For a square matrix, the transpose of the cofactor matrix
선형 대수(linear algebra) 에서, 정사각 행렬(square matrix) A 의 수반 (adjugate ) 또는 고전적 인접 (classical adjoint )은 여인수 행렬(cofactor matrix) 의 전치이고 adj(A ) 로 표시됩니다.[1] [2] 그것은 역시 때때로 부가 행렬 (adjunct matrix ),[3] [4] 또는 "인접(adjoint)"이라고 알려져 있지만,[5] 오늘날 후자의 용어는 통상적으로 행렬에 대해 켤레 전치(conjugate transpose) 인 인접 연산자(adjoint operator) 인 다른 개념을 참조합니다.
행렬과 그것의 수반 행렬의 곱은 그 대각선 엔트리가 원래 행렬의 행렬식(determinant) 인 대각 행렬 (주요 대각선에 없는 엔트리는 0임)을 제공합니다:
A
adj
(
A
)
=
det
(
A
)
I
,
{\displaystyle \mathbf {A} \operatorname {adj} (\mathbf {A} )=\det(\mathbf {A} )\mathbf {I} ,}
여기서 I 는 A 와 같은 크기의 항등 행렬(identity matrix) 입니다. 결과적으로, 역-가능 행렬(invertible matrix) 의 곱셈 역은 그 수반을 행렬식으로 나눔으로써 구할 수 있습니다.
Definition
A 의 수반 (adjugate )은 A 의 여인수 행렬(cofactor matrix) C 의 전치(transpose) 입니다:
adj
(
A
)
=
C
T
.
{\displaystyle \operatorname {adj} (\mathbf {A} )=\mathbf {C} ^{\mathsf {T}}.}
더 자세히, R 이 단위 교환 링(commutative ring) 이고 A 가 R 에서 엔트리를 갖는 n × n 행렬이라고 가정합니다. M ij 로 표시되는 A 의 (i , j ) -소행렬식 (minor ) 은 A 의 행 i 와 열 j 를 제거한 결과로 생긴 (n − 1) × (n − 1) 행렬의 행렬식(determinant) 입니다. A 의 여인수 행렬(cofactor matrix) 은 그 (i , j ) 엔트리가 A 의 (i , j ) 여인수 (cofactor ) 인 n × n 행렬 C 이며, 이는 (i , j ) -소행렬식 곱하기 부호 인수입니다:
C
=
(
(
−
1
)
i
+
j
M
i
j
)
1
≤
i
,
j
≤
n
.
{\displaystyle \mathbf {C} =\left((-1)^{i+j}\mathbf {M} _{ij}\right)_{1\leq i,j\leq n}.}
A 의 수반은 C 의 전치, 즉, 그 (i , j ) 엔트리가 A 의 (j , i ) 여인수인 n × n 행렬입니다:
adj
(
A
)
=
C
T
=
(
(
−
1
)
i
+
j
M
j
i
)
1
≤
i
,
j
≤
n
.
{\displaystyle \operatorname {adj} (\mathbf {A} )=\mathbf {C} ^{\mathsf {T}}=\left((-1)^{i+j}\mathbf {M} _{ji}\right)_{1\leq i,j\leq n}.}
Important consequence
수반은 A 와 그 수반의 곱이 그 대각선 엔트리가 행렬식 det(A ) 인 대각 행렬(diagonal matrix) 을 산출하도록 정의됩니다. 즉,
A
adj
(
A
)
=
adj
(
A
)
A
=
det
(
A
)
I
,
{\displaystyle \mathbf {A} \operatorname {adj} (\mathbf {A} )=\operatorname {adj} (\mathbf {A} )\mathbf {A} =\det(\mathbf {A} )\mathbf {I} ,}
여기서 I 는 n × n 항등 행렬(identity matrix) 입니다. 이것은 행렬식의 라플라스 전개(Laplace expansion) 의 결과입니다.
위 공식은 행렬 대수학의 기본 결과 중 하나, 즉 A 가 역가능(invertible) 인 것과 det(A ) 가 R 의 역가능 원소(invertible element) 인 것은 필요충분(iff) 조건임을 의미합니다. 이것이 성립할 때, 위 방정식은 다음을 산출합니다:
adj
(
A
)
=
det
(
A
)
A
−
1
,
A
−
1
=
det
(
A
)
−
1
adj
(
A
)
.
{\displaystyle {\begin{aligned}\operatorname {adj} (\mathbf {A} )&=\det(\mathbf {A} )\mathbf {A} ^{-1},\\\mathbf {A} ^{-1}&=\det(\mathbf {A} )^{-1}\operatorname {adj} (\mathbf {A} ).\end{aligned}}}
Examples
1 × 1 generic matrix
0 x 0 행렬의 행렬식은 1이기 때문에, 임의의 1 × 1 행렬 (복소수 스칼라)의 수반은
I
=
[
1
]
{\displaystyle \mathbf {I} ={\begin{bmatrix}1\end{bmatrix}}}
입니다.
A
adj
(
A
)
=
A
I
=
(
det
A
)
I
{\displaystyle \mathbf {A} \operatorname {adj} (\mathbf {A} )=\mathbf {A} \mathbf {I} =(\det \mathbf {A} )\mathbf {I} }
임을 관찰하십시오.
2 × 2 generic matrix
다음 2 × 2 행렬의 수반은
A
=
[
a
b
c
d
]
{\displaystyle \mathbf {A} ={\begin{bmatrix}a&b\\c&d\end{bmatrix}}}
다음과 같습니다:
adj
(
A
)
=
[
d
−
b
−
c
a
]
.
{\displaystyle \operatorname {adj} (\mathbf {A} )={\begin{bmatrix}d&-b\\-c&a\end{bmatrix}}.}
직접 계산에 의해,
A
adj
(
A
)
=
[
a
d
−
b
c
0
0
a
d
−
b
c
]
=
(
det
A
)
I
.
{\displaystyle \mathbf {A} \operatorname {adj} (\mathbf {A} )={\begin{bmatrix}ad-bc&0\\0&ad-bc\end{bmatrix}}=(\det \mathbf {A} )\mathbf {I} .}
이 경우에서, det (adj (A )) = det (A )이고 따라서 adj (adj (A )) = A 라는 것도 참입니다.
3 × 3 generic matrix
다음 3 × 3 행렬을 생각해 보십시오
A
=
[
a
11
a
12
a
13
a
21
a
22
a
23
a
31
a
32
a
33
]
.
{\displaystyle \mathbf {A} ={\begin{bmatrix}a_{11}&a_{12}&a_{13}\\a_{21}&a_{22}&a_{23}\\a_{31}&a_{32}&a_{33}\end{bmatrix}}.}
그것의 여인수 행렬은 다음과 같습니다:
C
=
[
+
|
a
22
a
23
a
32
a
33
|
−
|
a
21
a
23
a
31
a
33
|
+
|
a
21
a
22
a
31
a
32
|
−
|
a
12
a
13
a
32
a
33
|
+
|
a
11
a
13
a
31
a
33
|
−
|
a
11
a
12
a
31
a
32
|
+
|
a
12
a
13
a
22
a
23
|
−
|
a
11
a
13
a
21
a
23
|
+
|
a
11
a
12
a
21
a
22
|
]
,
{\displaystyle \mathbf {C} ={\begin{bmatrix}+{\begin{vmatrix}a_{22}&a_{23}\\a_{32}&a_{33}\end{vmatrix}}&-{\begin{vmatrix}a_{21}&a_{23}\\a_{31}&a_{33}\end{vmatrix}}&+{\begin{vmatrix}a_{21}&a_{22}\\a_{31}&a_{32}\end{vmatrix}}\\\\-{\begin{vmatrix}a_{12}&a_{13}\\a_{32}&a_{33}\end{vmatrix}}&+{\begin{vmatrix}a_{11}&a_{13}\\a_{31}&a_{33}\end{vmatrix}}&-{\begin{vmatrix}a_{11}&a_{12}\\a_{31}&a_{32}\end{vmatrix}}\\\\+{\begin{vmatrix}a_{12}&a_{13}\\a_{22}&a_{23}\end{vmatrix}}&-{\begin{vmatrix}a_{11}&a_{13}\\a_{21}&a_{23}\end{vmatrix}}&+{\begin{vmatrix}a_{11}&a_{12}\\a_{21}&a_{22}\end{vmatrix}}\end{bmatrix}},}
여기서
|
a
i
m
a
i
n
a
j
m
a
j
n
|
=
det
[
a
i
m
a
i
n
a
j
m
a
j
n
]
.
{\displaystyle {\begin{vmatrix}a_{im}&a_{in}\\a_{jm}&a_{jn}\end{vmatrix}}=\det \!{\begin{bmatrix}a_{im}&a_{in}\\a_{jm}&a_{jn}\end{bmatrix}}.}
그것은 수반은 그것의 여인수 행렬의 전치입니다,
adj
(
A
)
=
C
T
=
[
+
|
a
22
a
23
a
32
a
33
|
−
|
a
12
a
13
a
32
a
33
|
+
|
a
12
a
13
a
22
a
23
|
−
|
a
21
a
23
a
31
a
33
|
+
|
a
11
a
13
a
31
a
33
|
−
|
a
11
a
13
a
21
a
23
|
+
|
a
21
a
22
a
31
a
32
|
−
|
a
11
a
12
a
31
a
32
|
+
|
a
11
a
12
a
21
a
22
|
]
.
{\displaystyle \operatorname {adj} (\mathbf {A} )=\mathbf {C} ^{\mathsf {T}}={\begin{bmatrix}+{\begin{vmatrix}a_{22}&a_{23}\\a_{32}&a_{33}\end{vmatrix}}&-{\begin{vmatrix}a_{12}&a_{13}\\a_{32}&a_{33}\end{vmatrix}}&+{\begin{vmatrix}a_{12}&a_{13}\\a_{22}&a_{23}\end{vmatrix}}\\&&\\-{\begin{vmatrix}a_{21}&a_{23}\\a_{31}&a_{33}\end{vmatrix}}&+{\begin{vmatrix}a_{11}&a_{13}\\a_{31}&a_{33}\end{vmatrix}}&-{\begin{vmatrix}a_{11}&a_{13}\\a_{21}&a_{23}\end{vmatrix}}\\&&\\+{\begin{vmatrix}a_{21}&a_{22}\\a_{31}&a_{32}\end{vmatrix}}&-{\begin{vmatrix}a_{11}&a_{12}\\a_{31}&a_{32}\end{vmatrix}}&+{\begin{vmatrix}a_{11}&a_{12}\\a_{21}&a_{22}\end{vmatrix}}\end{bmatrix}}.}
3 × 3 numeric matrix
구체적인 예제로, 다음을 가집니다:
adj
[
−
3
2
−
5
−
1
0
−
2
3
−
4
1
]
=
[
−
8
18
−
4
−
5
12
−
1
4
−
6
2
]
.
{\displaystyle \operatorname {adj} \!{\begin{bmatrix}-3&2&-5\\-1&0&-2\\3&-4&1\end{bmatrix}}={\begin{bmatrix}-8&18&-4\\-5&12&-1\\4&-6&2\end{bmatrix}}.}
수반은 행렬식의 역(inverse) 곱하기 행렬식, −6 인지 확인하는 것은 쉽습니다.
수반의 두 번째 행, 세 번째 열에서 −1 은 다음과 같이 계산되었습니다. 수반의 (2,3) 엔트리는 A 의 (3,2) 여인수입니다. 이 여인수는 원래 행렬 A 의 세 번째 행과 두 번째 열을 삭제함으로써 얻은 부분행렬(submatrix) 을 사용하여 계산됩니다:
[
−
3
−
5
−
1
−
2
]
.
{\displaystyle {\begin{bmatrix}-3&-5\\-1&-2\end{bmatrix}}.}
(3,2) 여인수는 부호 곱하기 이 부분행렬의 행렬식입니다:
(
−
1
)
3
+
2
det
[
−
3
−
5
−
1
−
2
]
=
−
(
−
3
⋅
−
2
−
−
5
⋅
−
1
)
=
−
1
,
{\displaystyle (-1)^{3+2}\operatorname {det} \!{\begin{bmatrix}-3&-5\\-1&-2\end{bmatrix}}=-(-3\cdot -2--5\cdot -1)=-1,}
그리고 이것은 수반의 (2,3) 엔트리입니다.
Properties
임의의 n × n 행렬 A 에 대해, 기본 계산은 수반이 다음과 같은 속성을 가지고 있음을 보여줍니다:
adj
(
I
)
=
I
{\displaystyle \operatorname {adj} (\mathbf {I} )=\mathbf {I} }
, 여기서
I
{\displaystyle \mathbf {I} }
는 항등 행렬(identity matrix) 입니다.
adj
(
0
)
=
0
{\displaystyle \operatorname {adj} (\mathbf {0} )=\mathbf {0} }
, 여기서
0
{\displaystyle \mathbf {0} }
은 영 행렬(zero matrix) 이며,
n
=
1
{\displaystyle n=1}
이면
adj
(
0
)
=
I
{\displaystyle \operatorname {adj} (\mathbf {0} )=\mathbf {I} }
임을 제외합니다.
adj
(
c
A
)
=
c
n
−
1
adj
(
A
)
{\displaystyle \operatorname {adj} (c\mathbf {A} )=c^{n-1}\operatorname {adj} (\mathbf {A} )}
, 여기서 c 는 임의의 스칼라입니다.
adj
(
A
T
)
=
adj
(
A
)
T
{\displaystyle \operatorname {adj} (\mathbf {A} ^{\mathsf {T}})=\operatorname {adj} (\mathbf {A} )^{\mathsf {T}}}
.
det
(
adj
(
A
)
)
=
(
det
A
)
n
−
1
{\displaystyle \det(\operatorname {adj} (\mathbf {A} ))=(\det \mathbf {A} )^{n-1}}
.
만약 A 가 역가능이면,
adj
(
A
)
=
(
det
A
)
A
−
1
{\displaystyle \operatorname {adj} (\mathbf {A} )=(\det \mathbf {A} )\mathbf {A} ^{-1}}
입니다. 다음임이 따라옵니다:
adj(A ) 는 역 (det A )−1 A 을 갖는 역가능입니다.
adj(A −1 ) = adj(A )−1 .
adj(A ) 는 A 에서 엔트리별 다항식(polynomial) 입니다. 특히, 실수(real) 또는 복소수에 걸쳐, 수반은 A 의 엔트리의 매끄러운 함수(smooth function) 입니다.
복소수에 걸쳐,
adj
(
A
¯
)
=
adj
(
A
)
¯
{\displaystyle \operatorname {adj} ({\overline {\mathbf {A} }})={\overline {\operatorname {adj} (\mathbf {A} )}}}
, 여기서 막대는 복소수 켤레(complex conjugation) 를 나타냅니다.
adj
(
A
∗
)
=
adj
(
A
)
∗
{\displaystyle \operatorname {adj} (\mathbf {A} ^{*})=\operatorname {adj} (\mathbf {A} )^{*}}
, 여기서 별표는 켤레 전치(conjugate transpose) 를 나타냅니다.
B 가 또 다은 n × n 행렬이라고 가정합니다. 그런-다음
adj
(
A
B
)
=
adj
(
B
)
adj
(
A
)
.
{\displaystyle \operatorname {adj} (\mathbf {AB} )=\operatorname {adj} (\mathbf {B} )\operatorname {adj} (\mathbf {A} ).}
이는 세 가지 방법으로 입증 될 수 있습니다. 한 가지 방법은, 임의의 교환 링에 대해 유효하며, 코시–비네 공식(Cauchy–Binet formula) 을 사용한 직접 계산입니다. 두 번째 방법은, 실수 또는 복소수에 대해 유효하며, 먼저 역가능 행렬 A 와 B 에 대해 다음임을 관찰하는 것입니다:
adj
(
B
)
adj
(
A
)
=
(
det
B
)
B
−
1
(
det
A
)
A
−
1
=
(
det
A
B
)
(
A
B
)
−
1
=
adj
(
A
B
)
.
{\displaystyle \operatorname {adj} (\mathbf {B} )\operatorname {adj} (\mathbf {A} )=(\det \mathbf {B} )\mathbf {B} ^{-1}(\det \mathbf {A} )\mathbf {A} ^{-1}=(\det \mathbf {AB} )(\mathbf {AB} )^{-1}=\operatorname {adj} (\mathbf {AB} ).}
모든 각 비-역가능 행렬은 역가능 행렬의 극한이기 때문에, 수반의 연속성(continuity) 은 그런-다음 A 또는 B 중 하나가 역가능이 아닐 때 공식이 참으로 유지된다는 것을 의미합니다.
