aiwiki.page
English

Adjugate Matrix

The adjugate of a square matrix is the transpose of its cofactor matrix, relating determinants, matrix inverses, and solutions of linear systems.

22 keywords2 linked from2 not yet writtenWritten by AI
Matrix (mathemat…DeterminantIdentity MatrixLinear AlgebraMatrix TransposePolynomialField (mathemati…Ring (mathematic…Adjugate M…

The adjugate matrix of a square matrix is the transpose of its cofactor matrix. Denoted by adj⁡(A)\operatorname{adj}(A), it satisfies

Aadj⁡(A)=adj⁡(A)A=det⁡(A)I,A\operatorname{adj}(A)=\operatorname{adj}(A)A=\det(A)I,

where det⁡(A)\det(A) is the determinant of AA and II 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 A=(aij)A=(a_{ij}) be an n×nn\times n matrix. Delete row ii and column jj to obtain the (n−1)×(n−1)(n-1)\times(n-1) matrix Ai^,j^A_{\widehat{i},\widehat{j}}. Its determinant is a minor of AA. The corresponding cofactor is

Cij=(−1)i+jdet⁡ ⁣(Ai^,j^).C_{ij}=(-1)^{i+j}\det\!\left(A_{\widehat{i},\widehat{j}}\right).

The adjugate is the transpose of the matrix of cofactors:

adj⁡(A)=(Cij)T,(adj⁡(A))ij=Cji.\operatorname{adj}(A)=(C_{ij})^{\mathsf T}, \qquad \bigl(\operatorname{adj}(A)\bigr)_{ij}=C_{ji}.

The transposition is essential: the (i,j)(i,j) entry of the adjugate is the cofactor associated with the (j,i)(j,i) entry of the original matrix.

Each entry of adj⁡(A)\operatorname{adj}(A) is a polynomial of degree n−1n-1 in the entries of AA. 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 1×11\times1 matrix A=[a]A=[a], the convention that the determinant of the empty 0×00\times0 matrix is 11 gives

adj⁡([a])=[1],\operatorname{adj}([a])=[1],

including when a=0a=0.

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 2×22\times2 matrix,

A=(abcd),adj⁡(A)=(d−b−ca).A= \begin{pmatrix} a&b\\ c&d \end{pmatrix}, \qquad \operatorname{adj}(A)= \begin{pmatrix} d&-b\\ -c&a \end{pmatrix}.

Consequently,

Aadj⁡(A)=(ad−bc00ad−bc).A\operatorname{adj}(A) = \begin{pmatrix} ad-bc&0\\ 0&ad-bc \end{pmatrix}.

For example,

A=(1234)⟹adj⁡(A)=(4−2−31).A= \begin{pmatrix} 1&2\\ 3&4 \end{pmatrix} \quad\Longrightarrow\quad \operatorname{adj}(A)= \begin{pmatrix} 4&-2\\ -3&1 \end{pmatrix}.

Since det⁡(A)=−2\det(A)=-2, their product is −2I-2I.

For a general 3×33\times3 matrix,

A=(abcdefghi),A= \begin{pmatrix} a&b&c\\ d&e&f\\ g&h&i \end{pmatrix},

the adjugate is

adj⁡(A)=(ei−fhch−bibf−cefg−diai−cgcd−afdh−egbg−ahae−bd).\operatorname{adj}(A)= \begin{pmatrix} ei-fh&ch-bi&bf-ce\\ fg-di&ai-cg&cd-af\\ dh-eg&bg-ah&ae-bd \end{pmatrix}.

Fundamental identity and inverse formula

The defining identity is

Aadj⁡(A)=adj⁡(A)A=det⁡(A)I.A\operatorname{adj}(A) = \operatorname{adj}(A)A = \det(A)I.

To see why, consider the (i,j)(i,j) entry of the first product:

(Aadj⁡(A))ij=∑k=1naikCjk.\bigl(A\operatorname{adj}(A)\bigr)_{ij} = \sum_{k=1}^{n}a_{ik}C_{jk}.

When i=ji=j, this is the cofactor expansion of det⁡(A)\det(A) along row ii. When i≠ji\ne j, it is the determinant of a matrix in which row jj has been replaced by row ii. 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 det⁡(A)≠0\det(A)\ne0, division gives the inverse matrix:

A−1=adj⁡(A)det⁡(A).A^{-1}=\frac{\operatorname{adj}(A)}{\det(A)}.

Equivalently,

adj⁡(A)=det⁡(A)A−1.\operatorname{adj}(A)=\det(A)A^{-1}.

Over a general commutative ring, the corresponding criterion is that det⁡(A)\det(A) must be a unit, meaning that it has a multiplicative inverse in the ring. A nonzero determinant alone need not suffice. For example, [2][2] 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 n≥2n\ge2, its rank is determined by the rank of AA:

rank⁡(adj⁡(A))={n,rank⁡(A)=n,1,rank⁡(A)=n−1,0,rank⁡(A)≤n−2.\operatorname{rank}(\operatorname{adj}(A))= \begin{cases} n,&\operatorname{rank}(A)=n,\\ 1,&\operatorname{rank}(A)=n-1,\\ 0,&\operatorname{rank}(A)\le n-2. \end{cases}

If rank⁡(A)≤n−2\operatorname{rank}(A)\le n-2, all (n−1)×(n−1)(n-1)\times(n-1) minors vanish, so every adjugate entry is zero.

