1
$\begingroup$

I have a smooth function $f(x,y)$ that is unbounded in every direction. In other words, if we choose a direction $(a,b)\in S^1$ and keep moving along the curve $(ta,tb)$, then $$\lim_{t\to\infty}f(ta,tb)=\infty.$$

Can we conclude that there exists some disc of radius $R$ such that outside of this disc, the function $f(x,y)> 1$ (alternatively, $f(x,y)\geq 1$; see below)?

Note: The "obvious" way to approach this problem is to define a function $g(\theta)$ on $S^1$, where $$g(\theta)=\inf\limits_{t_0} \{t_0:f(t\theta)\geq 1\text{ for all }t\geq t_0\}.$$ If this is a continuous function, then using the compactness of $S^1$, we can conclude that the function attains a maximum at some point on $S^1$, and then use that maximum value to define $R$.

However, based on a couple of examples that I've constructed, this function is not turning out to be continuous. However, changing the requirement from $f(x,y)>1$ to $f(x,y)\geq 1$ seems to do the trick. Is this true in general?

$\endgroup$
1
  • $\begingroup$ Note that whatever assumptions on f do imply the thesis that {f≥1} is bounded, as soon as they also hold for f/M, they also imply {f≥M) is bounded, that is $f(z)\to\infty$ as $\|z\|\to+\infty$ $\endgroup$ Commented Jul 18 at 21:50

1 Answer 1

4
$\begingroup$

A counterexample: $f(x,y):=x^2y^2$ if $xy\ne0$ and $f(x,y):=|x|+|y|$ otherwise.


The OP has changed the question to additionally require that $f$ be smooth. Here is a smooth counterexample: $f(x,y):=x^2y^2+(x^2+y^2)e^{-x^2y^2(x^2+y^2)}$.

Indeed, take any $(a,b)\in S^1$. If $ab\ne0$, then $f(ta,tb)\ge t^2(ab)^2\to\infty$ as $t\to\infty$. If $a=0$, then $b\ne0$ and $f(ta,tb)=t^2b^2\to\infty$ as $t\to\infty$. The case $b=0$ is similar. However, if $0<x\to\infty$ and $y=x^{-3/2}$, then $$f(x,y)=\frac1x+\Big(x^2+\frac1{x^3}\Big) \exp\Big(-x-\frac1{x^4}\Big)\to0.$$

$\endgroup$
3
  • $\begingroup$ Thank you. I suppose the basic idea behind the smooth example is that for points where $x^2 y^2$ is small, but $(x^2+y^2)$ is large enough to ensure that $(x^2+y^2)e^{-x^2 y^2(x^2+y^2)}$ is effectively $0$, the function is equal to $x^2 y^2$, and is hence small? $\endgroup$ Commented Jul 18 at 17:44
  • $\begingroup$ @RyanHendricks : Right. The smooth counterexample is just a modification of the non-smooth one. $\endgroup$ Commented Jul 18 at 18:21
  • $\begingroup$ I have now asked a follow up question, with the additional restriction that $f(x,y)$ is a polynomial- mathoverflow.net/questions/475327/… $\endgroup$ Commented Jul 18 at 20:48

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .