Professional Documents
Culture Documents
Partial Ordering of Block Matrices in Minkowski Space: Keywords: MSC 2010 No.
Partial Ordering of Block Matrices in Minkowski Space: Keywords: MSC 2010 No.
1,2
Department of Mathematics
Annamalai University
Annamalainagar- 608002
Tamilnadu, India
1 krishna_swamy2004@yahoo.co.in; 2 mpvelmurugan18@gmail.com;
Received: ; Accepted:
Abstract
In this paper, we study the partial orderings of block matrices and the submatrix partial orderings,
we also present the results of star orderings in Minkowski space.
Keywords: Matrix partial orderings; Moore-Penrose inverse; block matrix; Minkowski adjoint;
Minkowski inverse; Minkowski space
MSC 2010 No.: 15A09, 15A45, 15A57
1. Introduction
Throughout this paper, Let us denote the set of complex matrices as Cm×n and Cn
represent complex n-tuples. The components of this complex vector in Cn is represented
as u = (u0 , u1 , u2 , ..., un−1 ). Let G be the Minkowski metric tensor defined by Gu =
(u0 , −u1 , −u2 , ..., −un−1 ). Clearly the Minkowski metric matrix is given by
!
1 0
G= (1)
0 −In−1
G = G∗ and G2 = In . In[Renardy (1996)], defined Minkowski inner product on Cn by (u, v) =
[u, Gv], where [., .] denotes the conventional Hilbert space inner product. M denotes the Minkowski
space, which is a space with Minkowski inner product.
In 2000 Meenakshi [Meenakshi (2000)] presented the concept of Minkowski inverse of a matrix
represented as A ∈ Cm×n . Also presented a unique solution to the following four matrix equations:
AXA = A, XAX = X, (AX)∼ = AX, (XA)∼ = XA (2)
where A denotes the Minkowski adjoint of the matrix A in M.
∼
However, the Minkowski inverse of a matrix does not exists always as in Moore-Penrose inverse
of a matrix. The proved that the Minkowski inverse of a matrix A ∈ Cm×n exists if and only if
1
2 D. Krishnaswamy and M. velmurugan
In this section, we present the results on the star partial orderings in Minkowski space.
Theorem 2.1.
∼ ∼
Let P1 , R1 ∈ Cm×n and Q1 , S1 ∈ Cm×k be star-ordered as P1 ≤ R1 , Q1 ≤ S1 . If R(P1 ) = R(Q1 ),
∼
then G P1 Q1 ≤ G R1 S1 .
Proof.
∼ ∼
On account of (3) and (4), since P1 ≤ R1 , Q1 ≤ S1 and R(P1 ) = R(Q1 ), so
∼
(i)P1 ≤ R1 ⇔ (P1 G)∼ GP1 = (P1 G)∼ GR1 and P1 G(GP1 )∼ = R1 G(GP1 )∼
∼
(ii)Q1 ≤ S1 ⇔ (Q1 G)∼ GQ1 = (Q1 G)∼ GS1 and Q1 G(GQ1 )∼ = S1 G(GQ1 )∼
P1 G(GP1 )∼ = R1 G(GP1 )∼
P1 G = R1 G(GP1 )∼ ((GP1 )∼ )
m
= R1 G(GP1 )∼ ((GP1 )
m )∼
P1 G = R1 G((GP1 )
GP1 )∼
m (9)
Q1 G(GQ1 )∼ = S1 G(GQ1 )∼
Q1 G = S1 G(GQ1 )∼ ((GQ1 )∼ )
m
= S1 G(GQ1 )∼ ((GQ1 )
)∼
m
Q1 G = S1 G((GQ1 )
m (GQ1 ))∼ . (10)
Consider,! !
GP1∼ GP1∼ GP1 GP1∼ GQ1
GP1 GQ1 =
GQ∼1 GQ∼1 GP1 GQ∼
1 GQ1
!
∼ (P1 G)∼ GP1 (P1 G)∼ GQ1
G P1 Q1 G P1 Q1 =
(Q1 G)∼ GP1 (Q1 G)∼ GQ1
(P1 G)∼ ((Q1 G)(Q1 G)
m )∼ GS1
!
(P1 G)∼ GR1
= [Using (7) and (8)]
(Q1 G)∼ ((P1 G)(P1 G)
m )∼ GR1 (Q1 G)∼ GS1
((Q1 G)(Q1 G)
m P1 G)∼ GS1
!
(P1 G)∼ GR1
=
((P1 G)(P1 G)
m Q1 G)∼ GR1 (Q1 G)∼ GS1
!
(P1 G)∼ GR1 (P1 G)∼ GS1
=
(Q1 G)∼ GR1 (Q1 G)∼ GS1
!
(P1 G)∼
= GR 1 GS 1
(Q1 G)∼
!
GP1∼
= GR 1 GS 1
GQ∼ 1
!
P1∼
=G G R 1 S 1
Q∼
1 ∼
= G P1 Q1 G R1 S1 (11)
Consider, ! !
P ∼G P GP ∼G P GQ ∼G
1 1 1 1 1
P1 G Q1 G =
Q∼1 G Q 1 GP1
∼ G Q GQ∼ G
1 1
!
∼ P1 G(GP1 ) ∼ P1 G(GQ1 )∼
P1 Q1 G P1 Q1 G =
Q1 G(GP1 )∼ Q1 G(GQ1 )∼
R1 G((GP1 )
m (GP1 ))∼ (GQ1 )∼
!
R1 G(GP1 )∼
= [Using (9) and (10)]
S1 G((GQ1 )
m (GQ1 ))∼ (GP1 )∼ S1 G(GQ1 )∼
4 D. Krishnaswamy and M. velmurugan
R1 G((GQ1 )(GP1 )
m (GP1 ))∼
!
R1 G(GP1 )∼
=
S1 G((GP1 )(GQ1 )
m (GQ1 ))∼ S1 G(GQ1 )∼
!
R1 G(GP1 )∼ R1 G(GQ1 )∼
=
S1 G(GP1 )∼ S1 G(GQ1 )∼
!
(GP )∼
1
= R1 G S1 G
(GQ1 )∼
!
P ∼G
1
= R1 G S1 G
Q∼ 1G
!
P1∼
= R1 S1 G G
Q∼1
∼
= R1 S1 G P1 Q1 G
pre and post multiplying
by G ∼
= G R1 S1 G P1 Q1 (12)
From (11) and
(12),we have
∼
Therefore G P1 Q1 ≤ G R1 S1 .
Hence proved.
Theorem 2.2.
Let P1 , R1 ∈ Cm×n and Q1 , S1 ∈ Cm×k be star-ordered as P1 ∼ ≤ R1 , Q1 ∼ ≤ S1 . If R(P1 ) = R(Q1 ),
then G(P1 Q1 )∼ ≤ G(R1 S1 ).
Proof.
i)P1 ∼ ≤ Q1 ⇔ (P1 G)∼ GP1 = (P1 G)∼ GQ1 (or(GP1 )
m )GP1 = (GP1 )
m GQ1 and
R(P1 ) ⊆ R(Q1 ).
(ii)P1 ∼ ≤ R1 ⇔ (P1 G)∼ GP1 = (P1 G)∼ GR1 (or(GP1 )
m )GP1 = (GP1 )
m GR1 and
R(P1 ) ⊆ R(R1 ).
(iii)Q1 ∼ ≤ S1 ⇔ (Q1 G)∼ GQ1 = (Q1 G)∼ GS1 (or(GQ1 )
m )GQ1 = (GQ1 )
m GS1 and
R(Q1 ) ⊆ R(S1 ).
Consider,! !
GP1∼ GP1∼ GP1 GP1∼ GQ1
GP1 GQ 1 =
GQ∼1 GQ∼1 GP1 GQ∼
1 GQ1
!
∼ (P1 G)∼ GP1 (P1 G)∼ GQ1
G P1 Q1 G P1 Q1 =
(Q1 G)∼ GP1 (Q1 G)∼ GQ1
(P1 G)∼ ((Q1 G)(Q1 G)
m )∼ GS1
!
(P1 G)∼ GR1
= [Using (7) and (8)]
(Q1 G)∼ ((P1 G)(P1 G)
m )∼ GR1 (Q1 G)∼ GS1
((Q1 G)(Q1 G)
m P1 G)∼ GS1
!
(P1 G)∼ GR1
=
((P1 G)(P1 G)
m Q1 G)∼ GR1 (Q1 G)∼ GS1
!
(P1 G)∼ GR1 (P1 G)∼ GS1
=
(Q1 G)∼ GR1 (Q1 G)∼ GS1
AAM: Intern. J., Vol. , Issue 5
!
(P1 G)∼
= GR 1 GS 1
(Q1 G)∼
!
GP1∼
= GR 1 GS 1
GQ∼ 1
!
P1∼
=G G R S
1 1
Q∼
1 ∼
= G P1 Q1 G R1 S1
On the otherhand, on account of (5), from the conditions P1 ∼ ≤ R1 and Q1 ∼ ≤ S1 , we have
R(P1 ) ⊆ R(R1 ) and R(Q1 ) ⊆ R(S1 ), which imply that R(P1 Q1 ) ⊆ R(R1 S1 ). According to (5), we
have G(P1 Q1 )∼ ≤ G(R1 S1 ).
Theorem 2.3.
∼ ∼
Let P1 , R1 ∈ Cm×n and Q1 , S1 ∈ Cm×k be star-ordered as G(P1 Q1 ) ≤ G(R1 S1 ). If P1 ≤ R1 (or
∼ ∼ ∼ ∼ ∼
Q1 ≤ S1 ), then Q1 ≤ S1 (or P1 ≤ R1 ). Moreover, the condition P1 ≤ R1 (or Q1 ≤ S1 ) can be replaced
by P1 ≤ ∼R1 (or Q1 ≤ ∼S1 ).
Proof.
Proof of Theorem 2.3 follows from Theorem 2.1.
Corollary 2.4.
∼ ∼
Let P1 , R1 ∈ Cm×n and!Q1 , S1 ∈ C!k×n be star-ordered as P1 ≤ R1 , Q1 ≤ S1 . If R((P1 G)∼ ) =
P1 ∼ R1
R((Q1 G)∼ ), then G ≤G .
Q1 S1
Proof.
It is an immediate consequence of proof of theorem 2.1.
Corollary 2.5.
Let P1 , R1 ∈ Cm×n and Q , S ∈ Ck×n be star-ordered as P1 ≤ ∼R1 , Q1 ≤ ∼S1 . If R((P1 G)∼ ) =
!1 1 !
P1 R1
R((Q1 G)∼ ), then G ≤ ∼G .
Q1 S1
Proof.
Given,
(i)P1 ≤ ∼R1 ⇔ (P1 G)(GP1 )∼ = R1 G(GP1 )∼ (or P1 G(P1 G)
m = Q1 G(P1 G)
m ) and
R((P1 G) ) ⊆ R((R1 G) ).
∼ ∼
! !∼ !
P1 P1 P1 G(GP1 )∼ P1 G(GQ1 )∼
G G=
Q1 Q1 Q1 G(GP1 )∼ Q1 G(GQ1 )∼
R1 G((GP1 )
m (GP1 ))∼ (GQ1 )∼
!
R1 G(GP1 )∼
= [Using (9) and (10)]
S1 G((GQ1 )
m (GQ1 ))∼ (GP1 )∼ S1 G(GQ1 )∼
R1 G((GQ1 )(GP1 )
m (GP1 ))∼
!
R1 G(GP1 )∼
=
S1 G((GP1 )(GQ1 )
m (GQ1 ))∼ S1 G(GQ1 )∼
!
R1 G(GP1 )∼ R1 G(GQ1 )∼
=
S1 G(GP1 )∼ S1 G(GQ1 )∼
!
R1 G
= (GP1 )∼ (GQ1 )∼
S1 G
!
R1 G ∼
= P1 G Q∼1G
S1 G
! !∼
R1 P1
= G G
S1 Q1
pre and post multiplying
! by
!∼G
R1 P1
=G G .
S1 Q1
On the otherhand, on account of equation (6), from the conditions P1 ≤ ∼R1 and Q1 ≤ ∼S ! 1,
P1
we have R((P1 G)∼ ) ⊆ R((R1 G)∼ ) and R((Q1 G)∼ ) ⊆ R((S1 G)∼ ), which imply that R ⊆
Q1
!
R1
R .
S1
! !
P1 R1
According to equation (6), we have G ≤ ∼G .
Q1 S1
Hence the proof.
Corollary 2.6. ! !
P1 ∼ R1
Let P1 , R1 ∈ Cm×n and Q1 , S1 ∈ Ck×n be star-ordered as G ≤ G . If P1 ∼ ≤ R1 (or
Q1 S1
∼ ∼
Q1 ∼ ≤ S1 ), then Q1 ≤ S1 (or P1 ≤ R1 ).
Proof.
The proof follows from Theorem 2.3.
Theorem 2.7.
