Let F be a field and n,p≥1 be natural numbers. A matrixA∈Fn×p over F is a collection of np scalars aij∈F indexed by a row indexi=1,…,n and a column indexj=1,…,p. The order of A is n×p, where n is the number of rows and p is the number of columns. A matrix is written in rectangular form.
Let A∈Fn×p be a matrix. The matrix A is square if and only if n=p. The scalars aii of a square matrix are its diagonal elements.
A column vectora∈Fn is a matrix of order n×1 and is written in column form.
a=a1a2⋮an
A row vectora∈F1×p is a matrix of order 1×p and is written in row form.
a=[a1a2…ap]
Let A∈Fn×p be a matrix. The columns of A are the p column vectors a1,…,ap∈Fn formed from each column of A, and the rows of A are the n row vectors formed from each row of A.
Let A,B∈Fn×p be matrices of the same order. Matrix addition is the matrix A+B∈Fn×p whose elements are the element-wise sums.
(A+B)ij:=aij+bij
Let A∈Fn×p be a matrix and λ∈F be a scalar. Matrix scalar multiplication is the matrix λA∈Fn×p whose elements are the elements of A multiplied by λ.
(λA)ij:=λaij
The set Fn×p with matrix addition and matrix scalar multiplication is a vector space over F.
Let A∈Fn×p and B∈Fp×m be matrices such that A has as many columns as B has rows. Matrix multiplication is the matrix AB∈Fn×m whose element in row i and column j is the sum of the products of the elements of row i of A with the elements of column j of B.
Matrix multiplication has the following properties:
Associativity: (AB)C=A(BC)∀A∈Fn×p,B∈Fp×m,C∈Fm×q.
Distributivity: A(B+C)=AB+AC and (A+D)B=AB+DB∀A∈Fn×p,B∈Fp×m,C∈Fp×m,D∈Fn×p.
Non-commutativity: in general AB=BA, even when both products are defined.
Let n≥1. The identity matrixIn∈Fn×n is the square matrix whose diagonal elements are 1 and whose other elements are 0. The identity matrix has the following property for every A∈Fn×p:
InA=AIp=A
Let A∈Fp×p be a square matrix and n∈N. The matrix powerAn∈Fp×p is defined recursively by A0=Ip and, for n≥1,
An:=An−1A.
Let A∈Fn×p be a matrix. The transpose of A is the matrix A⊺∈Fp×n whose rows are the columns of A.
Let A∈Fn×p be a matrix. The column space of A is the subspace of Fn spanned by the columns of A. The row space of A is the subspace of F1×p spanned by the rows of A. The column space and the row space of a matrix have equal dimension.
Let A∈Fn×p be a matrix. The rank of A, written rankA, is the dimension of the column space of A. The rank of A equals the maximum number of linearly independent columns of A, which equals the maximum number of linearly independent rows of A. The rank of A is at most min{n,p}.
Let A∈Fn×p be a matrix with rank r. The matrix A is full rank if and only if r=min{n,p}. The matrix A is rank deficient if and only if r<min{n,p}, in which case its rank deficiency is min{n,p}−r.
Let A∈Fn×n be a square matrix. The matrix A is a diagonal matrix if and only if every element of A that is not a diagonal element is 0.
Let A∈Fn×p be a matrix. The matrix A is a null matrix if and only if every element of A is 0.
Let A∈Fn×n be a square matrix. The matrix A is a symmetric matrix if and only if A⊺=A. The matrix A is a skew-symmetric matrix if and only if A⊺=−A.
Let A∈Fn×n be a square matrix. The matrix A is an upper triangular matrix if and only if every element below the diagonal is 0, that is aij=0∀i>j. The matrix A is a lower triangular matrix if and only if every element above the diagonal is 0, that is aij=0∀i<j.
Let A∈Fn×n be a square matrix. The matrix A is invertible if and only if there exists a matrix A−1∈Fn×n such that AA−1=A−1A=In. The matrix A−1 is the inverse of A. If an inverse exists then it is unique.
Let A∈Fn×n be a square matrix. The matrix A is singular if and only if it is not invertible.
Let A∈Fn×n be a square matrix. The following conditions are equivalent, and each is equivalent to A being invertible:
A has full rank;
the columns of A form a basis of Fn;
the transpose A⊺ is invertible;
the equation Av=0 has only the solution v=0.
The matrix inverse has the following properties for invertible matrices A,B∈Fn×n and non-zero scalars λ∈F:
(λA)−1=λ−1A−1(AB)−1=B−1A−1(A⊺)−1=(A−1)⊺
Let A∈Rn×n be a real square matrix. The matrix A is an orthogonal matrix if and only if the product of A with its transpose is the identity matrix.
AA⊺=A⊺A=In
Equivalently, A is orthogonal if and only if its columns form an orthonormal set in Rn with respect to the inner product ⟨u,v⟩=u⊺v, if and only if its rows form an orthonormal set. Orthogonal matrices have the following properties:
the inverse is the transpose: A−1=A⊺;
the product AB of orthogonal matrices A and B is orthogonal.
Let A∈Rn×p be a real matrix with n=p. The matrix A is a semi-orthogonal matrix if and only if either n>p and the columns of A form an orthonormal set, or n<p and the rows of A form an orthonormal set.
Let A∈Cn×p be a complex matrix. The conjugate transpose of A is the matrix A∗∈Cp×n whose rows are the columns of A with each element replaced by its complex conjugate.
Let A∈Cn×n be a complex square matrix. The matrix A is a Hermitian matrix if and only if A∗=A, that is aij=aji. The diagonal elements of a Hermitian matrix are real. A real symmetric matrix is Hermitian.
Let A∈Cn×n be a complex square matrix. The matrix A is a unitary matrix if and only if the product of A with its conjugate transpose is the identity matrix.
AA∗=A∗A=In
A real orthogonal matrix is unitary.
Let A∈Cn×n be a Hermitian matrix, so that v∗Av is real for every v∈Cn. The matrix A is positive definite if and only if v∗Av>0∀v∈Cn∖{0}. The matrix A is negative definite if and only if v∗Av<0∀v∈Cn∖{0}. The matrix A is positive semi-definite if and only if v∗Av≥0∀v∈Cn. The matrix A is negative semi-definite if and only if v∗Av≤0∀v∈Cn. For a real symmetric matrix the same definitions apply with v⊺Av for v∈Rn.