How find maximum this integral $$\int_{0}^{y}\sqrt{x^4+\left(y-y^2\right)^2}dx$$ where $y\in \left[0,1\right]$?
-
Could the fundamental theorem of calculus be of some help ? – Claude Leibovici Dec 27 '14 at 17:55
-
Not really, you still have a nasty integral left over because $y$ is inside the integral too. – Empy2 Dec 27 '14 at 17:59
-
It is better to prove that $f(y)$ is an increasing function, hence the maximum $\frac{1}{3}$ is attained in $y=1$. – Jack D'Aurizio Dec 27 '14 at 18:08
-
ok, how prove $f(y)=\int_{0}^{y}\sqrt{x^4+\left(y-y^2\right)^2}dx$ is an increasing function? – piteer Dec 27 '14 at 18:13
-
@piteer: $g(y)=y(1-y)$ is increasing over $I=[0,\frac{1}{2}]$ hence $f$ is increasing over $I$. Can you complete this argument and prove that $f$ is increasing over $[1/2,1]$, too? – Jack D'Aurizio Dec 27 '14 at 18:31
-
@piteer: notice that it is sufficient to prove that for any $y\in[1/2,1]$ we have $f(y)<f(1)=\frac{1}{3}$. – Jack D'Aurizio Dec 27 '14 at 18:37
4 Answers
Since $g(y)=y(1-y)$ is increasing over $I=\left[0,\frac{1}{2}\right]$, so it is $f$. Assume now $y\in\left[\frac{1}{2},1\right]$.
We have: $$ f'(y)=-\int_{0}^{y}\frac{(y-y^2)(2y-1)}{\sqrt{x^4+(y-y^2)^2}}dx+\sqrt{y^4+(y-y^2)^2},\tag{1}$$ but in virtue of the mean value theorem: $$ \int_{0}^{y}\frac{(y-y^2)(2y-1)}{\sqrt{x^4+(y-y^2)^2}}dx \leq y(2y-1)\tag{2}$$ so it is sufficient to prove that: $$ y^2+(1-y)^2 \geq (2y-1)^2 \tag{3}$$ for any $y\in\left[\frac{1}{2},1\right]$, to have that $f$ is increasing over $[0,1]$. However, $(3)$ is equivalent to $y(1-y)\geq 0$, that is trivial. This gives that the maximum of $f(y)$ over $[0,1]$ is attained in $y=1$, and the maximum is $\frac{1}{3}$.
- 353,855
By differentiation under the integral sign we get \begin{align*}\left(\int_0^y\sqrt{x^4+(y-y^2)^2}\,dx\right)'&=\int_0^y\frac{2(y-y^2)(1-2y)}{2\sqrt{x^4+(y-y^2)^2}}dx+\sqrt{y^4+(y-y^2)^2}\\&=\sqrt{y^4+(y-y^2)^2}+\int_0^y\frac{y(1-y)(1-2y)}{\sqrt{x^4+(y-y^2)^2}}dx,\end{align*} which is positive for $y\in(0,\frac12)$. Else suppose $y\in(\frac12,1)$, then it's positive if $$\int_0^y\frac{(2y-1)(y-y^2)}{\sqrt{x^4+(y-y^2)^2}}dx=-\int_0^y\frac{y(1-y)(1-2y)}{\sqrt{x^4+(y-y^2)^2}}dx<\sqrt{y^4+(y-y^2)^2}=y\sqrt{y^2+(1-y)^2}.$$ Clearly, $$\int_0^y\frac{(2y-1)(y-y^2)}{\sqrt{x^4+(y-y^2)^2}}dx<y\cdot\frac{(2y-1)(y-y^2)}{\sqrt{(y-y^2)^2}}=y(2y-1),$$ so it remains to check that $$2y-1\le\sqrt{y^2+(1-y)^2}\iff 0\le y^2+(1-y)^2-(2y-1)^2=-2y^2+2y=2y(1-y)$$ So we can conclude the maximum is attained for $y=1$, where it's $$\int_0^1\sqrt{x^4}\,dx=\int_0^1 x^2\, dx=\frac13.$$
- 16,422
-
Whoops, I have just posted a solution that is almost identical to yours, I did not notice you wrote it. Hope you don't mind. (+1), anyway. – Jack D'Aurizio Dec 27 '14 at 20:46
-
@JackD'Aurizio Well, that sometimes happens. As long as it's independent, it's perfectly fine, so have a +1 too :) – user2345215 Dec 27 '14 at 20:54
HINT:
By taking derivative with respect to $y$ and setting it to be equal to $0$ we have : $$\int_{0}^{y}\frac{(y-y^2)(1-2y)}{\sqrt{x^4+(y-y^2)^2}}dx+\sqrt{y^4+(y-y^2)^2}=0$$ Hence: $$\int_{0}^{y}\frac{dx}{\sqrt{x^4+(y-y^2)^2}}dx=\frac{\sqrt{y^4+(y-y^2)^2}}{(y-y^2)(2y-1)}$$
- 2,379
-
-
@piteer: nothing, since equality cannot hold for $y\in[0,1]$. The given function is increasing over $[0,1]$, hence the maximum is attained in the right endpoint, not in a stationary point. – Jack D'Aurizio Dec 27 '14 at 18:29
$$\begin{align}\frac{d}{dy}(f(y)) &= \frac{d}{dy} \Bigg(\int_{0}^{y}\sqrt{x^4+(y-y^2)^2}dx\Bigg) \\&= \int_{0}^{y}\frac{(y -y^2)(1-2y)}{\sqrt{x^4+(y-y^2)^2}} dx + \sqrt{y^4+(y-y^2)^2} . 1\end{align}$$
Now $$\begin{align}&\int_{0}^{y}\frac{(y -y^2)(1-2y)}{\sqrt{x^4+(y-y^2)^2}} dx + \sqrt{y^4+(y-y^2)^2} = 0 \\&\Rightarrow \int_{0}^{y}\frac{(y -y^2)(1-2y)}{\sqrt{x^4+(y-y^2)^2}} dx = - \sqrt{y^4+(y-y^2)^2} \\&\Rightarrow \int_{0}^{y}\frac{1}{\sqrt{x^4+(y-y^2)^2}} dx = \frac{\sqrt{y^4+(y-y^2)^2}}{(y-y^2)(2y-1)} \end{align}$$
Check the sign of $\frac{d}{dy}f(y)$ to see if the function $f(y)$ is decreasing, increasing.
- 17,571
-
-
-
I cannot see an obvious way to derive $f'(y)>0$ from your calculations. – Jack D'Aurizio Dec 27 '14 at 18:32
-
WolframAlpha solves the integral in terms of an elliptic integral, but I don't know how to solve an equation with elliptic integrals in it. – Empy2 Dec 27 '14 at 18:32
-
@Michael: also because there are no solutions in $[0,1]$. The maximum is achieved in an endpoint, not in a stationary point. – Jack D'Aurizio Dec 27 '14 at 18:48