5 Some Inverses on k−Idempotent

Fuzzy Matrices

In matrix Algebra, the concept of generalized inverse presents a very inter-

esting area of research. The generalized inverse (g−inverse or γ−inverse)is a
generalization of the inverse of matrix for either singular or a non-square matrix.
The γ−inverse has innumerable applications in control theory, robotics, singnal
processing and associative memories. Generalized inverses of a m × n matrix A
over a field F is a n×m matrix G over F such that GB is a solution of the system
AX = B of linear equations whenever B is such that this system is consistent.
Conventionally, a γ−inverse of A is denoted by A− .

However this fails for a fuzzy matrix, that is, for A ∈ Fmn under the max-
min fuzzy operations, the matrix equation AXA = A need not have a solution
X. If A has a generalized inverse (γ−inverse), then A is said to be regular.

This chapter is a detailed discussion of various γ−inverses assoiciated

with a regular matrix and obtain characterization of the set of all γ−inverses for
idempotent zκ matrices. It is found that the square of an idempotent zκ matrix
behaves as a inverse for various inverses. Also, it has been discussed that some
of the inverse of an idempotent zκ matrix is also κ− idempotent.

A study on k−idempotent fuzzy matrices 55

 5 Some Inverses on k−Idempotent Fuzzy Matrices

5.1 Generalized inverse of idempotent zκ Matrices

In this section, a γ−inverse of an idempotent zκ matrix A is found. Moreover,

Some fundamental results are derived and some basic properties using idempo-
tent zκ matrix have been obtained.

Definition 5.1.1 For A ∈ Fnk , A− −

k is said to be a γ−inverse of A if AAk A = A

for all A−
k ∈ A{1}.

 
1 0.5
Example 5.1.2 Let A = ∈ Fn2 and the γ−inverse is given by
0.5 1
 
− 0.5 1
Ak = such that
1 0
   
− 1 0.5 0.5 1 1 0.5
AAk A =
0.5 1 1 0 0.5 1
 
1 0.5
= = A.
0.5 1

Theorem 5.1.3 If A− k 2
k is a γ−inverse of A ∈ Fn , then γ−inverse of A is

KA−
k K.

Proof A2 (KA− 2 2 −
k K)A = A KAk KA
2

= KAA−
k AK

= KAK

= A2 .

Therefore, KA− 2
k K ∈ A {1}.

Proposition 5.1.4 If A− k −
k is a γ−inverse of A ∈ Fn , then Ak K = KA and

KA−
k = AK.

A study on k−idempotent fuzzy matrices 56

 5 Some Inverses on k−Idempotent Fuzzy Matrices

Proof KA[A− −
k K]KA = KAAk A

= KA

AK[KA− −
k ]AK = AAk AK

= AK.

Proposition 5.1.5 If A− k 3 − −
k is a γ−inverse of A ∈ Fn , then A = AAk = Ak A.

Proof A3 (AA− 3 3 −
k )A = A AAk AA
2

= A3 AA2

= (A3 )2

= A3 .

Lemma 5.1.6 If A ∈ Fnk , then A2 ∈ A{1}.

Proof AA2 A = A4

that is AA2 A = A.

Thus A2 is the γ−inverse of A.

Theorem 5.1.7 Let A ∈ Fnk and X ∈ A{1}, then AX and XA are idempotent
zκ matrices.

Proof Consider, K(AX)2 K = K(AX)(AX)K

= K(AXA)XK

= KAKX

= A2 X

= AX.

A study on k−idempotent fuzzy matrices 57

 5 Some Inverses on k−Idempotent Fuzzy Matrices

Also, K(XA)2 K = K(XA)(XA)K

= KX(AXA)K

= XKAK

= XA2

= XA.

Hence, AX and XA are idempotent zκ matrices.

   
1 0 1 1
Example 5.1.8 If A = then there exists a X = ∈ A{1} such
0 0 1 0
that AXA = A.
   
1 1 1 0
Here AX = and XA =
0 0 1 0
    
2 0 1 1 1 1 1 0 1
Also K(AX) K =
1 0 0 0 0 0 1 0
 
1 1
= = AX
0 0
    
0 1 1 0 1 0 0 1
and K(XA)2 K =
1 0 1 0 1 0 1 0
 
1 0
= = XA
1 0
therefore, AX and XA are κ−idempotent.

Definition 5.1.9 For a subspace W of Vn , if (Vn )X = W where X is the idem-

potent zκ matrix then W is retract of Vn .

