Let $\phi: G \to \overline{G}$ be a surjective homomorphism with kernel $K$. I am wondering if there is a bijection from the collection of subgroups of $G$ that contain $K$, $\{ S: S \leq G, S \supseteq K \}$, and the collection of subgroups of $\overline{G}$, $\{\overline{S}: \overline{S} \leq \overline{G} \}$.

My attempt:

Let $A := \{ S: S \leq G, S \supseteq K \}, B := \{\overline{S}: \overline{S} \leq \overline{G} \}$. I considered the function $f: B \to A$ that sends $\overline{S}$ to $\phi^{-1}(\overline{S})$, the preimage of $\overline{S}$ under $\phi$. We need to check three things: that this function is well-defined, that it is surjective, and that it is injective.

[TL;DR: I think I have showed well-defined-ness and surjectivity, but I'm not sure about injectivity.]

Well-defined: Given a subgroup $\overline{S}$ of $\overline{G}$, we need to show that its preimage (under $\phi$), call it $S$, is a subgroup of $G$ that contains $K$. This would show that $f$ is indeed well-defined. $S$ contains $K$ because given $g \in K$, we have $\phi(g) = \overline{e} \in \overline{S}$. Since $K$ is nonempty, so is $S$, and so to show that $S$ is a subgroup, all that remains is to show closure under the operation and closure under inverses. Thus, let $x, y \in S$. Then $\phi(xy) = \phi(x)\phi(y) \in \overline{S}$, which shows closure under the operation. Finally, $\phi(x^{-1}) = \phi(x)^{-1} \in \overline{S}$, which shows closure under inverses.

Surjective: Given $S \leq G, S \supseteq K$, we want to show that $S$ is the preimage of some subgroup $\overline{S}$ of $\overline{G}$. I claim that the image of $S$, $\phi(S)$, satisfies this. That is, I claim that $\phi(S)$ is a subgroup of $\overline{G}$ and $\phi^{-1}(\phi(S)) = S$. First, we prove that $\phi(S)$ is a subgroup of $\overline{G}$. Well, $\phi(S)$ is nonempty because it contains $\overline{e} = \phi(e)$. It is closed under the operation because given $s, t \in S$, we have $\phi(s)\phi(t) = \phi(st) \in \phi(S)$. It is closed under inverses because $\phi(s)^{-1} = \phi(s^{-1}) \in \phi(S)$. Next, we show that $(\phi^{-1}(\phi(S)) = S$. The containment $S \subseteq (\phi^{-1}(\phi(S))$ is a standard fact about images/preimages. As for $(\phi^{-1}(\phi(S)) \subseteq S$, suppose $x$ is in the LHS, so that $\phi(x) \in \phi(S)$, so that $\phi(x) = \phi(s)$ for some $s \in S$. So $\overline{e} = \phi(x)\phi(s)^{-1} = \phi(xs^{-1})$, hence $xs^{-1} \in K$. Since $K \subseteq S$ by assumption, this means $xs^{-1} \in S$, so $x \in S$.

Injective: This is where I am stuck. I think I should show that no subgroup $S$ of $G$ that contains $K$ can be the preimage of two different subgroups of $\overline{G}$, which I've tried to do but am not sure how.

  • 1
    $\begingroup$ Injectivity is already true on the level of subsets because the map is a surjection. Let $x$ be in one subgroup but not the other (or subset). Then $x=\phi(g)$ for some $g$ by surjectivity. Thus $g$ is contained in one preimage, but not the other. $\endgroup$
    – jgon
    Commented Sep 11, 2020 at 23:04
  • 1
    $\begingroup$ @jgon Thank you, I forgot about using surjectivity of $\phi$. That helped me show injectivity of the map $f$ that I considered. (If $\overline{S_1} \neq \overline{S_2}$, say $x \in \overline{S_1} \backslash \overline{S_2}$, then by surjectivity of $\phi$ there is $g$ such that $x = \phi(g)$. In other words, $g \in \phi^{-1}(\overline{S_1})$, but $g \notin \phi^{-1}(\overline{S_2})$. So $\phi^{-1}(\overline{S_1}) \neq \phi^{-1}(\overline{S_2})$.) $\endgroup$
    – twosigma
    Commented Sep 12, 2020 at 0:43

2 Answers 2


Yes, this is true and part of the isomorphism theorems. Show that every subgroup containing $K$ is the preimage of its image (this property characterizes the subgroups containing $K$).


