LOVELY PROFESSIONAL

UNIVERSITY
TERM PAPER

OF
MATHEMATICS (101)
ON

SYLVESTE’S RANK INEQUALITY

SUBMITTED BY:-SABBIR ALAM REG NO:-11002829

ROLL NO :-A06 SEC :-E6001

DATE OF ALLOTMENT:-01-09-2010 DATE OF SUBMISSION:-15-11-2010

SUBMITTED TO:-Mrs. GURPREET KAUR

 *ACKNOWLEDGMENT*

THERE ARE MANY PERSONS TO WHOM I OWE THANKS FOR

THEIR HELP,SUPPORT,ADVICE AND SUGGESTIONS DURIND
THE VARIOUS STAGES OF THIS TERM PAPER.ABOVE ALL,I
THAK Mrs. GUPREET KAUR LECTURER OF MATHMETICS ,
WHO GUIDED ME FOR THE ENTIRE PROJECT.LAST BUT NOT
THE LEAST I THANK MY PARENTS WHO PROVIDED THEIR
MORAL SUPPORT AND GUIDED ME THROUGHOUT THIS TERM
PAPER.

THANKING YOU

SABBIRALAM
 *INTRODUCTION*
The column rank of a matrix A is the maximal number of linearly
independent columns of A. Likewise, the row rank is the maximal number of
linearly independent rows of A.
Since the column rank and the row rank are always equal, they are simply called
the rank of A. More abstractly, it is the dimension of the image of the linear
transformation that is multiplication by A.
Matrix theory on skew field is one of the basic direction in non-exchange
algebra research.And the rank of matrix on skew field is an important digital
feature of matrix. The inequality (equality)of rank of matrix is one of
the important issues newly discussed in the matrix theory. For the proof of
matrix-rank inequality(equality),the paper firstly makes use of some matrices
to construct block matrix ,then proves the inequality(equality)of rank of
matrix
by doing generalized elementary transformation to the block matrix above
,therefore some well-known inequalities can be proved by this method.

The Basic Idea.

Since the early 90’s, matrix inequalities (MIs) have become
very important in engineering, particularly, in control theory. If one has the
ability to convert the MIs arising in a particular problem to an LMI, then the
problem can be solved up to substantial size. The wide acceptance of LMIs
stems from the following
advantages:
1. if a control problem is posed as an LMI, then any local solution is a global
optimum;
2. efficient numerical LMI solvers are readily available;
3. once a control problem is posed as an LMI, adding constraints in the form of
LMIs results in a LMI problem. On the other hand, the LMI framework has the
following disadvantages:
1. there is no systematic way to produce LMIs for general classes of problems;
2. there is no way of knowing whether or not it is possible to reduce a system
problem to an LMI without actually doing it;
3. the user must possess the knowledge of manipulating LMIs, which takes
considerable training. Indeed, if one does not have the ability to deal with LMIs,
then it is not clear what one should do;
4. transformations via Schur complements can lead to a large LMI
representation.
SYLVESTER’S RANK INEQUALITY OF MATRICES:-

Let A and B be n × n matrices. A classical result about the rank

function is Sylvester’s inequality which states that the rank of the
product of AB is at most min{rank (A), rank (B)} and at least rank
(A)+rank (B)−n. A generalization of Sylvester’s inequality is
Frobenius’s inequality which states that
rank (AB) + rank (BC) ≤ rank (ABC) + rank (B).

Sylvester’s inequality:

ρ(A) + ρ(B) − n ≤ ρ(AB) ≤ min{ρ(A), ρ(B)}

For proving the Sylvester’s rank inequality, we use the matrix theory
on skew field. Skew field is a space in which we have elements that
have a multiplicative inverse.
Let K be a skew field ,Ks×n represent the group of the unit s×n matrix,In
represent the identity matrix on the skew field ,r(A) mean the rank of
matrix
1. Let A ∈ Ks×n ,B ∈ Kn×m, and if AB = 0, we can get
r(A) + r(B) ≤ n.
Proof Suppose B = (B1, B2, . . . , Bm) , from the known condition AB = 0
,we can get
A(B1, B2, . . . , Bm) = 0,
ABj = 0(j = 1, 2, . . . ,m).
And it also means these Bj(j = 1, 2, . . . ,m) are the solutions of right system
of
homogeneous linear equations AX = 0 ,while we have known that the
quantity
of the solutions in the solution-based system of AX = 0 is n−r(A), Therefore
we have r(B) ≤ n − r(A), namely
r(A) + r(B) ≤ n.

1. Let A  K mXn , B  K mXn , then

 A 0
r    r  A  r ( B)
 0 B 
2. Let A  K mXn , B  K mXn , then

A x
r    r  A  r ( B)
 0 B 
All these forms are from matrix theory and can be proved.

For form of inequality of matrices

n n

 r( A )   r(B )
i 1
i
i 1
i

Then block matrix of A is made as follows:

 A1 0 ... 0 
 
 0 A 2 ... 0 
M=  : : : 0 
 
0 0 ... An 


Then by elementary transformations we change it as:

 B1 x12 ... x1s 

 
 0 B 2 ... x 2 s 
 : : : : 
 
0 0 0 Bn 


n n

So,  r ( A )  r (M )   r ( B )
i 1
i
i 1
i

For proving:-
We first take n to the other side and we get,
r(AB) + n or r(AB) + nI

I 0 
This can be written as:  
0 AB  using the 2 proof.
From this we get the block matrix. Coversion of block matrix followed by
elementary transformations will give equation as in *. Hence we prove the
Sylvester’s rank inequality.

That is : r(AB) >= r(A) + r(B) –n.

(PROVED)

EXAMPLE-1:
Assume A ∈ Ks×n,B ∈ Kn×m , to prove the Sylvester inequality
r(AB) ≥ r(A)+r(B)-n.
Proof
The first equation is equivalent to
r(AB) + n ≥ r(A) + r(B)
that is

We construct block matrice

Example 2
Assume that A,B,C are multiplying-allowed matrices on skew
field K,to prove Frobeniua inequality
r(ABC) ≥ r(AB)+r(BC)−r(B). (2)
Proof :
The second equation is equivalent to
r(ABC) + r(B) ≥ r(AB) + r(BC)

Example3
Proved that A,B are two matrices on the center of skew field
K,and AB = BA , we can get
r(AB) ≤ r(A)+r(B)−r(A+B).
Proof The fifth equation is equivalent to
r(A) + r(B) ≥ r(AB) + r(A + B).
References:-
1.Songgui Wang ect, Inequality of matrix[M](the second press),
Science press,
Beijing(2006).
2. Tingming Wang, From Proof of inequality of matrix-rank by the
block matrix
construction[J], Higher mathematics(2008).
3. Wajin Zhuang , Guaidance of matrix theory on skew field[M],
Science press,
Beijing(2006).
4.WWW.WIKIPEDIA.COM