UITS Journal of Science & Engineering  Volume: 7, Issue: 1

ISSN: 2521-8107, January-2020

Computing Determinants of Block Matrices

Md. Yasin Ali1, Ismat Ara Khan2

ABSTRACT: In some studies of physics and applied mathematics, there

arise large size of matrices and calculating determinants of those
matrices are very complex. In this case we can partition on such
matrices into some blocks. After partitioning, the new matrix which
elements are those partitions is a block matrix. In this article, we have
studied and explored some formulae to compute the determinant of
block matrices. We have curbed our absorption in 2  2 block matrices,
where each blocks are any m n size, where m, n   .
Keywords: Block matrix, Block diagonal matrix, Schur complement,
Determinant.
1. INTRODUCTION
Block matrices appear frequently in physics and applied mathematics [1-
5]. Among those some of the determinants of these matrices are very
large, for example, a model of high density quark matter must include
color (3), flavor (2-6), and Dirac (4) indices, giving rise to a matrix
between size 24× 24 and 72 ×72. In this case, calculating determinants of
those matrices are very complex such as computational time and
technique. But, we can calculate the determinant easily if we partition
these matrices into some blocks. Silvester [6] has calculated the
determinant of m m block matrices. Block matrices also have been
studied by Molinari and Popescu [9, 10]. In this work we have studied and
investigated some properties of 2×2 block matrices. These properties of
2×2 block matrices can help to calculate the determinant of any large sizes
matrices.
The paper is organized as follows. In section 2 we have discussed about
basic definitions and notations which are used throughout this paper. In
section 3, we have studied and investigated some formulae to compute the
determinant of 2×2 block matrices with an example.

1
Assistant Professor, Department of Electrical and Electronic Engineering, UITS.
* Corresponding author: Email: [email protected]
2
Lecturer, Department of Electrical and Electronic Engineering, UITS.
Email: [email protected]
5
Computing Determinants of Block Matrices

2. PRELIMINARIES
Definition 2.1: [7] A block matrix (also called partitioned matrix) is a
matrix of the kind
 B C
A 
D E 
Where B, C, D and E are also matrices, called blocks. Basically, a block
matrix is obtained by cutting a matrix two times: one vertically and one
horizontally. Each of the four resulting pieces is a block.
Example 2.1 (a): We consider the matrix
 3 1 3
A  2 5 7 
1 2 3

We can partition it into four blocks as

3 1  3
2 5  7
A 
     
 
1 2  3

By taking
 3 1  3
B  , C   , D  1 3, E  3
2 5 7 
The above matrix can be written as
 B C
A 
D E 
Definition 2.2: [7] Block matrices whose off-diagonal blocks are all equal
B 0 
to zero are called block-diagonal. The matrix A    is a block
 0 E
diagonal where 0 is a zero matrix.
a b 
Definition 2.3: [7] Let A    where a, b, c, d are numbers, then the
c d 
a b
determinant of A is A   ad  bc.
c d

6
 UITS Journal of Science & Engineering  Volume: 7, Issue: 1

3. DETERMINANTS OF BLOCK MATRICES

B 0 
Proposition 3.1: Let A    be a block diagonal matrix, where B
 0 E
and E are of any n n and m m size where m  n; m, n  N and 0 is
B 0
zero matrix, then A   B E.
0 E
Proof: Let I be a n n matrix. Then
B 0  B 0  I n 0
 0 E   0 I  0 E 
   m 

Now by the product formula, we have

B 0 B 0 In 0
 
0 E 0 Im 0 E

BE

B C
Proposition 3.2: Let A   be a block matrix, if C  0 , or D  0
D E 
B 0 B C B 0 B C
that is A    or A    BE 
E 
, then
D E  0 D E 0 E
Proof: Trivial
 B C
Proposition 3.3: Let A    be a block matrix, if A  0 or E  0
D E 
and C , or D is square but not of same size, then
0 C B C
 DC 
D E D 0
Proof: Trivial
 B C
Proposition 3.4: Let A    be a block matrix, if A  0 or E  0
D E 
and C , or D is square and of same size, then
0 C B C
 D C 
D E D 0
Proof: Trivial
7
Computing Determinants of Block Matrices

 B C 1
Definition 3.5: [8] If A    , then S B  E  DB C,
 D E 
SC  D  EC 1B, S D  C  BE 1F , S E  B  CE 1D, are called the Schur
complement of B, C, D and E respectively.

 B C
Theorem 3.6: Let A    be a block matrix, where the block,
D E 
B, C, D and E are of any m n size where m, n  N . If B is non-singular
then A  B S B , where S B is the Schur complement of B and also if B
and E are of same size then A  BE  DC , if BD  DB ; and
A  EB  DC , if BC  CB .

 B C
Proof: Suppose A    and B is non-singular.
D E 
 I 0  B C   B C 
Therefore,  1     1 
 DB I   D E   0 E  DB C 

 I A  B E  DB1C

Thus A  B E  DB1C  B S B (3.1)

 BE  BDB1C

If B and E are of same size.

