I was reviewing for the A.P. calculus exam, when I got to maximization/minimization problems. I cam across this problem http://jwilson.coe.uga.edu/EMT668/EMAT6680.Folders/Howard/MaximumVol.Cone/MaxVol.CONE.slh.html (not here, in my calculus book), and solved it. I wondered what the solution would be if you could cut it into as many sectors as you want, and you turned all of them into cones. However, I was not sure how to solve it.
Here is the problem stated. You are given a paper disk. You can cut multiple lines from the edge of the disk to the center to leave you with multiple sectors. You then roll up each sector into a cone. How many sector should you cut the disk into, and how big should each sector be to maximize the total volume of all the cones?
The volume of a cone obtained from a sector with radius $r$ and normalised central angle $x=\alpha/(2\pi)$ is given by: $$ V(x)=x^2\sqrt{1-x^2}, $$ where $0\le x\le1$ and I discarded an inessential multiplicative constant of $\pi r^3/3$.
If we have more than two cones, we can ask whether unfolding two cones, and making from the sum of their sectors a single cone, leads to a cone with a volume larger than the sum of the original volumes. That happens for sector angles $x$ and $y$ such that $V(x+y)>V(x)+V(y)$, that is when: $$ F(x,y)={V(x+y)\over V(x)+V(y)} >1. $$ I plotted (with Mathematica) a contour-plot of $F(x,y)$:
As you can see, $F(x,y)\ge1$ if $x+y\le0.8$. But with three or more cones, the sum of the two smaller normalised central angles cannot be more than $2/3$. Hence merging the two smaller cones into a single cone will always lead to a greater total volume, and we then obtain the maximum volume with just two cones.
The solution for two cones is well known: the sum of the volumes is given by $f(x)=V(x)+V(1-x)$ whose derivative is $$ f'(x)=2 x \sqrt{1-x^2}-\frac{x^3}{\sqrt{1-x^2}} +\frac{(1-x)^3}{\sqrt{x(2-x)}}-2 (1-x)\sqrt{x(2-x)} . $$ Equating that to zero, and eliminating square roots, leads to a cubic equation for $(x-1/2)^2$ with three real solutions, in addition to the "trivial" solution $x=1/2$ (which corresponds to a local minimum). The maximum occurs for $x\approx 0.324014$ and its symmetric $1-x$, with $f(x)\approx 0.43606$.