What You Actually Mean When You Say Topology
Topology is the study of properties preserved under continuous deformations. That's the textbook answer, and it's not wrong. It's just not the part that matters when you're actually doing work with it. The thing that trips people up is that "continuous deformation" sounds like something you can picture, like stretching rubber. It isn't. It's a precise mathematical condition involving open sets and preimages, and the intuition comes later, if at all. Let me just get the definition out of the way so we can move on. A topological space is a set X together with a collection of subsets of X, called open sets, satisfying three conditions: The empty set and X itself are in . Arbitrary unions of members of are in . Finite intersections of members of are in .
That's it. That's the entire definition. Everything else in topology builds on this, and most of the interesting questions in the field come down to constructing the right topology for a problem or recognizing when two spaces are homeomorphic without being able to write down the explicit map between them. I spent a semester grading undergrad papers where students would state this definition correctly and then have no idea what to do with it afterward. That's the gap we need to address. The reason this definition exists is that it captures the notion of "nearness" without requiring a metric. You can talk about convergence, continuity, and compactness in spaces where distance doesn't make sense. That's the whole point. Metric spaces are a special case, not the general story. When I first learned this, it felt like mathematicians were just being difficult on purpose. It took me a while to see that the abstraction is the feature, not the bug.
How This Actually Shows Up In Practice
I'm going to give you a concrete example because the formal definition alone won't help you. Consider the real line with the standard topology, where open sets are unions of open intervals. Now consider the same set with the discrete topology, where every subset is open. Same underlying set. Completely different topological spaces. Functions that are continuous in one may not be in the other. Compactness behaves differently. Sequence convergence behaves differently. This is why the choice of topology matters more than the choice of set in most applications. Here's where it gets less intuitive. The cofinite topology on an infinite set is where the open sets are the empty set and complements of finite sets. Sequences converge to every point in this topology, which makes it useless for analysis but perfectly fine for certain algebraic geometry constructions. I ran into this when working through a problem set on spectral spaces, and my first reaction was that something was broken. Nothing was broken. The topology was doing exactly what it was supposed to do. You just have to stop expecting sequences to characterize convergence in general topological spaces. That leads me to a point most introductory courses gloss over: sequences are not sufficient to describe topology in the general case. You need nets or filters. If you're working with metric spaces or second-countable spaces, sequences are fine. Step outside that comfort zone and they fail you. I learned this the hard way when trying to prove that a certain function on a product space was continuous by checking sequential continuity. The product was uncountable, sequential continuity held, but the function wasn't actually continuous. Took me three days to spot the error. Nets would have made it immediate.
Compactness And The Things That Catch You
Compactness is probably the single most useful concept in topology, and it's also the one most commonly misunderstood. A space is compact if every open cover has a finite subcover. That's the definition. The power comes from what it lets you do: turn local information into global information. If a property holds in a neighborhood of every point and the space is compact, you can often deduce the property holds everywhere. But here's the counter-intuitive part that beginners miss: compactness is not about being bounded and closed in the way you learn in calculus. That's true in Euclidean space with the standard topology. In a general topological space, "bounded" isn't even defined. And a compact subset of a non-Hausdorff space doesn't have to be closed. I encountered this when dealing with the Zariski topology on algebraic varieties, where compact (quasi-compact, to be precise) sets are everywhere and closed sets are rare. The analogy to R^n breaks down completely, and clinging to it will mislead you. Another thing that bites people: the Heine-Borel theorem, which characterizes compact subsets of R^n as those that are closed and bounded, only works because R^n is a specific kind of space. Don't assume it generalizes. In a general metric space, compact implies complete and totally bounded, but the converse requires the space to be complete. This is a slightly different statement from Heine-Borel and applies to a wider range of situations, but it's still not universal.
Homeomorphisms And Why They Matter
Two spaces are homeomorphic if there's a continuous bijection between them with a continuous inverse. Homeomorphic spaces are topologically indistinguishable. Any topological property you can prove for one space is automatically true for the other. This is the central equivalence relation in topology, and recognizing homeomorphisms is both simpler and harder than it sounds. Simpler because you sometimes don't need to construct the explicit map. If you can prove that space A has a topological property that space B lacks, they're not homeomorphic. The standard tools are compactness, connectedness, fundamental groups, homology groups. Build a list of invariants and check them. This is how you show the sphere and the torus are different without writing down any maps. Harder because the converse doesn't work. Having the same invariants doesn't guarantee homeomorphism. There are spaces that share all the standard algebraic topological invariants but aren't homeomorphic. I ran into this explicitly when studying lens spaces in a graduate seminar. Lens spaces L(p,q) and L(p,q') can have identical homology groups and fundamental groups but be non-homeomorphic. The distinguishing invariant is a more subtle arithmetic quantity called the Reidemeister torsion. You won't see this in an introductory course. It shows up when you need to tell spaces apart that all the basic tools can't distinguish.
Constructing Topologies Without Breaking Your Brain
If you need to put a topology on a set and you don't already have one, there are standard ways to build it. The subspace topology is the easiest: if Y is a subset of X with topology , the subspace topology on Y consists of all sets of the form U Y where U is in . This is how we get the topology on [0,1] from the topology on R, and it's why [0,1) is both open and closed in itself even though it's neither open nor closed in R. The product topology is the next one you'll need. For a finite product, the basis consists of products of open sets. For an arbitrary product, you use the same idea but the basis elements can only restrict finitely many coordinates. This Tychonoff theorem proof is one of the reasons people either love or hate the axiom of choice. The product of compact spaces is compact, and the proof for arbitrary products requires it. Without choice, Tychonoff's theorem fails. This isn't a minor technicality. It means that in contexts where you can't use choice, compactness of products behaves very differently. Quotient topologies are where things get messy. You take a space and an equivalence relation, put the finest topology that makes the quotient map continuous. Functions out of the quotient space are continuous if and only if their composition with the quotient map is continuous. This is useful but easily misapplied. I once tried to construct a quotient space to model a configuration space for a mechanics problem, and the quotient topology turned out to not be Hausdorff. The space I was trying to build had points that couldn't be separated by open sets, which made the whole construction useless for the analysis I wanted to do. The workaround was to refine the equivalence relation so that the quotient became Hausdorff, but that required adding constraints that changed the problem slightly. Not ideal, but it worked.
Common Mistakes And How To Avoid Them
The biggest mistake I see is treating topology as a collection of definitions to memorize rather than a language for saying things precisely. The definition of a topological space is five lines. The consequences are vast. If you're studying topology and you can't immediately see why someone would care about the difference between first-countable and second-countable, you're approaching it wrong. Another mistake: assuming that because something is true for R^n it's true generally. It isn't. The real line is special. It's connected, locally connected, path-connected, metrizable, second-countable, locally compact, -compact, paracompact, normal, and a dozen other properties that many topological spaces lack. Don't let your intuitions from R^n carry you astray. And don't confuse related but distinct concepts. Compact and countably compact are different. Compact and limit point compact are different. In metric spaces they coincide, which is why the confusion persists. Sequential continuity and continuity are different in general. Continuous functions preserve compactness but not necessarily the converse. A continuous bijection from a compact space to a Hausdorff space is a homeomorphism, but drop either hypothesis and the inverse may fail to be continuous.
Where Topology Actually Goes After The Basics
Point-set topology is the foundation. Algebraic topology, differential topology, and geometric topology all build on it, but they diverge significantly. Algebraic topology assigns algebraic objects to topological spaces and turns geometric problems into algebraic ones. Differential topology studies smooth manifolds and smooth maps. Geometric topology focuses on low-dimensional spaces, especially surfaces and 3-manifolds, where the techniques are quite different from the high-dimensional case. If you're coming at this from analysis, you'll find functional analysis and topology are deeply connected. The weak topology on a Banach space is a prime example. It's coarser than the norm topology, which means fewer open sets, which means more convergent sequences, which is exactly what you need for existence proofs. The Banach-Alaoglu theorem, which states that the closed unit ball of a dual space is weak-* compact, is a topology result that makes analysis possible. Without it, you'd be stuck proving compactness in norm topology, which rarely works. If you're coming from algebra, you'll eventually meet schemes, and the Zariski topology will feel familiar but behave in ways that no metric intuition can predict. Closed sets are algebraic varieties. Open sets are complements of varieties. The space is almost never Hausdorff. It's compact in the sense that every open cover has a finite subcover. It's a topological space, but it doesn't resemble anything from your first course. That's normal. Scheme theory is where point-set topology becomes a tool rather than the subject itself.
Resources That Actually Help
Munkres' Topology is the standard undergraduate text. It's clear, thorough, and covers exactly what you need for a first course. The first half on point-set topology is solid. The second half on algebraic topology is good but condensed. If you want more detail on the algebraic side, Hatcher is better, though it assumes more mathematical maturity. For a different perspective, Willard's General Topology is more comprehensive but denser. It's reference-quality. I keep it on my shelf and consult it when Munkres isn't enough. Kelley's General Topology is older and more axiomatic, but it's precise and the exercises are genuinely challenging in a useful way. Online, the nLab is indispensable once you're past the introductory level. It's a wiki run by people who actually work in the area, and the entries on topics like weak topology, Stone-Čech compactification, and classifying spaces are better than most textbooks. The quality varies by entry, but the expert contributions are reliable.
There's no single best resource because topology branches so quickly. The point is to build a solid foundation in point-set topology and then follow the branch that matches your interests. The foundation matters more than the branch. Most of the confusion people have in advanced topology traces back to gaps in their understanding of open sets, continuity, and compactness at the basic level.
