{x\to0}\frac{\tan(x)}{x}=1$, which are shown geometrically in this answer, we have $$ \sin(x)-x=\sum_{k=0}^\infty2^k\sin(x/2^k)-2^{k+1}\sin(x/2^{k+1})\tag{4} $$ and $$ \tan(x)-x=\sum_{k=0}^\infty2^k\tan(x/2^k)-2^{k+1}\tan(x/2^{k+1})\tag{5} $$ By $(3)$ each term of $(4)$ is between $-\frac12-\epsilon$ and $-\frac12+\epsilon$ of the corresponding term of $(5)$.

Therefore, if $|x|\le\delta$ $$ \left|\,\frac{\sin(x)-x}{\tan(x)-x}+\frac12\,\right|\le\epsilon\tag{6} $$ We can restate $(6)$ as $$ \lim_{x\to0}\frac{x-\sin(x)}{x-\tan(x)}=-\frac12\tag{7} $$

robjohn
  • 345,667
  • Telescope the partial sums of the series. You use the limits to get the convergence of the partial sums. – OR. Sep 29 '13 at 18:03
  • 5
    @experimentX: for $(4)$ $$ \sum_{k=0}^{n-1}2^k\sin(x/2^k)-2^{k+1}\sin(x/2^{k+1})=\sin(x)-2^n\sin(x/2^n) $$ and $$ \lim_{n\to\infty}2^n\sin(x/2^n)=x $$ Similar for $(5)$, just replace $\sin$ with $\tan$. – robjohn Sep 29 '13 at 21:13
  • If $\tan(x)<0$ and we have $$-\varepsilon-\frac12\le\frac{2^k\sin\left(\frac{x}{2^k}\right)-2\cdot 2^k\sin\left(\frac{\frac{x}{2^k}}2\right)}{2^k\tan\left(\frac{x}{2^k}\right)-2\cdot 2^k\tan\left(\frac{\frac{x}{2^k}}2\right)}\le\varepsilon-\frac12$$ shouldn't we say each term of $(4)$ is between $-\frac12-\varepsilon$ and $-\frac12+\varepsilon$ of the absolute value of the corresponding term of $(5)$? – PinkyWay Dec 01 '20 at 20:54
  • 2
    @Invisible: When the bottom is negative, the top is positive and vice-versa. The ratio is close to $-\frac12$ when $|x|$ is small, positive or negative (both numerator and denominator are odd functions). – robjohn Dec 01 '20 at 22:04
10

$$\frac{x - \sin(x)}{x - \tan(x)} = \frac{x - \sin(x)}{x^3} \cdot \frac{x^3}{x - \tan(x)}$$

Let $x = 3y$ and $x\to 0 \implies y\to 0$ $$\lim_{x\to0} \frac{x - \sin(x)}{x^3} = L $$ $$L = \lim_{y\to0}\frac{3y - \sin(3y)}{(3y)^3} = \lim_{y\to0} \frac 3 {27} \frac{y - \sin(y)}{y^3} + \lim_{y\to0} \frac{4}{27} \frac{\sin^3(y)}{y^3} = \frac{1}{9} L + \frac 4{27} $$

This gives $$\lim_{x\to0}\frac{x - \sin(x)}{x^3} = \frac 1 6 $$

\begin{align*} L &= \lim_{y\to0}\frac{ 3y - \tan(3y)}{27 y^3} \\ &= \lim_{y\to0} \frac{1}{(1 - 3\tan^2(y ))} \cdot \frac{3y(1 - 3\tan^2(y )) - 3 \tan(y) + \tan^3(y)}{27y^3}\\ &= \lim_{y\to0} \frac{1}{(1 - 3\tan^2(y ))} \cdot \left( \frac 3 {27} \frac{y - \tan(y)}{y^3} + \frac 1 {27} \frac{\tan^3(y)}{y^3} - \frac 9 {27} \frac{y \tan^2(y)}{y^3 } \right )\\ &= \frac {3L}{27} + \frac 1 {27} - \frac 1 3 \\ \end{align*}

This gives other limit to be $-1/3$, put it up and get your limit.

S L
  • 11,731
  • 2
    I wonder if you assume implicitly that $L$ exists. – Siminore Sep 29 '13 at 12:54
  • 1
    @Siminore yes that I assumed :( I can't find any other way. I saw that one for sine in yahoo answer couple of months ago ... but I can't seem to find it. – S L Sep 29 '13 at 12:54
  • 3
    @experimentX: this is very cute, assuming the limit exists. (+1). Proving the limit exists is a bit more difficult. – robjohn Sep 29 '13 at 16:56
  • @robjohn thanks, but i had seen it done somewhere on yahoo answers. – S L Sep 29 '13 at 16:58
7

Hint

Use the Taylor series: $$\sin x=x-\frac{x^3}{6}+o(x^3)\quad \text{and}\quad \tan x=x+\frac{x^3}{3}+o(x^3)$$

