Volume 7, Issue 12, December – 2022 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165

Construction of a Positive Valued Scalar Function of

Strictly Rectangular Complex Matrices using the
Framework of Spacer Matrix Components and
Related Matrices
Debopam Ghosh

Abstract:- The Research paper presents a mathematical dimension, Reference matrix associated with the Embedded
method to construct a real, positive valued scalar function Matrix space corresponding to a strictly rectangular complex
of matrices belonging to the strictly rectangular complex matrix space.
matrix spaces. The formulated method leads to
association of a strictly rectangular complex matrix with Notations
a positive vector whose components sum to unity and
belongs to the real co-ordinate space of dimensionality  N denotes the set of all Natural numbers
equal to that of the embedding dimension associated with  C denotes the set of all Complex numbers
the input degrees of freedom of the rectangular complex
matrix space. These vector components act as the  R denotes the set of all Real numbers
respective weights to the ordered eigenvalues (ordered  I ww denotes the Identity matrix of order ‘w’
from largest to smallest) of a real, hermitian and positive
 C w denotes the complex coordinate space of order ‘w’
definite matrix, termed as the Reference matrix, which is 
uniquely determined by the Spacer component matrices  c denotes the complex conjugate of the complex
associated with the strictly rectangular complex matrix number ‘ c ‘
space. The resulting weighted sum of these eigenvalues,  M x y (C ) denotes the complex Matrix space of order ‘x’
expressed as an appropriate Inner product, is the
by ‘y’
numerical value which the scalar function assigns as the
output for the input complex strictly rectangular matrix  ‘ s ’ denotes the Embedding dimension
under consideration.  M ss (C ) denotes the Embedded Matrix Space
Keywords:- Spacer matrix based generalized matrix
multiplication, Spacer matrix components, Embedding

 1 
 
 2
   C ,   .  ,    1
d
 2 . .  d 1d , 1 , 2 ,...., d  C
 
. 
 d 
d 1
d f
 C d
,  C f
, Td  f  Tij  , Td  f  M d  f (C ) , therefore:  T     iTij  j
i 1 j 1

1  1  1  1  1 
1  1  1  1  1 
         
m   . , n   . , s   . , s  m   . , s  n   .
         
 .  .  .  .  .
1  1  1  1  1 
m1 n1 s1 ( s  m )1 ( s  n )1

IJISRT22DEC322 www.ijisrt.com 27
Volume 7, Issue 12, December – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165

 R s ( R) denotes the Real coordinate space of dimension ‘s’

 1 
 
 2
 R ( R)()  {   R ( R) |   .  ,  1 ,  2 ,...,  s  R,  i  0, i  1, 2,...., s}
s s

 
. 
  s 
s1

  R ( R) , therefore:    1  2 .
s
.  s 1s

 AH denotes the Hermitian conjugate of the matrix A

 ( X ww ) 1 denote the Proper Inverse of the Invertible matrix X ww , i.e. ( X ww ) 1 X ww  X ww ( X ww ) 1  I ww
X ww  M ww (C ) , such that X ww is hermitian, positive definite , then ( X ww ) 2  M ww (C ) , ( X ww )
1 1
 2
is
1 1
 ( X ww )1 , we also have: ( X ww ) 2  M ww (C) , ( X ww ) 2
1 1
hermitian, positive definite, such that: ( X ww ) 2 ( X ww ) 2

( X ww ) 2 ( X ww ) 2  X ww

1 1
is hermitian, positive definite, such that:

 Ax y denotes the Frobenius norm of the matrix Ax y

F
 max(a, b) denotes the maximum of the two inputs ‘a’ and ‘b’
 a  b denotes the absolute value of the difference between the two inputs ‘a’ and ‘b’
 ‘EVD’ is the abbreviation for “ Eigen value decomposition ”
 E.S (ˆ ) denotes the Eigenvalue spectrum of the matrix  ˆ
ss s s

 diag (v1 , v2 ,..., vw ) denotes a diagonal matrix of order ‘w’, whose diagonal elements along the main diagonal, from top
to bottom, are v1 , v2 ,..., vw respectively
 X  Y denotes that the set X is a proper or an improper subset of the set Y

