Centralizers, Normalizers, and the Center

Centralizers and normalizers are the load-bearing walls of finite group theory: the The Class Equation literally counts conjugacy classes using the orbit-stabilizer relationship $|\text{class of }x| = [G : C_G(x)]$, and Sylow's theorems compute the number of Sylow $p$-subgroups via $n_p = [G : N_G(P)]$ for a Sylow subgroup $P$. In representation theory, the center $Z(G)$ governs which irreducible representations exist and how characters decompose, since $Z(G)$ acts by scalars on any irreducible representation (Schur's Lemma). Beyond pure algebra, centralizers appear in crystallography and physics as the symmetry operations commuting with a chosen axis or defect in a crystal lattice, directly analogous to $C_G(x)$ for a fixed symmetry $x$. In quantum mechanics, the centralizer of a Hamiltonian's symmetry group inside a larger symmetry group determines which additional conserved quantities (selection rules) survive a perturbation. In cryptography and coding theory, normalizer subgroups of permutation groups underlie the automorphism groups of codes, and centralizer computations are a standard primitive in computational group theory software (GAP, Magma) used to analyze large finite groups algorithmically.

Ask the tutor: Can you show why $[G : C_G(x)]$ always equals the size of $x$'s conjugacy class, using orbit-stabilizer?

Think of every element $x \in G$ as demanding a private audience: "which elements of $G$ commute with me?" The answer, $C_G(x)$, is a subgroup that measures how rigid $x$ is under conjugation. If $x$ is central — if it commutes with everyone — then $C_G(x) = G$. At the opposite extreme, in a highly non-abelian group, $C_G(x)$ can be as small as $\{e, x\}$ (or even just $\langle x \rangle$), meaning almost nothing else commutes with $x$. The normalizer relaxes the demand from "fix me" to "keep my set intact." If $S$ is a subgroup $H$, then $N_G(H)$ is the largest subgroup of $G$ in which $H$ sits as a normal subgroup — conjugation by any $g \in N_G(H)$ sends $H$ back onto itself, even though individual elements of $H$ may get scrambled among themselves. This is exactly the bridge to the notion of Normal Subgroups: a subgroup $H \trianglelefteq G$ precisely when $N_G(H) = G$. The dihedral group $D_4$ of symmetries of a square (order $8$) makes this vivid. Let $r$ be rotation by $90°$ and $s$ a reflection, with $srs^{-1} = r^{-1}$. The centralizer of $r$ consists exactly of the rotations $\{e, r, r^2, r^3\}$, since reflections invert $r$ rather than fix it. But the normalizer of the subgroup $\langle s \rangle = \{e, s\}$ can be strictly larger than its centralizer in general subgroups — though here, since reflections don't commute with rotations, we must check case by case; the key qualitative point is that "normalizing" a set is a weaker request than "centralizing" it, and the gap between the two subgroups is a direct measurement of how elements are permuted rather than fixed by conjugation.

Example 1: Centralizer and normalizer in $S_4$. Let $G = S_4$, the symmetric group on $\{1,2,3,4\}$, $|G| = 24$. Consider the element $x = (1\,2\,3)$, a $3$-cycle, and the subgroup $H = \langle x \rangle = \{e, (1\,2\,3), (1\,3\,2)\}$, of order $3$. Computing $C_G(x)$. An element $g \in S_4$ centralizes $x$ iff $g x g^{-1} = x$. Recall that conjugation by $g$ relabels the cycle: if $x = (1\,2\,3)$, then $gxg^{-1} = (g(1)\,g(2)\,g(3))$. We need $(g(1)\,g(2)\,g(3)) = (1\,2\,3)$ as cycles, which forces $g$ to fix $4$ (since $x$ fixes $4$, and any conjugate of $x$ fixes $g(4)$; for the conjugate to equal $x$, which fixes only $4$, we need $g(4)=4$) and to send $(1,2,3)$ to a cyclic rotation of itself. The rotations of $(1,2,3)$ realizable by a bijection fixing $4$ are exactly the powers of $x$ itself: $g|_{\{1,2,3\}} \in \{e, x, x^2\}$ as permutations of $\{1,2,3\}$. Hence

