find extrema of $f(x,y,z)=z$ with domain

142 Views Asked by At

$D=\{(y^2+z^2)/6\le x,x^2+y^2=z^2+16\}$ in $\mathbb R$

on $f:D \to R ,f(x,y,z)=z$

  • First off just with a brief look at my function I can say that there are no critical points ($f_x=0,f_y=0,f_z=1$), (morover its an open set which means that the max and min cannot occur ? correct me if I'm wrong)
  • I can now analyze the domain $D$, what I can say is that :

$(y^2+z^2)/6\le x\Longrightarrow$ It's a kind of Paraboloid

$x^2+y^2=z^2+16 \Longrightarrow$ It's a Hyperboloid of One Sheet

If I want to find the boundary of $D$, I need to put those two equations in a system, finding: $x^2-2z^2+6x-16=0$ which is the intersection between those two surfaces.

  • Now in order to find possible min/max points, I can use the Lagrange multiplier system with $x^2-2z^2+6x-16=0$ as a constraint.

$$\lambda (2x+6)=0$$

$$0=0$$

$$1+\lambda (-2z)=0$$

$$x^2-2z^2+6x-16=0$$

  • From the first equation I can say that it is TRUE for $\lambda = 0$ or $x=-3$ , but $\lambda$ cannot be zero becouse It doesn't satisfy the third equation . What I can do instead is using $x=-3$ in the forth equation , but here is the problem : I remain with $-z^2=25$ and I conclude that I didn't find any points. Even if I use the rhird equation finding $z$ and putting it inside the forth equation it still gives me something like $\lambda^2 $ equal to a negative number.

  • Where did I make the mistake? (maybe the boundary ?)

2

There are 2 best solutions below

1
On BEST ANSWER

The domain $D$ is a two-dimensional wobbly disc cut out from the hyperboloid. The boundary of this disc is the curve $\gamma$ of intersection between the hyperboloid and the paraboloid. This curve is determined by two constraints (and not by one, as you seem to believe).

Drawing a figure showing the projection to the $(x,z)$-plane suggests that the point on $\gamma$ with maximal $z$-coordinate is on the symmetry plane $y=0$. There are two such points, namely $P_\pm:=\bigl(8,0, \pm4\sqrt{3}\bigr)$. These two points would then lead to the extremal values $\pm4\sqrt{3}$ of $f\restriction D$.

Doing it the hard way we first remark that the hyperboloid has no horizontal tangent planes. It follows that necessarily the extrema of $f\restriction D$ are taken on $\partial D=\gamma$. We therefore set up the Lagrangian $$\Phi:=z-\lambda(6x-y^2-z^2)-\mu(x^2+y^2-z^2-16)$$ and then obtain the equations $$\eqalign{\Phi_x&=-6\lambda-2\mu x=0 \cr \Phi_y&=2\lambda y-2\mu y=0\cr \Phi_z&=1+2\lambda z+2\mu z=0\ .\cr}$$ From the second equation it follows that $y=0$ or $\lambda=\mu$. The case $y=0$ leads to the points $P_\pm$ we have already found. When $\lambda=\mu$ we are left with the equations $$ \lambda(3+x)=0,\quad 4\lambda z=-1\ .$$ This enforces $\lambda\ne0$, hence $x=-3$. But there are no points in $D$ with $x=-3$.

To sum it up: The extrema of $f\restriction D$ are taken in the points $P_\pm$.

2
On

Using a slack variable $\epsilon$ to convert the inequality into an equality we have

$$ L(z,y,z,\lambda,\mu,\epsilon) = z+\lambda\left(\frac 16(y^2+z^2)-x+\epsilon^2\right)+\mu(x^2+y^2-z^2-16) $$

the stationary points are the solutions for

$$ \nabla L = \left\{ \begin{array}{rcl} 2 \mu x-\lambda &=&0 \\ \frac{\lambda y}{3}+2 \mu y&=&0 \\ \frac{\lambda z}{3}-2 \mu z+1&=&0 \\ \epsilon ^2-x+\frac{1}{6} \left(y^2+z^2\right)&=&0 \\ x^2+y^2-z^2-16&=&0 \\ 2 \epsilon \lambda &=&0 \\ \end{array} \right. $$

which are

$$ \left[ \begin{array}{cccccc} x & y & z & \lambda & \mu & \epsilon \\ 8 & 0 & -4 \sqrt{3} & \frac{2 \sqrt{3}}{5} & \frac{\sqrt{3}}{40} & 0 \\ 8 & 0 & 4 \sqrt{3} & -\frac{2 \sqrt{3}}{5} & -\frac{\sqrt{3}}{40} & 0 \\ \end{array} \right] $$

As we can observe the two stationary points are with $\epsilon = 0$ meaning that the restriction is actuating and they are at the feasible region border.

Attached a plot showing in red the extrema locations as well as the intersection curve between the borders of the two restrictions in blue.

enter image description here

Attached the MATHEMATICA script to produce the plot h = (y^2 + z^2)/6 - x; g = (x^2 + y^2 - z^2 - 16); p1 = {8, 0, -4 Sqrt[3]}; p2 = {8, 0, 4 Sqrt[3]}; gr1 = Graphics3D[{Red, Sphere[p1, 0.3]}]; gr2 = Graphics3D[{Red, Sphere[p2, 0.3]}]; gr0 = ContourPlot3D[{h == 0, g == 0}, {x, -10, 10}, {y, -10, 10}, {z, -10, 10}, MeshFunctions -> {Function[{x, y, z, f}, h - g]}, MeshStyle -> {{Thick, Blue}}, Mesh -> {{0}}, ContourStyle -> Directive[Orange, Opacity[0.5], Specularity[White, 30]], PlotPoints -> 40]; Show[gr0, gr1, gr2, PlotRange -> All] `