I. INTRODUCTION

The Spacer matrix and the Spacer component matrices allow many of the mathematical operations that are only possible on
square matrices to be extended to strictly rectangular matrices. The mathematical framework based on Spacer matrix and its
component matrices allows for a generalized matrix multiplication scheme involving the spacer units as the compensating and
connecting units to be defined [6, 7, 11], subspaces of strictly rectangular complex matrix spaces in terms of the powers of the
matrices ( matrices connected by the spacer units) belonging to the complex matrix spaces to be constructed [13], correlation
between co-ordinate vectors belonging to non-compatible co-ordinate spaces to be quantified [8, 12]

The present research article formulates a mathematical scheme to construct a positive valued scalar function of strictly
rectangular complex matrices using the mathematical elements constructed from the spacer component matrices Pn s and Qsm . In
the formulated method, the information on the input dimensions ‘m’ and ‘n’ is used to determine the embedding dimension ‘s’,
which is followed by construction of the respective spacer matrix components and the corresponding Orthonormal matrices Wsn

and ˆ .
Gsm , the real, hermitian and positive semi-definite matrix ss and the Reference matrix  s s

The Input strictly rectangular matrix Amn is used to construct the hermitian, positive definite matrix ˆ ss . Using the matrix
ˆ ss , the information on the eigenvalues and associated eigenvectors of the Reference matrix  ˆ , the reference vector  and
s s

the distribution vector associated with the matrix Amn , denoted by  ( Amn ) , is formulated. The scalar function of the matrix

Amn , denoted as  ( Amn ) , is expressed as the inner product of the vectors  and  ( Amn ) .

The article presents the mathematical framework and definitions of a set of associated mathematical elements. The
methodology is numerically demonstrated using numerical samples from the complex matrix spaces of dimensions (m=2, n=1) and

IJISRT22DEC322 www.ijisrt.com 28
Volume 7, Issue 12, December – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
(m=2, n=3). The article concludes with a discussion of the numerical results obtained and insights about future directions of the
follow up research studies.

II. MATHEMATICAL FRAMEWORK

 m  N , n  N and m  n , s  max(m, n)  m  n , therefore s  N , s  m and s  n

The Spacer component matrices [6, 7, 8, 9, 10, 11, 12, 13] , Pn s and Qsm , are defined as follows:

1 0 . . 0 1n 1n . . 1
n
0 1 . . 0 1n 1n . . 1 

 1   n

Pns   I nn ( ) n sn   . . . . . . . . . . 
 n  n s  
 . . . . . . . . . . 
0 0 . . 1 n 1n
1 . . 1 
n  n s

1 0 . . 0 
0 1 . . 0 
 
 . . . . . 
 
 . . . . . 
 I mm  0
  0 . . 1 
Qsm  1  
( ) s  m m   m1 1
m . . 1
m
 m  s m  1m 1 
1 . .
 m m

 . . . . . 
 . . . . . 
 
 m
1 1
m . . 1 
m  s m

Wsn  ( Pns ) H [( PP H ) 2 ]nn , therefore W H W  I nn ,

1

Gsm  (Qsm )[(QH Q) 2 ]mm , therefore G H G  I mm

1

 Construction of the Reference Matrix ˆ

 s s
1 1 ˆ  ( 1 ) I  ( 1 )   ( 1 ) I  ( 1 )( P H P)  ( 1 )(QQ H )
 ss  ( )( P H P) ss  ( )(QQ H ) ss ,  s s s s s s s s s s s s ,
2 2 2 2 2 4 4
ˆ  M (C ) ,
 ss ss
ˆ
 is real, Hermitian and Positive definite
s s

