6

How can we show that $\int_0^1\int_{1-y}^1\sqrt{(x-1)(y-1)(x+y-1)}\mathrm dx\mathrm dy=\frac{2\pi}{105}$ ?

Desmos says it's true.

The inner indefinite integral is not nice. And strangely, when I plug in the inner limits of integration $x=1$ and $x=1-y$, I get $0-0=0$, contradicting the Desmos result. (Edit: I should cancel factors in numerator and denominator, then I don't get $0-0$. But then I get $\int_0^1 -\frac14 y^2\sqrt{y-1}\sinh^{-1}(i) \mathrm dy$ and I don't know how to continue.)

Based on another question of mine about a double integral, I tried substituting $a=x-1$ and $b=y-1$, which leads to $\int_{-1}^0\int_{-1-b}^0\sqrt{ab(a+b+1)}\mathrm da\mathrm db$, which has another inner indefinite integral that is not nice.

Context: I asked myself, "A unit stick is broken at two uniformly random points; given that the three pieces form a triangle (which has probability $\frac14$), what is the expected area of the triangle? The answer is $\frac12$ times the integral in this question (which is supported by experimental trials, and also this article). This question was inspired by another question about breaking a stick at random points.

Dan
  • 22,158
  • 2
    I suspect that you will receive a more positive response from MathSE reviewers if you show all of your work, directly in your posting, in a very clear, well organized manner. If you tried more than one approach, please include the additional approaches. Further, assuming that the problem was assigned from a book/class, there must have been previously worked examples, theorems, or solved problems, that led up to this problem. I also suggest that you regard these as your tools, and include a summary of any tools that might be pertinent, directly in your posting. – user2661923 Jan 12 '24 at 14:51
  • 1
    $a=x-1$ is a good substitution here so it's unclear why you had no success afterwards. What did you obtain? As a hint to finish the integral, try to think of a substitution that isolates both variables (after $a=x-1$) - e.g. to obtain something like $\int \int f(x) g(y) dxdy$ – Zacky Jan 12 '24 at 15:20
  • 1
    @Zacky I have editted, showing my result after substitution. Thanks for the hint; I'm not sure what further substitution would isolate both variables. – Dan Jan 12 '24 at 15:26
  • 1
    I don't get $ 0 $ when I plug in either $ x = 1 $ or $ x = 1 - y $ to the indefinite integral that you got from Wolfram Alpha. If I plug these directly into Wolfram's expression, I got $ 0 / 0 $, and if I cancel the factors first (which Wolfram refuses to do automatically, probably because of it defaults to complex numbers), then I get $ 0 $ for neither of them (and I get different results so their difference isn't $ 0 $ either). So while this may not be the best way to approach it, you don't get $ 0 $. – Toby Bartels Jan 12 '24 at 15:37
  • 1
    @Dan Aah I got it now, thanks for sharing. You did the two substitutions in the same time, I thought you meant separately.$$\int_0^1\int_{1-y}^1 \sqrt{(x-1)(y-1)(x-1+y)}dxdy\overset{x-1=a}=\int_0^1 \int_{-y}^0 \sqrt{a(y-1)(a+y)}dady$$ And now I had in mind the substitution $-a=xy$ (which isolates $a$ and $y$ from $a+y$), to get: $$\int_0^1\int_0^1 \sqrt{x(1-x)} y^2\sqrt{1-y} dxdy$$ – Zacky Jan 12 '24 at 21:11
  • @Dan: It seems you answered this old question, in which the OP posted their own answer but didn't manage to evaluate the integral analytically. Perhaps you want to post an answer or comment there? – joriki Jan 13 '24 at 14:39
  • 1
    @joriki I've left a comment there linking to this question. – Dan Jan 13 '24 at 14:52
  • 1
    @Dan: I now also came across this related question. The last comment under that question links to this paper (Expected Areas of Randomly Generated Triangles of Fixed Perimeter by Douglass, Fitzgerald and Mihalik ($2002$), $\Pi$ME Journal, Vol. $11$, No. $7$, pp. $365$–$371$). – joriki Jan 13 '24 at 16:14

6 Answers6

4

Inner integral does not yield $0-0$. In fact, it is the difference of these two: $$ {\frac {{y}^{2}\arcsin(1)}{16}\sqrt {1-y}} \\ \text{and} \\ {\frac {{y}^{2}\arcsin(-1)}{16}\sqrt {1-y}} $$ They are not the same.


Added.

The inner integral.
In your attempt, you wrote $\sqrt{y-1}$, square-root of a negative number. Do you really want to do that? You wrote $\sinh^{-1}(i)$ which is $i \sin^{-1}(1)=i\pi/2$.

