3
$\begingroup$

Evaluate $\lim_{n\to\infty}\int_0^{\frac {\pi}{3}}\frac {\sin^nx}{\sin^nx+\cos^nx}dx$

I tried using the substitution $u=\frac {\pi}{2}-x$ and maybe thought this function might be symmetrical in some way and got: $$\lim_{n\to\infty}\int_{\frac {\pi}{6}}^{\frac {\pi}{2}}\frac {\cos^nx}{\sin^nx+\cos^nx}dx$$

but it's not really helping me... any other ideas?

$\endgroup$
  • 2
    $\begingroup$ Hint: if $a, b > 0$, then $$ \lim_{n\to\infty} \frac{a^n}{a^n+b^n} = \begin{cases} 1, & a>b \\ \frac{1}{2}, & a = b \\ 0, & a < b \end{cases} $$ $\endgroup$ – Sangchul Lee Jun 7 '18 at 6:13
  • $\begingroup$ @SangchulLee Yeah but $\cos x$ and $\sin x$ to $\infty$... I don't think they work like that $\endgroup$ – C. Cristi Jun 7 '18 at 6:16
  • 1
    $\begingroup$ @SangchulLee Obviously useful, but is it sufficient? The limit function is not continuous, so convergence is not uniform. Can we still justify integrating the pointwise limit? If dominated convergence is a known tool, then... $\endgroup$ – Jyrki Lahtonen Jun 7 '18 at 6:17
  • 1
    $\begingroup$ @JyrkiLahtonen, That is a valid objection. My initial guess was that this is a calculus-level question so that certain technicality may be deliberately ignored. If a basic level of analysis is also available, then writing $I_n$ for the integral and using the monotonicity of the integrand of $I_n$, with $a_\pm=\frac{\pi}{4}\pm\epsilon$ we have $$ \left(\frac{\pi}{3}-a_+\right)\frac{\sin^n(a_+)}{\sin^n(a_+)+\cos^n(a_+)}\leq I_n\leq a_-\frac{\sin^n(a_-)}{\sin^n(a_-)+\cos^n(a_-)}+\frac{\pi}{3}-a_-.$$ This gives a bound on limsup/inf and then letting $\epsilon\to0$ yields the answer. $\endgroup$ – Sangchul Lee Jun 7 '18 at 6:28
  • $\begingroup$ @C.Cristi, My intention was that you may take a leap of faith to interchange the order of integral and limit$$\lim_{n\to\infty}\int_{0}^{\frac{\pi}{3}}\frac{\sin^n x}{\sin^n x+\cos^n x}\,dx=\int_{0}^{\frac{\pi}{3}}\lim_{n\to\infty}\frac{\sin^n x}{\sin^n x+\cos^n x}\,dx$$ and then evaluate the inner limit. Of course, this is indeed a huge leap and deserved some justification for mathematical rigor. $\endgroup$ – Sangchul Lee Jun 7 '18 at 6:37
4
$\begingroup$

A heuristic solution. Interchanging the order of integral and limit, we have

$$ \lim_{n\to\infty} \int_{0}^{\frac{\pi}{3}} \frac{\sin^n x}{\sin^n x+\cos^n x} \, dx = \int_{0}^{\frac{\pi}{3}} \lim_{n\to\infty} \frac{\sin^n x}{\sin^n x+\cos^n x} \, dx. $$

Now depending on the relative size of $\sin x$ and $\cos x$, we have

$$ \lim_{n\to\infty}\frac{\sin^n x}{\sin^n x+\cos^n x} = \begin{cases} 0, & \text{if } 0 \leq x < \frac{\pi}{4} \\ \frac{1}{2}, & \text{if } x = \frac{\pi}{4} \\ 1, & \text{if } \frac{\pi}{4} < x \leq \frac{\pi}{3}. \end{cases} \tag{1} $$

So it follows that the limit is $\int_{\frac{\pi}{4}}^{\frac{\pi}{3}} dx = \frac{\pi}{12}$.

Remark. Indeed we took a huge leap by assuming that the integral and the limit can be switched. This is indeed possible in our case, although a direct justification required some advance results such as dominated convergence theorem.

A preliminary analysis level solution. Fix any sufficiently small $\epsilon > 0$ and consider $a=\frac{\pi}{4}-\epsilon$ and $b = \frac{\pi}{4}+\epsilon$. If we write $I_n = \int_{0}^{\frac{\pi}{3}} \frac{\sin^n x}{\sin^n x+\cos^n x} \, dx$, then the integrand is monotone increasing in $x$ and hence

\begin{align*} I_n &\geq \int_{b}^{\frac{\pi}{3}} \frac{\sin^n x}{\sin^n x+\cos^n x} \, dx \\ &\geq \int_{b}^{\frac{\pi}{3}} \frac{\sin^n b}{\sin^n b+\cos^n b} \, dx \\ &= \left(\frac{\pi}{3}-b\right)\frac{\sin^n b}{\sin^n b+\cos^n b} \end{align*}

This gives

$$ \liminf_{n\to\infty} I_n \geq \lim_{n\to\infty} \left(\frac{\pi}{3}-b\right)\frac{\sin^n b}{\sin^n b+\cos^n b} = \frac{\pi}{12}-\epsilon.$$

But since the LHS of the above inequality is a constant independent of $\epsilon$, letting $\epsilon \downarrow 0$ proves that $\liminf_{n\to\infty} I_n \geq \frac{\pi}{12}$. Similarly,

\begin{align*} I_n &\leq \int_{0}^{a} \frac{\sin^n x}{\sin^n x+\cos^n x} \, dx + \int_{a}^{\frac{\pi}{3}} \frac{\sin^n x}{\sin^n x+\cos^n x} \, dx \\ &\geq \int_{0}^{a} \frac{\sin^n a}{\sin^n a+\cos^n a} \, dx + \int_{a}^{\frac{\pi}{3}} dx \\ &= a\frac{\sin^n a}{\sin^n b+\cos^n a} + \left(\frac{\pi}{3}-a\right) \end{align*}

and hence

$$ \limsup_{n\to\infty} I_n \leq \lim_{n\to\infty} \left[ a\frac{\sin^n a}{\sin^n b+\cos^n a} + \left(\frac{\pi}{3}-a\right) \right] = \frac{\pi}{12}+\epsilon. $$

Letting $\epsilon \downarrow 0$, we obtain $\limsup_{n\to\infty} I_n \leq \frac{\pi}{12}$. These together tell that

$$\liminf_{n\to\infty} I_n = \limsup_{n\to\infty} I_n = \frac{\pi}{12}$$

and therefore the limit of $I_n$ exists and has the value $\frac{\pi}{12}$.


Addendum. The following demonstrates graphs of the integrand for different values of $n$.

$\hspace{5em}$graphs of the integrand

The graph is already quite close to $\text{(1)}$ when $n=100$, which provides a sanity check.

$\endgroup$
  • $\begingroup$ What happened with the cases when $x$ is in between $0\leq x \leq \frac {\pi}{4}$? $\endgroup$ – C. Cristi Jun 7 '18 at 7:08
  • $\begingroup$ @C.Cristi, Are you fine with the claim $$a,b>0\quad\Rightarrow\quad\lim_{n\to\infty} \frac{a^n}{a^n+b^n} = \begin{cases} 1, & a>b \\ \frac{1}{2}, & a = b \\ 0, & a < b \end{cases}?$$ If then, simply consider $a=\sin x$ and $b=\cos x$ for each given $x\in(0,\frac{\pi}{4}]$. (The case $x=0$ is trivial.) $\endgroup$ – Sangchul Lee Jun 7 '18 at 7:10

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.