The main aim of this section will be to prove the following theorem.
Suppose that and that is homeomorphic to for some .
If then is just equal to and so is empty and .
If then and for all .
If then and all other homology groups are trivial.
We can combine cases (b) and (c) in Theorem 21.1 by saying that . For the most obvious case of the theorem, we can express as , so
The space is then homeomorphic to , with
We can define maps by and . We then have , and we have a homotopy between and the identity given by . Thus, in this case is actually homotopy equivalent to . However, that is not true in general. For example, let be a knotted circle in . We can identify with by stereographic projection, and thus think of as a subspace of . The theorem tells us that in this case has the same homology as , but it can be shown that is not homotopy equivalent to .
We will also deduce the following result, for which the case is called the Jordan Curve Theorem:
Suppose that , and let be an injective continuous map. Then the complementary set has precisely two path components, one bounded and the other unbounded.
It is easy to see that the Jordan Curve Theorem is true for maps that are reasonably simple. However, it is hard to prove in the general case, where may wiggle in an extremely complicated way and can also be fractal or otherwise badly behaved.
We will work up to Theorem 21.1 by first considering in cases where is homeomorphic to (or equivalently, ) rather than .
We say that a space is acyclic if and for all .
We have seen previously that all contractible spaces are acyclic. Conversely, most commonly occurring acyclic spaces are contractible, but there are some exceptions.
Let be a space. Let denote the one-point space, so we have a constant map . Then is acyclic iff the map is an isomorphism.
If is an isomorphism then is isomorphic to i.e. and for , which means that is acyclic. Suppose instead we start withthe assumption that is acyclic. For the groups and are both zero, so the map is automatically an isomorphism. Next, we know that . As is acyclic we have so so is nonempty and path connected. We can therefore choose and we have . We also have and so is an isomorphism. ∎
Let be a subset of that is homeomorphic to for some . Then is acyclic.
Before proving the theorem we will prove a simpler lemma. This will not directly contribute to the theorem, but will introduce some relevant ideas.
Suppose that and are closed subsets of such that the sets , and are all acyclic. Then the set is also acyclic.
Put and . These are open subsets of with and . Thus, our assumptions are that , and are acyclic, and we need to prove that is acyclic. We have a Mayer-Vietoris sequence
If then the first and third terms are zero so as expected. If we instead have an exact sequence
and it is not hard to deduce that . ∎
In Theorem 21.7, we assume that is homeomorphic to . This means that we can choose an injective continuous map with . We will make some constructions in that context.
Let be an injective continuous map. By a slice we mean a set of the form with . The width of such a slice is . A coslice is the complement in of a slice; we define the width of a coslice to be the same as the width of the corresponding slice.
We could attempt to prove Theorem 21.7 as follows. We argue by induction on , the case being easy. For , we divide into a large number of very thin slices. Any slice of width has acyclic complement, by the induction hypothesis. Our slices have very small width, so it seems reasonable to assume that they will also have acyclic complement. The intersection of two adjacent slices is a slice of width , and so has acyclic complement. Using Lemma 21.8, we deduce that the union of two adjacent slices again has acyclic complement. By repeating this procedure, we see that the union of any block of adjacent slices has acyclic complement. In particular, the full set has acyclic complement, as required.
The problem with this approach is that our “reasonable assumption” is hard to justify directly. Our actual proof of Theorem 21.7 will use many of the same ideas, but arranged in a slightly different way.
Suppose that and are closed subsets of such that the set is acyclic. Note that we have inclusions
Suppose that is nonzero; then either is nonzero in , or is nonzero in .
We use the same notation and the same Mayer-Vietoris sequence as in Lemma 21.8:
We again have , so is injective, so . However, is just , so either or must be nonzero. ∎
Let be a topological space, and let be a sequence of open sets with and . Then any chain is contained in for some .
We can express as for some integers and continuous maps . Note that the sets are open in , and the union of all these sets is . As is compact, it must be covered by a finite subcollection of the sets . As these sets are nested inside each other, this just means that there exists with . Equivalently, we have or . Thus, if we put then we have for all , and so . ∎
In the context of Lemma 21.12, suppose we have an element which maps to zero in . Then already maps to zero in for some .
Choose a cycle such that . As maps to zero in , there must be a chain with in . By the lemma, the element lies in for some . We can therefore interpret the equation as an equation in , showing that in as required. ∎
If then is just a single point, so is homeomorphic to by stereographic projection. This means that is contractible and therefore acyclic. For we will argue by induction. Choose a homeomorphism .
Suppose (for a contradiction) that we have a nonzero element for some . If we also assume that in . We will define a sequence of slices of width such that has nonzero image in for all .
We start with . Suppose we have already defined . We can write as , where and are slices of width , and the set has width and so is homeomorphic to . Our induction hypothesis says that has acyclic complement, so Lemma 21.11 tells us that must have nonzero image in or in . We choose or as appropriate to ensure that has nonzero image in .
