p-Groups and Their Properties

The nontrivial-center theorem is the single most-used lemma in finite group theory: it is the base case for induction arguments throughout the Sylow Theorems, where every finite group is studied via its maximal p-subgroups. The chain of normal subgroups of every order $p^i$ is the prototype for the definition of Nilpotent Groups, a class of groups defined exactly by possessing such a "central series," of which p-groups are the founding examples. In cryptography, groups of prime-power order (and their subgroup lattices) appear in the design and analysis of discrete-log-based schemes over finite fields and elliptic curves, where understanding p-Sylow structure of the underlying group order affects security parameter choices. In physics and crystallography, p-groups and their normal series appear when decomposing symmetry groups of periodic structures into solvable towers of prime-order extensions. In computer science, the algorithmic study of p-groups (polycyclic presentations, coset enumeration) underlies computational group theory software such as GAP and Magma, which rely on exactly the chain-of-normal-subgroups structure proven here to represent p-groups efficiently for computation.

Ask the tutor: How does the "full chain of normal subgroups" in a p-group generalize to the definition of a nilpotent group's central series?

Think of a finite group as a machine acting on itself by conjugation: each element $g$ moves every other element $x$ to $gxg^{-1}$. The orbits of this action are the conjugacy classes, and the Class Equation partitions $|G|$ into the sizes of these orbits. Crucially, orbit size always divides $|G|$ — for a p-group, every orbit size is $1$ or a multiple of $p$. Now imagine sorting the elements of a p-group into bins by conjugacy class. Bins of size $1$ correspond exactly to central elements — those that commute with everything. All other bins have size divisible by $p$. If we look at $|G| \pmod p$, the multiple-of-$p$ bins contribute nothing, so the count of size-1 bins must also be a multiple of $p$. Since the identity always sits alone in its own bin, there is at least one such bin — forcing at least $p$ of them. This is the seed from which every other property in this article grows: the center is never trivial (unless $G$ itself is trivial). Once you know the center is nontrivial, you can pick an element of order $p$ living inside it (guaranteed by Cauchy's Theorem applied to $Z(G)$), quotient it out, and repeat the whole argument on the smaller group $G/Z(G)$-adjacent pieces. This recursive "peel off a central $\mathbb{Z}/p\mathbb{Z}$" process is exactly how the full chain of normal subgroups of every order $p^i$ gets built.

Example 1: A nonabelian group of order $p^3$ — the center is strictly smaller than $G$, and we exhibit the full chain. Let $p = 3$ and let $G$ be the Heisenberg group over $\mathbb{F}_3$,

under matrix multiplication, with $|G| = 3^3 = 27$ (three free parameters $a,b,c \in \{0,1,2\}$). Computing the center. Direct computation of the matrix product shows

For a matrix with parameters $(a,b,c)$ to be central, it must commute with the generators $(1,0,0)$ and $(0,0,1)$ (i.e. with $a=1,b=c=0$ and $a=b=0,c=1$). Requiring the commutator to vanish forces $a = 0$ and $c = 0$ (checking against both generators), leaving $b$ free. So

This confirms the theorem: $Z(G) \ne \{e\}$ (it has order $3 = p$, a genuine but proper subgroup — note $G$ itself is nonabelian, since e.g. taking $a=1,c=0,b=0$ versus $a=0,c=1,b=0$ gives different products depending on order, as the formula's $ac'$ cross term shows). Building the chain. Set $H_1 = Z(G)$, order $3$. Then $G/H_1$ has order $9 = 3^2$; by the Corollary above, groups of order $p^2$ are abelian, so every subgroup of $G/H_1$ is automatically normal. Take $\bar K_1 \le G/H_1$ any subgroup of order $3$ (it exists by Cauchy's theorem applied to the abelian group $G/H_1 \cong \mathbb{F}_3^2$), e.g. the image of matrices with $b$ free and $a=0$ fixed, $c$ ranging — concretely let

One checks $H_2$ is a subgroup (closed under the product formula with $a=a'=0$) containing $H_1$, and $H_2 \trianglelefteq G$: conjugating a matrix in $H_2$ (with $a=0$) by an arbitrary element of $G$ (with parameter $a$) preserves the "$a=0$" property up to the group law's structure — indeed a direct computation of $g h g^{-1}$ for $h \in H_2, g\in G$ shows the resulting matrix again has zero in the $a$-slot, since the $a$-entries only add under multiplication and $H_2$'s condition $a=0$ is preserved by conjugation exactly because the group commutator lands back in $Z(G)\le H_2$. Finally $H_3 = G$. This gives the chain

exactly as the theorem guarantees, with $H_1 = Z(G)$ genuinely nontrivial but genuinely proper — a concrete, asymmetric (non-cyclic, nonabelian) illustration.

Example 2: Order $p^2 = 49$ with $p = 7$ — classifying the possibilities completely. Let $|G| = 49$. By the Corollary, $G$ is abelian. We now pin down exactly which groups arise, using Cauchy's theorem and the structure of finite abelian groups (built from cyclic pieces). Case A. Suppose some element $g \in G$ has order $49$. Then $G = \langle g \rangle \cong \mathbb{Z}/49\mathbb{Z}$, cyclic of order $49$. Case B. Suppose no element has order $49$; then by Lagrange every non-identity element has order dividing $49$, so order $1$ or $7$ (order $49$ excluded by assumption). Pick $g_1 \ne e$; it has order $7$, and $\langle g_1\rangle$ has order $7$, index $7$ in $G$. Pick $g_2 \notin \langle g_1 \rangle$ (exists since $|\langle g_1\rangle| = 7 < 49$); $g_2$ also has order $7$. Since $G$ is abelian, the subset

is a subgroup (product of two subgroups of an abelian group is always a subgroup), and it has size $49$ because $\langle g_1\rangle \cap \langle g_2 \rangle$ must be trivial: any common nonidentity element would generate all of $\langle g_1\rangle$ (order $7$, prime, so no proper nontrivial subgroups), forcing $\langle g_1\rangle = \langle g_2\rangle$, contradicting $g_2 \notin \langle g_1\rangle$. So $|\langle g_1\rangle \cdot \langle g_2\rangle| = 7 \cdot 7 / 1 = 49 = |G|$, giving $G = \langle g_1\rangle \times \langle g_2\rangle \cong \mathbb{Z}/7\mathbb{Z}\times\mathbb{Z}/7\mathbb{Z}$. So exactly two groups of order $49$ exist up to isomorphism: $\mathbb{Z}/49\mathbb{Z}$ and $\mathbb{Z}/7\mathbb{Z} \times \mathbb{Z}/7\mathbb{Z}$ — illustrating that "abelian of order $p^2$" is a genuinely useful classification, not just an existence statement.