CarCraAGT.dvi @ Applied General Topology c© Universidad Politécnica de Valencia Volume 9, No. 1, 2008 pp. 15-19 Cancellation of 3-Point Topological Spaces S. Carter and F. J. Craveiro de Carvalho ∗ Abstract. The cancellation problem, which goes back to S. Ulam [2], is formulated as follows: Given topological spaces X, Y, Z, under what circumstances does X × Z ≈ Y × Z (≈ meaning homeomorphic to) imply X ≈ Y ? In [1] it is proved that, for T0 topological spaces and denoting by S the Sierpinski space, if X × S ≈ Y × S then X ≈ Y . This note concerns all nine (up to homeomorphism) 3-point spaces, which are given in [4]. 2000 AMS Classification: 54B10 Keywords: Homeomorphism, Cancellation problem, 3-point spaces. 1. Two cancellation results Below X and Y denote T1 topological spaces. Proposition 1.1. Let S be a topological space with a unique closed singleton {p}. If there is a homeomorphism φ : X×S → Y ×S then φ(X×{p}) = Y ×{p}. Proof. We shall show that φ(X×{p}) ⊂ Y ×{p} which, using similar arguments, will be enough to prove that φ(X × {p}) = Y × {p} and, consequently, that X ≈ Y . Let us suppose that for some x ∈ X,y ∈ Y and q ∈ S \ {p} we have φ(x,p) = (y,q). Then {(y,q)} is closed and, therefore, (Y × S) \ {(y,q)} is open. Let r belong to the topological closure of {q},r 6= q. Then (y,r) ∈ (Y ×S) \ {(y,q)} and we must have open sets Uy,Ur, containing y and r, respectively, such that Uy × Ur ⊂ (Y × S) \ {(y,q)}. We reach a contradiction since (y,q) belongs to Uy × Ur. � ∗The second named author gratefully acknowledges financial support from Fundação para a Ciência e Tecnologia, Lisboa, Portugal. 16 S. Carter and F. J. Craveiro de Carvalho An example of such an S is obtained as follows. Let S be a set with 4 elements at least. Let a,b ∈ S and denote by S1 the complement of the subset they form. Take then as basis for a topology on S the set {{a},{a,b},S1}. If S happens to have just 4 points then it is the only minimal, universal space with such a number of elements [3]. Proposition 1.2. Let S be a topological space with a dense, open singleton {p} and such that, for every q ∈ S \ {p}, the topological closure of {q} is finite. If there is a homeomorphism φ : X × S → Y × S then φ(X × {p}) = Y × {p}. Proof. Let {p} be an open, dense singleton in S. We will show that φ(X × {p}) = Y × {p} which, as observed before, is enough to conclude that X ≈ Y . Assume that for some x ∈ X,y ∈ Y and q 6= p we have φ(x,p) = (y,q). Consider the closed set {y}×{q}, the bar denoting closure, its image φ−1({y}× {q}), which is also closed, and suppose that {q} has s elements. Also, observe that p /∈ {q}. Since (x,p) belongs to φ−1({y}×{q}) and this set has s elements, there is an r in {q} such that (x,r) does not belong to this set. There are then open sets Ux,Ur, containing x and r, respectively, with Ux×Ur ⊂ (X×S)\φ −1({y}×{q}). We have a contradiction since (x,p) ∈ Ux × Ur. � An example for S can be the following Door space. Let S be a set and fix p ∈ S. Define U ⊂ S to be open if it is empty or contains p. 2. 3-point spaces We go on assuming that X,Y are T1 topological spaces though such assum- ption is not used in Propositions 2.1 and 2.2 below. If we now consider S = {a,b,c} to be one of the 3-point spaces [4], we see that Propositions 1.1 and 1.2 of §1 allow us to deduce immediately that S can be cancelled except in the following cases - S is discrete, - S has {{a},{b},{a,c}} as a topological basis, - S is trivial. If S is discrete the situation is not as simple as one might be led to think. Let us take the following example. Let S = Z, here Z stands for the integers with the discrete topology, and consider the discrete spaces X = {0, 1, . . . ,n − 1},n ≥ 2,Y = {0}. Now define φ : {0, 1, . . . ,n − 1} × Z → {0} × Z by φ(x,r) = (0,nr + x). This map is a homeomorphism and however Z cannot be cancelled. We can say something when the spaces X,Y have a finite number of con- nected components. Proposition 2.1. Let S be a finite discrete space and assume that X has a finite number of connected components. If X × S ≈ Y × S then X ≈ Y . Cancellation of 3-Point Topological Spaces 17 Proof. The connected components of X × S or Y × S are of the type X′ × {x},Y ′ ×{y}, where X′,Y ′ are components of X and Y , respectively. It follows that Y has the same number of components as X. Let us consider in the sets of connected components of X and connected components of Y the homeomorphism equivalence relation and take an equi- valence class of components of X, say {X1, . . . ,Xk}. The subspace k⋃ i=1 Xi × S has kn components, where n is the cardinal of S. The same happens with φ( k⋃ i=1 Xi × S), where φ is a homeomorphism between X × S and Y × S. Let p ∈ S. For every i = 1, . . . ,k, φ(Xi × {p}) = Yi × {qi}, where the qi’s belong to S and the Yi’s are components of Y homeomorphic to the Xi’s. Assume that the equivalence class to which the Yi’s belong is {Y1, . . . ,Yl}. Then φ( k⋃ i=1 Xi × {p}) ⊂ l⋃ j=1 Yj × S. Consequently, also φ( k⋃ i=1 Xi × S) ⊂ l⋃ j=1 Yj × S. Using the inverse homeomorphism φ−1, we are led to conclude that the re- verse inclusion holds and, therefore, φ( k⋃ i=1 Xi × S) = l⋃ j=1 Yj × S. So