1

Solving an optimization problem in multiple variables, I had to examine a function $$ f(x,y,z)=x^2+2yz $$ defined on a ball $$ \{ (x,y,z)\enspace|\enspace x^2+y^2+z^2\leq1 \}. $$ The boundary is then described by the sphere $x^2+y^2+z^2=1$. According to my textbook, one can search for maxima on the boundary by rewriting $x^2+y^2+z^2=1$ as $x^2=1-y^2-z^2$, then substituting this into $f$ to obtain $$ f(x,y,z)=1-y^2-z^2+2yz,\quad y^2+z^2\leq1 $$ Now what I don't understand is, how come we need to specify that $y^2+z^2\leq1$? The unit sphere is described by $x^2+y^2+z^2=1$ with no need to specify any restrictions on $y$ or $z$, but once we rewrite this as $x^2=1-y^2-z^2$ and plug it into the function, suddenly it becomes necessary to specify that $y^2+z^2\leq1$. Why?

Andrea
  • 1,909
  • 3
  • 18
  • 31

2 Answers2

0

Because $0 \leq x^2 = 1-y^2-z^2$.

Siminore
  • 35,136
0

When you write $$x^2+y^2+z^2=1$$

The set of solutions implicitly has the restriction that $y^2+z^2\le1$ because if this condition is violated there are no solutions in $\mathbb R$.

$$y^2+z^2\gt1$$

$$x^2+y^2+z^2\gt1+x^2$$

$$1\gt1+x^2\implies x^2\lt0\implies x\in \phi$$

Guy
  • 8,857
  • 1
  • 28
  • 57
  • Right, but if it's implicit from the beginning, why do I have to write it explicitly later? Shouldn't it still be implicit? – Andrea Apr 04 '14 at 09:53
  • 1
    @Andreas In $f(x,y,z)$ there is no "$\color{red}{=}$". It is simply an expression. You need to point out, that the inputs into $f$ come from an "$\color{red}{=}$"(which is constrained by some inequalities) – Guy Apr 04 '14 at 10:01
  • To make myself clearer, $f(x,y,z)$ is defined for all $x,y,z$, but you are only using that definition on the points on the sphere. $f(100,100,100)$ has a very well defined value, but that value is never obtained for the points on the sphere. – Guy Apr 04 '14 at 10:02
  • $f$, clearly doesn't have a global maxima, but there is a maximum value for points within or on the sphere. – Guy Apr 04 '14 at 10:03
  • Okay, I think I understand. I assumed that, because we inserted $x^2+y^2+z^2=1$ into the function, the function would implicitly contain the equality $x^2+y^2+z^2=1$. I see now that it doesn't. – Andrea Apr 04 '14 at 10:06
  • @Andreas yes you're right. It doesn't contain the equality. You need to explicitly enforce that yourself. – Guy Apr 04 '14 at 10:08