Let me suggest a perhaps more direct approach. I will use a slightly different notation. Consider a surjective group morphism $f \colon G \to G'$ with kernel $\mathrm{Ker}f=K$. Introduce first the collections of subgroups:

$$\begin{align*} \mathscr{A}&\colon=\{H \leqslant G|\ K \leqslant H\}\\ \mathscr{B}&\colon=\left\{H'|\ H' \leqslant G'\right\}, \end{align*}$$ regarded as ordered sets when equipped with the natural order given by inclusion.

Recall the fact that for any $H \leqslant G$ it holds that $f[H] \leqslant G'$ (simply by virtue of $f$ being a morphism) and also in the inverse direction for any $H' \leqslant G'$ it holds that $K=f^{-1}\left[\left\{1_{G'}\right\}\right] \leqslant f^{-1}\left[H'\right] \leqslant G$ (once again since $f$ is a morphism and since trivially $\{1_{G'}\} \leqslant H'$ and taking preimages implements a map which is isotonic with respect to inclusion).

This means that you can define the following two maps: $$\begin{align*} \Phi \colon \mathscr{A} &\to \mathscr{B}\\ \Phi(H)\colon&=f[H]\\ \Psi \colon \mathscr{B} &\to \mathscr{A}\\ \Psi\left(H'\right)\colon&=f^{-1}\left[H'\right], \end{align*}$$ maps which are easily seen to be isotonic (increasing) with respect to inclusion.

It will suffice to establish the relations $\Psi \circ \Phi=\mathbf{1}_{\mathscr{A}}$ respectively $\Phi \circ \Psi=\mathbf{1}_{\mathscr{B}}$ in order to show that the two maps are mutually inverse order automorphisms and hence that $\mathscr{A} \approx \mathscr{B} \ (\mathbf{Ord})$ (by which I mean that $\mathscr{A}$ and $\mathscr{B}$ are isomorphic objects in the category $\mathbf{Ord}$ of ordered sets, fact which is known as the correspondence theorem for subgroups; there is also a version for normal subgroups as well as many other variations).

Let us proceed to establish the forementioned relations. Since $f$ is surjective, we have that $f\left[f^{-1}[Y]\right]=Y$ for any subset $Y \subseteq G'$, a general property characterising surjectivity for any map. This applies in particular to any subgroup $H' \in \mathscr{B}$ and entails the relation $\Phi \circ \Psi=\mathbf{1}_{\mathscr{B}}$.

As for the dual relation, we recall the general fact that $f^{-1}\left[f[X]\right]=KX=XK$ for any subset $X \subseteq G$ (in general, for arbitrary subsets $X, Y \subseteq G$ by $XY:=\{xy\}_{\substack{x \in X\\y \in Y}}$ I am referring to the subset product). In particular, for a subgroup $H \in \mathscr{A}$ we will have $f^{-1}\left[f[H]\right]=HK=H$ (since $K \subseteq H$ it follows that $H=H\{1_G\} \subseteq HK \subseteq HH \subseteq H$). This signifies that $\Psi \circ \Phi=\mathbf{1}_{\mathscr{A}}$.

I would like to leave you with the remark that in some instances it is possible to exhibit in a very natural manner the inverse map -- with immediate proof that it is indeed the inverse -- rather than to just argue for the bijectivity of the originally given map.


You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .