A problem on finding the nearest points to the origin on the intersection of two surfaces

676 Views Asked by At

Suppose we are to find the points nearest to the origin on the curve of intersection of the two surfaces $g^{-1}_{1}\{ 0 \}$ and $g_{2}^{-1}\{ 0 \}$, where $g_{1}: (x, y, z) \mapsto x^{2} - xy + y^{2} - z^{2} - 1$ and $g_{2}: (x, y, z) \mapsto x^{2} + y^{2} - 1.$ Since on $g_{1}^{-1}\{ 0 \} \cap g_{2}^{-1}\{ 0 \}$ we have $xy + z^{2} = 0,$ we may define $g: (x, y, z) \mapsto xy + z^{2},$ so that $g^{-1}\{ 0 \} = g_{1}^{-1}\{ 0 \} \cap g_{2}^{-1}\{ 0 \}.$ Then we are to minimize the function $f: (x, y, z) \mapsto x^{2} + y^{2} + z^{2}$ on $g^{-1}\{ 0 \}.$

If $(x, y, z) \in g^{-1}\{ 0 \}$ such that $f(x, y, z)$ is a local extremum and $\nabla g(x, y, z) \neq 0$, then there is a $\lambda \in \mathbb{R}$ such that $$\lambda \nabla f(x, y, z) = (2\lambda x, 2\lambda y, 2\lambda z) = \nabla g(x, y, z) = (y, x, 2z)$$

However, it seems to me that the vector equation above in combination with the constraint $g(x, y, z) = 0$ lead to $x = y = z = 0.$ Where did I commit an error?

1

There are 1 best solutions below

11
On BEST ANSWER

I didn't really see any error in your computation. However you shouldn't drop one constraint. Instead you should keep the two simpler constraints

$$g_2(x,y,z)=x^2+y^2-1\\ g(x,y,z)=xy+z^2$$

Use two parameters for Lagrange multiplier:

$$L=f(x,y,z)+\lambda g_2(x,y,z)+\mu g(x,y,z)$$

Then set up:

$$\nabla f+\lambda \nabla g_2 +\nabla g=0$$

together with the two constraints.

Try to continue from there. There are several cases to consider. You will get nontrivial solutions.

To answer your question about the geometry: You can try letting $x,y,z$ be constants and visualize the cross-sections. They seem like parabola and hyperbola, and shrinking toward the origin. Now including the cylinder $x^2+y^2=1$, you'll get a closed curve (probably two). Notice that your objective function has level surfaces as spheres. Shrinking or enlarging them to touch the curve would give some maxima and\or minima.