By construction we have for some sequence with . It follows that the numbers converge to a limit , and that the set is just . This is homeomorphic to and so has acyclic complement by our induction hypothesis. In particular, the element must map to zero in . (In the case , we are using the assumption here.) However, we can regard as the union of the nested open sets , so must map to zero in for some , which contradicts our construction of .
This contradiction shows that no element as described above can exist. In other words, for we have , and also the kernel of the map is zero, so is injective. On the other hand, is contractible but is not, so cannot be equal to , so we can choose a point , and then . This shows that is also surjective, so it is an isomorphism as required. ∎
Suppose that is homeomorphic to for some . If then this just means that consists of two points. If this clearly means that as claimed. If then we recall that is homeomorphic to , so removing two points gives which is homotopy equivalent to and so has the same homology as , as claimed.
We now suppose that , and argue by induction on . We can write as the union of two hemispheres, with intersection . Correspondingly, we can write as , where and are homeomorphic to (or ) and is homeomorphic to . Theorem 21.7 tells us that the sets and are acyclic. The induction hypothesis tells us that the space has the same homology as . We need to show that the space has the same homology as (or that if ). For we note that so the Mayer-Vietoris sequence
shows that is an isomorphism. This means that for all with , but that if (or equivalently ) then .
This just leaves a few exceptional cases to consider. First suppose that , so , so . We then have a Mayer-Vietoris sequence
and it follows easily that as required.
Now suppose instead that . In this case we must show that has the same homology as , or in other words that and for . The case is covered by our main discussion above. The induction hypothesis tells us that has the same homology as , so the Mayer-Vietoris sequence
becomes
We can describe as the free abelian group generated by . Using this we see that is essentially the addition map , with kernel generated by . We can choose an element with , then it is not hard to deduce that is a basis for . This means that , or in other words that has precisely two path components.
Finally suppose that . Here the induction hypothesis shows that has the same homology as , so in particular it has two path components. We must show that . The Mayer-Vietoris sequence
becomes
Choose and , so and . The sequence shows that is surjective, but that is only possible if lies in one path component of and lies in the other. That implies that is actually an isomorphism, and then exactness shows that . However, we know that is the free abelian group on , so , so as claimed. ∎
If is an open subset of , then every path component of is also an open subset of . Similarly, if is an open subset of , then every path component of is also an open subset of .
First let be open in , and suppose we have a point , with path component say. Suppose that , so there is a path from to in . As is open, we can find a radius such that . By reducing if necessary, we can assume that . For any we have a linear path from to in . As we see that this does not pass through , so we can define ; this gives a path from to in . This stays within so it stays within . This means we have a path from to in , so . This proves that . As was arbitrary, this proves that is open as claimed.
The argument for is similar but easier. ∎
Let be an injective continuous map, where . First note that is compact, so it is bounded and closed in . It will be harmless to multiply by a small positive constant, so we can assume that for all .
We now recall a few more details about stereographic projection. We identify with , so . We put and . We have a homeomorphism given by . This also gives a homeomorphism , where . Theorem 21.1 tells us that has the same homology as , so in particular , so has precisely two path components. Let be the path component containing , and let be the other path component, so is the disjoint union of and . For we therefore have , so . This shows that both and are acyclic.
Because for all we see that the last coordinate of is always negative. Thus, the whole upper hemisphere of is contained in . Moreover, is clearly path connected and contains so . It follows that is contained in the lower hemisphere, and so the corresponding subset of is bounded.
We next claim that the set is path connected. We will prove this using the Mayer-Vietoris sequence
Here is just , and is acyclic. Thus, the first and fourth groups above are and . The space is the same as , which is homeomorphic to and so homotopy equivalent to . In particular, as we are assuming that , we know that is connected. Thus, the second group in our sequence is . It is also clear that is contractible and so . We therefore have an exact sequence
This is only consistent if , which means that is path connected.
We now see that is the disjoint union of path connected sets and , so these are the path components. It is clear that is bounded and is unbounded. ∎
Let be continuous and injective. Then is open in .
Put . As is compact, it is closed in , so is open. Theorem 21.1 tells us that , so has two path components. Let be the path component containing , and let be the other one. The sets and are open in by Lemma 21.14. Now put and . Using the injectivity of we see that and are disjoint and . Given we have a path from to in the set ; this shows that is path connected. As we see that . Next, Theorem 21.7 tells us that is acyclic and therefore also path connected. If there was a path in joining some point in to some point in , then we could conclude that the whole space was path connected. However, we know that has two path components, so no such path can exist.
Now consider a point . As is the path component of , we can find a path with and . Here and no path in can cross from to so the point must also lie in . This proves that . We have already remarked that is open, and it follows that is open, as claimed. ∎
Let be an open subset of , and let be a continuous injective map. Then is also open.
As usual we can identify with the complement of a point in , which is an open subset of . It will therefore be enough to show that is open in . Consider a point . As is open, we can choose such that is contained in . We then have a continuous injective map given by . The proposition tells us that the set is open, and it is clear that . It follows that is the union of all these open sets , and thus that is open. ∎