Intuitively, a homeomorphism is a way of mapping two spaces without any tearing or gluing together. Thus, I would expect the formal definition of homeomorphism in terms of continuous functions to be intuitive as well.

Specifically, the book I am reading says (translated from Chinese):

A homeomorphism is a mapping $f$ with the properties of

  1. $f$ is both surjective (don't create new points [in Chinese: 不产生新点]) and injective (without overlapping [in Chinese: 不出现重叠现象]);
  2. $f$ is continuous (without tearing [in Chinese: 不撕裂]);
  3. $f^{-1}$ is continuous (without gluing [in Chinese: 不粘连]).

The first two properties are quite intuitive to me. However, I am confused about the third one.

  1. What is gluing geometrically? What is the difference of it from non-injection (thus, overlapping) of a mapping? How is it related to the continuity of $f^{-1}$?

Consider the frequently-used example which demonstrates that the mapping $$f: [0,1) \to S^{1}: f(t) = e^{i 2 \pi t}$$ , where $S^{1}$ is the unit circle in the complex plane, is not a homeomorphism.
With $f^{-1}$ being not continuous, I would expect some gluing phenomenon in the deformation between $[0,1)$ and $S^{1}$.

  1. What is the gluing phenomenon in the above example?
  • $\begingroup$ 0 is being glued to the right side of the interval [0,1) $\endgroup$ Sep 21, 2014 at 14:19

1 Answer 1


Continuity of $f$ can be seen as "if you move just a bit, the image through $f$ will move a bit too".

In your case $f^{-1}$ being not continuous would imply that you can find two points on the codomain "close enough" such that their $f$-counter image are actually far away.

Think of a rectangle and glue two opposite sides together, and think of the obvious function $f$ mapping the rectangle onto this "bracelet" (suppose you restrain the domain so that $f$ is injective). If you take two points on the latter (one on one side of the glue line, the other on the other one), their $f$-counter image would be actually far from each other.

This is exactly what happens in your second question. Take the points on the circle $A=e^{i 2\pi \theta}$ and $B=e^{i 2\pi(-\theta)}$ with $0<\theta<1$ "small enough". On the circle A and B are really close, but their counter image on the interval $[0,1)$ would be $\theta$ and $1-\theta$, not really close.

  • $\begingroup$ Nice explanation. Do you mean removing an edge of the rectangle by restraining the domain? $\endgroup$
    – hengxin
    Sep 21, 2014 at 14:46
  • $\begingroup$ Exactly. Topologically speaking you would actually have the domain being a quotient of sets, but yeah you can think of removing one edge. $\endgroup$ Sep 21, 2014 at 14:47
  • $\begingroup$ Is it OK to regard "gluing" (of $f$ on $[0,1)$ or your rectangle) as the same with "tearing" (of $f^{-1}$ on $S^{-1}$ or your bracelet)? After all, both "gluing" and "tearing" are about continuity. Briefly, the "gluing" from the perspective of $f$ is the "tearing" from the perspective of $f^{-1}$. Similarly, the "tearing" from the perspective of $f$ is the "gluing" from the perspective of $f^{-1}$. Is this understanding reasonable? $\endgroup$
    – hengxin
    Sep 22, 2014 at 1:26
  • $\begingroup$ Yeah, it's legit. As long as $f$ is bijective, you can see it in both points of view: you "glue" the rectangle to a bracelet, and you "rip" the bracelet to a rectangle, it's just a question of domains and codomains. $\endgroup$ Sep 22, 2014 at 12:28

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.