2

Suppose that $-1<a<1$ and consider $$g(a)=\int_0^{π/2}\ln\left(\frac{1+a\cos x}{1-a\cos x}\right)\frac{\mathrm dx}{\cos x}.$$

How can I find the value of $g(a)$?

I believe the denominator, $\cos x$, may be problematic when it equals $0$. How would I deal with it?

Joel
  • 16,256
lzane
  • 101
  • Sorry, I found where I evaluated this: http://math.stackexchange.com/questions/571570/evaluating-integral-int-0-frac-pi2-log-left-frac1a-cosx1-a-cosx/571596#571596 – Ron Gordon Jun 09 '15 at 16:44
  • @RonGordon thanks for your reminding – lzane Jun 09 '15 at 16:47
  • @RonGordon: The trouble with these duplicate questions is that sometimes there is no way to search whether they have already been asked before posting. What would be the search terms? In this very case, for instance, would you flag this question as being a duplicate? – Alex M. Jun 09 '15 at 16:50
  • @AlexM.: This is one of the trickiest and most frustrating aspects of this site. How are people supposed to know if a problem has been solved exactly already? It is for this reason that I am not a fan of closing questions as being duplicates. However, in this case, this problem is an exact duplicate. I not only should flag - I think I can just close it, as I should have. The good news is that you should be able to just move your answer to the original problem - I think your method had not been done, but you need to make sure of that. – Ron Gordon Jun 09 '15 at 16:55
  • @RonGordon: Is there a way of moving answers, other than by copying and then pasting them? Wow, how does one do this? – Alex M. Jun 09 '15 at 17:07
  • @AlexM.: uhhh...copy and paste. Not that hard. Easier if you open two instances of the webpage. – Ron Gordon Jun 09 '15 at 17:08
  • @RonGordon We should probably improve the other question. – Random Variable Jun 09 '15 at 17:16
  • @RandomVariable: Have at it. You can change the condition on $a$ for starters. – Ron Gordon Jun 09 '15 at 17:18
  • @RonGordon What should the condition on $a$ be? It seems correct to me. – Random Variable Jun 09 '15 at 17:24
  • @RandomVariable: $|a| \le 1$ – Ron Gordon Jun 09 '15 at 17:24
  • @RonGordon: Remarcable. It also finds an even older version. – Alex M. Jun 09 '15 at 17:24
  • @AlexM. That even older thread asks about a specific approach and only that approach is explained. – Random Variable Jun 09 '15 at 17:38
  • Interesting how much attention many questions (as well the answers) got in the old time compared to nowadays. – tired Jun 09 '15 at 21:15
  • @Dr.MVI think you have already given me a wonderful amswer, thankyou – lzane Jun 10 '15 at 23:59
  • @zane Thank you!! That made my day. – Mark Viola Jun 11 '15 at 00:03
  • @Dr.MV I'm new here. Can I select more than one correct answers? I found when I select another , the original one is miss – lzane Jun 11 '15 at 01:52
  • @zane You can give as many up votes as you like, but you can only designate one "best" answer. – Mark Viola Jun 11 '15 at 02:12

4 Answers4

2

Derive with respect to $a$ and get

$$g'(a) = \int \limits _0 ^{\frac \pi 2} \frac {1 - a \cos x} {1 + a \cos x} \frac {\cos x (1 - a \cos x) + \cos x (1 + a \cos x) } {(1 - a \cos x)^2} \frac 1 {\cos x} \Bbb d x = 2 \int \limits _0 ^{\frac \pi 2} \frac 1 {1 - a^2 \cos ^2 x} .$$

Now, make the substitution $t = \tan x$, obtaining $\Bbb d x = \frac 1 {1 + t^2}$ and $\cos ^2 x = \frac 1 {1 + t^2}$, therefore you continue:

$$ g'(a) = 2 \int \limits _0 ^\infty \frac 1 {1 - \frac {a^2} {1+t^2}} \frac 1 {1+t^2} \Bbb d t = 2 \int \limits _0 ^\infty \frac 1 {(1-a^2) + t^2} = \frac 2 {\sqrt {1-a^2} } \arctan \frac t {\sqrt {1-a^2} } \Big| _0 ^\infty = \frac \pi {\sqrt {1-a^2} } .$$

Integrating now back with respect to $a$ you get:

$$g(a) = \pi \int \frac 1 {\sqrt {1-a^2}} \Bbb d a = \pi \arcsin a + C ,$$

with $C$ an integration constant.