Lemma 5.1.10 For A ∈ Fn , A is regular fuzzy matrix if, and only if there exists
an idempotent zκ matrix H such that R(H) = R(A).

Lemma 5.1.11 For A ∈ Fn , there exists an idempotent zκ matrix H such that

ρ(H) = ρ(A) if, and only if A is a regular fuzzy matrix.

Theorem 5.1.12 R(A) is a retract of Vn if, and only if A is regular.

A study on k−idempotent fuzzy matrices 58

 5 Some Inverses on k−Idempotent Fuzzy Matrices

Proof Consider R(A) = (Vn )X.

We have X is κ− idempotent.

By Lemma 5.1.10, A is regular.

Conversely, Consider A as regular.

Then there exists an κ−idempotent X.

Therefore R(X) = R(A) with KX 2 K = X.

Hence R(A) = (Vn )X.

Definition 5.1.13 For A ∈ Fn , there exists a semi γ−inverse X for A such

that AXA = A and XAX = X for all X ∈ A{1, 2}.

Theorem 5.1.14 If A ∈ Fnk , then A2 is both γ−inverse and semi γ−inverse

of A.

Proof By Lemma 5.1.6, A2 ∈ A{1}.

Now, A2 AA2 = A5

=A

Therefore A2 ∈ A{1, 2}.

5.2 µ−Penrose inverse of idemotent zκ matrices

In this section, with the existence of least square γ−inverse and minimum norm
γ−inverse, µ− penrose inverse for idempotent zκ matrix are found. A nec-
essary condition for the µ−Penrose inverse of an idempotent zκ matrix to be
k−idempotent is also derived.

Definition 5.2.1 For an idempotent zκ matrix A, a matrix G is called least

square γ−inverse of A if AGA = A and (AG)T = AG for all X ∈ A{1, 3}.

A study on k−idempotent fuzzy matrices 59

 5 Some Inverses on k−Idempotent Fuzzy Matrices

Definition 5.2.2 For an idempotent zκ matrix A, a matrix G is called mini-

mum norm γ−inverse of A if AGA = A and (GA)T = GA for all X ∈ A{1, 4}.

By Theorem 5.1.3, we will get the following results obviously.

(1,3)
Theorem 5.2.3 If Ak is a {1, 3}−inverse of an idempotent zκ matrix A,
(1,3) (1,3)
then [A2 ]k = KAk K.

(1,4)
Theorem 5.2.4 If Ak is a {1, 4}−inverse of an idempotent zκ matrix A,
(1,4) (1,4)
then [A2 ]k = KAk K.

Definition 5.2.5 For an idempotent zκ matrix A, if the fuzzy matrix P is both

{1, 3}−inverse and {1, 4}−inverse , then P is called µ− Penrose inverse of A
which is denoted as A+
k.

By combining the Theorems 5.1.3, 5.2.3 and 5.2.4, we easily prove the
following theorem.

Theorem 5.2.6 If A+
k is a µ− Penrose inverse of an idempotent zκ matrix A,

then the µ− Penrose inverse of A2 is KA+

k K.

Theorem 5.2.7 If A ∈ Fnk , then (A3 )+

k commutes with the κ−permuted matrix

K.

Proof If A is an idempotent zκ matrix, then KA = AK

KA3 = A3 K

(KA3 )+ 3 +
k = (A K)k

Kk+ (A3 )+ 3 + +
k = (A )k Kk

K(A3 )+ 3 +
k = (A )k K .

Theorem 5.2.8 If A ∈ Fnk , then A+ k

k ∈ Fn also.

A study on k−idempotent fuzzy matrices 60

 5 Some Inverses on k−Idempotent Fuzzy Matrices

Proof We have

A2 (A+ 2 2 + +
k ) A = (AA)(Ak Ak )(AA)

= (AAA+ +
k )(Ak AA)

= AA+ +
k A.AAk A

= AA
= A. (5.1)

and (A+ 2 2 + 2 + + + +
k ) A (Ak ) = (Ak Ak )(AA)(Ak Ak )

= (A+ + + +
k Ak A)(AAk Ak )

= (A+ + + +
k AAk )(Ak AAk )

= A+ +
k Ak

= (A+ 2
k) . (5.2)

From equation 5.1 and 5.2, we have

(A2 )+ + 2
k = (Ak )

Now A+ 2 +
k = (KA K)k

= K(A2 )+
kK

= K(A+ 2
k ) K.

Hence A+
k is idempotent zκ matrix.

By Definition 5.2.5, we directly observe the following theorem.

