# Direct product of groups

In mathematics, specifically in group theory, the **direct product** is an operation that takes two groups *G* and *H* and constructs a new group, usually denoted *G* × *H*. This operation is the group-theoretic analogue of the Cartesian product of sets and is one of several important notions of direct product in mathematics.

Given groups *G *(with operation *) and *H* (with operation ∆), the **direct product** *G* × *H* is defined as follows:

The resulting algebraic object satisfies the axioms for a group. Specifically:

Let *G* and *H* be groups, let *P* = *G* × *H*, and consider the following two subsets of *P*:

Both of these are in fact subgroups of *P*, the first being isomorphic to *G*, and the second being isomorphic to *H*. If we identify these with *G* and *H*, respectively, then we can think of the direct product *P* as containing the original groups *G* and *H* as subgroups.

These subgroups of *P* have the following three important properties:
(Saying again that we identify *G*′ and *H*′ with *G* and *H*, respectively.)

Together, these three properties completely determine the algebraic structure of the direct product *P*. That is, if *P* is *any* group having subgroups *G* and *H* that satisfy the properties above, then *P* is necessarily isomorphic to the direct product of *G* and *H*. In this situation, *P* is sometimes referred to as the **internal direct product** of its subgroups *G* and *H*.

In some contexts, the third property above is replaced by the following:

This property is equivalent to property 3, since the elements of two normal subgroups with trivial intersection necessarily commute, a fact which can be deduced by considering the commutator [*g*,*h*] of any *g* in *G*, *h* in *H*.

The algebraic structure of *G* × *H* can be used to give a presentation for the direct product in terms of the presentations of *G* and *H*. Specifically, suppose that

As mentioned above, the subgroups *G* and *H* are normal in *G* × *H*. Specifically, define functions *π _{G}*:

*G*×

*H*→

*G*and

*π*:

_{H}*G*×

*H*→

*H*by

Then *π _{G}* and

*π*are homomorphisms, known as

_{H}**projection homomorphisms**, whose kernels are

*H*and

*G*, respectively.

It follows that *G* × *H* is an extension of *G* by *H* (or vice versa). In the case where *G* × *H* is a finite group, it follows that the composition factors of *G* × *H* are precisely the union of the composition factors of *G* and the composition factors of *H*.

The direct product *G* × *H* can be characterized by the following universal property. Let *π _{G}*:

*G*×

*H*→

*G*and

*π*:

_{H}*G*×

*H*→

*H*be the projection homomorphisms. Then for any group

*P*and any homomorphisms ƒ

_{G}:

*P*→

*G*and ƒ

_{H}:

*P*→

*H*, there exists a unique homomorphism ƒ:

*P*→

*G*×

*H*making the following diagram commute:

This is a special case of the universal property for products in category theory.

If *A* and *B* are normal, then *A* × *B* is a normal subgroup of *G* × *H*. Moreover, the quotient of the direct products is isomorphic to the direct product of the quotients:

Note that it is not true in general that every subgroup of *G* × *H* is the product of a subgroup of *G* with a subgroup of *H*. For example, if *G* is any non-trivial group, then the product *G* × *G* has a diagonal subgroup

The subgroups of direct products are described by Goursat's lemma. Other subgroups include fiber products of *G* and *H*.

Two elements (*g*_{1}, *h*_{1}) and (*g*_{2}, *h*_{2}) are conjugate in *G* × *H* if and only if *g*_{1} and *g*_{2} are conjugate in *G* and *h*_{1} and *h*_{2} are conjugate in *H*. It follows that each conjugacy class in *G* × *H* is simply the Cartesian product of a conjugacy class in *G* and a conjugacy class in *H*.

Along the same lines, if (*g*, *h*) ∈ *G* × *H*, the centralizer of (*g*, *h*) is simply the product of the centralizers of *g* and *h*:

Similarly, the center of *G* × *H* is the product of the centers of *G* and *H*:

Normalizers behave in a more complex manner since not all subgroups of direct products themselves decompose as direct products.

If *α* is an automorphism of *G* and *β* is an automorphism of *H*, then the product function *α* × *β*: *G* × *H* → *G* × *H* defined by

is an automorphism of *G* × *H*. It follows that Aut(*G* × *H*) has a subgroup isomorphic
to the direct product Aut(*G*) × Aut(*H*).

It is not true in general that every automorphism of *G* × *H* has the above form. (That is, Aut(*G*) × Aut(*H*) is often a proper subgroup of Aut(*G* × *H*).) For example, if *G* is any group, then there exists an automorphism *σ* of *G* × *G* that switches the two factors, i.e.

For another example, the automorphism group of **Z** × **Z** is *GL*(2, **Z**), the group of all 2 × 2 matrices with integer entries and determinant, ±1. This automorphism group is infinite, but only finitely many of the automorphisms have the form given above.

In general, every endomorphism of *G* × *H* can be written as a 2 × 2 matrix

where *α* is an endomorphism of *G*, *δ* is an endomorphism of *H*, and *β*: *H* → *G* and *γ*: *G* → *H* are homomorphisms. Such a matrix must have the property that every element in the image of *α* commutes with every element in the image of *β*, and every element in the image of *γ* commutes with every element in the image of *δ*.

When *G* and *H* are indecomposable, centerless groups, then the automorphism group is relatively straightforward, being Aut(*G*) × Aut(*H*) if *G* and *H* are not isomorphic, and Aut(*G*) wr 2 if *G* ≅ *H*, wr denotes the wreath product. This is part of the Krull–Schmidt theorem, and holds more generally for finite direct products.

It is possible to take the direct product of more than two groups at once. Given a finite sequence *G*_{1}, ..., *G*_{n} of groups, the **direct product**

This has many of the same properties as the direct product of two groups, and can be characterized algebraically in a similar way.

It is also possible to take the direct product of an infinite number of groups. For an infinite sequence *G*_{1}, *G*_{2}, ... of groups, this can be defined just like the finite direct product of above, with elements of the infinite direct product being infinite tuples.

Recall that a group *P* with subgroups *G* and *H* is isomorphic to the direct product of *G* and *H* as long as it satisfies the following three conditions:

A **semidirect product** of *G* and *H* is obtained by relaxing the third condition, so that only one of the two subgroups *G*, *H* is required to be normal. The resulting product still consists of ordered pairs (*g*, *h*), but with a slightly more complicated rule for multiplication.

It is also possible to relax the third condition entirely, requiring neither of the two subgroups to be normal. In this case, the group *P* is referred to as a **Zappa–Szép product** of *G* and *H*.

The **free product** of *G* and *H*, usually denoted *G* ∗ *H*, is similar to the direct product, except that the subgroups *G* and *H* of *G* ∗ *H* are not required to commute. That is, if

Unlike the direct product, elements of the free product cannot be represented by ordered pairs. In fact, the free product of any two nontrivial groups is infinite. The free product is actually the coproduct in the category of groups.

If *G* and *H* are groups, a **subdirect product** of *G* and *H* is any subgroup of *G* × *H* which maps surjectively onto *G* and *H* under the projection homomorphisms. By Goursat's lemma, every subdirect product is a fiber product.

Let *G*, *H*, and *Q* be groups, and let *φ*: *G* → *Q* and *χ*: *H* → *Q* be homomorphisms. The **fiber product** of *G* and *H* over *Q*, also known as a pullback, is the following subgroup of *G* × *H*:

If *φ*: *G* → *Q* and *χ*: *H* → *Q* are epimorphisms, then this is a subdirect product.