Let P1 , Q1 ∈ Cm×n , R1 ∈ Cm×k and S1 ∈ Ck×n . Then
∼ ∼ ∼
(1)If P1 ≤ Q1 and R(R1 ) ⊆ R(P1 ), then G(P1 R1 ) ≤ G(Q1 R1 ) and G(R1 P1 ) ≤ G(R1 Q1 ).
∼ ∼ ∼
Moreover, both G(P1 R1 ) ≤ G(Q1 R1 ) and G(R1 P1 ) ≤ G(R1 Q1 ) imply P1 ≤ Q1 , even though
R(R1 ) 6⊂ R(P1 ).
(2)P1 ∼ ≤ Q1 and R(R1 ) ⊆ R(P1 ), then G(P1 R1 )∼ ≤ G(Q1 R1 ) and (R1 P1 )∼ ≤ (R1 Q1 ).
AAM: Intern. J., Vol. , Issue 7
! ! ! !
∼ P1 ∼ Q1 S1 ∼ S1
(3)If P1 ≤ Q1 and R((S1 G)∼ )
⊆ R((P1 G)∼ ),
then G ≤G and G ≤G .
S1 S1 P1 Q1
! ! ! !
P1 ∼ Q1 S1 ∼ S1 ∼
Moreover, both G ≤ G and G ≤ G imply P1 ≤ Q1 , even though
S1 S1 P1 Q1
R((S1 G)∼ ) 6⊂ R((P1 G)∼ ). ! !
P Q
(4)If P1 ≤ ∼Q1 and R((S1 G)∼ ) ⊆ R((P1 G)∼ ), then G 1 ≤ ∼G 1
and
S1 S1
! !
S1 S1
G ≤ ∼G .
P1 Q1
Proof.
(1) Given,
∼
P1 ≤ Q1 ⇔ (P1 G)∼ GP1 = (P1 G)∼ GQ1 and P1 G(GP1 )∼ = Q1 G(GP1 )∼ .
P1 G(GP1 )∼ = Q1 G(GP1 )∼
P1 G = Q1 G(GP1 )∼ ((GP1 )∼ )
m
m
= Q1 G(GP1 ) ((GP1 ) )∼
∼
P1 G = Q1 G((GP1 )
m (GP1 ))∼ . (14)
Consider,! !
GP1∼ GP1∼ GP1 GP1∼ GR1
GP1 GR1 =
GR1∼ GR1∼ GP1 GR1∼ GR1
!
∼ (P1 G)∼ GP1 (P1 G)∼ GR1
G P1 R1 G P1 R1 =
(R1 G)∼ GP1 (CG)∼ GR1
!
(P1 G)∼ GQ1 (P1 G)∼ GR1
= (Using (13))
(R1 G)∼ ((P1 G)(P1 G)
m )∼ GQ1 (R1 G)∼ GR1
!
(P1 G)∼ GQ1 (P1 G)∼ GR1
=
((P1 G)(P1 G)
m (R1 G))∼ GQ1 (R1 G)∼ GR1
!
(P1 G)∼ GQ1 (P1 G)∼ GR1
=
(R1 G)∼ GQ1 (R1 G)∼ GR1
!
(P1 G)∼
= GQ1 GR1
(R1 G) ∼
!
GP1∼
= GQ 1 GR 1
GQ∼ 1
8 D. Krishnaswamy and M. velmurugan
∼
= G P1 R1 G Q1 R1 (15)
Consider, ! !
P ∼G P1 GP1∼ G P1 GR1∼ G
1
P1 G R1 G =
R1∼ G R1 GP1∼ G R1 GR1∼ G
!
∼ P1 G(GP1 )∼ P1 G(GR1 )∼
P1 R1 G P1 R1 G =
R1 G(GP1 )∼ R1 G(GR1 )∼
Q1 G((GP1 )
m GP1 )∼ (GR1 )∼
!
Q1 G(GP1 )∼
= [Using (14)]
R1 G(GP1 )∼ R1 G(GR1 )∼
Q1 G((GR1 )(GP1 )
m (GP1 ))∼
!
Q1 G(GP1 )∼
=
R1 G(GP1 )∼ R1 G(GR1 )∼
!
Q1 G(GP1 )∼ Q1 G(GR1 )∼
=
R1 G(GP1 )∼ R1 G(GR1 )∼
!
(GP )∼
1
= Q 1 G R1 G
(GR1 )∼
!
