5
$\begingroup$

The problem is:

Let $A$ and $B$ disjoint closed subsets of a smooth manifold $M$. Show that there exist a function $f \in C^\infty(M)$ such that $A=f^{-1}(0)$ and $B=f^{-1}(1)$.

I thought this problem could be solved by using the Smooth Urysohn's Lemma, but can't see how to guarantee that $0 < f(x) < 1$ for all $x\in M\backslash (A\cup B)$. Tips?

$\endgroup$
3
  • $\begingroup$ By the way, in your question you did not say that $0\le f\le 1$, although I certainly agree that this can be arranged. $\endgroup$ May 15 '17 at 22:49
  • $\begingroup$ @TedShifrin The Urysohn's Lemma say that the Codomain is $[0,1]$. I was referring to that. But if you know how to obtain a function (with any Codomain) satisfying what I want, please let me know. $\endgroup$
    – Filburt
    May 16 '17 at 23:31
  • $\begingroup$ The Urysohn Lemma is not good enough here. It only gives you continuous functions. What are you referring to as the Smooth Urysohn Lemma? $\endgroup$ May 16 '17 at 23:34
2
$\begingroup$

I originally misread the question and thought you wanted $A\subset f^{-1}(0)$ and $B\subset f^{-1}(1)$. This can be arranged, as I suggested, with a simple partition of unity argument:

Let $\{\phi_i\}$ be a partition of unity subordinate to the open cover $\mathscr U = \{X-A,X-B\}$ of $X$. Let $$\phi_A = \sum_{i: \,\text{supp }\phi_i\subset X-B} \phi_i \qquad\text{and}\qquad \phi_B = \sum_{i:\, \text{supp }\phi_i\subset X-A} \phi_i.$$ Note that $\phi_B(x)=0$ when $x\in A$, $\phi_B(x)=1$ when $x\in B$ (since $\phi_A+\phi_B=1$), and, of course, $0\le\phi_B\le 1$. Thus, the function $f=\phi_B$ will do.

To get the full result you seek we need to be a bit trickier. As @zhw suggested, it will suffice to find a smooth function $g$ with $g^{-1}(0)=A$ and $g\ge 0$. For then we similarly find a smooth function $h$ with $h^{-1}(0)=B$ and $h\ge 0$ and take $f=g/(g+h)$.

Here's a sketch of how you can obtain $g$. Let $\{U_i\}$ be a countable open cover of $M-A$ by sets diffeomorphic to an open ball with the property that sub-balls $U'_i\subset \overline{U'_i}\subset U_i$ also cover $M-A$. Let $g_i\colon M\to [0,1]$ be a smooth function with $g_i>0$ on $U'_i$ and $g_i=0$ on $M-U'_i$. Then the partial derivatives of $g_1,g_2,\dots,g_i$ of all orders $\le i$ are bounded (in magnitude) above on $M$, say, by $k_i>0$. Define $$g = \sum_i \frac{g_i}{k_i2^i},$$ and check that $g\in C^\infty$ and $g^{-1}(0)=A$.

$\endgroup$
1
$\begingroup$

Partial answer: Suppose you can find $f,g\in C^\infty(M)$ such that $f=0$ on $A,$ $f>0$ elsewhere, and $g=0$ on $B,$ $g>0$ elsewhere. Then $f/(f+g) = 0$ on $A,$ $f/(f+g) = 1$ on $B,$ and $0<f/(f+g)<1$ elsewhere.

$\endgroup$
5
  • $\begingroup$ How could I find such smooth functions? $\endgroup$
    – Filburt
    May 14 '17 at 1:41
  • $\begingroup$ Do you know about partitions of unity subordinate to an open cover of a manifold? If not, look them up. $\endgroup$ May 15 '17 at 22:41
  • $\begingroup$ @TedShifrin Yes, the Urysohn Lemma is used to prove the existence of them. Could this solve the problem? $\endgroup$
    – Filburt
    May 16 '17 at 23:32
  • $\begingroup$ I mean smooth ($C^\infty$) partitions of unity here. $\endgroup$ May 16 '17 at 23:34
  • $\begingroup$ @TedShifrin Yes, I know. Could this solve the problem? $\endgroup$
    – Filburt
    May 17 '17 at 5:12

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.