a subgroup of order $3$. Computing $N_G(H)$ where $H = \langle x \rangle$. Since $H$ is generated by $x$, $g \in N_G(H)$ iff $gxg^{-1} \in H$ (conjugation sends the generator to some element of $H$, and since $H$ has order $3$, either $gxg^{-1}=e$, impossible as $x\ne e$ has order 3 and conjugation preserves order, or $gxg^{-1} \in \{x, x^2\}$). Both possibilities are achievable: $g = e$ gives $gxg^{-1}=x$; and taking $g = (1\,2)$ (a transposition fixing $3,4$... let's check directly) — actually take $g=(2\,3)$: then $gxg^{-1} = (g(1)\,g(2)\,g(3)) = (1\,3\,2) = x^2 \in H$. So $(2\,3) \in N_G(H) \setminus C_G(H)$. To find $N_G(H)$ exactly, note $g \in N_G(H)$ iff $g$ permutes $\{1,2,3\}$ among themselves (and consequently fixes $4$, or sends $4$ elsewhere while permuting $\{1,2,3\}\to\{1,2,3\}$ — but a bijection of a 4-element set sending a 3-element subset into itself must fix the complement too, since it's injective). Precisely: $gxg^{-1} = (g(1)\,g(2)\,g(3))$ lies in $H = \{e,(1\,2\,3),(1\,3\,2)\}$ exactly when $\{g(1),g(2),g(3)\} = \{1,2,3\}$, i.e. $g$ restricted to $\{1,2,3\}$ is a bijection of that set (forcing $g(4)=4$ too). The permutations of $\{1,2,3,4\}$ fixing $4$ and permuting $\{1,2,3\}$ arbitrarily form a copy of $S_3$ sitting inside $S_4$. Hence

Indeed $C_G(x) = H$ has order $3$ and sits inside $N_G(H)$ of order $6$ with index $2$ — consistent with $C_G(H) \trianglelefteq N_G(H)$ from part (4) of the theorem, since here $C_G(H) = C_G(x) = H$ (every element centralizing the generator $x$ centralizes all of $H = \langle x\rangle$) and a subgroup of index $2$ is always normal.

Example 2: The center of $D_5$ vs. $D_4$ (parity matters). Let $D_n = \langle r, s \mid r^n = e, s^2 = e, srs^{-1} = r^{-1} \rangle$ be the dihedral group of order $2n$, symmetries of a regular $n$-gon. Claim: $Z(D_n) = \{e\}$ if $n$ is odd, and $Z(D_n) = \{e, r^{n/2}\}$ if $n$ is even. We verify this for $n = 5$ (odd) and $n=4$ (even) concretely, an asymmetric pair chosen precisely because the answer differs qualitatively. Case $n = 5$, $|D_5| = 10$. Suppose $g = r^k \in Z(D_5)$ for some $0 \le k \le 4$. Centrality requires $g$ to commute with $s$: $r^k s = s r^k$. But $s r^k = r^{-k} s$ (from $srs^{-1}=r^{-1}$, so $sr^k s^{-1} = r^{-k}$, i.e. $sr^k = r^{-k}s$). So we need $r^k s = r^{-k} s$, i.e. $r^k = r^{-k}$, i.e. $r^{2k} = e$. Since $r$ has order $5$, this forces $5 \mid 2k$, and since $\gcd(2,5)=1$, $5 \mid k$, so $k \equiv 0 \pmod 5$, giving only $g = e$. One checks separately that no element of the form $r^k s$ is central (such an element doesn't even commute with $r$ unless $r$ has order $\le 2$, which it doesn't here). So

Case $n = 4$, $|D_4| = 8$. By the identical computation, $r^k \in Z(D_4)$ requires $r^{2k} = e$, i.e. $4 \mid 2k$, i.e. $2 \mid k$. So $k \in \{0, 2\}$, giving candidates $e$ and $r^2$. We must also check $r^2$ commutes with all of $D_4$, not just $s$: for any $r^j s^\epsilon$, $r^2$ commutes with rotations trivially, and $r^2 (r^j s) = r^{j+2}s$ while $(r^js)r^2 = r^j (s r^2) = r^j r^{-2} s = r^{j-2}s$; since $r$ has order $4$, $r^{j+2} = r^{j-2}$ (as $4 \mid 4$), so indeed these agree. Hence $r^2$ is central. So

This confirms the general pattern: for odd $n$ the center is trivial (rotation by $r^{2k}=e$ forces $k=0$ since $\gcd(2,n)=1$), while for even $n$ the "half-turn" $r^{n/2}$ is a genuine nontrivial central element, since it satisfies $2\cdot(n/2) = n \equiv 0$. This is exactly the kind of asymmetric, parameter-dependent computation that shows why one must verify the divisibility condition rather than guess.

Cauchy's work on substitutions (his term for permutations) treated conjugation and commuting elements concretely, cycle by cycle, long before "centralizer" or "normalizer" existed as vocabulary. The formal names and the abstract group-theoretic framework were consolidated later, chiefly through the efforts of Camille Jordan, Ferdinand Georg Frobenius, and William Burnside, who needed precise language to organize the emerging theory of finite groups, particularly to state and prove Sylow's theorems and the class equation.