ˆ :
EVD of the matrix  ˆ  U D (U )H where U   u
 u2 . . us  ,
s s ss ss ss ss s s  1
(U ss ) H  (U ss ) 1 , Dss  diag (1 , 2 ,...., s ) such that 1  2  ....s  0
s
We define: E.S ( ˆ )   o    , ˆ  {  R |      } ,
ˆ )  { ,  ,....,  } , Trace(
ss 1 2 s s s j s 1
j 1

1 s s j
  ( )  j ,   j ( ) therefore we have  ˆ ,   ˆ
s j 1 j 1 o

 Construction of the Reference vector  and the Scaled Reference vector 

IJISRT22DEC322 www.ijisrt.com 29
Volume 7, Issue 12, December – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165

 1   1 
   
 2 1  2 j 
  .  ,   ( o )   .  therefore:  j  ( s )  ( oj ) , j  1, 2,...., s
     
.  .  j 1
j
 s    s 
s1 s1

  Rs ( R)() ,   Rs ( R)() , s  1

 Amn  M mn (C ) , we define Bss  M ss (C ) such that : Bss  Gsm Amn (Wsn ) H
Thus, trace( BH B)  trace( BB H )  ( Bss F )2  ( Amn F )2
1 1 1 1
 ss  ( ) I ss  ( )( B H B) ss  ( )( BB H ) ss , ss  M ss (C ) , trace(  ss )  ( )( s  ( Amn F ) 2 )
2 4 4 2
1 1 1 1
ˆ ss  M ss (C ) , ˆ ss ( )  s s  ( )( I ss  ( )( B H B) ss  ( )( BB H ) ss )
trace(  ss ) s  ( Amn F ) 2
2 2
Thus, ˆ ss is Hermitian, Positive definite and has unit trace
 u1 ˆ u1 u1 ˆ u2 . . u1 ˆ us 
 
 u2 ˆ u1 u2 ˆ u2 . . u2 ˆ us 
ss  M ss (C ) ,  ss  (U ss ) H ˆ ssU ss  . . . . . 
 
. . . . . 
 u ˆ u1 us ˆ u2 . . us  us 
ˆ
 s
Thus,  s s is Hermitian, Positive definite and has unit trace

 Construction of the Distribution vector  ( Amn )

 u1 ˆ u1 
 
 u2 ˆ u2 
Amn  M mn (C ) ,  ( Amn )  R ( R)() ,  ( Amn )  .
s
 , s  ( Amn )  1
 
. 
 u ˆ us  s1
 s
 Analytical Expression of the scalar function  ( Amn )
s
 () : M mn (C) ˆ , such that: Amn  M mn (C ) ,  ( Amn )    ( Amn )    j u j ˆ u j
j 1

We define the following subsets of the Matrix space M mn (C ) :

o Sˆ[M mn (C) |    ]  {Amn  M mn (C) |  ( Amn )   }

o Sˆ[ M mn (C ) |    ]  { Amn  M mn (C ) |  ( Amn )   }
1 1
o Sˆ[ M mn (C ) |  ( Amn )  ( ) s ]  { Amn  M mn (C ) |  ( Amn )  ( ) s }
s s
o ˆ
S[M mn (C ) |  ( Amn )   ]  { Amn  M mn (C ) |  ( Amn )   }

Therefore, we have the following:

IJISRT22DEC322 www.ijisrt.com 30
Volume 7, Issue 12, December – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
1
o Sˆ[ M mn (C ) |  ( Amn )  ( ) s ]  Sˆ[ M mn (C ) |    ] and
s
Sˆ[ M mn (C ) |  ( Amn )   ]  Sˆ[M mn (C ) |    ]

 The case of Amn  0mn

1 1
Amn  0mn , this implies: ˆ ss   ss  ( ) I ss therefore:  ( Amn  0mn )  ( ) s which implies:
s s
1
0mn  Sˆ[ M mn (C ) |  ( Amn )  ( ) s ]
s
Numerical Case Studies

 m  2, n  1 therefore s  3

 
1 0 
 2 3
  1 2  3
1   
P13  1 1 1 , Q32  0 1  , W31 
1
1 , G32   2  3 2  3  ,
  3  2 3 
 1  1  2 2 
 1 
 2 2 
 1 1 1 
 
 4 2 3 8 2 3   3 2 6 
33  ( )  2 4 3 , 
1 ˆ  ( 1 ) 2 8 3 , U   1  1 1 
4
33
8   3 3  3 2 6  ,
 3 3 3  3 3 7   
 1 2 
 3 0 
 6 
13   13 
8  23 
   