이전 공식의 따름정리(corollary) 는 임의의 비-음 정수 k 에 대해 다음이라는 것입니다:
adj
(
A
k
)
=
adj
(
A
)
k
.
{\displaystyle \operatorname {adj} (\mathbf {A} ^{k})=\operatorname {adj} (\mathbf {A} )^{k}.}
만약 A 가 역가능이면, 위의 공식은 역시 음수 k 에 대해 유지됩니다.
다음 항등식으로부터
(
A
+
B
)
adj
(
A
+
B
)
B
=
det
(
A
+
B
)
B
=
B
adj
(
A
+
B
)
(
A
+
B
)
,
{\displaystyle (\mathbf {A} +\mathbf {B} )\operatorname {adj} (\mathbf {A} +\mathbf {B} )\mathbf {B} =\det(\mathbf {A} +\mathbf {B} )\mathbf {B} =\mathbf {B} \operatorname {adj} (\mathbf {A} +\mathbf {B} )(\mathbf {A} +\mathbf {B} ),}
우리는 다음을 추론합니다:
A
adj
(
A
+
B
)
B
=
B
adj
(
A
+
B
)
A
.
{\displaystyle \mathbf {A} \operatorname {adj} (\mathbf {A} +\mathbf {B} )\mathbf {B} =\mathbf {B} \operatorname {adj} (\mathbf {A} +\mathbf {B} )\mathbf {A} .}
A 가 B 와 교환한다 고 가정합니다. 왼쪽과 오른쪽의 항등식 AB = BA 에 adj(A ) 를 곱하면 다음임을 입증합니다:
det
(
A
)
adj
(
A
)
B
=
det
(
A
)
B
adj
(
A
)
.
{\displaystyle \det(\mathbf {A} )\operatorname {adj} (\mathbf {A} )\mathbf {B} =\det(\mathbf {A} )\mathbf {B} \operatorname {adj} (\mathbf {A} ).}
만약 A 가 역가능이면, 이것은 adj(A ) 도 B 와 교환한다는 것을 의미합니다. 실수 또는 복소수에 걸쳐, 연속성은 A 가 역가능이 아닐 때조차도 adj(A ) 가 B 와 교환한다는 것을 의미합니다.
마지막으로, 두 번째 증명보다 더 일반적인 증명이 있으며, 이는 n × n 행렬이 적어도 2n + 1 원소 (예를 들어, 정수 모듈로 11에 걸쳐 5 × 5 행렬)을 갖는 필드(field) 에 걸쳐 엔트리를 갖기만 하면 됩니다. det(A +t I ) 는 많아야 차수 n 을 갖는 t 에서 다항식이므로, 그것은 많아야 n 개의 근(roots) 을 가집니다. adj((A +t I )(B )) 의 ij 번째 엔트리는 많아야 차수 n 의 다항식이고, adj(A +t I ) adj(B ) 에 대해서도 마찬가지임에 주목하십시오. ij 번째 엔트리에서 이들 두 다항식은 A +t I 가 역가능인 필드의 적어도 n + 1 원소를 가지고, 역가능 행렬에 대해 항등식을 입증했으므로 적어도 n + 1개 점에서 일치합니다. n + 1개의 점에 일치하는 차수 n 의 다항식은 동일해야 합니다 (그것들을 서로 빼고 많아야 차수 n 의 다항식에 대해 n + 1개의 근을 가집니다 – 그것들의 차이가 동일하게 0이 아닌 한 모순입니다). 두 다항식은 동일하므로, 그것들은 모든 각 t 의 값에 대해 같은 값을 가집니다. 따라서, 그것들은 t = 0일 때 같은 값을 취합니다.
위의 속성과 기타 기본 계산을 사용하여, A 가 다음 속성 중 하나를 가지면, adj A 도 마찬가지라는 것을 간단하게 보여줄 수 있습니다:
만약 A 가 역가능이면, 위에서 언급한 대로, A 의 행렬식과 역의 관점에서 adj(A ) 에 대한 공식이 있습니다. A 가 역가능이 아닐 때, 수반은 다르지만 밀접하게 관련된 공식을 만족시킵니다.
만약 rk(A ) ≤ n − 2 이면, adj(A ) = 0 입니다.
만약 rk(A ) = n − 1 이면, rk(adj(A )) = 1 입니다. (일부 소행렬식은 비-영이므로, adj(A ) 는 비-영이고 따라서 랭크(rank) 적어도 1을 가집니다; 항등식 adj(A ) A = 0 은 adj(A ) 의 널공간(nullspace) 의 차원(dimension) 이 적어도 n − 1 임을 의미하므로, 그것의 랭크는 많아야 1입니다.) 따라서 adj(A ) = α xy T 가 되며, 여기서 α 는 스칼라이고 x 와 y 는 Ax = 0 와 A T y = 0 를 만족하는 벡터입니다.
Column substitution and Cramer's rule
A 를 열 벡터(column vectors) 로 분할합니다:
A
=
[
a
1
⋯
a
n
]
.
{\displaystyle \mathbf {A} ={\begin{bmatrix}\mathbf {a} _{1}&\cdots &\mathbf {a} _{n}\end{bmatrix}}.}
b 를 크기 n 의 열 벡터라고 놓습니다. 1 ≤ i ≤ n 를 고정하고 A 의 열 i 를 b 로 대체함으로써 형성된 행렬을 생각해 보십시오:
(
A
←
i
b
)
=
def
[
a
1
⋯
a
i
−
1
b
a
i
+
1
⋯
a
n
]
.
