6
$\begingroup$

I am stuck with the integration

$$ I=\int\ln\left(\frac{L}{2}+\sqrt{\frac{L^2}{4}+y^2+z^2}\right)\ \mathrm dy $$

I got this from the question from the book

"Field and wave electromagnetics, Cheng, 2nd, Problem 3-18.

I tried to solve this equation using method of integration by parts, but my equation got worse.

I know the answer by Wolfram Alpha, but I can't get how.

$\endgroup$
  • $\begingroup$ If you can do $\int \ln(a+(b+cx^2)^{1/2}) dx$, then you can specialize to your context ($a=L/2$,$b=L^2/4+z^2$). Judging by Wolfram's output, it looks like the first step is the "standard" logarithm integration by parts, with $u=\log(a+(b+cx^2)^{1/2})$ and $dv=dx$. Can you work this step out and then show us what you get? $\endgroup$ – Ian Jun 11 '15 at 3:07
  • $\begingroup$ Alternately, you might begin with a trigonometric substitution: given a right triangle with legs $\sqrt{b}$ and $\sqrt{c}x$, the hypotenuse has length $(b+cx^2)^{1/2}$. So $(b+cx^2)^{1/2}=\sqrt{b} \sec(\theta)$ is a reasonable substitution to start with. Since the final answer involves inverse tangents, this approach is probably also productive. $\endgroup$ – Ian Jun 11 '15 at 3:11
  • $\begingroup$ Integration isn't my strongest side, but an idea came to mind to add the integral $\int{\ln\left(-\frac{L}{2}+\sqrt{\frac{L^2}{4}+y^2+z^2}\right)dy}$, such that the resulting integral can be simplified to $\int{\ln\left(y^2+z^2\right)dy}$. I am not sure how to separate this integral into its original form though. $\endgroup$ – Kwin van der Veen Jun 11 '15 at 3:11
  • $\begingroup$ @Dr.MV Your answer seems fine with me. I can proceed my problem with the help of your aid, without requiring any other materials. $\endgroup$ – user65452 Jun 11 '15 at 10:55
3
$\begingroup$

Following @ClaudeLeibovici, we have

$$I=y\log\left(\frac{L}2+\sqrt{\left(\frac{L}{2}\right)^2+y^2+z^2}\right)-J$$

where

$$\begin{align} J&=\int\frac{y^2}{\left(\frac{L}2+\sqrt{\left(\frac{L}{2}\right)^2+y^2+z^2}\right)\sqrt{\left(\frac{L}{2}\right)^2+y^2+z^2}}dy\\\\ &=-\int\frac{y^2\left(\frac{L}2-\sqrt{\left(\frac{L}{2}\right)^2+y^2+z^2}\right)}{(y^2+z^2)\sqrt{\left(\frac{L}{2}\right)^2+y^2+z^2}}dy\\\\ &=-\frac{L}{2}\int\frac{y^2}{(y^2+z^2)\sqrt{\left(\frac{L}{2}\right)^2+y^2+z^2}}dy+\int\frac{y^2}{(y^2+z^2)}dy\\\\ &=-\frac{L}{2}\int\frac{1}{\sqrt{\left(\frac{L}{2}\right)^2+y^2+z^2}}dy+\frac{L}{2}z^2\int\frac{1}{(y^2+z^2)\sqrt{\left(\frac{L}{2}\right)^2+y^2+z^2}}dy+y-z\arctan(y/z) \\\\ &=K_1+K_2+y-z\arctan(y/z) \end{align}$$

where in $(1)$

$$K_1=-\frac{L}{2}\int\frac{1}{\sqrt{\left(\frac{L}{2}\right)^2+y^2+z^2}}dy$$

and

$$K_2=\frac{L}{2}z^2\int\frac{1}{(y^2+z^2)\sqrt{\left(\frac{L}{2}\right)^2+y^2+z^2}}dy$$


We can easily evaluate $K_1$ by making the substitution $y=\sqrt{\left(\frac{L}{2}\right)^2+z^2}\tan t$. Then,

$$K_1=-\frac{L}{2}\log\left(y+\sqrt{\left(\frac{L}{2}\right)^2+y^2+z^2}\right)$$

where the term $\sqrt{\left(\frac{L}{2}\right)^2+z^2}$ is an integration constant that we omitted.


We effect the same substitution for $K_2$. Then, we have

$$\begin{align} K_2&=\frac{L}{2}z^2\int\frac{1}{(y^2+z^2)\sqrt{\left(\frac{L}{2}\right)^2+y^2+z^2}}dy\\\\ &=\frac{L}{2}z^2\int \frac{\cos t}{\left(\frac{L}{2}\right)^2+z^2\sin^2t}dt\\\\ &=z\arctan\left(\frac{(L/2)\sin t}{z}\right)\\\\ &=z\arctan\left(\frac{(L/2)y}{z\sqrt{\left(\frac{L}{2}\right)^2+y^2+z^2}}\right)\\\\ \end{align}$$


Putting it all together reveals

$$\begin{align} I&=y\log\left(\frac{L}2+\sqrt{\left(\frac{L}{2}\right)^2+y^2+z^2}\right)\\\\ &-y+z\arctan(y/z)\\\\ &+\frac{L}{2}\log\left(y+\sqrt{\left(\frac{L}{2}\right)^2+y^2+z^2}\right)\\\\ &-z\arctan\left(\frac{(L/2)y}{z\sqrt{\left(\frac{L}{2}\right)^2+y^2+z^2}}\right) \end{align}$$

$\endgroup$
  • $\begingroup$ brilliant... it's just simple substitution and product splitting. $\endgroup$ – user65452 Jun 11 '15 at 10:18
  • $\begingroup$ @user65452 Wow! Thank you! $\endgroup$ – Mark Viola Jun 11 '15 at 13:24
  • $\begingroup$ I noticed one minor mistake. In evaluating $K_2$, I got $\frac{L}{2}z^2\int \frac{\cos t}{\left(\frac{L}{2}\right)^2\sin^2t+z^2}dt$. But its result is same, $z\arctan\left(\frac{(L/2)\sin t}{z}\right)$. Can you check? $\endgroup$ – user65452 Jun 17 '15 at 3:19
  • $\begingroup$ @user65452 That isn't actually a mistake. The term that you wrote is merely an integration constant. It depends only on $z$, not $y$. $\endgroup$ – Mark Viola Jun 18 '15 at 1:42
1
$\begingroup$

Considering $$I=\int\log\left(\frac{L}{2}+\sqrt{\frac{L^2}{4}+y^2+z^2}\right)\ \mathrm dy$$ what I would first do is to get rid of the logarithm by a first integration by parts $$u=\log\left(\frac{L}{2}+\sqrt{\frac{L^2}{4}+y^2+z^2}\right)$$ $$du=\frac{y\, dy}{\sqrt{\frac{L^2}{4}+y^2+z^2} \left(\frac{L}{2}+\sqrt{\frac{L^2}{4}+y^2+z^2}\right)}$$ $v'=dy$, $v=y$ which makes $$I=y \log\left(\frac{L}{2}+\sqrt{\frac{L^2}{4}+y^2+z^2}\right)- J$$ where $$J=\int\frac{y^2\, dy}{\sqrt{\frac{L^2}{4}+y^2+z^2} \left(\frac{L}{2}+\sqrt{\frac{L^2}{4}+y^2+z^2}\right)}$$ Now, an apparent change of variable could be $$\frac{L^2}{4}+y^2+z^2=t^2$$ $$y=\frac{1}{2} \sqrt{4 t^2-(L^2+4 z^2)}$$ $$dy=\frac{2 t}{\sqrt{4 t^2-(L^2+4 z^2)}}$$ which makes $$J=\int \frac{\sqrt{4 t^2-(L^2+4 z^2)}}{2t+L}\,dt$$ Again, $2t+L=w$, $t=\frac{w-L}{2}$, $dt=\frac{dw}{2}$ make $$J=\frac 12 \int\frac{\sqrt{w^2-2 L w-4 z^2}}{ w} dw$$ At this point, we find integrals which are given in the Table of Integrals, Series, and Products by I.S. Gradshteyn and I.M. Ryzhik (in the $7^{th}$ edition, look at section $2.267$).

It is effectively quite tedious !

$\endgroup$
  • $\begingroup$ Thank you for that book, Table of Integrals, Series, and Products. It is truly helpful! $\endgroup$ – user65452 Jun 11 '15 at 4:54
  • $\begingroup$ You are very welcome ! This book is very useful when you are as lazy as I am ! Be careful : it can help but it does not teach you !! Cheers. $\endgroup$ – Claude Leibovici Jun 11 '15 at 4:58
  • $\begingroup$ @ClaudeLeibovici It wasn't quite as bad as one might have suspected. $\endgroup$ – Mark Viola Jun 11 '15 at 5:46
  • $\begingroup$ @Dr.MV. It could have been worse but this was really tedious. $\endgroup$ – Claude Leibovici Jun 11 '15 at 10:06
0
$\begingroup$

This is incomplete. But it's close.

We want $I=\int\ln\left(\frac{L}{2}+\sqrt{\frac{L^2}{4}+y^2+z^2}\right)\ \mathrm dy $

If we look at $\frac{L}{2}+\sqrt{\frac{L^2}{4}+y^2+z^2} =\frac{L+\sqrt{L^2+4(y^2+z^2)}}{2} $, this is one of the roots of the quadratic $x^2-Lx-(y^2+z^2) =0$. Since the product of the roots is $-(y^2+z^2) $, if we define $J=\int\ln\left(\frac{L}{2}-\sqrt{\frac{L^2}{4}+y^2+z^2}\right)\ \mathrm dy $, then $I+J =\int\ln(-(y^2+z^2))dy $.