   3
,     , E.S (
6 ˆ )  {  13 ,   3 ,   1} , ˆ  {  R | 1    13} ,
4  23  33 1
8
2
4
3
2 2 8
1  
  4
 2  31  23  31

23 221
 ,  
24 184
Example 1:
7
16 
 
1   5  ,  ( A )  137
A21    , therefore we have ( A21 ) 2
 1 ,  ( A ) 
0
F 21
16  21
128
 
1
 4  31

IJISRT22DEC322 www.ijisrt.com 31
Volume 7, Issue 12, December – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
Example 2:
7
16 
 
0
 ( A21 )    ,  ( A21 ) 
5 137
A21    , therefore we have: ( A21 F ) 2  1 ,
1  16  128
 
1
 4  31
Example 3:
7
 1  16 
  
2
)  1 ,  ( A21 )   
5 137
A21    , therefore we have: ( A21 2
,  ( A21 ) 
 i  F 16  128
  
2  1
 4  31

Example 4:
7
 1  16 
 2   
A21    , therefore we have: ( A21 ) 2
 1 ,  ( A )   5  ,  ( A )  137
 i  F 21
16  21
128
  2   
1
 4  31

Example 5:
1
 1  2
  
2
)  1 ,  ( A21 )    ,  ( A21 ) 
1 9
A21    , therefore we have: ( A21 2

 1  F 4 8
 1
2   
 4  31
Example 6:
3
 1  8
 2   
)  1 ,  ( A21 )    ,
3 65
A21    , therefore we have: ( A21 2
 ( A21 ) 
 1  F 8 64
  2   
1
 4  31
 m  2, n  3 therefore s  4

IJISRT22DEC322 www.ijisrt.com 32
Volume 7, Issue 12, December – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165

 
 1 1 0 
1 0 0    4  3 2  3 2  3 
3    
  0 1
  1  2  3 4  3 2  3 
 0
1
P34

1 0
3
, Q42    , W43  ( 6 )  
 1 1   2  3 2  3 4  3
 
0 1  2 2  

0 1    3 3 3 
3  1 1
 
 2 2

1  2 1  2  12 0 3 5  24 0 3 5 
     
1 1  2 1  2  1  0 12 3 5 ˆ 1  0 24 3 5 
G42 ( )  ,  44  (12 )  3 3 9 5 ,  44  ( 24 )  3 3 21 5 
2 2 1 1     
   5 5 5 5  5 5 5 17 
1 1 

1 1 1 1 
2 2 6 12 
 
1 1 1 1 
2 
2 6 12  ˆ )  {  4 ,   1,   3 1
U 44   , E.S ( 44 1 2 3 , 4  } ,
1 2 1  3 4 2
2 0 
6 12 
 
1 0 0 
3 
 2 12 
 16 
4  43 
3  
   12 
1   1 4 43 517
  3 ,    43  , ˆ  {  R |    } ,   ,  
  9 2 3 48 516
4  43 
1 6
   
 2  41
 43  41

Example 1:
7
10 
 
1
1 1 1 10  139
A23   , therefore we have: ( A23 )2  6 ,  ( A23 )    ,  ( A23 ) 
1 1 1  F
1 120
10 
1
 
10  41
Example 2:

IJISRT22DEC322 www.ijisrt.com 33
Volume 7, Issue 12, December – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165

7
 30 
 
1
 1 1 1   355
A23   , therefore we have: ( A23 )2  6 ,  ( A23 )   2  ,  ( A23 ) 
 1 1 1  F
1 360
6
1
 
 10  41

Example 3:
7
 30 
 
2
1 i i  5 173
A23   2
     ,  ( A23 ) 
i 
, therefore we have: ( A23 ) 6 , ( A23 )
 i
F
1 4 180
 15 
1
 