k⋃ i=1 Xi × S and l⋃ j=1 Yj × S have the same number of components and it follows that k = l. From each component class in X choose a representative and use φ to es- tablish a homeomorphism between that representative and a component in Y . These homeomorphisms can then be used to conclude that every component of X is homeomorphic to a component of Y . Since components are closed and finite in number, X is homeomorphic to Y . � Proposition 2.2. Let X and Y be topological spaces with the same finite num- ber of connected components and S be a discrete space. Assume, moreover, that neither space has two homeomorphic components. If X × S ≈ Y × S then X ≈ Y . Proof. Let Xi, i = 1, . . . ,n, be the components of X and fix p ∈ S. If φ is a homeomorphism between X × S and Y × S then there are qi ∈ S,i = 1, . . . ,n, such that φ(Xi × {p}) = Yi × {qi}, i = 1, . . . ,n, where, due to our assumption on the non-existence of homeomorphic components, the Yi’s are the components of Y . Hence φ induces a homeomorphism φi : Xi → Yi, i = 1, . . . ,n. Again, since the number of components is finite and they are closed, the φi’s can be used to obtain a homeomorphism between X and Y . � Proposition 2.3. Let S have {{a},{b},{a,c}} as basis. If φ : X ×S → Y ×S is a homeomorphism then φ(X × {b}) = Y × {b}. 18 S. Carter and F. J. Craveiro de Carvalho Proof. Let πS : Y × S → S denote the standard projection. The image πS (φ(X × {b})) is open and, therefore, it is either {b} or contains a. Assume that for some x ∈ X,y ∈ Y we have φ(x,b) = (y,a). The subset {(x,b)} is closed and, consequently, the same happens with {(y,a)}. Hence (Y × S) \ {(y,a)} is open and contains (y,c). We must then have an open neighbourhood Uy of y such that Uy × {a,c} ⊂ (Y × S) \ {(y,a)}. Again we have a contradiction and φ(X × {b}) = Y × {b}. � To conclude the proof that a non-discrete 3-point space can be cancelled it only remains to deal with the case where S is trivial. Above we have an example of a homeomorphism φ : X × S → Y × S which does take a slice X ×{x} onto a slice Y ×{y}. More examples can be obtained. Take X = Y , with at least 2 elements, a trivial space S with also, at least, 2 elements and let ψ : S → S be a fixed point free bijection. Fix x0 ∈ X and define φ : X × S → X × S by φ(x,s) = (x,s), for x 6= x0, and φ(x0,s) = (x0,ψ(s)). Then φ is a bijection and φ({x} × S) = {x} × S, for x ∈ X. Since open sets in X × S are of the form U × S, U open in X, and φ(U × S) = U × S, φ is a homeomorphism. Obviously no slice X × {x} is mapped onto a similar slice. Proposition 2.4. Let S be a finite trivial space. If X × S ≈ Y × S then X ≈ Y . Proof. Open (closed) sets in X × S and Y × S are of the form U × S, where U is open (closed). We are going to define f : X → Y as follows. Let x ∈ X. Then {x} is closed and so are {x} × S and φ({x} × S), where φ : X × S → Y × S is a homeomorphism. Hence φ({x}×S) = C ×S, for some closed set C in Y . Since S is finite, C is a singleton and we make {f(x)} = C. This way we obtain an f which is a bijection since we began with a bijective φ. If C is closed in X, φ(C × S) = f(C) × S is closed in Y × S. Consequently f(C) is closed in Y . Therefore f is closed and f−1 is continuous. Taking φ−1, we would conclude that f is continuous the same way. � We can now state. Theorem 2.5. For X and Y T1 topological spaces and S a non-discrete 3-point topological space, if X × S ≈ Y × S then X ≈ Y . 3. A particular case We will no longer assume X,Y to be T1 and will suppose that S has a unique isolated point a. Moreover, the singleton {a} will be assumed to be closed. That is, for instance, the case where S = {a,b,c} and {{a},{b,c}} is an open basis. Cancellation of 3-Point Topological Spaces 19 Proposition 3.1. Let S have a unique isolated point a. Assume that {a} is closed. For X,Y connected with, at least, an isolated point each, if φ : X×S → Y × S is a homeomorphism then φ(X × {a}) = Y × {a}. Proof. Let πS : Y × S → S denote the standard projection, as before. The image πS (φ(X ×{a})) is open and connected. Therefore it is either {a} or some open, connected subset of S, which naturally does not contain a. Let the latter be the case. If x ∈ X is an isolated point then {(x,a)} is open and the same happens to its image under πS ◦ φ. This is impossible because {a} is the unique open singleton of S. � Examples of spaces satisfying the conditions of Proposition 3.1 are, again, some Door spaces. Let Z be a set. Fix p ∈ Z and define U ⊂ Z to be open if U = Z or p /∈ U. References [1] B. Banaschewski and R. Lowen, A cancellation law for partially ordered sets and T0 spaces, Proc. Amer. Math. Soc. 132 (2004). [2] R. H. Fox, On a problem of S. Ulam concerning cartesian products, Fund. Math. 27 (1947). [3] K. D. Magill Jr, Universal topological spaces, Amer. Math. Monthly 95 (1988). [4] J. R. Munkres, Topology, a first course, Prentice-Hall, Inc., 1975. Received July 2006 Accepted November 2006 S. Carter (s.carter@leeds.ac.uk) School of Mathematics, University of Leeds, Leeds LS2 9JT, U.K. F. J. Craveiro de Carvalho (fjcc@mat.uc.pt) Departamento de Matemática, Universidade de Coimbra, 3001-454 Coimbra, PORTUGAL