1
$\begingroup$

I'm currently trying to solve the integral: $$ I(\vec{a},\vec{b})=4\pi\int\limits_0^1\frac{\mathrm{d}u}{1-(\vec{a}u+\vec{b}(1-u))^2}, $$ but I can't seem to find a good starting point. I know that if $\vec{a}$ and $\vec{b}$ are scalars this integral would be easilly solved by a partial fraction decomposition (where the relation $x^2-y^2=(x-y)(x+y)$ is really helpful), yielding an answer in terms of the natural logarithm.

Right now however I am stuck with a vector relation, which doesn't make the calculations any easier. I tried the partial fraction decomposition, this doesn't make things easier. I also tried writing out the denominator, but this also isn't that helpful.

I was wondering if there were general ways to tackle these kind of problems and/or if there are any tips for this integral ?

The vectors $\vec{a}$ and $\vec{b}$ are 3D, this can for example be seen by the fact that the factor $4\pi$ follows from the 3D solid angle $$ 4\pi=\int\limits_0^{2\pi}\mathrm{d}\varphi\int\limits_{0}^\pi\sin(\theta)\mathrm{d}\theta. $$

$\endgroup$
4
  • $\begingroup$ It just wants you to do it component wise, so if a and b are 3-vectors, you have 3 integrals to do! One for their first, then second, then third components $\endgroup$
    – Alec Teal
    Apr 8, 2015 at 7:01
  • 1
    $\begingroup$ @AlecTeal, this is a scalar integral; note the squaring in the denominator, so that the only the norm of that vector is used. $\endgroup$ Apr 8, 2015 at 7:05
  • $\begingroup$ @AlecTeal, and even if it were vectors, how would you define 1/(1-vector) ? As far as I know this is a pretty undefined relationship. How do you even sum a scalar and a vector ? $\endgroup$
    – Nick
    Apr 8, 2015 at 7:41
  • $\begingroup$ @AlexMeiburg didn't spot the squared, that makes it a bit of a push, but none the less stuff like $\int\vec{u}dt$ can mean componentwise. $\endgroup$
    – Alec Teal
    Apr 8, 2015 at 8:05

2 Answers 2

1
$\begingroup$

As you said, integrals are best tackled with scalars. Thus the most straightforward way to solve these types of problems (unless there's some other nice geometric property about the vectors to make it easier) is to deal with the individual components. Instead of $\vec a$ and $\vec b$, deal with $a_x, a_y, b_x, b_y$. Then we have $$(\vec a u + \vec b(1-u))^2 = (a_x u+b_x(1-u),a_y u+b_y(1-u))^2 = b_x^2 + b_y^2 + 2(a_x b_x - b_x^2 + a_y b_y - b_y^2) u + (a_x^2 + a_y^2 - 2 a_x b_x + b_x^2 - 2 a_y b_y + b_y^2) u^2$$ Which is -- from the perspective of $u$, the variable we're integrating -- just a quadratic. So the integral becomes $$\int \frac{du}{(1-b_x^2+b_y^2) + 2(b_x^2 + b_y^2 - a_x b_x - a_y b_y) u + (2 a_x b_x + 2 a_y b_y - b_x^2 - b_y^2 - a_x^2 - a_y^2 ) u^2}$$ or perhaps more succinctly (and geometrically) $$\int \frac{du}{(1-|b|^2) + 2(|b|^2-a\cdot b)u + (2a\cdot b-|b|^2-|a|^2)u^2}$$ Then you can apply that $$\int_0^1 \frac{du}{x + y u + z u^2} = \frac{2\left(\arctan\left(\frac{y+2z}{\sqrt{4xz-y^2}}\right)-\arctan\left(\frac{y}{\sqrt{4xz-y^2}}\right)\right)}{\sqrt{4xz-y^2}}$$ and plug in those values for $x, y, z$ to get your integral.

$\endgroup$
3
  • $\begingroup$ that was indeed the way that I was thinking. Now the vectors are (unfortunately) 3D, so I'm not really looking forward to the extensive bookkeeping (edited the question). Your first equality (the vector product) has a square to many ;). $\endgroup$
    – Nick
    Apr 8, 2015 at 7:16
  • $\begingroup$ I'm going to spare you the pain of writing it all out for 3D and simply accept ;). I worked in the solution with the scalar quantities $A=1-\vec{a}\cdot\vec{b}$ and $B=\sqrt{(1-a^2)(1-b^2)}$, this saved me a lot of effort. $\endgroup$
    – Nick
    Apr 8, 2015 at 9:22
  • $\begingroup$ That sounds like it's probably more elegant. :) Side note though, the above still works with 3D just as well, though, once I rewrote it in terms of norms/dots. This is because you can choose your "x" and "y" directions to be in the plane defined by $\vec a$ and $\vec b$, and then your z component is 0. In a general sense, an equation proven about $n$ vectors in $n$ space is also true for any higher space. $\endgroup$ Apr 8, 2015 at 18:08
0
$\begingroup$

First write your norm $F(u) = \left\Vert u \vec{a} u + (1-u)\vec{b}\right\Vert^2$ in the form of a quartic in $u$, as $A u^2 + B u + C$ (Use the expression of vectors $\vec{a}$, $\vec{b}$ in orthogonal coordinates if needed). Then it should be straightforward (there is a general, well-known method for integrating $1/\mathrm{quartic}$).

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .