4
$\begingroup$

Suppose we are given a smooth function $f\colon \mathbb{R}^n \rightarrow \mathbb{R}$ and some number $c$. What can be said about the preimage $f^{-1}(c)$.

There's the theorem on regular preimages, asserting that if $\nabla f$ is nowhere vanishing on $M=f^{-1}(c),$ then $M$ is a smooth submanifold of dimension $n-1$.

But now lets assume that one does not have a regular preimage. Can one say anything about the preimage here (except what would already follow from continuity), or is this a hopeless case?

$\endgroup$
9
  • 1
    $\begingroup$ The preimage $f^{-1}(c)$ is of course closed but can be rather wild. For instance, for $n=1$, an arbitrary closed subset in $\mathbb R$ has such a form. $\endgroup$ Oct 11, 2013 at 21:13
  • $\begingroup$ If you mean in the case of $n=1$, let $M=f^{-1}(0)$, where $f:{\mathbb R}\to{\mathbb R}$ is $C^\infty$.Then the complement of $M$ is given by $f^2(x)>0$ $\dots$ $\endgroup$ Oct 11, 2013 at 21:29
  • 1
    $\begingroup$ Well, the OP is changing his/her question more quickly than I am typing. I give up. $\endgroup$ Oct 11, 2013 at 21:38
  • 4
    $\begingroup$ Anyway, the case is hopeless. The zero set of a $C^\infty$ function $f:\mathbb R^n\to\mathbb R$ can be any closed set, as Sasha said already (there is nothing special about dimension $1$ here). Should we close? $\endgroup$
    – fedja
    Oct 11, 2013 at 21:41
  • 2
    $\begingroup$ @fedja I guess, not many professionals know your answer. In this sense, the question is helpful. $\endgroup$ Oct 11, 2013 at 21:52

1 Answer 1

5
$\begingroup$

I do not mean to spoil the exercise, but note that if $\phi$ is any smooth function on a Banach space $E$, with bounded derivatives of any order (say $\|D^j\phi\|_\infty < \infty$ for any $j\ge0$), then $f(x):=\sum_{k=1}^\infty d_k^k \phi\big(\frac{x-a_k}{d_k} \big)$ certainly defines a smooth function for any choice of a sequence of vectors $ a_k\in E$, and of numbers $0 < d_k \le 1/2$. To be able to make the zero-set of $f$ whatever closed set $C\subset E$ you like, just requires $E$ to be separable, and possess a smooth $\phi$ with bounded derivatives and bounded support (and non-zero, of course, hence wlog also non-negative).

$\endgroup$
1
  • 1
    $\begingroup$ Nice! (For $n=1$, the idea was similar, but your exposition is great and does not look as spoiling.) $\endgroup$ Oct 12, 2013 at 15:33

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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