International Journal of Engineering Research & Science (IJOER)

ISSN: [2395-6992]

[Vol-2, Issue-5 May- 2016]

Study on Some Properties of Anti-centrosymmetric Matrices

Wenhui Lan1, Junqing Wang2
School of Science, Tianjin Polytechnic University, Tianjin, 300387, China

Abstract In this paper, the anti-centrosymmetric matrices have been researched. According to the structural
characteristics of the anti-centrosymmetric matrix, some new methods have been used to prove the necessary and sufficient
conditions of a matrix being anti-centrosymmetric and its properties of eigenvalue and eigenvector; the nonsingularity of the
anti-centrosymmetric matrices have been discussed,that the odd order anti-centrosymmetric matrix is singular has been
obtained, and two methods of computing inverse of the matrices(even order) have been given.
Keywords Anti-centrosymmetric Matrix, Centrosymmetric Matrix, Eigenvalue, Inverse Matrix.

I.

INTRODUCTION

Centrosymmetric and anti-centrosymmetric matrices are two kinds of important special matrices, which are widely used in
the fields of information theory, numerical analysis, linear system theory and so on. At present, some significant
achievements have been acquired from the research on the structure and properties, eigenvalues and inverse of
centrosymmetric and anti-centrosymmetric matrices. For example, in reference[3], Tan Ruimei used the definition of anticentrosymmetric matrix to prove some new conclusions on adjoint matrix, eigenvalue and eigenvector of it; in reference[6],
Liu Lianfu discussed the method of computing inverse of anticentrosymmetric matrices in the light of the structure and
representation of it. In this paper, based on the previous literatures, some systematic summary research has been made and
some new proof methods have been shown for anti-centrosymmetric matrices.
Definition 1. If A ( aij ) C

nn

Definition 2. If A ( aij ) C

nn

an1i,n1 j , i, j 1, 2,, n ,then A is called centrosymmetric matrix.

For centrosymmetric matrix, the elements in line r from the end are precisely those in line r from the beginning which are
nn
arranged in the reverse order. The set of the whole n-order centrosymmetric matrices is denoted as CSR .
and it meets aij

an1i ,n1 j , i, j 1, 2,, n ,then A is called anti-centrosymmetric

matrix. For anti-centrosymmetric matrix, the elements in line r from the end are precisely the opposite number of those in
line r from the beginning which are arranged in the reverse order. The set of the whole n-order anti-centrosymmetric matrices
nn
is denoted as ACSR .
and it meets aij

(en , , e1 ) is called n-order permutation matrix. Obviously,

Definition 3. The matrix J n

Jn Jn In .

a11 a12

a21 a22
Definition 4. If A

am1 am 2

a1n
amn

an 1,n
a2 n

, then A

amn
a1n

am,n 1
an 1,n 1

a1,n 1

am1

an 1,1

a11

is denoted as the inversion of matrix A .

Lemma 1.

A ACSRnn if and only if A J n AJ n .

Lemma2. If A ACSR

2 n2 n

,then A can be expressed as

Page | 15

International Journal of Engineering Research & Science (IJOER)

ISSN: [2395-6992]

A
A 1
J n A2
and if A ACSR

(2 n1)(2 n1)

[Vol-2, Issue-5 May- 2016]

A2 J n

J n A1 J n

, then A can be expressed as

A1

A= T
J A
n 2

0
J n

A2 J n

T J n
J n A1 J n

among which, A1 , A2 are n-order square matrices; J n is n-order permutation matrix; , are n dimensional column vectors.
Lemma 3. A ACSR

nn

Lemma 4. A ACSR

nn

; A CSR
if and only if A A

nn

as well as A ACSR

nn

if and only if A O .

if and only if

);
(1)When n 2k , k Z , A ( A1 , A
1
*

).
(2))When n 2k 1, k Z , A ( A1 , , A
1
*

among which, is the column vector of A in line k 1 , A1 is constituted by the front k column elements of A with the

is the inversion of A .
original order and A
1
1
Proof: When n 2k 1 , let A ACSR

nn

( A1 , , A2 ) .From Lemma 3,

A ( A1 , , A2 ) A ( A2 , , A1 )
and from Definition 2 and Definition 4,

aij a2 k 2i ,2 k 2 j and , then

a11 a2 k 1,2 k 1 , , a1k a2 k 1,k 2 ,

a1,k 2 a2 k 1,k , , a1n a2 k 1,1

a21 a2 k ,2 k 1 , , a2 k a2 k ,k 2 ,

a2,k 2 a2 k ,k , , a2 n a2 k ,1

a2 k ,1 a2 n ,

a2 k ,k = a2,k 2 ,

a2 k 1,1 a1n , , a2 k 1,k a1,k 2

a2 k ,k 2

a2 k , , a2 k ,2 k 1 a21

a2 k 1,k 2 a1k , , a2 k 1,2 k 1 a11

, therefore, A ( A , , A ) .
That is A1 A
2
1
1
Similarly, the conclusion(1) can be proved.