 10  41

III. DISCUSSION AND CONCLUSION In example 6, the weightage associated with the first two
leading eigen-spaces become equal, which effectively leads
The mathematical framework presented in the article to a lower value of the scalar function in example 6 as
essentially provides a method to construct a distribution compared to that for the examples 1, 2, 3 and 4.
vector  ( Amn ) , which is a real positive valued vector,
In the case of the numerical examples from the complex
whose elements sum up to unity, for every matrix Amn matrix space of dimensions (m=2, n=3), it can be observed
belonging to the strictly rectangular complex matrix space that the value of the scalar function decreases successively
from example 1 to 2 and from 2 to example 3. There is
M mn (C ) . The components of this vector are the diagonal reduction in weightage of the leading eigen-space of the
elements of the matrix  ss , which incorporate information reference matrix and unequal increment in the weightage
associated with the next two successive eigen-spaces, going
from the matrix Amn and the eigenvector information from example 1 to example 2. However, on going from
ˆ . Thus, the scalar
associated with the reference matrix 
example 2 to example 3, there is no change in weightage
s s associated with the eigen-spaces associated with the largest
function  ( Amn ) effectively provides a weighted average and the smallest eigenvalue of the reference matrix; only the
of the ordered eigenvalues of the reference matrix; these weightage associated with the second leading eigen-space is
weights are components of the distribution vector associated reduced while that of the third leading eigen-space is
increased while preserving their sum contribution in both the
with the matrix Amn , and hence have implicit dependence on examples.
the matrix elements that constitute the matrix Amn .
The presented mathematical framework allows any
iterative evolution scheme defined on strictly rectangular
In the case of the numerical examples from the complex
complex matrix spaces to be mapped on to an evolution
matrix space of dimensions (m=2, n=1), we can observe that
scheme involving positive vectors of unit sum belonging to
the examples 1, 2, 3, 4 are mapped to the same distribution s
vector and hence, to the same numerical value for the scalar the co-ordinate space R ( R) which can therefore, be
function. In case of the example 5 there is increased visualized as being mediated by Markov type transition
weightage along the eigen-space associated with the largest matrices. Follow up studies dedicated to understanding the
eigenvalue of the reference matrix and reduction in the interrelationship between the evolution schemes on the Input
weightage corresponding to the next largest eigenvalue of the rectangular matrix space and the resulting mapped evolution
reference matrix, in comparison to the respective values for sequences defined on R ( R)() , in light of the
s
the distribution vector corresponding to the examples 1,2,3,4.
This leads to an overall larger value of the scalar function in mathematical framework presented in this article, is expected
the example 5 compared to that of the examples 1, 2, 3 and 4. to produce a deeper understanding of the framework and its
limitations and scope of applicability in domains of research