{\displaystyle (\mathbf {A} {\stackrel {i}{\leftarrow }}\mathbf {b} )\ {\stackrel {\text{def}}{=}}\ {\begin{bmatrix}\mathbf {a} _{1}&\cdots &\mathbf {a} _{i-1}&\mathbf {b} &\mathbf {a} _{i+1}&\cdots &\mathbf {a} _{n}\end{bmatrix}}.}
라플라스는 i 열을 따라 이 행렬의 행렬식을 전개합니다. 결과는 곱 adj(A )b 의 엔터리 i 입니다. 다양한 가능한 i 에 대해 이들 행렬식을 수집하면 열 벡터의 상등을 산출합니다:
(
det
(
A
←
i
b
)
)
i
=
1
n
=
adj
(
A
)
b
.
{\displaystyle \left(\det(\mathbf {A} {\stackrel {i}{\leftarrow }}\mathbf {b} )\right)_{i=1}^{n}=\operatorname {adj} (\mathbf {A} )\mathbf {b} .}
이 공식은 다음과 같은 구체적인 결과를 가집니다. 다음 선형 방정식 시스템 을 생각해 보십시오:
A
x
=
b
.
{\displaystyle \mathbf {A} \mathbf {x} =\mathbf {b} .}
A 가 비-특이(non-singular) 라고 가정합니다. 왼쪽의 이 시스템에 adj(A ) 를 곱하고 행렬식으로 나누면 다음을 산출합니다:
x
=
adj
(
A
)
b
det
A
.
{\displaystyle \mathbf {x} ={\frac {\operatorname {adj} (\mathbf {A} )\mathbf {b} }{\det \mathbf {A} }}.}
이 상황에 이전 공식을 적용하면 크라메르의 규칙 (Cramer's rule )을 산출합니다:
x
i
=
det
(
A
←
i
b
)
det
A
,
{\displaystyle x_{i}={\frac {\det(\mathbf {A} {\stackrel {i}{\leftarrow }}\mathbf {b} )}{\det \mathbf {A} }},}
여기서 x i 는 x 의 i 번째 엔트리입니다.
Characteristic polynomial
A 의 특성 다항식(characteristic polynomial) 을 다음과 같이 놓습니다:
p
(
s
)
=
det
(
s
I
−
A
)
=
∑
i
=
0
n
p
i
s
i
∈
R
[
s
]
.
{\displaystyle p(s)=\det(s\mathbf {I} -\mathbf {A} )=\sum _{i=0}^{n}p_{i}s^{i}\in R[s].}
p 의 첫 번째 나누어진 차이(divided difference) 는 차수 n − 1 의 대칭 다항식(symmetric polynomial) 입니다:
Δ
p
(
s
,
t
)
=
p
(
s
)
−
p
(
t
)
s
−
t
=
∑
0
≤
j
+
k
<
n
p
j
+
k
+
1
s
j
t
k
∈
R
[
s
,
t
]
.
{\displaystyle \Delta p(s,t)={\frac {p(s)-p(t)}{s-t}}=\sum _{0\leq j+k<n}p_{j+k+1}s^{j}t^{k}\in R[s,t].}
s I − A 에 그 수반을 곱합니다. 케일리-해밀턴 정리(Cayley–Hamilton theorem) 에 의해 p (A ) = 0 이기 때문에, 일부 기본 조작이 드러납니다:
adj
(
s
I
−
A
)
=
Δ
p
(
s
I
,
A
)
.
{\displaystyle \operatorname {adj} (s\mathbf {I} -\mathbf {A} )=\Delta p(s\mathbf {I} ,\mathbf {A} ).}
특히, A 의 분해(resolvent) 는 다음으로 정의됩니다:
R
(
z
;
A
)
=
(
z
I
−
A
)
−
1
,
{\displaystyle R(z;\mathbf {A} )=(z\mathbf {I} -\mathbf {A} )^{-1},}
그리고 위의 공식에 의해, 이것은 다음과 같습니다:
R
(
z
;
A
)
=
Δ
p
(
z
I
,
A
)
p
(
z
)
.
{\displaystyle R(z;\mathbf {A} )={\frac {\Delta p(z\mathbf {I} ,\mathbf {A} )}{p(z)}}.}
Jacobi's formula
수반은 행렬식의 도함수(derivative) 에 대한 야코비의 공식(Jacobi's formula) 에도 나타납니다. 만약 A (t ) 가 연속적으로 미분-가능(continuously differentiable) 이면, 다음과 같습니다:
d
(
det
A
)
d
t
(
t
)
=
tr
(
adj
(
A
(
t
)
)
A
′
(
t
)
)
.
{\displaystyle {\frac {d(\det \mathbf {A} )}{dt}}(t)=\operatorname {tr} \left(\operatorname {adj} (\mathbf {A} (t))\mathbf {A} '(t)\right).}
따라서 행렬식의 전체 도함수(total derivative) 는 수반의 전치입니다:
d
(
det
A
)
A
0
=
adj
(
A
0
)
T
.
{\displaystyle d(\det \mathbf {A} )_{\mathbf {A} _{0}}=\operatorname {adj} (\mathbf {A} _{0})^{\mathsf {T}}.}
Cayley–Hamilton formula
p A (t ) 를 A 의 특성 다항식이라고 놓습니다. 케일리-해밀턴 정리(Cayley–Hamilton theorem) 는 다음이라고 말합니다:
p
A
(
A
)
=
0
.