Look back at the formula defining $g$ and note that $g(0) = 0$; on the other hand, according to our calculations, $g(0) = C$, so $C=0$ and finally

$$g(a) = \pi \arcsin a .$$

(Here $\log$ is the natural logarithm and we have used that $\log 1 = 0$.)

Edit:

The integral above is improper of the second type, apparently having a singularity in $\frac \pi 2$. In reality, the integrand is bounded and the integral converges because, applying l'Hospital's theorem,

$$\lim \limits _{x \to \frac \pi 2} \frac {\log \frac {1 + a \cos x} {1 - a \cos x}} {\cos x} = \lim \limits _{x \to \frac \pi 2} \frac {\frac {1 - a \cos x} {1 + a \cos x} \frac {- a \sin x (1 - a \cos x) - (1 + a\cos x) a \sin x} {(1-a \cos x)^2}} {- \sin x} = 2a .$$

Alex M.
  • 35,207
1

Let $I(a)$ be the function

$$I(a)=\int_0^{\pi/2}\log\left(\frac{1+a\cos x}{1-a\cos x}\right)\frac{dx}{\cos x}$$

Then, $I'(a)$ is given by

$$\begin{align} I'(a)&=\int_0^{\pi/2}\frac{dx}{1+a\cos x}dx+\int_0^{\pi/2}\frac{dx}{1-a\cos x}dx\\\\ &=\frac{2}{\sqrt{1-a^2}}\left(\arctan \left(\sqrt{\frac{1-a}{1+a}}\right)+\arctan\left(\sqrt{\frac{1+a}{1-a}}\right)\right) \tag 1\\\\ &=\pi{\sqrt{1-a^2}}\tag 2 \end{align}$$

where we used the Wierstrauss substitution to obtain $(1)$ and made use of the identity $\arctan x+\arctan (1/x)=\pi/2$ to obtain $(2)$.

Now, integrating $I'(a)$ from $(3)$ reveals that

$$I(a)= \pi \arcsin (a)+C$$

whereupon using $I(0)=0$ implies that

$$I(a)= \pi \arcsin (a)$$

Mark Viola
  • 179,405
0

The denominator diverges for $x\to\frac\pi2$. Let $\epsilon=\frac\pi2-x$. We have for $\epsilon\to0$, $\cos x=\sin\epsilon\approx\epsilon$, whereas $\ln(1+a\cos x)\approx a\epsilon$ and $\ln(1-a\cos x)\approx-a\epsilon$. Therefore, in $x=\frac\pi2$ the limit is $2a$ and the integral converges.

To compute $g(a)$ I suggest to keep in mind that $g(0)=0$ and derivate with respect to $a$.

Tom-Tom
  • 6,867
0

Unless I made a mistake $$ g'(a) = 2\int_0^{\pi/2}\frac{1}{1-a^2\cos^2 x}dx $$ Which may or may not be easier to integrate with respect to $x$ and then by $a$.

Chinny84
  • 14,186
  • 2
  • 22
  • 31
  • But when $x=π/2$ the denominator $cosx=0$ , in this case , can I also calculate g′(a)? I think maybe it should add something – lzane Jun 09 '15 at 16:05
  • The problem is you are looking at just the denominator. if you look at your integral you have $\lim_{x\to \pi/2} \frac{\ln (1)}{0}$ in the integrand so you have to check if the limit exists, right? – Chinny84 Jun 09 '15 at 16:08