The Universe of Discourse

Thu, 23 Nov 2017

Mathematical pettifoggery and pathological examples

Last week I said:

Mathematicians do tend to be the kind of people who quibble and pettifog over the tiniest details. This is because in mathematics, quibbling and pettifogging does work.

This example is technical, but I think I can explain it in a way that will make sense even for people who have no idea what the question is about. Don't worry if you don't understand the next paragraph.

In this math SE question: a user asks for an example of a connected topological space !!\langle X, \tau\rangle!! where there is a strictly finer topology !!\tau'!! for which !!\langle X, \tau'\rangle!! is disconnected. This is a very easy problem if you go about it the right way, and the right way follows a very typical pattern which is useful in many situations.

The pattern is “TURN IT UP TO 11!​!” In this case:

  1. Being disconnected means you can find some things with a certain property.
  2. If !!\tau'!! is finer than !!\tau!!, that means it has all the same things, plus even more things of the same type.
  3. If you could find those things in !!\tau!!, you can still find them in !!\tau'!! because !!\tau'!! has everything that !!\tau!! does.
  4. So although perhaps making !!\tau!! finer could turn a connected space into a disconnected one, by adding the things you needed, it definitely can't turn a disconnected space into a connected one, because the things will still be there.
  5. So a way to look for a connected space that becomes disconnected when !!\tau!! becomes finer is:

    Start with some connected space. Then make !!\tau!! fine as you possibly can and see if that is enough.

  6. If that works, you win. If not, you can look at the reason it didn't work, and maybe either fix up the space you started with, or else use that as the starting point in a proof that the thing you're looking for doesn't exist.

I emphasized the important point here. It is: Moving toward finer !!\tau!! can't hurt the situation and might help, so the first thing to try is to turn the fineness knob all the way up and see if that is enough to get what you want. Many situations in mathematics call for subtlety and delicate reasoning, but this is not one of those.

The technique here works perfectly. There is a topology !!\tau_d!! of maximum possible fineness, called the “discrete” topology, so that is the thing to try first. And indeed it answers the question as well as it can be answered: If !!\langle X, \tau\rangle!! is a connected space, and if there is any refinement !!\tau'!! for which !!\langle X, \tau'\rangle!! is disconnected, then !!\langle X, \tau_d\rangle!! will be disconnected. It doesn't even matter what connected space you start with, because !!\tau_d!! is always a refinement of !!\tau!!, and because !!\langle X, \tau_d\rangle!! is always disconnected, except in trivial cases. (When !!X!! has fewer than two points.)

Right after you learn the definition of what a topology is, you are presented with a bunch of examples. Some are typical examples, which showcase what the idea is really about: the “open sets” of the real line topologize the line, so that topology can be used as a tool for studying real analysis. But some are atypical examples, which showcase the extreme limits of the concept that are as different as possible from the typical examples. The discrete space is one of these. What's it for? It doesn't help with understanding the real numbers, that's for sure. It's a tool, it's the knob on the topology machine that turns the fineness all the way up.[1] If you want to prove that the machine does something or other for the real numbers, one way is to show that it always does that thing. And sometimes part of showing that it always does that thing is to show that it does that even if you turn the knob all the way to the right.

So often the first thing a mathematician will try is:

What happens if I turn the knob all the way to the right? If that doesn't blow up the machine, nothing will!

And that's why, when you ask a mathematician a question, often the first thing they will say is “ťhat fails when !!x=0!!” or “that fails when all the numbers are equal” or “ťhat fails when one number is very much bigger than the other” or “that fails when the space is discrete” or “that fails when the space has fewer than two points.” [2]

After the last article, Kyle Littler reminded me that I should not forget the important word “pathological”. One of the important parts of mathematical science is figuring out what the knobs are, how far they can go, what happens if you turn them all the way up, and what are the limits on how they can be set if we want the machine to behave more or less like the thing we are trying to study.

We have this certain knob for how many dents and bumps and spikes we can put on a sphere and have it still be a sphere, as long as we do not actually puncture or tear the surface. And we expected that no matter how far we turned this knob, the sphere would still divide space into two parts, a bounded inside and an unbounded outside, and that these regions should behave basically the same as they do when the sphere is smooth.[3]

But no, we are wrong, the knob goes farther than we thought. If we turn it to the “Alexander horned sphere” setting, smoke starts to come out of the machine and the red lights begin to blink.[4] Useful! Now if someone has some theory about how the machine will behave nicely if this and that knob are set properly, we might be able to add the useful observation “actually you also have to be careful not to turn that “dents bumps and spikes” knob too far.”

The word for these bizarre settings where some of the knobs are in the extreme positions is “pathological”. The Alexander sphere is a pathological embedding of !!S^2!! into !!\Bbb R^3!!.

[1] The leftmost setting on that knob, with the fineness turned all the way down, is called the “indiscrete topology” or the “trivial topology”.

[2] If you claim that any connected space can be disconnected by turning the “fineness” knob all the way to the right, a mathematican will immediately turn the “number of points” knob all the way to the left, and say “see, that only works for spaces with at least two points”. In a space with fewer than two points, even the discrete topology is connected.

[3]For example, if you tie your dog to a post outside the sphere, and let it wander around, its leash cannot get so tangled up with the sphere that you need to walk the dog backwards to untangle it. You can just slip the leash off the sphere.

[4] The dog can get its leash so tangled around the Alexander sphere that the only way to fix it is to untie the dog and start over. But if the “number of dimensions” knob is set to 2 instead of to 3, you can turn the “dents bumps and spikes” knob as far as you want and the leash can always be untangled without untying or moving the dog. Isn't that interesting? That is called the Jordan curve theorem.

[Other articles in category /math] permanent link