This article includes a list of general references, but it lacks sufficient corresponding inline citations. Please help to improve this article by introducing more precise citations. (August 2022) (Learn how and when to remove this template message)

In mathematics, particularly in matrix theory, a **permutation matrix** is a square binary matrix that has exactly one entry of 1 in each row and each column and 0s elsewhere. Each such matrix, say P, represents a permutation of m elements and, when used to multiply another matrix, say A, results in permuting the rows (when pre-multiplying, to form PA) or columns (when post-multiplying, to form AP) of the matrix A.

Given a permutation π of *m* elements,

represented in two-line form by

there are two natural ways to associate the permutation with a permutation matrix; namely, starting with the *m* × *m* identity matrix, *I*_{m}, either permute the columns or permute the rows, according to π. Both methods of defining permutation matrices appear in the literature and the properties expressed in one representation can be easily converted to the other representation. This article will primarily deal with just one of these representations and the other will only be mentioned when there is a difference to be aware of.

The *m × m* permutation matrix *P*_{π} = (*p*_{ij}) obtained by permuting the columns of the identity matrix *I*_{m}, that is, for each *i*, *p*_{ij} = 1 if *j* = π(*i*) and *p*_{ij} = 0 otherwise, will be referred to as the **column representation** in this article.^{[1]} Since the entries in row *i* are all 0 except that a 1 appears in column π(*i*), we may write

where , a **standard basis vector**, denotes a row vector of length *m* with 1 in the *j*th position and 0 in every other position.^{[2]}

For example, the permutation matrix *P*_{π} corresponding to the permutation is

Observe that the *j*th column of the *I*_{5} identity matrix now appears as the π(*j*)th column of *P*_{π}.

The other representation, obtained by permuting the rows of the identity matrix *I*_{m}, that is, for each *j*, *p*_{ij} = 1 if *i* = π(*j*) and *p*_{ij} = 0 otherwise, will be referred to as the **row representation**.

The column representation of a permutation matrix is used throughout this section, except when otherwise indicated.

Multiplying times a column vector **g** will permute the rows of the vector:

Repeated use of this result shows that if M is an appropriately sized matrix, the product, is just a permutation of the rows of M. However, observing that

for each k shows that the permutation of the rows is given by π

As permutation matrices are orthogonal matrices (that is, ), the inverse matrix exists and can be written as

Multiplying a row vector **h** times will permute the columns of the vector:

Again, repeated application of this result shows that post-multiplying a matrix M by the permutation matrix *P*_{π}, that is, *M P*_{π}, results in permuting the columns of M. Notice also that

Given two permutations π and σ of *m* elements, the corresponding permutation matrices *P*_{π} and *P*_{σ} acting on column vectors are composed with

The same matrices acting on row vectors (that is, post-multiplication) compose according to the same rule

To be clear, the above formulas use the prefix notation for permutation composition, that is,

Let be the permutation matrix corresponding to π in its row representation. The properties of this representation can be determined from those of the column representation since In particular,

From this it follows that

Similarly,

Permutation matrices can be characterized as the orthogonal matrices whose entries are all non-negative.^{[3]}

If (1) denotes the identity permutation, then *P*_{(1)} is the identity matrix.

Let *S _{n}* denote the symmetric group, or group of permutations, on {1,2,...,

The map *S*_{n} → GL(*n*, **Z**_{2}) that sends a permutation to its column representation is a faithful representation.

A permutation matrix is itself a doubly stochastic matrix, but it also plays a special role in the theory of these matrices. The Birkhoff–von Neumann theorem says that every doubly stochastic real matrix is a convex combination of permutation matrices of the same order and the permutation matrices are precisely the extreme points of the set of doubly stochastic matrices. That is, the Birkhoff polytope, the set of doubly stochastic matrices, is the convex hull of the set of permutation matrices.^{[4]}

The trace of a permutation matrix is the number of fixed points of the permutation. If the permutation has fixed points, so it can be written in cycle form as π = (*a*_{1})(*a*_{2})...(*a*_{k})σ where σ has no fixed points, then **e**_{a1},**e**_{a2},...,**e**_{ak} are eigenvectors of the permutation matrix.

To calculate the eigenvalues of a permutation matrix , write as a product of cycles, say, . Let the corresponding lengths of these cycles be , and let be the set of complex solutions of . The union of all s is the set of eigenvalues of the corresponding permutation matrix. The geometric multiplicity of each eigenvalue equals the number of s that contain it.^{[5]}

From group theory we know that any permutation may be written as a product of transpositions. Therefore, any permutation matrix *P* factors as a product of row-interchanging elementary matrices, each having determinant −1. Thus, the determinant of a permutation matrix *P* is the signature of the corresponding permutation.

When a matrix *M* is multiplied by a permutation matrix *P* on the left to make *PM*, the product is the result of permuting the rows of *M*. As a special case, if *M* is a column vector, then *PM* is the result of permuting the entries of *M*:

When instead *M* is multiplied by a permutation matrix on the right to make *MP*, the product is the result of permuting the columns of *M*. As a special case, if *M* is a row vector, then *MP* is the result of permuting the entries of *M*:

The permutation matrix *P*_{π} corresponding to the permutation is

Given a vector **g**,

A permutation matrix will always be in the form

where **e**_{ai} represents the *i*th basis vector (as a row) for **R**^{j}, and where

is the permutation form of the permutation matrix.

Now, in performing matrix multiplication, one essentially forms the dot product of each row of the first matrix with each column of the second. In this instance, we will be forming the dot product of each row of this matrix with the vector of elements we want to permute. That is, for example, **v **= (*g*_{0},...,*g*_{5})^{T},

**e**_{ai}·**v**=*g*_{ai}

So, the product of the permutation matrix with the vector **v** above, will be a vector in the form (*g*_{a1}, *g*_{a2}, ..., *g*_{aj}), and that this then is a permutation of **v** since we have said that the permutation form is

So, permutation matrices do indeed permute the order of elements in vectors multiplied with them.

- Costas array, a permutation matrix in which the displacement vectors between the entries are all distinct
- n-queens puzzle, a permutation matrix in which there is at most one entry in each diagonal and antidiagonal