However, the inner expression is negative, so I'll define $J=\int\ln\left(-\frac{L}{2}+\sqrt{\frac{L^2}{4}+y^2+z^2}\right)\ \mathrm dy $. We now get then $I+J =\int\ln(y^2+z^2)dy $. This is, according to Wolfram Alpha, $y (log(y^2+z^2)-2)+2 z tan^{-1}(y/z) $.

What we need now is $I-J$.

The expression in the $\ln$ would be

$\begin{array}\\ \frac{L+\sqrt{L^2+4(y^2+z^2)}}{-L+\sqrt{L^2+4(y^2+z^2)}} &=\frac{L+\sqrt{L^2+4(y^2+z^2)}}{-L+\sqrt{L^2+4(y^2+z^2)}}\\ &=\frac{L+\sqrt{L^2+4(y^2+z^2)}}{-L+\sqrt{L^2+4(y^2+z^2)}} \frac{L+\sqrt{L^2+4(y^2+z^2)}}{L+\sqrt{L^2+4(y^2+z^2)}}\\ &=\frac{L^2+L^2+4(y^2+z^2)+2L\sqrt{L^2+4(y^2+z^2)}}{4(y^2+z^2)}\\ &=\frac{L^2}{2(y^2+z^2)}+1+\frac{L\sqrt{L^2+4(y^2+z^2)}}{2(y^2+z^2)}\\ \end{array} $

Again, according to WA,

$\int \frac{\sqrt{x^2+a}}{x^2+b} dx = \frac{\sqrt{a-b} \tan^{-1}\left(\frac{x \sqrt{a-b}}{\sqrt{b} \sqrt{a+x^2}}\right)}{\sqrt{b}}+\ln(\sqrt{a+x^2}+x) $

This would be enough to get $I-J$, and, from these, $I$ amd $J$.

However, it's late and I'm tired, so I'll stop here.

$\endgroup$
0
$\begingroup$

We can rework the constants to write the integral in the form $$\int\ln\left(b\left(\sqrt{x^2+4}+a\right)\right)dx$$with $a<2$. The constant $b$ can be pulled out.

Set $x=t-\dfrac 1t$ to get $$\int\ln\left(\sqrt{x^2+4}+a\right)dx=\int\ln\left(t+\frac1t+a\right)\left(1+\frac1{t^2}\right)dt.$$

You can decompose and transform the second term by inverting $t$: $$\int\ln\left(t+\frac1t+a\right)\frac1{t^2}dt=-\int \ln\left(\frac1s+s+a\right)ds=\int\ln(s)\,ds-\int \ln\left(s^2+as+1\right)ds.$$ Now, by parts $$\int\ln\left(t^2+at+1\right)dt=\frac12\left(2t+a\right)\ln\left(t^2+at+1\right)-\frac12\int\frac{4t^2+4at+1}{t^2+at+1}dt\\ =\frac12\left(2t+a\right)\ln\left(t^2+at+1\right)-2t+\frac32\int\frac{dt}{t^2+at+1},$$ and finally $$\int\frac{dt}{t^2+at+1}=\frac4{4-a^2}\int\frac{dt}{\left(\dfrac{2t+a}{\sqrt{4-a^2}}\right)^2+1}=\frac1{\sqrt{4-a^2}}\arctan\left(\frac{2t+a}{\sqrt{4-a^2}}\right).$$ All together, $$I(t)=\int\ln\left(t+\frac1t+a\right)dt=\frac12\left(2t+a\right)\ln\left(t^2+at+1\right)-t+\frac3{2\sqrt{4-a^2}}\arctan\left(\frac{2t+a}{\sqrt{4-a^2}}\right)-t\ln(t),$$ and $$\int\ln\left(b\left(\sqrt{x^2+4}+a\right)\right)dx=I\left(-\frac x2+\sqrt{\frac{x^2}4+1}\right)-I\left(-\frac x2-\sqrt{\frac{x^2}4+1}\right)+\ln(b)x.$$

$\endgroup$
  • $\begingroup$ Quel boulot ! Nice solution. $\endgroup$ – Claude Leibovici Jun 11 '15 at 10:09
  • $\begingroup$ Le travail c'est la santé ! I tried to make is as simple as possible, but didn't really succeed :( $\endgroup$ – Yves Daoust Jun 11 '15 at 10:28
  • $\begingroup$ This is brilliant too. There are tones of variation in the method of integration I could use as reference, and this is quite a method. $\endgroup$ – user65452 Jun 11 '15 at 10:51
  • $\begingroup$ @user65452: thanks for the appraisal. I like the substitution $x=t-1/t$ with radicals. $\endgroup$ – Yves Daoust Jun 11 '15 at 10:55

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.