II.
Theorem 1. If A ACSR

nn

MAIN CONCLUSIONS

is the eigenvector of A and 0 is the corresponding eigenvalue, then 0 is also one of the

eigenvalues which belong to A , besides,

J n is the eigenvector which is corresponding to 0 .

Proof: From what is known,

0 I A 0
Because of Lemma 1,

J n (0 I ) J n J n AJ n 0
Page | 16

International Journal of Engineering Research & Science (IJOER)

ISSN: [2395-6992]

[Vol-2, Issue-5 May- 2016]

and that is

J n 0 I A J n 0 , 0 I A 0
Then

( ) 0 I A 0
and 0 is the eigenvalue of A .
Since

A 0 , then A 0 , and according to Lemma 1,

J n AJ n 0

Premultiplying both sides of the equation yields

J n J n AJ n (0 ) J n
From Definition 3,

A( J n ) ( 0 ) J n
which shows that J n is the eigenvector which is corresponding to 0 .
Theorem 2. If A, B ACSR

nn

, then kA ACSR nn , (k C ) and A B ACSRnn , thus all of the n-order anti-

centrosymmetric matrices on the complex field constitute a linear subspace of C

n n

Theorem 3. The trace of a n-order anti-centrosymmetric matrix is 0, and therefore the sum of the matrix's eigenvalues is 0.
Proof: In terms of Definition 2,

aij an1i ,n1 j , i, j 1, 2,, n

For odd order anti-centrosymmetric matrices,the center of its elements must be 0. Then the sum of the elements on the
diagonal is

a11 a22 ann 0

Because of all of the above,the trace of a n-order anti-centrosymmetric matrix is 0, and the sum of the eigenvalues is 0, too.
Theorem 4. If A is an n-order square matrix, then there exist the only B CSR

nn

and C ACSR

nn

, which can

indicate A B C .
Proof: Let

B
then B CSR

nn

and C ACSR

nn

A A
A A
,C
,
2
2

, the uniqueness will be proved in the following.

Suppose

A B1 C1 B2 C2
among which

B1 , B2 CSR nn and C1 , C2 ACSRnn ,

then
Page | 17

International Journal of Engineering Research & Science (IJOER)

ISSN: [2395-6992]

[Vol-2, Issue-5 May- 2016]

B1 B2 C1 C2
However,

B1 B2 CSR nn and C1 C2 ACSR nn , on the basis of Lemma 3,

B1 B2 C1 C2 O
so

B1 B2 , C1 C2
Theorem 5. The product of a anti-centrosymmetric matrix and a same order centrosymmetric matrix is a anticentrosymmetric matrix; the product of two anti-centrosymmetric matrices of the same order is a centrosymmetric matrix.
Proof: (i)If A ACSR

nn

and B CSR

nn

,then

A J n AJ n and B J n BJ n
hence

AB J n AJ n J n BJ n J n ( AB ) J n
which means that AB ACSR
(iiIf A, B ACSR

nn

nn

,then

A J n AJ n and B J n BJ n
hence

AB J n AJ n J n BJ n J n ( AB) J n
which means that

AB CSRnn .

Theorem 6. Odd order anti-centrosymmetric matrices are singular.

Proof: Suppose A ACSR

(2 n1)(2 n1)

and 1 , 2 , , n , n 1 is the n 1 eigenvalues of it.According to Theorem 3,

1 2 n n 1 0
From Theorem 1, there are n pairs eigenvalues which is opposite with each other in 1 , 2 , , n , n 1 , and there must be a
eigenvalue i

0 left. Then 0 must be a eigenvalue of A, which implicates that det A 0. Consequently, odd order anti-

centrosymmetric matrices are singular.

In this paper, the nonsingular anti-centrosymmetric matrices can only be considered in the field of even order matrices.
Theorem 7. The converse of a nonsingular anti-centrosymmetric matrix is anti-centrosymmetric.
1

Proof: From what is given, A J n AJ n and AA

I n , then

J n AA1J n J n AJ n J n A1J n A(J n A1J n ) I n

1

Above with the uniqueness of A , A

J n A1 J n can be deduced, and that is A1 is also anti-centrosymmetric.

Page | 18

International Journal of Engineering Research & Science (IJOER)

Theorem 8. If the nonsingular matrix A ACSR

nn

ISSN: [2395-6992]

[Vol-2, Issue-5 May- 2016]

and B ( A Ek ) , among which n 2k , k Z and Ek is the

*

first k columns of n-order identity matrix. Using elementary transformation to B , if A becomes identity matrix and
1

meanwhile Ek becomes matrix C , then A

(C , C ) .

