$G$ is a group. $Z(G) = \{g \in G: \forall x \in G (gx = xg) \}$ is called the center of $G$. It is easy to verify that $Z(G) \triangleleft G$. $\text{Inn}(G)$ is the inner automorphism group of $G$. As a consequence of the first isomorphism theorem, we have $$G / Z(G) \cong \text{Inn}(G).$$ The homomorphism involved here is defined as $a \in G \mapsto \sigma_{a} \in \text{Inn}(G)$ where $\sigma_{a}$ is a bijection from $G$ to $G$ with $\sigma_{a}(x) = axa^{-1}$. The details can be found here: Factor Group over Center Isomorphic to Inner Automorphism Group.

The isomorphism is too abstract. Could anyone offer some intuitive (or even visual) explanations?

  • 6
    $\begingroup$ Actually this isomorphism looks very concrete to me: every element $g\in G$ defines an automorphism of $G$ by conjugation. Two elements define the same automorphism exactly when they "differ" by a central element. $\endgroup$ Mar 19, 2014 at 13:51
  • 1
    $\begingroup$ What is the central elements? $\endgroup$
    – hengxin
    Mar 19, 2014 at 13:54
  • 2
    $\begingroup$ A central element is an element in the center $\endgroup$ Mar 19, 2014 at 13:55

2 Answers 2


Well, "conjugation by $a$" shuffles the elements of $G$: each $g$ goes to $aga^{-1}$, right?

There are two possibilities: (i) if $a$ happens to commute with every element of $g$, then $aga^{-1}$ is the same as $gaa^{-1} = g$, so the "shuffling" is just the identity. (ii) If not, then some element $g$ is sent to a different element, and the shuffling is nontrivial.

So not every element $a$ of $G$ leads to an interesting shuffle: only the ones that fail to commute with at least one other element. So you might think that the inner automorphisms would correspond to $G - Z(G)$. But in fact, if $a$ produces an interesting automorphism, $\phi$, but $b$ produces the identity, then $ab$ will also produce $\phi$. So really the right correspondence is with $G/Z(G)$.

Does that help at all?

Let me add a concrete example: Suppose that $G$ is the group of all rotations and dilations of 3-space, where the rotations leave the origin fixed, and by "dilations" I mean maps like $\mathbf v \mapsto c\mathbf v$, where $c$ is a nonzero scalar. Then the subgroup $D$ of dilations is the center of $G$: if you scale up, rotate, and then scale down, it's the same as just rotating. And if you scale up by $a$, then by $b$, then by $1/a$, it's the same as scaling just by $b$.

And $G/D$ looks just like $SO(3)$ (i.e., the group of origin-fixing rotations of 3-space, known as the "special orthogonal" group -- "orthogonal" because the matrices are orthogonal; "special" because their determinant is $+1$). Why does this correspond to the inner automorphisms of $G$? Because each rotation $R$ corresponds to an automorphism of $G$, namely, the change of basis induced by $R$.

I don't know if that helps clarify things or not, but it was worth a shot...

  • $\begingroup$ Yes, it does. Thanks. I will accept it if no other "competitive" answers appear (for some time). $\endgroup$
    – hengxin
    Mar 19, 2014 at 14:18
  • $\begingroup$ Nice example! Possible typos: the subgroup $D$ is dilations is the center of $G$? In addition, what does $SO$ stand for? $\endgroup$
    – hengxin
    Mar 20, 2014 at 8:13

I would suggest looking at it in the following way.

If we define $Aut(G)$ to be the group of automorphisms of $G$, i.e.

$$ Aut(G) = \{f : G \to G \mid f \text{ is a bijective homomorphism}\} $$

then there is a map $\phi : G \to Aut(G)$ given by

$$ \phi(g) = \big(h \mapsto ghg^{-1}\big) $$

that is, $\phi(g)$ is the automorphism given by conjugation by $g$.

Now, it is (hopefully) clear that the image of $G$ in $Aut(G)$ is exactly the set of inner automorphisms, i.e. $Inn(G)$. What is the kernel? Well, the kernel is exactly those $g \in G$ such that $h = ghg^{-1}$ for all $h \in G$. i.e. the kernel is $Z(G)$!

Thus it follows that $im(\phi) = G / \ker\phi$ (this is a restatement of the first isomorphism theorem), i.e. $Inn(G) = G / Z(G)$.


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.