# Intuition behind the concept of a topology

So I have tried to understand the basics of topology, but I have some trouble getting a good intuition for it. I know that the idea is supposed to be that we have various open sets telling us something about the "nearness" of the points in the space, and together the set of all open sets (the topology) tells us something about the structure of the whole space.

These concepts make a lot of sense to me for metric spaces themselves, where open sets are finite or infinite unions of open balls with various radii. In particular, if we have some point y and some ball centered at it, then this ball gives the notion of a "neighborhood" around y, and the smaller the ball is, the "closer" the points in it will be to y. I thought a lot about the notion of neighbourhoods and open sets for general spaces, and I concluded that, roughly, the same applies to all spaces: if you have some open set in the basis of the topology (like the open balls are in the case of metric spaces) containing a point y, then the "smaller" the set is (in other words, the smaller the number of elements is), the "closer" the points in the set can be imagined to be to y. This then says something about how open sets tell us about the "nearness" of points, as I mentioned in the first paragraph.

But then I had some trouble applying these ideas to find the topologies of even very simple sets containing a small number of elements, which shows that there is something wrong with my intuition above. So, then, what is a better way of thinking about topologies and how they contain information about the "nearness" of points in a set?

• There are a bunch of similar questions on this site for your viewing pleasure: math.stackexchange.com/q/3439540, math.stackexchange.com/questions/1598409/…, mathoverflow.net/questions/19152/….... Jul 6, 2020 at 19:39
• And then there's this math overflow question. Jul 6, 2020 at 19:43
• Yes, I have read a little bit under that math overflow question before, but I didn't find it useful, as it didn't say much I have already heard before (like the fact that topologies tell us about the nearness of points etc.). Jul 6, 2020 at 19:46
• Jul 6, 2020 at 20:24
• I'd say it was a long process until the currently applied definition cleared up. You can achieve the same by defining neighborhoods of a point as basic concept. But to catch the 'right' notion of topology, the axioms won't be that simple. Jul 7, 2020 at 18:21

To each their own, and let a thousand intuitions bloom, but here's one opinion:

I would not adopt the "distance" intuition because everything will go wrong later. For example, uniform continuity is not a topological property, it is a combination of metric and topological ideas; another example is non-Hausdorff spaces, where distinct points can't necessarily be placed in separate neighborhoods. As things get more advanced, you start to realize a lot of important properties and structures depend on metrics that are really the "nearness" concept, and not so much topologies. For example, I am answering this soft question because I am frustrated that I can't get upper hemi-continuity to work the way I want using a metric, because the sequential and topological definitions of uhc are equivalent, but the standard metric definition of uhc implies the topological one but not the converse.

The topological definition of continuity is: $$f$$ is continuous if the inverse image of every open set is open''. The whole point is to get rid of the metric, and think about properties of spaces that are preserved under continuous mappings, like connectedness, compactness, etc. The basic idea of topologies is to remove all the extra structure around the $$\varepsilon/\delta$$ or sequential definitions of continuity, and replace it with mappings between sets for which the topological definition is the "right" or only one. So if we throw out all of math and start over except for the idea of continuously stretching and shrinking and manipulating sets in continuous ways, which results are clearer and extend to more abstract spaces and what results do we lose? That's topology.

• But then why we say that a topology is the set of all open sets? In other words, it's an extra property on our original set, and not so much about continous mappings between two distinct sets. Jul 6, 2020 at 19:28
• The topology on $X$ is axiomatically defined, and for any particular set, there are potentially many different topologies. Even in a metric space, there are at least three: the discrete, indiscrete, and metric topology based on open balls. Once you've committed to the open sets, you have "implicitly" defined the continuous functions as those that take open sets back to open sets: a function or set of functions might be continuous in one topology, but not in another.
– user762914
Jul 6, 2020 at 19:34
• Okay, but then I don't understand the motivation for defining the topology like we do, with the open sets and their axioms (the union of open sets is open, and the finite interpretation of open sets is open). How does this notion "help us" better understand the set? Jul 6, 2020 at 19:38
• Part of it is so that DeMorgan's-type laws and the definition of continuity play well together, so operations like $\cap_{i=1}^n f^{-1}(A_i) = f^{-1}\left( \cap_{i=1}^n A_i\right)$. If some of those inverse images weren't in the topology or their union wasn't, there would be problems on a fundamental level with the definition ($f^{-1}(A)$ is open but you can break up $A$ in such a way that the inverse image is a union of sets that aren't open). Once you include closure, interior, and complement as operations, things become more interesting and complicated.
– user762914
Jul 6, 2020 at 19:49