{\displaystyle p_{\mathbf {A} }(\mathbf {A} )=\mathbf {0} .}
상수 항을 분리하고 방정식에 adj(A ) 를 곱하면 A 와 p A (t ) 의 계수에만 의존하는 수반에 대한 표현식을 제공합니다. 이들 계수는 완비 지수 벨 다항식(Bell polynomials) 을 사용하여 A 의 거듭제곱의 대각합(traces) 의 관점에서 명시적으로 표현될 수 있습니다. 결과 수식은 다음과 같습니다:
adj
(
A
)
=
∑
s
=
0
n
−
1
A
s
∑
k
1
,
k
2
,
…
,
k
n
−
1
∏
ℓ
=
1
n
−
1
(
−
1
)
k
ℓ
+
1
ℓ
k
ℓ
k
ℓ
!
tr
(
A
ℓ
)
k
ℓ
,
{\displaystyle \operatorname {adj} (\mathbf {A} )=\sum _{s=0}^{n-1}\mathbf {A} ^{s}\sum _{k_{1},k_{2},\ldots ,k_{n-1}}\prod _{\ell =1}^{n-1}{\frac {(-1)^{k_{\ell }+1}}{\ell ^{k_{\ell }}k_{\ell }!}}\operatorname {tr} (\mathbf {A} ^{\ell })^{k_{\ell }},}
여기서 n 은 A 의 차원이고, 합은 s 와 다음 선형 디오판토스 방정식(Diophantine equation) 을 만족시키는 kl ≥ 0 의 모든 수열에 걸쳐 취합니다:
s
+
∑
ℓ
=
1
n
−
1
ℓ
k
ℓ
=
n
−
1.
{\displaystyle s+\sum _{\ell =1}^{n-1}\ell k_{\ell }=n-1.}
2 × 2 경우에 대해, 이것은 다음을 제공합니다:
adj
(
A
)
=
I
2
(
tr
A
)
−
A
.
{\displaystyle \operatorname {adj} (\mathbf {A} )=\mathbf {I} _{2}(\operatorname {tr} \mathbf {A} )-\mathbf {A} .}
3 × 3 경우에 대해, 이것은 다음을 제공합니다:
adj
(
A
)
=
1
2
I
3
(
(
tr
A
)
2
−
tr
A
2
)
−
A
(
tr
A
)
+
A
2
.
{\displaystyle \operatorname {adj} (\mathbf {A} )={\frac {1}{2}}\mathbf {I} _{3}\!\left((\operatorname {tr} \mathbf {A} )^{2}-\operatorname {tr} \mathbf {A} ^{2}\right)-\mathbf {A} (\operatorname {tr} \mathbf {A} )+\mathbf {A} ^{2}.}
4 × 4 경우에 대해, 이것은 다음을 제공합니다:
adj
(
A
)
=
1
6
I
4
(
(
tr
A
)
3
−
3
tr
A
tr
A
2
+
2
tr
A
3
)
−
1
2
A
(
(
tr
A
)
2
−
tr
A
2
)
+
A
2
(
tr
A
)
−
A
3
.
{\displaystyle \operatorname {adj} (\mathbf {A} )={\frac {1}{6}}\mathbf {I} _{4}\!\left((\operatorname {tr} \mathbf {A} )^{3}-3\operatorname {tr} \mathbf {A} \operatorname {tr} \mathbf {A} ^{2}+2\operatorname {tr} \mathbf {A} ^{3}\right)-{\frac {1}{2}}\mathbf {A} \!\left((\operatorname {tr} \mathbf {A} )^{2}-\operatorname {tr} \mathbf {A} ^{2}\right)+\mathbf {A} ^{2}(\operatorname {tr} \mathbf {A} )-\mathbf {A} ^{3}.}
같은 공식은 A 의 특성 다항식을 효율적으로 결정하는 파데예프–르베리에 알고리듬(Faddeev–LeVerrier algorithm) 의 종료 단계에서 직접 따릅니다.
Relation to exterior algebras
수반은 외부 대수(exterior algebras) 를 사용하여 추상적인 용어로 볼 수 있습니다. V 를 n -차원 벡터 공간 이라고 놓습니다. 외부 곱(exterior product) 은 다음과 같은 쌍선형 쌍을 정의합니다:
V
×
∧
n
−
1
V
→
∧
n
V
.
{\displaystyle V\times \wedge ^{n-1}V\to \wedge ^{n}V.}
추상적으로,
∧
n
V
{\displaystyle \wedge ^{n}V}
는 R 과 동형적(isomorphic) 이고, 임의의 그러한 동형 아래에서 외부 곱은 완전 쌍화(perfect pairing) 를 이룹니다. 그러므로, 그것은 다음과 같은 동형을 산출합니다:
ϕ
:
V
→
≅
Hom
(
∧
n
−
1
V
,
∧
n
V
)
.
{\displaystyle \phi \colon V\ \xrightarrow {\cong } \ \operatorname {Hom} (\wedge ^{n-1}V,\wedge ^{n}V).}
명시적으로, 이 쌍은 v ∈ V 를
ϕ
v
{\displaystyle \phi _{\mathbf {v} }}
로 보내며, 여기서
ϕ
v
(
α
)
=
v
∧
α
.