 BE  DC if BD  DB (3.2)

Again we can write (3.1) as A  E  DB 1C B

 EB  DB 1CB

If B and E are of same size.

 BE  DC if BC  CB (3.3)

 B C
Theorem 3.7: Let A    be a block matrix, where the block,
D E 
B, C, D and E are of any m n size where m, n  N .

8
 UITS Journal of Science & Engineering  Volume: 7, Issue: 1

If C is non-singular and also if C and D are not of same size then

A  C SC , where SC is the Schur complement of C , again if C and D
are of same size then A   C SC and A  EB  CD if CE  EC;
also A  EB  DC if BC  CB.

 B C
Proof: Suppose A    and C is non-singular.
D E 
 I 0  B C   B C
Therefore,  
I   D E   D  EC B 0 
  
1 1
(3.4)
 EC
If C and D are not of same size then we obtain
 I A  D  EC 1B C

Thus A  C D  EC 1B  C SC (3.5)

Now from (3.4) if C and D are of same size then we obtain

 I A   D  EC 1B C

Thus A   C D  EC 1B   C SC (3.6)

 CEC 1B  CD

 EB  CD if CE  EC (3.7)

Again we can write (3.5) as A   D  EC 1B C

 EC 1BC  DC

 EB  DC if BC  CB (3.8)

 B C
Theorem 3.8: Let A    be a block matrix, where the block,
D E 
B, C, D and E are of any m n size where m, n  N . If B, C, D and E
are of all square matrices, then A  BE  DC if ; BD  DB;
A  EB  DC if BC  CB ; A  EB  CD if CE  EC; A  BE  CD
if DE  ED.
Proof: Trivial

9
Computing Determinants of Block Matrices

1 1 0 4 5 6 
3 4 5 9 4 6 
 
3 2 1 4 3 1 
Example 3.9: Let A    be a 6  6 order matrix.
0 0 0  1 0  2 
0 0 0 0 2 1 
 
0 0 0 0 5 2 
Now for computing determinant of A we can partition this matrix as
follows.
1 1 0  4 5 6 0 0 0 
Let B  3 4 5 , C  9 4 6 , D  0 0 0 and
3 2 1 4 3 1 0 0 0
  1 0  2
E   0 2 1  .
 0 5 2 

B C
Therefore, A   B E  6 1  6
0 E
4. CONCLUSIONS
A block matrix or a partitioned matrix is a partition of a matrix into
rectangular smaller matrices called blocks. Block matrices emerge often in
modern applications of linear algebra and it is not very difficult to
compute the determinant of those block matrices. But it is important, how
to implement the algorithms to compute the determinant of the block
matrices. In this article, we tough on a few thoughts and apparatus for
computing the determinant of block matrix. Particularly, we have proved
several propositions for computing determinant of 2 2 block matrices.
Also by using the Schur complement, some theorems for computing
determinant of 2  2 block matrices have been proved. This work will be
helpful to readers as well as researchers to calculate determinant of any
large sizes of matrices.

REFERENCES
[1] Abuki H, Baym G, Hatsuda T, Yamamoto N, (2010) Nambu–
Jona-Lasinio model of dense three-flavor matter with axial
anomaly: The low temperature critical point and BEC-BCS
diquark crossover, Phys. Rev. D, 81, pp. 125010-125018.

10
 UITS Journal of Science & Engineering  Volume: 7, Issue: 1

[2] Roessner S, Ratti C, Weise W, (2007) Polyakov loop, diquarks

and the two-flavour phase diagram, Phys. Rev. D, 75(3), pp.
034007-034024.
[3] Sasaki CFriman B, Redlich K, (2007) Susceptibilities and the
phase structure of a chiral model with Polyakov loops, Phys. Rev.
D, 75, pp. 074013-074029.
[4] Cliffe K A, Garratt T J, Spence A, (1993) Eigen values of block
matrices arising from problems in fluid mechanics, SIAM J.
Matrix Anal. Appl., 15(4), pp. 1310-1318.
[5] Murman E M, Abarbanel S S, eds.,(1984) Progress and
Supercomputing in Computational Fluid Dynamics, Proceedings
of U. S.-Israel Workshop, Birkh¨auser Boston, Inc., Germany.
[6] Silvester J R, (2000) Determinants of block matrices, Math. Gaz.,
84(501), pp. 460-467.
[7] Anton H, Rorres C., (1999) Elementary linear algebra, Eighth
edition, Wiley and Sons, USA.
[8] Zhang F, ed., (2005) The Schur Complement and Its
Applications, Springer, New York.
[9] Molinari L G, (2008) Determinants of block tridiagonal matrices,
Linear Algebra Appl., 429(8-9), pp. 2221-2226.
[10] Popescu O, Rose C, Popescu D C, (2004) Maximizing the
determinant for a special class of block-partitioned matrices,
Math. Prob. in Eng., 1, pp. 49-61.

11