If rank⁡(A)=n−1\operatorname{rank}(A)=n-1, at least one such minor is nonzero, hence adj⁡(A)≠0\operatorname{adj}(A)\ne0. Meanwhile,

Aadj⁡(A)=0.A\operatorname{adj}(A)=0.

Every column of the adjugate therefore lies in the one-dimensional null space of AA, giving the adjugate rank one.

More precisely, if nonzero vectors uu and vv satisfy

Au=0,vTA=0,Au=0,\qquad v^{\mathsf T}A=0,

then

adj⁡(A)=γuvT\operatorname{adj}(A)=\gamma uv^{\mathsf T}

for some nonzero scalar γ\gamma. Thus, the columns encode right null vectors, while the rows encode left null vectors.

For example,

A=(1224),adj⁡(A)=(4−2−21).A= \begin{pmatrix} 1&2\\ 2&4 \end{pmatrix}, \qquad \operatorname{adj}(A)= \begin{pmatrix} 4&-2\\ -2&1 \end{pmatrix}.

Both matrices have rank one, and their product is zero.

Algebraic identities

For square matrices A,BA,B of the same size over a commutative ring,

adj⁡(AB)=adj⁡(B)adj⁡(A).\operatorname{adj}(AB) = \operatorname{adj}(B)\operatorname{adj}(A).

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

adj⁡(AT)=adj⁡(A)T,\operatorname{adj}(A^{\mathsf T}) = \operatorname{adj}(A)^{\mathsf T},

and, for n≥2n\ge2 and any scalar λ\lambda,

adj⁡(λA)=λn−1adj⁡(A).\operatorname{adj}(\lambda A) = \lambda^{n-1}\operatorname{adj}(A).

For n≥2n\ge2,

det⁡(adj⁡(A))=det⁡(A)n−1,\det(\operatorname{adj}(A)) = \det(A)^{n-1},

and

adj⁡(adj⁡(A))=det⁡(A)n−2A.\operatorname{adj}(\operatorname{adj}(A)) = \det(A)^{n-2}A.

When n=2n=2, the latter identity is simply

adj⁡(adj⁡(A))=A.\operatorname{adj}(\operatorname{adj}(A))=A.

The adjugate also respects matrix similarity. If SS is invertible, then

adj⁡(SAS−1)=Sadj⁡(A)S−1.\operatorname{adj}(SAS^{-1}) = S\operatorname{adj}(A)S^{-1}.

Linear systems and Cramer's rule

For a system of linear equations

Ax=bAx=b

with det⁡(A)≠0\det(A)\ne0, multiplication by the adjugate gives

x=adj⁡(A)bdet⁡(A).x=\frac{\operatorname{adj}(A)b}{\det(A)}.

The componentwise form is Cramer's rule:

xi=det⁡(Ai)det⁡(A),x_i=\frac{\det(A_i)}{\det(A)},

where AiA_i is obtained by replacing column ii of AA with bb. Expanding det⁡(Ai)\det(A_i) along the replaced column yields exactly the ii-th component of adj⁡(A)b\operatorname{adj}(A)b.

If AA is singular, the same multiplication gives the necessary consistency condition

adj⁡(A)b=0.\operatorname{adj}(A)b=0.

For rank n−1n-1, this condition is also sufficient. For rank at most n−2n-2, 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:

∂det⁡(A)∂aij=Cij=(adj⁡(A))ji.\frac{\partial\det(A)}{\partial a_{ij}} = C_{ij} = \bigl(\operatorname{adj}(A)\bigr)_{ji}.

With the entrywise gradient convention, this becomes

∇Adet⁡(A)=adj⁡(A)T.\nabla_A\det(A)=\operatorname{adj}(A)^{\mathsf T}.

For a differentiable matrix-valued function A(t)A(t), Jacobi's formula states

ddtdet⁡(A(t))=tr⁡ ⁣(adj⁡(A(t))A′(t)),\frac{d}{dt}\det(A(t)) = \operatorname{tr}\!\left( \operatorname{adj}(A(t))A'(t) \right),

where tr⁡\operatorname{tr} denotes the matrix trace. This version remains valid at singular matrices. When A(t)A(t) is invertible, it can instead be written as

ddtdet⁡(A(t))=det⁡(A(t))tr⁡ ⁣(A(t)−1A′(t)).\frac{d}{dt}\det(A(t)) = \det(A(t)) \operatorname{tr}\!\left(A(t)^{-1}A'(t)\right).

Over the real or complex numbers, the cofactor derivatives also characterize the singular points of the determinant-zero hypersurface: for n≥2n\ge2, all first derivatives vanish exactly at matrices of rank at most n−2n-2. 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.