I have a doubt when using this method to find min/max of a function. When I can find two (or more solutions) of system equation, so I can easily know it has max and min.

But, the problem is: if the system equation I solve just have only one solution. So, I can know it max or min (by replace arbitrary value to function to compare). For example is min. So, Can I say this function doesn't has max ?

Thanks :)


If you are looking a continuous function, say $f(x,y)$ (there could be more variables) over a compact set (closed and bounded), then there will always be a min and a max. But the min and/or max may occur on the boundary of the region, in which case setting partial derivatives equal to $0$ may (and usually will) fail to find it.

This is the analogue in dimensions $2$ and higher of the fact that if we are looking at a continuous function over a closed interval, the min and/or max may occur at an endpoint.

For example, if we are trying to find the min or max of $x^2+y^2$ on or inside the square with corners $(1,1)$, $(-1,1)$, $(-1,-1)$, and $(1,-1)$, setting partial derivatives equal to $0$ finds the minimum, which occurs at $(0,0)$, but does not find the maximum, which is at the four corners of the region.

If we have a function say $f(x,y)$ with continuous partial derivatives, and we are interested in finding the min and max of $f$ over all of $\mathbb{R}^2$, there may not be such a min or max. If there is a single place at which the partials are both $0$, then at least one of min or max fails to exist. But even if the partials are both $0$ at several places, there may not be a min or max.

For example, let $f(x,y)=x^3-3x^2+y^3-3y^2$. Then the partials are equal to $0$ at $(0,0)$, $(0,2)$, $(2,0)$, and $(2,2)$ but $f$ has neither a min nor a max on $\mathbb{R}^2$.

  • $\begingroup$ can you give me an example, please. My teacher says the thing opposite :( that we can sure that no max for that function (in above example) $\endgroup$ – hqt Jun 8 '12 at 6:47
  • $\begingroup$ @AndréNicolas: That's not strictly true; since there are inequality constraints, in this case you would need to look at 9 possibilities for a solution. You have chosen the one with no constraints active. Looking at all the other possibilities will indeed throw up the corner points (in fact all corner points, and the mid-points of the edges too). The Lagrange conditions are necessary. $\endgroup$ – copper.hat Jun 8 '12 at 7:37
  • $\begingroup$ Sorry. Can I ask you, when I read on Wiki, constraint is an equation. and when I ask my teacher, my teacher say when constraint is inequality, we use another method. $\endgroup$ – hqt Jun 8 '12 at 7:54
  • $\begingroup$ There are many different sorts of constraints. Equality constraints and inequality constraints are two types... $\endgroup$ – copper.hat Jun 8 '12 at 8:12
  • $\begingroup$ Nitpick: the maximum of $x^2+y^2$ on $[-1,1]\times[-1,1]$ also occurs at $(-1,1)$ and $(1,-1)$. $\endgroup$ – Rahul Jun 8 '12 at 17:43

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.