P ∼G
1
= Q 1 G R1 G
R1∼ G
∼
= Q1 R1 G P1 R1 G
pre and post
multiplying
∼ by G ∼
G P1 R1 G P1 R1 = G Q1 R1 G P1 R1 . (16)
From
equations
∼ (15) and
(16), we have
G P1 R1 ≤ G Q1 R1 . (17)
Similarly,
∼
G R1 P1 ≤ G R1 Q1 (18)
∼
Combining (17) and (18) implies that P1 ≤ Q1 , even though R(R1 ) 6⊂ R(P1 ).
Hence proved.
(2) Given, P1 ∼ ≤ Q1 and R(R1 ) ⊆ R(P1 ).
P1 ∼ ≤ Q1 ⇔ (P1G)∼ GP
m )GP1 = (GP1 )
m GQ1 and R(P1 ) ⊆ R(Q1 ).
1 = (P1 G) 1 (or(GP1 )
∼
GQ
To prove that G P1 R1 ∼ ≤ G Q1 R1 :
Consider,! !
GP1∼ GP1∼ GP1 GP1∼ GR1
GP1 GR1 =
GR1∼ GR1∼ GP1 GR1∼ GR1
!
∼ (P1 G)∼ GP1 (P1 G)∼ GR1
G P1 R1 G P1 R1 =
(R1 G)∼ GP1 (R1 G)∼ GR1
!
(P1 G)∼ GQ1 (P1 G)∼ GR1
= [Using (13)]
(R1 G)∼ ((P1 G)(P1 G)
m )∼ GQ1 (R1 G)∼ GR1
!
(P1 G)∼ GQ1 (P1 G)∼ GR1
=
((P1 G)(P1 G)
m (R1 G))∼ GQ1 (R1 G)∼ GR1
AAM: Intern. J., Vol. , Issue 9
!
(P1 G)∼ GQ1 (P1 G)∼ GR1
=
(R1 G)∼ GQ1 (R1 G)∼ GR1
!
(P1 G)∼
= GQ 1 GR 1
(R1 G)∼
!
GP1∼
= GQ 1 GR 1
GQ∼
1 ∼
= G P1 R1 G Q1 R1
On the otherhand, on account
of equation
(5), from
the conditions P1 ∼ ≤ Q1 , we have R(P1 ) ⊆
R(Q1 ) which imply that R P1 R1 ⊆ R Q1 R1 .
According to equation (5), we have G P1 R1 ∼ ≤ G Q1 R1 .
Similarly,
G R1 P1 ∼ ≤ G R1 Q1 .
Hence proved.
∼
! R((S1 G)! ) ⊆ R((P1 G) ).
(3) Given, P1 ≤ Q1 and ∼ ∼
P1 ∼ Q1
To prove that G ≤G .
S1 S1
∼
P1 ≤ Q1 ⇔ (P1 G)∼ GP1 = (P1 G)∼ GQ1 and P1 G(GP1 )∼ = Q1 G(GP1 )∼ .
Now consider, ! !
GP
1 GP1∼ GP1 GP1∼ GS1
GP1∼ GS1∼ =
GS1 GS1∼ GP1 GS1∼ GS1
!∼ ! !
P1 P1 (P1 G)∼ GP1 (P1 G)∼ GS1
G G =
S1 S1 (S1 G)∼ GP1 (S1 G)∼ GS1
!
(P1 G)∼ GQ1 (P1 G)∼ GS1
= [Using (13)]
(S1 G) ((P1 G)(P1 G)
∼ m )∼ GQ1 (S1 G)∼ GS1
!
(P1 G)∼ GQ1 (P1 G)∼ GS1
=
((P1 G)(P1 G)
m (S1 G))∼ GQ1 (S1 G)∼ GS1
!
(P1 G)∼ GQ1 (P1 G)∼ GS1
=
(S1 G)∼ GQ1 (S1 G)∼ GS1
!
GQ
1
= (P1 G)∼ (S1 G)∼
GS1
!
GQ
1
= GP1∼ GQ∼ 1
GS1
!∼ !
P1 Q1
=G G (19)
S1 S1
Now consider,
! !
P1 G ∼ P1 GP1∼ G P1 GS1∼ G
P1 G S1∼ G =
S1 G S1 GP1∼ G S1 GS1∼ G
10 D. Krishnaswamy and M. velmurugan
! !∼ !
P1 P1 P1 G(GP1 )∼ P1 G(GS1 )∼
G G=
S1 S1 S1 G(GP1 )∼ S1 G(GS1 )∼
Q1 G((GP1 )
m GP1 )∼ (GS1 )∼
!
