4

Reviewing old homework sets for a class and I came across this integral:

$$\displaystyle \int_0^1 \int_0^{1-y} \cos\Big( \frac{x-y}{x+y} \Big) \; dx dy,$$

which the question suggests to evaluate using a change of coordinates; however, I haven't a clue where to begin to identify a useful change of coordinates.

I tried $u = x-y$ and $v = x+y$, but then wasn't sure how I'd convert the domain of integration.

After that, I looked to the given limits for inspiration and noticed that $0<x<1-y$ could be rewritten $y < x+y < 1$, so tried $u = x+y$ and $v = y$, which yielded

$$\displaystyle \int_0^1 \int_y^{1} \cos\Big( \frac{ u - 2y }{u} \Big) \; du\,dy,$$

but that doesn't seem any simpler than the original, to me.

Any advice would be appreciated!

Rax Adaam
  • 1,176

2 Answers2

3

Note that by $u = x-y$ and $v = x+y$

$$du\,dv=|J| \,dx \,dy \iff du\,dv=\begin{vmatrix} 1 & -1 \\ 1 & 1 \end{vmatrix}dx\,dy\iff dx\,dy=\frac12du\,dv$$

then

$$\int_0^1 \int_0^{1-y} \cos\Big( \frac{x-y}{x+y} \Big) \; dx dy=\frac12\int_0^1 dv \int_{-v}^{v} \cos\Big( \frac{u}{v} \Big) \; du$$

enter image description here

user
  • 154,566
  • 1
    Very nice! but I think the second picture should have $v=1$, not $u = 1$, no? The integral as you wrote it (w.r.t. $dv$ first) isn't any easier otherwise, I think... I basically got the same, except rotated by 90 degrees CCW. – Rax Adaam Jan 25 '18 at 22:02
  • @RaxAdaam ops sorry I've inverted the variables! I fix – user Jan 25 '18 at 22:09
2

Your original substitution was right. The original domain of integration is the triangle bounded by $x=0$, $y=0$, and $x+y=1$.

Since $x = \frac12(u+v)$, $y=\frac12(v-u)$, and $x+y=v$, these boundary curves correspond to the curves $v=-u$, $v=u$, and $v=1$ in the $uv$-plane.

Can you finish now? (You need to stop to consider whether you prefer to integrate $du\,dv$ or $dv\,du$.)

Ted Shifrin
  • 115,160
  • I am getting everything except how you found $v = u$; the integration afterwards seems simple if I first integrate $u$; however, the (systematic) process for identifying the boundary curves is still ambiguous for me... – Rax Adaam Jan 25 '18 at 21:56
  • never mind - I see it now: it's literally the equations of the boundary curves + the transformations. Thank you very much for your help! – Rax Adaam Jan 25 '18 at 21:58
  • 1
    You're very welcome. By the way, if you want to see more examples (and trickier ones), you might look at some of my lectures — in particular, MATH 3510 Days 14 and 15. In general, this is a bit more theoretical of a course, but there are plenty of applications and computations, as well. – Ted Shifrin Jan 25 '18 at 22:07