  • Didn't learn about this yet. I have no idea how to use it... – IndyZa Sep 29 '13 at 12:17
  • That was my initial reaction, @IndyZa. I assumed your class was at the beginning of the first course. I do not know what your teacher expects you to do. I suggest you inquire and let us know :) – Ted Shifrin Sep 29 '13 at 12:21
5

$$ L=\lim_{x\to0} \frac{x-\sin x}{x-\tan x}=\lim_{x\to0}\frac{x-\sin x}{x\cos x-\sin x}\cos x = \lim_{x\to0}\frac{2x-\sin2x}{2x\cos2x-\sin2x}\cos2x\\ = \lim_{x\to0}\frac{x-\cos x\sin x}{x(1-2\sin^2x)-\cos x\sin x}\cos2x=\lim_{x\to0}\frac{x-\cos x\sin x}{x-\cos x\sin x-2x\sin^2x}\cos2x $$ Which, noting that $\lim_{x\to0}\cos2x=1$, we can then write as $$ \lim_{x\to0}\frac{1}{1-\frac{2x\sin^2x}{x-\cos x\sin x}} = \frac{1}{1-2\lim_{x\to0}\frac{x\sin^2x}{x-\cos x\sin x}}=\frac{1}{1-2M} $$ Now, we turn our attention to that new limit... $$ \frac1M=\lim_{x\to0}\frac{x-\cos x\sin x}{x\sin^2x}=\lim_{x\to0}\frac{1-\cos x\frac{\sin x}x}{1-\cos^2x}=1+\lim_{x\to0}\frac{1-\frac{\sin x}{x\cos x}}{1-\cos^2x}\cos^2 x\\ =1+\lim_{x\to0}\frac{x-\tan x}{x\sin^2x} $$ But we also have $$ \frac1M = \lim_{x\to 0} \frac{2x-\sin2x}{2x\sin^2x}=2\lim_{x\to 0} \frac{2x-\sin2x}{2x(1-\cos2x)}=2\lim_{x\to0}\frac{x-\sin x}{x(1-\cos x)}\\ =2\lim_{x\to0}\frac{x-\sin x}{x(1-\cos^2 x)}(1+\cos x)=4\lim_{x\to0}\frac{x-\sin x}{x(1-\cos^2 x)}=4\lim_{x\to0}\frac{x-\sin x}{x\sin^2x}\\ =4\lim_{x\to0}\frac{x-\sin x}{x-\tan x}\cdot\frac{x-\tan x}{x\sin^2x} $$ So, we have $$ \frac1M = 4L\left(\frac1M-1\right) $$ or $1=4L(1-M)$... but $L=\frac{1}{1-2M}$ (or $1=L(1-2M)$).

Therefore, we have that $$ 1-2=4L-4LM-2L+4LM = 2L = -1 $$ Therefore, $L=-\frac12$. No use of $\lim_{x\to0}\frac{\sin x}x$ required.

Glen O
  • 12,425
  • +1. Quite a tour de force that I would never expect a beginning calculus student to do. – Ted Shifrin Sep 29 '13 at 15:13
  • 2
    This assumes that $\lim\lim_{x\to0}\frac{x\sin^2(x)}{x-\cos(x)\sin(x)}$ exists. Otherwise, it works. (+1) – robjohn Sep 29 '13 at 17:14
  • @TedShifrin: Oh, I agree. I just noticed the potential to find it without needing $\frac{\sin x}x$, and figured I'd provide it. To be honest, I can't see a solution approach that one would expect a beginning calculus student to do. – Glen O Sep 30 '13 at 01:12
  • @robjohn: As with experimentX's solution, I've assumed the existence of the limits (for sake of conciseness). – Glen O Sep 30 '13 at 01:20
3

Lemma $\bf1$: Suppose that both $\prod\limits_{k=1}^\infty(1-a_k)$ and $\sum\limits_{k=1}^\infty a_k$ converge. Then $$ \bbox[5px,border:2px solid #C0A000]{\lim_{n\to\infty}\frac{1-\prod\limits_{k=n}^\infty(1-a_k)}{\sum\limits_{k=n}^\infty a_k}=1}\tag1 $$ Proof: $$ \begin{align} \lim_{n\to\infty}\frac{1-\prod\limits_{k=n}^\infty(1-a_k)}{\sum\limits_{k=n}^\infty a_k} &=\lim_{n\to\infty}\frac{\prod\limits_{k=n+1}^\infty(1-a_k)-\prod\limits_{k=n}^\infty(1-a_k)}{\sum\limits_{k=n}^\infty a_k-\sum\limits_{k=n+1}^\infty a_k}\tag{1a}\\ &=\lim_{n\to\infty}\frac{a_n\prod\limits_{k=n+1}^\infty(1-a_k)}{a_n}\tag{1b}\\[15pt] &=\lim_{n\to\infty}\prod\limits_{k=n+1}^\infty(1-a_k)\tag{1c}\\[18pt] &=1\tag{1c} \end{align} $$ Explanation:
$\text{(1a)}$: Stolz-Cesàro
$\text{(1b)}$: simplify numerator and denominator
$\text{(1c)}$: cancel numerator and denominator
$\text{(1d)}$: the product converges