Q1 G(GP1 )∼
= [Using (14)]
S1 G(GP1 )∼ S1 G(GS1 )∼
Q1 G((GS1 )(GP1 )
m (GP1 ))∼
!
Q1 G(GP1 )∼
=
S1 G(GP1 )∼ S1 G(GS1 )∼
!
Q1 G(GP1 )∼ Q1 G(GS1 )∼
=
S1 G(GP1 )∼ S1 G(GS1 )∼
!
Q1 G
= (GP1 )∼ (GS1 )∼
S1 G
!
Q1 G ∼
= P1 G S1∼ G
S1 G
! !∼
Q1 P1
= G G
S1 S1
!
Q1 G
= (GP1 )∼ (GS1 )∼
S1 G
!
Q1 G ∼
= P1 G S1∼ G
S1 G
! !∼
Q1 P1
= G G
S1 S1
pre and post!multiplying
!∼ by G ! !∼
P1 P1 Q1 P1
G G =G G .
S1 S1 S1 S1
! (6), from!the condition P1 ≤ ∼Q1 , we have R((P1 G) ) ⊆
On the otherhand, on account of equation ∼
P1 Q1
R((Q1 G)∼ ) which imply that R1 ⊆ R1 .
S1 S1
! !
P1 Q1
According to equation (6), we have G ≤ ∼G .
S1 S1
Similarly,
! !
S1 S1
G ≤ ∼G .
P1 Q1
Hence the proof.
3. Conclusion
In this paper, we have concluded the algebraic structure of the star partial ordering, left and right
star partial ordering of block matrices in Minkowski space.
REFERENCES
Baksalary, J.K., Hauke, J. and Styan, G.P.H. (1994). On Some Distributional Properties of
Quadratic Forms in Normal variables and on Some Associated Matrix Partial orderings, Mul-
tivariate Analysis and Its Applications, Vol. 24, pp. 111-121.
Bayram Bala, Manaf Dzh. Manafov and Abdullah Kablan. (2019). Inverse spectral problems
for spectral Data and Two Spectra of N by N Tridiagonal Almost-Symmetric matrices,
Appl.Appl.math, Vol. 14, No.2, pp. 1132-1144.
Ben-Israel, A., Greville, T.N.E. (2003). Generalized Inverses: Theory and Applications, Springer,
New York.
Baksalary, J.K. and Puntanen, S. (1990). Characterizations of the Best Linear Unbiased Estimator
in the General Gauss Markov Model with the Use of Matrix Partial Orderings, Linear Algebra
and its Applications, Vol. 127, pp. 363-370.
Benitez, J., Liu, X., Zhong, J. (2010). Some results on matrix partial orderings and reverse order
law, Electronic J. Linear Algebra, Vol. 20, pp. 254-273.
12 D. Krishnaswamy and M. velmurugan
Drazin, M.P. (1978). Natural structures on semigroups with involution, Bull. Am. Math. Soc., Vol.
84, pp. 139-141.
Hartwig, R.E. (1980). How to partially order regular elements, Math. Jpn., Vol. 25, pp. 1-13.
Kavyasree, P.R. and Surender Reddy, B.(2019). Some results of double sequences in 2-Normed
and n-Normed spaces, Appl.Appl.math, Special issue No.4, pp. 97-105.
Meenakshi, A.R. (2000). Generalized Inverses of Matrices in Minkowski Space, Proceedings of
National Seminar on Algebra and Its Applications, Vol. 1, pp. 1-14 .
Meenakshi, A.R. (2000). Range symmetric Matrices in Minkowski Space, Bulletin of the
Malaysian Mathematical Sciences Society, Vol. 23, pp. 45-52 .
Mitra, S.K. (1987). On Groups Inverses and the Sharp Order, Linear Algebra and its Applications,
Vol. 92, pp. 17-37.
Puntanen, S. and Styan, G.P.H. (2015). Best Linear Unbiased Estimation in Linear Models, Stat-
Prob:The Encyclopedia Sponsored by Statistics and Probability Societies.
Renardy, M. (1996). Singular value decomposition in Minkowski space, Linear Algebra and its
Applications, Vol. 236, pp. 53-58.
Stepniak, C. (1987). Ordering of Nonnagative definite Matrices with Application to Comparison
of Linear Models, Linear Algebra and its Applications, Vol. 70, pp. 67-71.