Using only real nubers, the inner integral is $$ {\frac {{y}^{2}\pi}{8}\sqrt {1-y}} . $$

Are you really saying you don't know how to do $$ \frac {\pi}{8}\int_0^1 {y}^{2}\sqrt {1-y}\;dy $$

GEdgar
  • 111,679
  • Thanks, I found the mistake I was making, and have editted. But I still don't know how to finish the problem. – Dan Jan 12 '24 at 15:57
3

Continue with $$I= \int_{-1}^0\int_{-1-b}^0\sqrt{ab(a+b+1)}dadb $$ along with the variable changes $a={u+v}$ and $b={u-v}$, to decouple the double integral

\begin{align} I= \int_{-\frac1{2}}^0\int^{-u}_u 2\sqrt{(u^2-v^2)(2u+1)}dv du \end{align} Then, utilize $\int_{-u}^u \sqrt{u^2–v^2}dv=\frac\pi2 u^2$ to obtain

$$I=\pi \int_{-\frac1{2}}^0 {u^2} \sqrt{2u+1}\ du=\frac{2\pi}{105} $$

Quanto
  • 97,352
3

Use the following result which is very interested: $$\int_{a}^{b} \sqrt{(b-x)(x-a)} \mathrm{d}x=\frac{\pi(b-a)^2}{8}.\tag 1$$ For detail we can refer Area of semi-disk.

Let $$D=\{(x,y)\mid0\leq y\leq1,1-y\leq x\leq1\},$$ $$f(x,y)=\sqrt{(1-x)(1-y)(x+y-1)},\quad (x,y)\in D,$$ then $$I=\int_0^1\int_{1-y}^1\sqrt{(x-1)(y-1)(x+y-1)}\mathrm dx\mathrm dy =\iint_{D}f(x,y)\mathrm dx\mathrm dy.$$ By $(1)$, we get $$\int_{1-y}^1\sqrt{(1-x)[x-(1-y)]}\ \mathrm dx=\frac{\pi y^2}{8},$$ so $$I=\int_0^1\sqrt{1-y}\cdot\frac{\pi y^2}{8}\mathrm dy =\frac{2\pi}{105}.$$

Riemann
  • 7,203
1

Self-answering.

With help from Wolfram for the inner anti-derivative, and using $\sinh^{-1}A=\log(A+\sqrt{A^2+1})$, we have:

$\begin{align} \int_{1-y}^1\sqrt{(x-1)(y-1)(x+y-1)}\mathrm dx&=\left[-\frac14y^2\sqrt{y-1}\ln\left({\sqrt{\frac{x-1}{y}}+\sqrt{\frac{x+y-1}{y}}}\right)\right]_{1-y}^1\\ &=-\frac14y^2\sqrt{y-1}\ln{i}\\ &=-\frac{\pi i}{8}y^2\sqrt{y-1}\\ &=\frac{\pi}{8}y^2\sqrt{1-y} \end{align}$

Then the integral in the OP is

$$\frac{\pi}{8}\int_0^1y^2\sqrt{1-y}\mathrm dy=\frac{\pi}{8}\left[-\frac{2}{105}(1-y)^{3/2}(15y^2+12y+8)\right]_0^1=\frac{2\pi}{105}$$

Dan
  • 22,158
1

At the begining, if we write the inner integral as $$\sqrt{1-y}\int_{1-y}^1\sqrt{(1-x)(y+x-1)}dx$$ then by $1-x=u$ $$\sqrt{1-y}\int_{0}^{y}\sqrt{yu-u^2}du$$ and by $\frac y2-u=\frac y2\sin v$, we have $$\frac14y^2\sqrt{1-y}\int_{\frac{-\pi}2}^{\frac{\pi} 2}\cos^2vdv=\frac\pi 8 y ^2\sqrt{1-y}$$ Trigonometric way we know, works in real world. I mean no i.

Bob Dobbs
  • 10,988
1

$$ \begin{aligned} I & =\int_0^1 \int_{1-y}^1 \sqrt{(1-x)(1-y)(x+y-1)} d x d y \\ & =\int_0^1 \sqrt{1-y} \int_{1-y}^1 \sqrt{(1-x)(x+y-1)} d x d y \\ & =\int_0^1 \sqrt{1-y} \cdot\frac{y^2}{2} \int_0^{\frac{\pi}{2}} \sin ^2 2 \theta d \theta,\quad \textrm{via the substitution } y\sin \theta=1-x\\&= \frac{\pi}{8} \int_0^1 y^2 \sqrt{1-y}\ d y\\& =\frac{2 \pi}{105} \quad (\textrm{ via integration by parts twice }) \end{aligned} $$

Lai
  • 20,421