Proof: In view of Theorem 7, if A ACSR

nn

and A is nonsingular, then A

-1

which A1 is the matrix which is constituted by the first k columns of A , and

ACSRnn . Let A1 ( A1 , A2 ) , among

A2 is the matrix which is constituted by the

end k columns of A . The same block-divided method can be used to I n , and let A( A1 , A2 ) ( Ek , G) .
Now,

( AA1 , AA2 ) ( Ek , G)
so

AA1 Ek
If A becomes I n by elementary transformation, then there exist elementary matrices P1 , P2 ,, Pm ,which can infer

PP
1 2 Pm A I n .
Consequently,
1
A1 PP
1 2 Pm and A1 A Ek PP
1 2 Pm Ek .

From all of the above, if elementary transformation is used to B ( A Ek ) , and A

1

becomes the identity matrix, then Ek becomes the first k columns of A . Finally, according to Lemma 4(2), the conclusion
can be obtained.
Theorem 9. If A ACSR

2 n2 n

A1
J n A2

and that is A

A2 J n
, then A is nonsingular if and only if
J n A1 J n

B A1 A2 and C A1 A2 are all nonsingular, besides,

1
1
(C 1 B 1 ) J n
1 B C
A

2 J n (C 1 B 1 ) J n ( B 1 C 1 ) J n
1

among which

A1 , A2 are n-order square matrices, and J n is n-order permutation matrix.

Proof: Let U

1 In

2 Jn

In
, then
Jn
O C A1
U T AU

B O J n A2

and A B C . Therefore,

A2 J n
A
J n AVJ
1
n

A is nonsingular if and only if B and C are nonsingular.

Moreover,

Page | 19

International Journal of Engineering Research & Science (IJOER)

A1

J n A2

ISSN: [2395-6992]

[Vol-2, Issue-5 May- 2016]

A2 J n 1 B 1 C 1
(C 1 B 1 ) J n I n

J n A1 J n 2 J n (C 1 B 1 ) J n ( B 1 C 1 ) J n O

In

so
1
1
(C 1 B 1 ) J n
1 B C
A1

2 J n (C 1 B 1 ) J n ( B 1 C 1 ) J n

3 7 3 1

2 3 1 1
1

Example: If A
, then A ?
1 1 3 2

1 3 7 3
Solve: 1. According to Theorem 8,

3 7

2 3
( A E )
1 1

1 3
1 3

0 1
0 3

0 2

3 1
1 1
3 2
7 3
7 3
5 5 2
15
7
18
8

1
0
0
0

0
0
0
1

0 1

1 1

0 2

0 3
0 1

0 0

1 0

0 0

3 7 3
1 3 2
3 1 1
7 3 1
0
1
0
0

0
0
1
0

0
0
0
1

0
1
3

5 8 5 4
(I , C)
18 3 4

0
2

0
0
0
1

3
2
0
1

5 8 5 4 3 4 1 8
1

then A (C , C )
1 8 3 4
5 4 5 8

2
3
1
0
2. According to Theorem 9,

3 7
1 3
3 1
A1
, A2
, A2 J

2 3
1 1
1 1
then

2 4
4 10
A1 A2
B , A1 A2
C
3 4
1 2
and

1 1 1 5
1
B1
,C

3 4 1 2
1 2 2
so

Page | 20

International Journal of Engineering Research & Science (IJOER)

ISSN: [2395-6992]

[Vol-2, Issue-5 May- 2016]

3
2
0
1

(C B ) J 5 8 5 4 3 4 1 8
1 B C
A1

5 4 5 8
2 J (C 1 B 1 ) J ( B 1 C 1 ) J 1 8 3 4

2
3
1
0
1

REFERENCES
[1] Xianfeng Wu, Xiaolin Zhang. Some Properties of Centre Skew Symmetric Matrix[J]. Journal of Science of Teachers' College and
University: 2011,3163-5
[2] Fang Chen. A Simple Method of the Calculation of Centro-symmetric Matrices and Anti Centro-symmetric Matrices Inversion in
Engineering[J]. Journal of Sichuan University of Science & Engineering(Natural Science Edition): 2006,19136-38
[3] Ruimei Tan, Xueli Xue. Probe on Some Properties of Centroskewsymmetric Matrices[J]. College Mathematics: 2010, 264173175
[4] Yongan Zhou, Tao Qin. Some Properties on Centrosymmetric Matrices[J]. Mathematics in Practice and Theory: 2007,
378186-188
[5] Bowei Tang. Reverse Order Law of Generalized Inverse and the Properties of Centrosymmetric Matrix[D]. Master Degree Thesis of
Shanxi Normal University:2013,23-26
[6] Lianfu Liu, Changbo Zheng. Reversibility of Center Symmetric Matrices[J]. Journal of Shenyang Normal University(Natural
Science): 2012,30123-26

Page | 21