Problem $8$ in Chapter $1$ of Spivak's A Comprehensive Introduction to Differential Geometry, Vol. 1 reads:

8. For this problem, assume

  1. (The Generalized Jordan Curve Theorem) If $A\subset \mathbb{R}^n$ is homeomorphic to $S^{n-1},$ then $\mathbb{R}^n-A$ has $2$ components, and $A$ is the boundary of each.
  2. If $B\subset\mathbb{R}^n$ is homeomorphic to $D^n=\{x\in\mathbb{R}^n:d(x,0)\le 1\},$ then $\mathbb{R}^n-B$ is connected.

(a) One component of $\mathbb{R}^n-A$ (the "outside of $A$") is unbounded, and the other (the "inside of $A$") is bounded.

(b) If $U\subset\mathbb{R}^n$ is open, $A\subset U$ is homeomorphic to $S^{n-1}$ and $f:U\to \mathbb{R}^n$ is one-to-one and continuous (so that $f$ is a homeomorphism on $A$), then $f(\text{inside of }A)=\text{inside of }f(A)$.

(c) Prove Invariance of Domain, i.e. if $U\subset \mathbb{R^n}$ is open and $f:U\to\mathbb{R}^n$ is one-to-one and continuous, then $f(U)\subset \mathbb{R}^n$ is open.

I have solved (a) and (c), which I will present here.

Proof of (a): $A$ is homeomorphic to $S^{n-1},$ and so is compact. It follows that $A$ is bounded. Let $B$ be a closed ball containing $A.$ Then $\mathbb{R}^n-B\subset \mathbb{R}^n-A$ is connected and clearly unbounded. It must lie in a connected component of $\mathbb{R}^n-A,$ and so one of the components of the latter set is unbounded. It must then be that the other connected component of $\mathbb{R}^n-A$ is contained in $B,$ so that this connected component is bounded.

Proof of (c): Let $y\in f(U)$ and consider the preimage of $y$ under $f$, call it $x$, in $U$. Then there is some closed ball $B$ about $x$ in $U$. The boundary $A$ of $B$ is homeomorphic to $S^{n-1}$ and $x$ lies in the inside of $A$, so by part (b) $f(\text{inside of }A)=\text{inside of }f(A)$ is an open ball which contains $y$. We have thus found a neighborhood of $y$ contained in $f(U)$, so $f(U)$ is open in $\mathbb{R}^n$.

I do not know how to solve (b). I am completely and utterly stuck. I'm clearly missing something obvious. Any hint or even just the solution would be appreciated.

  • $\begingroup$ I think I can show this if the following "fact" is true: If $U\subset $ inside of $A$ is open with $\partial U = A$, then $U = $inside of $A$. The problem is I don't know if this is true... $\endgroup$ – user25959 Nov 20 '18 at 21:30
  • $\begingroup$ As it is stated in your question, (b) does not make sense. As an example consider $U = \mathbb{R}^2 \setminus \{ 0 \}$, $A = S^1$. The inside of $A$ is not contained in $U$, hence $f(\text{inside of } A)$ is not defined. So you will need the additional assumption that $\text{inside of } A \subset U$. $\endgroup$ – Paul Frost Dec 31 '18 at 15:43
  • $\begingroup$ ^^ yes I agree, and when doing the problem I took that assumption for granted since It was needed to do the problem. $\endgroup$ – D. Brogan Dec 31 '18 at 17:38

For $n = 1$ both 1. and 2. are wrong, so we must assume $n > 1$.

Let us first analyze the proof of (a). Here we only need 1. That the complement of a standard closed ball $B = D(0,r)$ is connected and unbounded is a trivial fact which does not use the full strength of 2.

Based on (a), for $A\subset \mathbb{R}^n$ homeomorphic to $S^{n-1}$, let $\text{ins}(A)$ denote the inside of $A$ and $\text{outs}(A)$ the outside of $A$.

As it is stated in your question, (b) does not make sense. As an example consider $U = \mathbb{R}^n \setminus \{ 0\}$ and $A = S^{n-1}$. The inside of $A$ is not contained in $U$, hence $f(\text{ins}(A))$ is not defined. So you will need the additional assumption that $\text{ins}(A) \subset U$.

We shall prove only a special case of (b) which is sufficient to show (c):

Let $B = D(x,r)$ be a closed ball contained in $U$ and $A$ its boundary (so that $\text{ins}(A)$ is the open ball with radius $r$ around $x$). Then $f(\text{ins}(A)) = \text{ins}(f(A))$.

Obviously $f(B)$ is compact, hence $C = \mathbb{R}^n \setminus f(B) \subset \mathbb{R}^n \setminus f(A)$ is unbounded. Using 2. we see that $C \subset \text{outs}(f(A))$.

We can write $\mathbb{R}^n \setminus f(A) = C \cup f(\text{ins}(A)) = \text{outs}(f(A)) \cup \text{ins}(f(A))$. These are two (potentially different) decompositions of $\mathbb{R}^n \setminus f(A)$ into disjoint sets.

$f(\text{ins}(A))$ is a connected subset of $\mathbb{R}^n \setminus f(A)$, hence contained either in $\text{ins}(f(A))$ or in $\text{outs}(f(A))$. Assume it is contained in $\text{outs}(f(A))$. Then $\mathbb{R}^n \setminus f(A) = C \cup f(\text{ins}(A)) \subset\text{outs}(f(A))$ which is impossible. Therefore $f(\text{ins}(A)) \subset \text{ins}(f(A))$. But now $$C \cup f(\text{ins}(A)) = \text{outs}(f(A)) \cup \text{ins}(f(A)) ,$$ $$C \subset \text{outs}(f(A)), f(\text{ins}(A)) \subset \text{ins}(f(A)) .$$ This implies $$C = \text{outs}(f(A)), f(\text{ins}(A)) = \text{ins}(f(A)) .$$


I would suggest to replace (b) by its essence:

Let $f : D^n \to \mathbb{R}^n$ be a continuous injection. Then $f(\text{ins}(S^{n-1})) = \text{ins}(f(S^{n-1}))$.

Note that for an arbitrary $A\subset \mathbb{R}^n$ homeomorphic to $S^{n-1}$, the set $B(A) = \overline{\text{ins}(A)} = \text{ins}(A) \cup A$ is closed and bounded, i.e. compact, but in general not homeomorphic to $D^n$. Thus the above proof breaks down. To get an example for this phenomenon take the Alexander horned sphere $S \subset \mathbb{R}^3$. Define $A = \phi(S)$, where $\phi : \mathbb{R}^3 \setminus \{ 0 \} \to \mathbb{R}^3 \setminus \{ 0 \}, \phi(x) = \dfrac{x}{\lVert x \rVert ^2}$. Then $B(A)$ is not homeomorphic to $D^3$.


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.