A simple question. Let $ f:X\to Y $ be a function and let $ E_f:=\{(x, y): f (x)=f (y)\}\subset X\times X $. What is the name of the set $ E(f) $? It would be nice to have some reference also. It seems to be a well known notion/construction in set theory or topology, but I couldn't find anything about it.

Another question is whether the following statement is true: If $ X$ is a connected topological space and the function $ f$ is continuous then $ E_f $ is connected.

  • 1
    $\begingroup$ In algebra, if f is a homomorphism, then E is a congruence. If f is not a homomorphism, E can still be an equivalence relation of interest. $\endgroup$ – The Masked Avenger Dec 28 '14 at 19:51
  • 2
    $\begingroup$ Another term for this is equalizer (ncatlab.org/nlab/show/equalizer) where the two parallel morphisms coincide. $\endgroup$ – Vidit Nanda Dec 28 '14 at 20:44
  • 5
    $\begingroup$ @ViditNanda it's not the equalizer, it's the kernel pair. $\endgroup$ – user40276 Dec 30 '14 at 0:21

It's called the kernel or kernel pair of $f$. It is used all over the place in category theory, for example to describe the useful notion of regular category where one sets up Galois connections which in one direction sends a morphism $f: X \to Y$ to the pair of projections $\pi_1, \pi_2: \ker(f) \rightrightarrows X$, and in the other sends a parallel pair $K \rightrightarrows X$ to its coequalizer $X \to Q$. The image of $f: X \to Y$ in such categories is then the coequalizer of its kernel pair.


E(f) does not have to be connected even when $\ X\ $ is.

Example: Consider $\ S^1 := \{z\in\mathbb C : |z| = 1\}\ $ -- the unit circle; and also $\ f:S^1\rightarrow S^1\ $ such that:

$$\forall_{z\in S^1}\ f(z):= z^2$$

Then $\ E(f) = \{(u\ v)\in S^1\times S^1 : u^2=v^2\}\ $ is not conected.

REMARK:   If $\ f:X\rightarrow Y\ $ is such that $\ X\ $ is connected, and $\ f^{-1}(y)\ $ is connected for every $\ y\in Y\ $, then $\ E(f)\ $ is connected

Example--just a variation of the above one:   E(f) is disjoint for $\ f:\mathbb R\rightarrow\mathbb R^2 $ given by: $\ \forall_{x\in\mathbb R}\ f(x):= \exp(\imath\cdot x)$.

  • 1
    $\begingroup$ A nice example! What is about $ f:\mathbb{R}^n\to\mathbb{R}$. My original problem comes from a rather concrete $ f:\mathbb{R}^4\to\mathbb{R} $ for which the statement is true, even in a considerably larger class of functions. $\endgroup$ – Vladimir Tkachev Dec 29 '14 at 20:06
  • 1
    $\begingroup$ @VladimirTkachev -- you got nice result! (I tried for more of something positive but didn't get anything). $\endgroup$ – Włodzimierz Holsztyński Dec 30 '14 at 7:35
  • $\begingroup$ The examples I have in mind, all of them have the property that "there exists at least one $ a\in \mathbb { R } $ such that $ f^{-1}(a) $ is connected". $\endgroup$ – Vladimir Tkachev Dec 30 '14 at 11:37
  • $\begingroup$ P.S. Notice that the target space is $\mathbb {R} $, it might be crucial. $\endgroup$ – Vladimir Tkachev Dec 30 '14 at 11:44

$E_f$ is an equivalence relation on $X$ and conversely for every equivalence relation $R$, you can construct the topological quotient $X/R$ and for the induced map $f:X\to X/R$, $E_f=R$. There are unconnected equivalence relations on connected spaces that contradicts the above statement.

For example $X=[0,1], R=\{(x,x):x\in X\}\cup\{(0,1),(1,0)\}, X/R = S^1$.


I think this is called fibered product (or rather a special case thereof) $X\times_YX$. More generally, one would have two maps $X\to Z$, $Y\to Z$ and fibered product $X\times_ZY$.


@VladimirTkachev, in addition to the general case of maps $\ f:X\rightarrow Y,\ $ was also concerned with the special case of $\ Y:=\mathbb R,\ $ i.e. with $\ X\rightarrow\mathbb R.\ $ I'll give an example (negative) in the extra simple case of $\ f:\mathbb I\rightarrow\mathbb R,\ $ where $\ \mathbb I\ $ is a closed interval.

EXAMPLE   Let $\ f:[-1;1]\rightarrow\mathbb R\ $ be defined by:

$$\forall_{x\in[-1;1]}\quad f(x) := (x+1)\cdot x\cdot(x-1)$$

Then $\ E(f)\ $ is disconnected. Indeed, $\ (-1\,\ 1)\in E(f)\ $ is an isolated point of $\ E(f).\ $ Of course $\ (1\,\ -\!1)\ $ is another isolated point like this, and there are no other isolated points in $\ E(f)$.

PROOF   We have 3 cases:

  • if $\ (x\ y)\in E(f)\ $ is such that $\ f(x)=f(y) < 0\ $ then both $\ x\ y>0\ $ are positive;
  • $\ E(f)\cap f^{-1}(0)\ =\ \{-1\,\ 0\,\ 1\}^2$
  • if $\ (x\ y)\in E(f)\ $ is such that $\ f(x)=f(y) > 0\ $ then both $\ x\ y<0\ $ are negative.

Thus the distance (say, Euclidean) from $\ (-1\,\ 1)\ $ or $\ (1\,\ -\!1)\ $ to any other point $\ (x\ y)\in E(f)\ $ is at least $\ 1$.

End of PROOF

(I'll add some comments/extensions).

  • $\begingroup$ Thank you for good examples! Probably it should be made more precise what kind of $X$ and $ f$ are relevant. I've also considered a similar cubic polynomial, but with the domain of definition being the whole $ X =\mathbb {R} $ instead of an interval. In that case, $ E (f)$ is connected (being the union of an oval and a diagonal line passing through the oval). The example you construct can be thought as a cutoff of the set $E(f)$ above. Thus, some 'completteness' of X is needed. Another relevant example is $ f=\sin x\sin y:\,\mathbb {R}^2\to\mathbb {R}$ for which $ E(f) $ is connected again. $\endgroup$ – Vladimir Tkachev Jan 3 '15 at 13:59

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.