If a function has a local extremum at the point $(x_0,y_0)$ along every line, then does the function have a point of extremum?

197 Views Asked by At

Say we have a 2 variable function $f(x,y)$. At the point $P(x_0,y_0)$ it has a local minima along every line. What I mean by this is, if we choose any line $L$ that passes the point $P$ on the $xy$ plane, then the curve $f(L)$ will have a local minima at the point $P$.

My question is whether the function will have a local minima at the point?

I would say that $f$ will also have a local minima at the point $P$ because I am picturing the function to be like a bowl. The center of the bowl being the point $P$. The semicircles that we get as the curves of $f(L)$ all have a minima at the point $P$. I don't see another picture for this function that satisfies the condition. And according to this $f$ does infact have a minima at the point $P$.

But as tested this intuitive deduction on some functions, I came across this one:
$$ f(x,y) = 5 y ^ 4 − 6 x y ^ 2 + x ^ 2 $$ Notice that if we take the curve along the line $y=mx$. We get $$ g(x)=f(x,mx)=5m^4x^4-6m^2x^3+x^2 $$

So, $$g'(x)=20m^4x^3-18m^2x^2+2x =0$$ Obviously $x=0$ is a solution.
Now $$ g''(x)= 60m^4x^2-36m^2x + 2$$ $$ g''(0) > 0$$ This says that for the point $P(0,0)$ we get that $g(x)$ has a minima at $P$ for every value of $m$.

But for $f(x,y)$, let us consider $t>0$. Notice that
$$ f(0,t) > 0 $$ And $$ f(2t^2,t)<0 $$ We will always get points in the neighborhood of $f$ some of which are negative and some of which are positive. This means that even though $g(x)$ has a minima at $0$, the function $f$ doesnot.

2

There are 2 best solutions below

3
On BEST ANSWER

No, $f$ need not have a local minimum at that point. Define $f(x,x^2)=x^3$ for $x\in \mathbb R,$ and set $f=0$ everywhere else. Then on each line through $(0,0),$ $f$ has a local minimum value of $0$ at $(0,0).$ However $(0,0)$ is not a local minimum point of $f$ as every neighborhood of $(0,0)$ contains points where $f<0.$

This example is actually differentiable at $(0,0).$

1
On

The paradox is explained by the fact that the function is negative (i.e. below the saddle point) in an area delimited by two parabolas (see Ted's comment), which is such that no straight line through the origin crosses it in a neighborhood of the origin.