{\displaystyle \phi _{\mathbf {v} }(\alpha )=\mathbf {v} \wedge \alpha .}
T : V → V 가 선형 변환(linear transformation) 이라고 가정합니다. T 의 (n − 1) 번째 외부 거듭제곱에 의한 당김은 Hom 공간의 사상을 유도합니다. T 의 수반 (adjugate )은 다음과 같은 합성입니다:
V
→
ϕ
Hom
(
∧
n
−
1
V
,
∧
n
V
)
→
(
∧
n
−
1
T
)
∗
Hom
(
∧
n
−
1
V
,
∧
n
V
)
→
ϕ
−
1
V
.
{\displaystyle V\ \xrightarrow {\phi } \ \operatorname {Hom} (\wedge ^{n-1}V,\wedge ^{n}V)\ \xrightarrow {(\wedge ^{n-1}T)^{*}} \ \operatorname {Hom} (\wedge ^{n-1}V,\wedge ^{n}V)\ \xrightarrow {\phi ^{-1}} \ V.}
만약 V = R n 이 정식의 기저(canonical basis) e 1 , …, e n 이 부여되고, 이 기저(basis) 에서 T 의 행렬이 A 이면, T 의 수반은 A 의 수반입니다. 이유를 확인하기 위해,
∧
n
−
1
R
n
{\displaystyle \wedge ^{n-1}\mathbf {R} ^{n}}
에 기저를 제공합니다:
{
e
1
∧
⋯
∧
e
^
k
∧
⋯
∧
e
n
}
k
=
1
n
.
{\displaystyle \{\mathbf {e} _{1}\wedge \dots \wedge {\hat {\mathbf {e} }}_{k}\wedge \dots \wedge \mathbf {e} _{n}\}_{k=1}^{n}.}
R n 의 기본 벡터 e i 를 고정합니다.
ϕ
{\displaystyle \phi }
아래에서 e i 의 이미지는 기본 벡터를 보내는 위치에 따라 결정됩니다:
ϕ
e
i
(
e
1
∧
⋯
∧
e
^
k
∧
⋯
∧
e
n
)
=
{
(
−
1
)
i
−
1
e
1
∧
⋯
∧
e
n
,
if
k
=
i
,
0
otherwise.
{\displaystyle \phi _{\mathbf {e} _{i}}(\mathbf {e} _{1}\wedge \dots \wedge {\hat {\mathbf {e} }}_{k}\wedge \dots \wedge \mathbf {e} _{n})={\begin{cases}(-1)^{i-1}\mathbf {e} _{1}\wedge \dots \wedge \mathbf {e} _{n},&{\text{if}}\ k=i,\\0&{\text{otherwise.}}\end{cases}}}
기본 벡터 위에, T 의 (n − 1) 번째 외부 거듭제곱은 다음과 같습니다:
e
1
∧
⋯
∧
e
^
j
∧
⋯
∧
e
n
↦
∑
k
=
1
n
(
det
A
j
k
)
e
1
∧
⋯
∧
e
^
k
∧
⋯
∧
e
n
.
{\displaystyle \mathbf {e} _{1}\wedge \dots \wedge {\hat {\mathbf {e} }}_{j}\wedge \dots \wedge \mathbf {e} _{n}\mapsto \sum _{k=1}^{n}(\det A_{jk})\mathbf {e} _{1}\wedge \dots \wedge {\hat {\mathbf {e} }}_{k}\wedge \dots \wedge \mathbf {e} _{n}.}
이들 각 항은 k = i 항을 제외하고
ϕ
e
i
{\displaystyle \phi _{\mathbf {e} _{i}}}
아래에서 영으로 매핑됩니다. 그러므로,
ϕ
e
i
{\displaystyle \phi _{\mathbf {e} _{i}}}
의 당김은 다음에 대한 선형 변환입니다:
e
1
∧
⋯
∧
e
^
j
∧
⋯
∧
e
n
↦
(
−
1
)
i
−
1
(
det
A
j
i
)
e
1
∧
⋯
∧
e
n
,
{\displaystyle \mathbf {e} _{1}\wedge \dots \wedge {\hat {\mathbf {e} }}_{j}\wedge \dots \wedge \mathbf {e} _{n}\mapsto (-1)^{i-1}(\det A_{ji})\mathbf {e} _{1}\wedge \dots \wedge \mathbf {e} _{n},}
즉, 그것은 다음과 같습니다:
∑
j
=
1
n
(
−
1
)
i
+
j
(
det
A
j
i
)
ϕ
e
j
.
{\displaystyle \sum _{j=1}^{n}(-1)^{i+j}(\det A_{ji})\phi _{\mathbf {e} _{j}}.}
ϕ
{\displaystyle \phi }
의 역을 적용하면 T 의 수반이 다음에 대한 선형 변환임을 알 수 있습니다:
e
i
↦
∑
j
=
1
n
(
−
1
)
i
+
j
(
det
A
j
i
)
e
j
.
{\displaystyle \mathbf {e} _{i}\mapsto \sum _{j=1}^{n}(-1)^{i+j}(\det A_{ji})\mathbf {e} _{j}.}
결과적으로, 그 행렬 표현은 A 의 수반입니다.
만약 V 가 안의 곱(inner product) 과 부피 형식이 부여되면, 맵 φ 는 더 분해될 수 있습니다. 이 경우에서, φ 는 호지 별 연산자(Hodge star operator) 와 이중화의 합성으로 이해될 수 있습니다. 구체적으로, 만약 ω 가 부피 형식이면, 그것은 안의 곱과 함께 동형을 결정합니다:
ω
∨
:
∧
n
V
→
R
.
{\displaystyle \omega ^{\vee }\colon \wedge ^{n}V\to \mathbf {R} .}
이것은 다음 동형을 유도합니다:
Hom
(
∧
n
−
1
R
n
,
∧
n
R
n
)
≅
∧
n
−
1
(
R
n
)
∨
.