$\large\square$


Lemma $\bf2$: Suppose that both $\prod\limits_{k=1}^\infty(1+a_k)$ and $\sum\limits_{k=1}^\infty a_k$ converge. Then $$ \bbox[5px,border:2px solid #C0A000]{\lim_{n\to\infty}\frac{1-\prod\limits_{k=n}^\infty(1+a_k)^{-1}}{\sum\limits_{k=n}^\infty a_k}=1}\tag2 $$ Proof: $$ \begin{align} \lim_{n\to\infty}\frac{1-\prod\limits_{k=n}^\infty(1+a_k)^{-1}}{\sum\limits_{k=n}^\infty a_k} &=\lim_{n\to\infty}\frac{\prod\limits_{k=n+1}^\infty(1+a_k)^{-1}-\prod\limits_{k=n}^\infty(1+a_k)^{-1}}{\sum\limits_{k=n}^\infty a_k-\sum\limits_{k=n+1}^\infty a_k}\tag{2a}\\ &=\lim_{n\to\infty}\frac{a_n\prod\limits_{k=n}^\infty(1+a_k)^{-1}}{a_n}\tag{2b}\\[15pt] &=\lim_{n\to\infty}\prod\limits_{k=n}^\infty(1+a_k)^{-1}\tag{2c}\\[18pt] &=1\tag2 \end{align} $$ Explanation:
$\text{(2a)}$: Stolz-Cesàro
$\text{(2b)}$: simplify numerator and denominator
$\text{(2c)}$: cancel numerator and denominator
$\text{(2d)}$: the product converges

$\large\square$


Induction, $\lim\limits_{x\to0}\frac{\sin(x)}x=1$, and $\frac{\sin^2(x)}{4\sin^2(x/2)}=\left(1-\sin^2(x/2)\right)$ prove

Lemma $\bf{3}$: $$ \bbox[5px,border:2px solid #C0A000]{\frac{\sin^2(x)}{x^2}=\prod_{k=1}^\infty\left(1-\sin^2\left(x/2^k\right)\right)}\tag3 $$


Induction, $\lim\limits_{x\to0}\frac{\tan(x)}x=1$, and $\frac{\tan(x)}{2\tan(x/2)}=\left(1-\tan^2(x/2)\right)^{-1}$ prove

Lemma $\bf{4}$: $$ \bbox[5px,border:2px solid #C0A000]{\frac{\tan(x)}{x}=\prod_{k=1}^\infty\left(1-\tan^2\left(x/2^k\right)\right)^{-1}}\tag4 $$


The Limit $$ \begin{align} \lim_{x\to0}\frac{x-\sin(x)}{x-\tan(x)} &=\frac12\lim_{x\to0}\left(1+\frac{\sin(x)}{x}\right)\frac{1-\frac{\sin(x)}{x}}{1-\frac{\tan(x)}x}\tag{5a}\\ &=\frac12\lim_{x\to0}\frac{1-\frac{\sin^2(x)}{x^2}}{1-\frac{\tan(x)}x}\tag{5b}\\ &=\frac12\lim_{n\to\infty}\frac{1-\frac{\sin^2\left(x/2^n\right)}{\left(x/2^n\right)^2}}{1-\frac{\tan\left(x/2^n\right)}{x/2^n}}\tag{5c}\\ &=\frac12\lim_{n\to\infty}\frac{1-\prod\limits_{k=n+1}^\infty\left(1-\sin^2\left(x/2^k\right)\right)}{1-\prod\limits_{k=n+1}^\infty\left(1-\tan^2\left(x/2^k\right)\right)^{-1}}\tag{5d}\\ &=-\frac12\lim_{n\to\infty}\frac{\sum\limits_{k=n+1}^\infty\sin^2\left(x/2^k\right)}{\sum\limits_{k=n+1}^\infty\tan^2\left(x/2^k\right)}\tag{5e}\\[6pt] &=-\frac12\lim_{n\to\infty}\frac{\sin^2\left(x/2^n\right)}{\tan^2\left(x/2^n\right)}\tag{5f}\\[18pt] &=-\frac12\tag{5g} \end{align} $$ Explanation:
$\text{(5a)}$: $\lim\limits_{x\to0}\frac12\left(1+\frac{\sin(x)}x\right)=1$
$\text{(5b)}$: algebra
$\text{(5c)}$: rewrite the limit
$\text{(5d)}$: Lemma $3$ and Lemma $4$
$\text{(5e)}$: Lemma $1$ and Lemma $2$
$\text{(5f)}$: Stolz-Cesàro
$\text{(5g)}$: $\lim\limits_{x\to0}\cos^2(x)=1$

robjohn
  • 345,667