Find the limit: $$\lim_\limits{x\to 0^+}{\left( e^{\frac{1}{\sin x}}-e^{\frac{1}{x}}\right)}$$

Using graph inspection, I have found the limit to be $+\infty$, but I cannot prove this in any way (I tried factorizing, using DLH)... Can anyone give a hint about that? The limit should be done without any approximations, because we haven't been taught those yet.

  • 1
    $\begingroup$ With "approximations", do you include Taylor series? $\endgroup$ – Peter Woolfitt Nov 26 '15 at 18:05
  • $\begingroup$ @PeterWoolfitt Yes... $\endgroup$ – Jason Nov 26 '15 at 18:06
  • 2
    $\begingroup$ oh well - Taylor series seems to be the apparent approach to me $\endgroup$ – Peter Woolfitt Nov 26 '15 at 18:07
  • $\begingroup$ Could you simplify $$e^a-e^b$$ ? $\endgroup$ – Stefan Nov 26 '15 at 18:07
  • $\begingroup$ @Stefan I have already tried to do so in many ways by extracting $e^{1/x}$ and $e^{1/\sin x}$ in separate approaches, but both failed (even with DLH!!!)... $\endgroup$ – Jason Nov 26 '15 at 18:10

First, you can show that $$\lim_{x\to 0^+}\left(\frac{1}{\sin x}-\frac{1}{x}\right)=0.$$ This shows that $$\lim_{x\to 0^+}\frac{e^{\frac{1}{\sin x}-\frac{1}{x}}-1}{\frac{1}{\sin x}-\frac{1}{x}}=1.$$ Now, write $$e^{1/\sin x}-e^{1/x}=e^{1/x}\left(\frac{1}{\sin x}-\frac{1}{x}\right)\frac{e^{\frac{1}{\sin x}-\frac{1}{x}}-1}{\frac{1}{\sin x}-\frac{1}{x}},$$ and letting $x\to 0^+$ gives \begin{align*}\lim_{x\to 0^+}\left(e^{1/\sin x}-e^{1/x}\right)&=\lim_{x\to 0}e^{1/x}\left(\frac{1}{\sin x}-\frac{1}{x}\right)\lim_{x\to 0^+}\frac{e^{\frac{1}{\sin x}-\frac{1}{x}}-1}{\frac{1}{\sin x}-\frac{1}{x}}=\lim_{x\to 0^+}e^{1/x}\left(\frac{1}{\sin x}-\frac{1}{x}\right)\\ &=\lim_{x\to 0^+}\frac{e^{1/x}(x-\sin x)}{x\sin x}=\lim_{x\to 0^+}\frac{e^{1/x}(x-\sin x)}{x^2}, \end{align*} since $\lim_{x\to 0}\frac{\sin x}{x}=1$. Continuing the computation, this last limit is equal to $$\lim_{x\to 0^+}xe^{1/x}\lim_{x\to 0^+}\frac{x-\sin x}{x^3}=\frac{1}{6}\lim_{x\to 0^+}xe^{1/x}=\frac{1}{6}\lim_{u\to\infty}\frac{e^u}{u}, $$ after performing the substitution $u=\frac{1}{x}$. The last limit is $\infty$, so the limit you ask is equal to $\infty$.

| cite | improve this answer | |
  • $\begingroup$ It should be $x\sin x$ in the denominator, not $x^2$, initially, athough you can use $\sin x/x\to 1$ to get $x^2$ there. $\endgroup$ – Thomas Andrews Nov 26 '15 at 18:23
  • $\begingroup$ Right, I'll add this. $\endgroup$ – detnvvp Nov 26 '15 at 18:23
  • $\begingroup$ For the second equation you equated two limits which cannot be done because of the indeterminate form... $\endgroup$ – Jason Nov 26 '15 at 18:26
  • $\begingroup$ Which one? No limit is indeterminate here. $\endgroup$ – detnvvp Nov 26 '15 at 18:27
  • $\begingroup$ OK, your answer has been edited to illustrate what I had been asking! All good now. $\endgroup$ – Jason Nov 26 '15 at 18:35

By the Mean Value Theorem applied to $f(x)=e^{1/x}$ with $f'(x)=-x^{-2}e^{1/x}$, we have $$e^{1/\sin x}-e^{1/x}=f(\sin x)-f(x)=(x-\sin x)f'(\xi)=\frac{x-\sin x}{\xi^2}\cdot e^{1/\xi} $$ with $\xi$ between $x$ and $\sin x$.

We can find $a>0$ such that for all small enough positive $x$ we have $$\tag1 \sin x < x -ax^3 $$ and hence $$\frac{x-\sin x}{\xi^2}>\frac{x-\sin x}{x^2}>ax.$$ Thus for small $x$ with $t:=\frac 1t$ $$e^{1/\sin x}-e^{1/x}=\frac{x-\sin x}{\xi^2}\cdot e^{1/\xi} >axe^{1/x}=a\cdot \frac{e^t}{t}\to +\infty$$ because the exponential grows stronger than any polynomial (You might use $e^x\ge 1+x$ for all $x$ $\implies e^t=(e^{t/2})^2\ge(1+\frac t2)^2=1+t+\frac14t^2$ for all $t>-2$).

How can we show $(1)$? Pick any $a$ with $0<a<\frac16$ and let $g(x)=x-ax^3-\sin x$. Then $g'(x)=1-3ax^2-\cos x$, $g''(x)=-6ax+\sin x$, $g''(x)=-6a+\cos x$, so $g'''$ is strictly decreasing from positive $1-6a$ to negative $-6a$ on $[0,\frac\pi2]$ and has a unique root $x_0\in(0,\frac\pi2)$. Thus $g''$ is strictly increasing in the interval $[0,x_0]$. Then says $g''(x)> g''(0)=0$ for all $x\in (0,x_0]$, so staht $g'$ is strictly increasing in $[0,x_0]$. Thus $g'(x)>g'(0)=0$ for all $x\in(0,x_0]$, so that $g$ is strictly increasing on $[0,x_0]$. Thus $g(x)>g(0)=0$ for $0<x\le x_0$, in other words: $(1)$.

| cite | improve this answer | |

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.