{\displaystyle \operatorname {Hom} (\wedge ^{n-1}\mathbf {R} ^{n},\wedge ^{n}\mathbf {R} ^{n})\cong \wedge ^{n-1}(\mathbf {R} ^{n})^{\vee }.}
R n 에서 벡터 v 는 다음 선형 함수형에 해당합니다:
(
α
↦
ω
∨
(
v
∧
α
)
)
∈
∧
n
−
1
(
R
n
)
∨
.
{\displaystyle (\alpha \mapsto \omega ^{\vee }(\mathbf {v} \wedge \alpha ))\in \wedge ^{n-1}(\mathbf {R} ^{n})^{\vee }.}
호지 별 연산자의 정의에 의해, 이 선형 함수형은 *v 와 이중입니다. 즉, ω∨ ∘ φ 는 v ↦ *v ∨ 와 같습니다.
Higher adjugates
A 를 n × n 행렬이라고 놓고, r ≥ 0 을 고정합니다. A 의 r 번째 더 높은 수반 (r th higher adjugate )은 adjr A 로 표시되는
(
n
r
)
×
(
n
r
)
{\textstyle {\binom {n}{r}}\!\times \!{\binom {n}{r}}}
행렬이며, 그 엔트리는 {1, ..., m } 의 크기 r 부분집합 I 와 J 에 의해 인덱싱됩니다. I c 와 J c 는 각각 I 와 J 의 여집합 을 나타낸다고 놓습니다. 역시
A
I
c
,
J
c
{\displaystyle \mathbf {A} _{I^{c},J^{c}}}
는 그 인덱스가 각각 I c 와 J c 에 있는 행과 열을 포함하는 A 의 부분행렬을 나타낸다고 놓습니다. 그런-다음 adjr A 의 (I , J ) 엔트리는 다음입니다:
(
−
1
)
σ
(
I
)
+
σ
(
J
)
det
A
J
c
,
I
c
,
{\displaystyle (-1)^{\sigma (I)+\sigma (J)}\det \mathbf {A} _{J^{c},I^{c}},}
여기서 σ(I ) 와 σ(J ) 는 각각 I 와 J 의 원소의 합입니다.
더 높은 수반의 기본 속성은 다음을 포함합니다:
adj0 (A ) = det A .
adj1 (A ) = adj A .
adjn (A ) = 1 .
adjr (BA ) = adjr (A ) adjr (B ) .
adj
r
(
A
)
C
r
(
A
)
=
C
r
(
A
)
adj
r
(
A
)
=
(
det
A
)
I
(
n
r
)
{\displaystyle \operatorname {adj} _{r}(\mathbf {A} )C_{r}(\mathbf {A} )=C_{r}(\mathbf {A} )\operatorname {adj} _{r}(\mathbf {A} )=(\det \mathbf {A} )I_{\binom {n}{r}}}
, where C r (A ) denotes the r th compound matrix .
더 높은 수반은
V
{\displaystyle V}
∧
r
V
{\displaystyle \wedge ^{r}V}
와
∧
n
−
1
V
{\displaystyle \wedge ^{n-1}V}
를 각각
∧
r
V
{\displaystyle \wedge ^{r}V}
와
∧
n
−
r
V
{\displaystyle \wedge ^{n-r}V}
로 대체하여 보통의 수반과 유사한 방식으로 추상적인 대수적 용어로 정의될 수 있습니다.
Iterated adjugates
역가능 행렬 A 의 수반을 k 번 반복적 으로 취하면 다음을 산출합니다:
adj
⋯
adj
⏞
k
(
A
)
=
det
(
A
)
(
n
−
1
)
k
−
(
−
1
)
k
n
A
(
−
1
)
k
,
{\displaystyle \overbrace {\operatorname {adj} \dotsm \operatorname {adj} } ^{k}(\mathbf {A} )=\det(\mathbf {A} )^{\frac {(n-1)^{k}-(-1)^{k}}{n}}\mathbf {A} ^{(-1)^{k}},}
det
(
adj
⋯
adj
⏞
k
(
A
)
)
=
det
(
A
)
(
n
−
1
)
k
.
{\displaystyle \det(\overbrace {\operatorname {adj} \dotsm \operatorname {adj} } ^{k}(\mathbf {A} ))=\det(\mathbf {A} )^{(n-1)^{k}}.}
예를 들어,
adj
(
adj
(
A
)
)
=
det
(
A
)
n
−
2
A
.
{\displaystyle \operatorname {adj} (\operatorname {adj} (\mathbf {A} ))=\det(\mathbf {A} )^{n-2}\mathbf {A} .}
det
(
adj
(
adj
(
A
)
)
)
=
det
(
A
)
(
n
−
1
)
2
.
{\displaystyle \det(\operatorname {adj} (\operatorname {adj} (\mathbf {A} )))=\det(\mathbf {A} )^{(n-1)^{2}}.}
See also
References
Bibliography
Roger A. Horn and Charles R. Johnson (2013), Matrix Analysis , Second Edition. Cambridge University Press, ISBN 978-0-521-54823-6
Roger A. Horn and Charles R. Johnson (1991), Topics in Matrix Analysis . Cambridge University Press, ISBN 978-0-521-46713-1
External links
Matrix Reference Manual
Online matrix calculator (determinant, track, inverse, adjoint, transpose) Compute Adjugate matrix up to order 8
"Adjugate of { { a, b, c }, { d, e, f }, { g, h, i } }" . Wolfram Alpha .{{cite web }}
: CS1 maint: url-status (link )