0
$\begingroup$

As far as I can tell, the notion of a topological basis is motivated by the following fact: given a collection of sets $\mathscr{B}$ satisfying the axioms of a topological basis, the set consisting of all arbitrary unions of sets in $\mathscr{B}$, i.e. the set: $$\left\{ U : U = \bigcup\limits_{B \in\mathcal{B}\subseteq \mathscr{B}} B \right\} $$ satisfies the axioms of a topology.

The fact above states the topological basis axioms are sufficient for this property (that the new collection of all arbitrary unions of sets in the old collection satisfies the axioms of a topology).

What I want to show is that the axioms for a topological basis are necessary to have this property.

1 Answers 1

1

Attempt. To be clear, I should recall what the axioms for a topological basis are.

A collection $\mathscr{B} \subseteq \mathscr{P}(X)$ of subsets of a set $X$ is a topological basis if it satisfies the following two properties:

  1. $X = \bigcup\limits_{B \in \mathscr{B}} B\,,$ i.e. the sets in $\mathscr{B}$ cover $X$.
  2. Given: $B_1, B_2 \in \mathscr{B}$. Then for all $x \in B_1 \cap B_2$, there exists a set $B_x \in \mathscr{B}$ such that: $$x \in B_x \subseteq B_1 \cap B_2 \,. $$

To recall, what I want to show is: if a collection of sets $\mathscr{A} \subseteq \mathscr{P}(X)$ satisfies the property given above, i.e. if the set of all arbitrary unions of sets in the collection $\mathscr{A}$ $$ \left\{ U : U = \bigcup\limits_{A \in\mathcal{A}\subseteq \mathscr{A}} A \right\}$$ satisfies the axioms of a topology on $X$, then $\mathscr{A}$ satisfies the axioms of a topological basis.

So assume that this property is satisfied, specifically this means that:

  1. $$\emptyset, X \in \left\{ U : U = \bigcup\limits_{A \in\mathcal{A}\subseteq \mathscr{A}} A \right\} $$
  2. $$T_1, T_2 \in \left\{ U : U = \bigcup\limits_{A \in\mathcal{A}\subseteq \mathscr{A}} A \right\} \implies T_1 \cap T_2 \in \left\{ U : U = \bigcup\limits_{A \in\mathcal{A}\subseteq \mathscr{A}} A \right\} $$
  3. $$\forall i \in I, T_i \in \left\{ U : U = \bigcup\limits_{A \in\mathcal{A}\subseteq \mathscr{A}} A \right\} \implies \bigcup\limits_{i \in I} T_i \in \left\{ U : U = \bigcup\limits_{A \in\mathcal{A}\subseteq \mathscr{A}} A \right\} $$

Clearly, for any $\mathcal{A} \subseteq \mathscr{A}$, we have that: $$\bigcup\limits_{A \in \mathcal{A}} A \subseteq \bigcup\limits_{A \in \mathscr{A}} A \,. $$ By our assumption, for some $\mathcal{A} \subseteq \mathscr{A}$, we have that $X = \bigcup\limits_{A \in \mathcal{A}} A$, so combining the above two statements, we have at once: $$X \subseteq \bigcup\limits_{A \in \mathscr{A}} A \,. $$ The reverse inclusion is trivial (since $\mathscr{A} \subseteq \mathscr{P}(X)$), and thus the first of the two axioms for a topological basis must be satisfied by $\mathscr{A}$.

To verify the second of the two axioms, let $A_1, A_2 \in \mathscr{A}$ be given. Consider an arbitrary $x \in A_1 \cap A_2$ (note that we can not necessarily assume that $A_1 \cap A_2 \in \mathscr{A}$, otherwise we would be done trivially).

