The adjugate matrix of a square matrix is the transpose of its cofactor matrix. Denoted by , it satisfies
where is the determinant of and is the identity matrix. This identity gives a formula for the inverse of an invertible matrix, but the adjugate itself is defined even when the matrix is singular. It is a standard construction in linear algebra, particularly in determinant identities, symbolic computation, and the study of singular matrices.
Definition and terminology
Let be an matrix. Delete row and column to obtain the matrix . Its determinant is a minor of . The corresponding cofactor is
The adjugate is the transpose of the matrix of cofactors:
The transposition is essential: the entry of the adjugate is the cofactor associated with the entry of the original matrix.
Each entry of is a polynomial of degree in the entries of . Thus, its definition involves no division. It applies to matrices over a field, such as the real or complex numbers, and more generally over any commutative ring with identity.
For a matrix , the convention that the determinant of the empty matrix is gives
including when .
The older term classical adjoint is common in textbooks. It should not be confused with the conjugate transpose, also called the adjoint in complex inner-product spaces. These are different operations.
Examples
For a matrix,
Consequently,
For example,
Since , their product is .
For a general matrix,
the adjugate is
Fundamental identity and inverse formula
The defining identity is
To see why, consider the entry of the first product:
When , this is the cofactor expansion of along row . When , it is the determinant of a matrix in which row has been replaced by row . That matrix has two identical rows, so its determinant is zero. Expansion by columns proves the identity with the factors reversed.
Over a field, if , division gives the inverse matrix:
Equivalently,
Over a general commutative ring, the corresponding criterion is that must be a unit, meaning that it has a multiplicative inverse in the ring. A nonzero determinant alone need not suffice. For example, has no inverse over the integers, although it is invertible over the rational numbers.
Singular matrices and rank
Unlike the inverse, the adjugate remains defined at singular matrices. For , its rank is determined by the rank of :
If , all minors vanish, so every adjugate entry is zero.
If , at least one such minor is nonzero, hence . Meanwhile,
Every column of the adjugate therefore lies in the one-dimensional null space of , giving the adjugate rank one.
More precisely, if nonzero vectors and satisfy
then
for some nonzero scalar . Thus, the columns encode right null vectors, while the rows encode left null vectors.
For example,
Both matrices have rank one, and their product is zero.
Algebraic identities
For square matrices of the same size over a commutative ring,
The order of the factors is reversed, as in the inverse-of-a-product formula. This identity also holds when either matrix is singular.
Other useful identities include
and, for and any scalar ,
For ,
and
When , the latter identity is simply
The adjugate also respects matrix similarity. If is invertible, then
Linear systems and Cramer's rule
For a system of linear equations
with , multiplication by the adjugate gives
The componentwise form is Cramer's rule:
where is obtained by replacing column of with . Expanding along the replaced column yields exactly the -th component of .
If is singular, the same multiplication gives the necessary consistency condition
For rank , this condition is also sufficient. For rank at most , the adjugate vanishes, so it provides no information about consistency.
Derivatives of the determinant
The adjugate describes how the determinant changes with individual matrix entries:
With the entrywise gradient convention, this becomes
For a differentiable matrix-valued function , Jacobi's formula states
where denotes the matrix trace. This version remains valid at singular matrices. When is invertible, it can instead be written as
Over the real or complex numbers, the cofactor derivatives also characterize the singular points of the determinant-zero hypersurface: for , all first derivatives vanish exactly at matrices of rank at most . This connects the adjugate with algebraic geometry.
Computation and numerical limitations
Computing every cofactor separately is straightforward for small matrices but becomes inefficient as dimension increases. Recursive determinant expansion is especially costly.
In numerical linear algebra, the adjugate formula is generally not the preferred method for solving systems or computing inverses. Factorization methods such as LU decomposition typically require less work and offer better control of rounding error. Separately computed cofactors and determinants can involve cancellation, overflow, or underflow, particularly near singularity.
These computational limitations do not diminish the adjugate's exact algebraic role. Its entries remain polynomial functions of the matrix entries, making it useful in symbolic calculations and in identities that must remain valid when an inverse does not exist.