Direct Products and Semidirect Products

Understanding how to build new groups from known ones is a central theme in group theory. Direct products offer a straightforward way to combine groups, while semidirect products provide a more flexible framework that captures the essence of many important groups, such as dihedral and affine groups. Mastering these constructions is essential for classifying groups of finite order and understanding their internal structure.

The Direct Product of Groups

The most elementary method of combining two groups is via their direct product. Given two groups $G$ and $H$, their (external) direct product, denoted $G \times H$, is the set of ordered pairs $\{(g, h) \mid g \in G, h \in H\}$ with the group operation defined componentwise: $(g_1, h_1)(g_2, h_2) = (g_1 g_2, h_1 h_2)$. This construction yields a new group whose identity is $(e_G, e_H)$ and where inverses are given by $(g, h)^{-1} = (g^{-1}, h^{-1})$.

For example, the direct product $\mathbb{Z}_2 \times \mathbb{Z}_2$ gives the Klein four-group $V_4$, a group of order 4 where every non-identity element has order 2. Crucially, the subsets $\tilde{G} = \{(g, e_H)\}$ and $\tilde{H} = \{(e_G, h)\}$ are normal subgroups of $G \times H$, and each is isomorphic to $G$ and $H$, respectively. Furthermore, $\tilde{G} \cap \tilde{H} = \{(e_G, e_H)\}$ and every element in $G \times H$ can be written uniquely as a product $\tilde{g}\tilde{h}$ with $\tilde{g} \in \tilde{G}$ and $\tilde{h} \in \tilde{H}$. This leads to the internal perspective.

Recognizing a Group as an Internal Direct Product

Often, you are presented with a single group $K$ and you need to determine if it decomposes as a direct product of its subgroups. A group $K$ is the (internal) direct product of its normal subgroups $N$ and $M$ if two conditions hold: first, $N \cap M = \{e\}$; second, $K = NM$, meaning every element $k \in K$ can be written as $k = nm$ for some $n \in N, m \in M$.

The power of this recognition theorem lies in its consequences. If $K = N \times M$ (internal), then $N$ and $M$ commute elementwise ($nm = mn$ for all $n \in N, m \in M$), and the representation $k=nm$ is unique. Structurally, $K$ is isomorphic to the external direct product $N \times M$. A classic application is showing that a finite cyclic group $C_{mn}$ is isomorphic to $C_m \times C_n$ if and only if $m$ and $n$ are coprime. This internal viewpoint is the key to determining when a group "splits" into simpler pieces.

Semidirect Products: A Controlled Generalization

The direct product requires that both subgroups be normal and commute. The semidirect product relaxes the first condition for one subgroup. Formally, a group $G$ is an (internal) semidirect product of a normal subgroup $N \trianglelefteq G$ by a subgroup $H \leq G$ if $G = NH$ and $N \cap H = \{e\}$. We write $G = N \rtimes H$. Here, $H$ acts on $N$ by conjugation: for $h \in H, n \in N$, the element $h n h^{-1}$ is in $N$. This action is a homomorphism $\phi: H \to \operatorname{Aut}(N)$, where $\phi_h(n) = h n h^{-1}$.

To construct an (external) semidirect product from scratch, you need a group $N$, a group $H$, and a homomorphism $\phi: H \to \operatorname{Aut}(N)$. The set $N \times H$ is then endowed with the group operation $$(n_1, h_1) \cdot (n_2, h_2) = (n_1 \phi_{h_1}(n_2), h_1 h_2).$$ If $\phi$ is the trivial homomorphism (sending every $h$ to the identity automorphism), this operation reduces to the standard direct product. Thus, the semidirect product $N \rtimes_{\phi} H$ generalizes the direct product by allowing a non-trivial, controlled interaction between $H$ and $N$.

Classifying Groups via Product Constructions

These product constructions are powerful tools for classifying groups of small order. The classification often follows a pattern: 1) Use the Sylow theorems to establish the existence and number of subgroups of certain prime-power orders. 2) Determine if these subgroups are normal. 3) Recognize the group as a direct or semidirect product based on these relationships.

Consider groups of order $pq$, where $p$ and $q$ are primes with $p < q$. The Sylow theorems force the Sylow $q$-subgroup $N \cong C_q$ to be normal. Let $H \cong C_p$ be a Sylow $p$-subgroup. The group is either a direct product $C_q \times C_p \cong C_{pq}$ (if $H$ is also normal) or a semidirect product $C_q \rtimes C_p$. The semidirect product is determined by a non-trivial homomorphism $\phi: C_p \to \operatorname{Aut}(C_q) \cong C_{q-1}$. Such a homomorphism exists if and only if $p$ divides $q-1$. This yields the complete classification: if $p \nmid (q-1)$, only the cyclic group exists; if $p \mid (q-1)$, there is also a non-abelian semidirect product.

A richer example is order 8. The abelian groups are direct products: $C_8$, $C_4 \times C_2$, and $C_2 \times C_2 \times C_2$. The non-abelian groups can be viewed as semidirect products. The dihedral group $D_8$ is $N \rtimes H$ where $N \cong C_4$ (the rotations) and $H \cong C_2$ (a reflection), with the action inverting elements of $N$. The quaternion group $Q_8$ is not a semidirect product of two non-trivial subgroups, as all its subgroups are normal but their intersections are not trivial—highlighting the limitations of this decomposition method.

Common Pitfalls

1. Assuming all subgroups in a product are normal. In a direct product $G \times H$, the copies $\tilde{G}$ and $\tilde{H}$ are always normal. However, in a semidirect product $N \rtimes H$, only $N$ is guaranteed to be normal in the full group. The subgroup $H$ is often not normal.
2. Confusing internal and external viewpoints. It is crucial to distinguish between building a new group from two separate groups (external) and recognizing a given group as a product of its own subgroups (internal). While isomorphic, the contexts for their use are different: construction versus analysis.
3. Misapplying the semidirect product construction. For an external semidirect product $N \rtimes_{\phi} H$, the map $\phi$ must be a homomorphism into $\operatorname{Aut}(N)$. A common error is to propose an action that is not an automorphism or that does not respect the group structure of $H$.
4. Overlooking the uniqueness of representation. In an internal direct product $K = N \times M$, every element writes uniquely as $nm$. In a general semidirect product, the representation $nh$ is still unique, but the multiplication is twisted by the action. Forgetting this can lead to incorrect calculations of element orders and group operations.

Summary

• The direct product $G \times H$ combines groups with a componentwise operation, resulting in two commuting, normal subgroups within the product.
• A group is an internal direct product of normal subgroups $N$ and $M$ if $N \cap M = \{e\}$ and $NM = G$; this implies $G \cong N \times M$.
• A semidirect product $N \rtimes H$ generalizes this by requiring only $N$ to be normal, with $H$ acting on $N$ via a homomorphism $\phi: H \to \operatorname{Aut}(N)$. The direct product is the special case where $\phi$ is trivial.
• These constructions are indispensable for classifying groups of finite order, such as groups of order $pq$, where the existence of non-abelian groups hinges on the existence of non-trivial semidirect products.
• Not every group decomposes as a non-trivial direct or semidirect product (e.g., $Q_8$), making the recognition theorems as important as the construction methods themselves.