For any $\mathcal{A}_1 \subseteq \mathscr{A}$ such that $A_1 \in \mathcal{A}_1$ and for any $\mathcal{A}_2 \subseteq \mathscr{A}$ such that $A_2 \in \mathcal{A}_2$, we have that: $$ \left(\bigcup\limits_{A \in \mathcal{A}_1} A \right) \cap \left( \bigcup\limits_{A \in \mathcal{A}_2} A \right) = \left( \bigcup\limits_{A \in \mathcal{A}_3} A \right) \,, $$ for some $A_3 \subseteq \mathscr{A}$, and likewise, reasoning in the other direction: $$\left(\bigcup\limits_{A \in \mathcal{A}_1} A \right) \cap \left( \bigcup\limits_{A \in \mathcal{A}_2} A \right) = \left(A_1 \cup \bigcup\limits_{A \in \mathcal{A}_1 \setminus \{A_1 \} } A \right) \cap \left( A_2 \cup \bigcup\limits_{A \in \mathcal{A}_2\setminus \{A_2 \} } A \right) \\ = (A_1 \cap A_2 ) \cup \bigcup\limits_{A \in \mathcal{A}_1 \setminus \{A_1 \} } (A \cap A_2) \cup \bigcup\limits_{A \in \mathcal{A}_2 \setminus \{A_2 \} } (A \cap A_1) \cup \left( \left(\bigcup\limits_{A \in \mathcal{A}_1 \setminus \{A_1 \} } A\right) \cap \left(\bigcup\limits_{A \in \mathcal{A}_2 \setminus \{A_2 \} } A \right) \right) \,. $$ Therefore we can conclude that, irrespective of the choice of $\mathcal{A}_1 \ni A_1, \mathcal{A}_2 \ni A_2$: $$(A_1 \cap A_2) \subseteq \bigcup\limits_{A \in \mathcal{A}_3} A \,. $$ In particular, we can choose $\mathcal{A}_1 = \{A_1 \}$ and $\mathcal{A}_2 = \{A_2 \}$ to get that: $$(A_1 \cap A_2) = \bigcup\limits_{A \in \mathcal{A}_3} A $$ for some $\mathcal{A}_3 \subseteq \mathscr{A}$. Therefore, there exists an $A_x \in \mathcal{A}_3$ such that $x \in A_x$ (irrespective of our choice of $x \in (A_1 \cap A_2)$). $A_x \in \mathcal{A}_3$ and $\mathcal{A}_3 \subseteq \mathscr{A}$ together imply that $A_x \in \mathscr{A}$, and clearly one has: $$A_x \subseteq \bigcup\limits_{A \in \mathcal{A}_3} A = (A_1 \cap A_2) \, , $$ so $A_x$ satisfies all of the required properties and the second axiom of a topological basis is verified. Thus being a topological basis is not only sufficient, but even necessary, for the collection of arbitrary unions to satisfy the axioms of a topology.


As Henno Brandsma astutely remarked in the comments, there is a simpler way to see not just the necessity, but both the necessity and sufficiency (the equivalence) at the same time, namely via an equivalent formulation of the second axiom of a topological basis.

2'. Given $B_1, B_2 \in \mathscr{B}$, there exists a $\mathcal{B} \subseteq \mathscr{B}$ such that: $$B_1 \cap B_2 = \bigcup\limits_{B \in \mathcal{B}} B \,.$$

Below I show briefly that 2. and 2'. are really equivalent.

Assume 2. Then we can write $$B_1 \cap B_2 = \bigcup\limits_{x \in B_1 \cap B_2} B_x \,. $$ Therefore 2'. holds. (I.e. take $\mathcal{B}=\{B_x : x \in B_1 \cap B_2 \}$.)

Assume 2'. Then for any $x \in B_1 \cap B_2$, there exists a $B_x \in \mathcal{B}$ such that $x \in B_x$ (assuming otherwise leads to a contradiction), and clearly we have, since $B_x \in \mathcal{B}$, $$B_x \subseteq \bigcup\limits_{B \in \mathcal{B}} B = B_1 \cap B_2\,. $$ Thus 2. holds.

  • 1
    axiom 1 is necessary as $X$ must be open. axiom 2 is equivalent to teh fact that the intersection of two base elements is a union of base elements, so it's necessary because base elements must be open and so their finite intersection is also open (so a union of base elements).2017-01-08
  • 0
    @HennoBrandsma What you suggest is definitely a better way to explain the second axiom. In particular, that formulation makes it much more evident why the collection of all arbitrary unions would be closed under finite intersections. I will try to remember to add a proof of the equivalence sometime later.2017-01-08