IJISRT22DEC322 www.ijisrt.com 34
Volume 7, Issue 12, December – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
problems arising in different areas of natural and applied [13]. Ghosh, Debopam, The formulation of Critical subspace
sciences. and Extended Critical subspace associated with
matrices of Complex Matrix spaces of order m by n,
REFERENCES where m≠n (Article DOI:
10.13140/RG.2.2.13862.60488) (2022)
[1]. Arfken, George B., and Weber, Hans J., Mathematical [14]. Graham, Alexander, Kronecker Products & Matrix
Methods for Physicists, 6th Edition, Academic Press Calculus with Applications, Dover Publications, Inc.
[2]. Brewer, J. W., Kronecker Products and Matrix Calculus [15]. Hardy, Lucien, Quantum Theory from Five Reasonable
in System Theory, IEEE Trans. on Circuits and Systems, Axioms, arXiv: quant-ph/0101012v4 (2001)
25, No.9, p 772-781 (1978) [16]. Hassani, Sadri, Mathematics Physics A Modern
[3]. Datta, B. N., Numerical Linear Algebra and Introduction to its Foundations, Springer
Applications, SIAM [17]. Hogben, Leslie, (Editor), Handbook of Linear Algebra,
[4]. Dorai, Kavita, Mahesh, T. S., Arvind and Anil Kumar, Chapman and Hall/CRC, Taylor and Francis Group
Quantum computation using NMR, Current Science, [18]. Jordan, Thomas. F., Quantum Mechanics in Simple
Vol.79, No. 10, p 1447-1458 (2000) Matrix Form, Dover Publications, Inc.
[5]. Ghosh, Debopam, A Tryst with Matrices: The Matrix [19]. Macklin, Philip A., Normal matrices for physicists,
Shell Model Formalism, 24by7 Publishing, India. American Journal of Physics, 52, 513(1984)
[6]. Ghosh, Debopam, A Generalized Matrix Multiplication [20]. Meyer, Carl. D., Matrix Analysis and Applied Linear
Scheme based on the Concept of Embedding Dimension Algebra, SIAM
and Associated Spacer Matrices, International Journal [21]. Nakahara, Mikio, and Ohmi, Tetsuo, Quantum
of Innovative Science and Research Technology, Computing: From Linear Algebra to
Volume 6, Issue 1, p. 1336 - 1343 (2021) [22]. Physical Realizations, CRC Press
[7]. Ghosh, Debopam, The Analytical Expressions for the [23]. Neudecker, H., Some Theorems on Matrix
Spacer Matrices associated with Complex Matrix Differentiation with special reference to Kronecker
spaces of order m by n, where m ≠ n, and other pertinent Matrix Products, J. Amer. Statist. Assoc., 64, p 953-
results , (Article DOI: 10.13140/RG.2.2.23283.45603) 963(1969)
(2021) [24]. Paris, Matteo G A, The modern tools of Quantum
[8]. Ghosh, Debopam, Construction of an analytical Mechanics : A tutorial on quantum states,
expression to quantify correlation between vectors measurements and operations, arXiv: 1110.6815v2 [
belonging to non-compatible complex coordinates quant-ph ] (2012)
spaces, using the spacer matrix components and [25]. Roth, W. E., On Direct Product Matrices, Bull. Amer.
associated matrices (Article DOI: Math. Soc., No. 40, p 461-468 (1944)
10.13140/RG.2.2.17857.68969) (2022) [26]. Sakurai, J. J., Modern Quantum Mechanics, Pearson
[9]. Ghosh, Debopam, Quantification of Intrinsic Overlap in Education, Inc
matrices belonging to strictly rectangular complex [27]. Steeb, Willi-Hans, and Hardy, Yorick, Problems and
matrix spaces, using the spacer matrix components and Solutions in Quantum Computing and Quantum
associated matrices (Article DOI: Information, World Scientific
10.13140/RG.2.2.18405.06886) (2022) [28]. Strang, Gilbert, Linear Algebra and its Applications,
[10]. Ghosh, Debopam , A Mathematical Scheme defined on Fourth Edition, Cengage Learning
strictly rectangular complex matrix spaces involving [29]. Sundarapandian, V., Numerical Linear Algebra, PHI
Frobenius norm preservation under the possibility of Learning Private Limited
matrix rank readjustment and internal redistribution of
Variance using Spacer component matrices and a
Completely Positive Trace Preserving transformation ,
International Journal of Innovative Science and
Research Technology, Volume 7, Issue 10, pp. 591 -
602 (2022)
[11]. Ghosh, Debopam, A Compilation of the Analytical
Expressions, Properties and related Results and
formulation of some additional mathematical elements
associated with the Spacer matrix components
corresponding to strictly rectangular Complex Matrix
Spaces (Article DOI: 10.13140/RG.2.2.31777.89446/1)
(2022)
[12]. Ghosh, Debopam, Quantifying Correlation between
vectors belonging to Non-Compatible Real Co-ordinate
Spaces using a Mathematical scheme based on Spacer
Matrix components(Article DOI:
10.13140/RG.2.2.26336.76805) (2022)

IJISRT22DEC322 www.ijisrt.com 35