Theorem 5.2.9 If A ∈ Fnk and A+ T

k exists, then A ∈ A{1}

Theorem 5.2.10 If every α−section of A ∈ Fnk has µ−Penrose inverse, then

AT is a µ−Penrose inverse of A.

A study on k−idempotent fuzzy matrices 61

 5 Some Inverses on k−Idempotent Fuzzy Matrices

Proof Let Aα be the α−section of A, then µ− Penrose inverse exists for all
α−sections.

Therefore A+ T
α = Aα

Now, AT = ( αAα )T
L
α∈φ
/ A

αATα )
L
=(
α∈φ
/ A

αA+
L
= α
α∈φ
/ A

= A+ .

Thus AT is the µ− Penrose inverse of A.

Definition 5.2.11 An idempotent zκ matrix A is said to be range κ−symmetric

if, and only if R(A) = R(AT ).
 
1 1
Example 5.2.12 1. Consider the fuzzy matrix A =
1 1
Here KA2 K = A and R(A) = R(AT ).

Therefore, A is range κ−symmetric.

 
1 0
2. Now let us take another fuzzy matrix A = .
1 0
Here KA2 K = A but R(A) 6= R(AT ).

Hence A is not range κ−symmetric.

Theorem 5.2.13 If A is a range κ−symmetric and K the κ−permuted matrix,

then R(KA) = R(AK).

Proof Since A is range κ−symmetric, AK = KA

it follows that R(KA) = R(AK).

Theorem 5.2.14 Let A ∈ Fnk be a range κ−symmetric and ρ(A) = r, then

ρ(A) = ρ(AT ).

A study on k−idempotent fuzzy matrices 62

 5 Some Inverses on k−Idempotent Fuzzy Matrices

Proof Since A is range κ−symmetric, R(A) = R(AT ).

By taking ρ(A) = r, it is obvious that

ρ(A) = R(AT ) and ρ(AT ) = R(A).

Hence ρ(A) = ρ(AT ).

Theorem 5.2.15 Let E be symmetric idempotent zκ matrix and A ∈ Fnk such

that R(A) = R(E). Then the following are equivalent.

1. A is range κ−symmetric.

2. A+
k exists and range κ−symmetric.

3. AE = EA.

Proof (1) ⇒ (2) :

Since A is range κ− symmetric,

R(A) = R(AT ) = ρ(A).

Since both column space and row space are equal, A+

k exists such that

A+ T
k = A .

Since R(A) = R(AT ) ⇒ R[(A)]T = R(AT )T

⇒ R[A+ T T
k ] = R[(A ) ]

Thus A+
k is range κ−symmetric.

(2) ⇒ (3) :

Since A+ + T +
k exists, Ak = A and E = Ak A is symmetric idempotent zκ

fuzzy matrix.

AE = AAT A = A ⇒ R(E) ⊇ R(A) and

A+
k A = E ⇒ R(A) ⊇ R(E).

A study on k−idempotent fuzzy matrices 63

 5 Some Inverses on k−Idempotent Fuzzy Matrices

Hence R(A) = R(E).

Since A+ T
k is range κ−symmetric, R(A ) = R(A) = R(E)

Since E is symmetric κ−idempotent, AT = AT E

(AT )T = (AT E)T

A = ET A

= EA.

Thus E = A+
k and AE = EA with R(A) = R(E).

(3) ⇒ (1) :

Since AE = EA = A,

EAT = AT E = AT
AT E = AT ⇒ R(AT ) ⊆ R(E) = R(A) (5.3)

and AT E = EAT ⇒ R(E) ⊇ R(EAT ) = R(A)

⇒ R(A) ⊇ R(AT ). (5.4)

From 5.3 and 5.4, ⇒ R(A) = R(AT ).

Thus, A is range κ−symmetric.

5.3 Spectral inverses on κ−idempotent fuzzy

matrices

In this section, the existence of #− inverse and δ− inverse for idempotent zκ

matrix have been discussed. The relation between the δ− inverse and #− inverse
of a square fuzzy matrix are determined ultimately.

Definition 5.3.1 For a κ−idempotent fuzzy matrix A, the #− inverse of A,

denoted as A# #
k is a commuting semi γ−inverse of A, that is AAk A = A,

A# # # # #
k AAk = Ak and AAk = Ak A.

A study on k−idempotent fuzzy matrices 64

 5 Some Inverses on k−Idempotent Fuzzy Matrices

Theorem 5.3.2 If A#
k is a #− inverse of an idempotent zκ matrix A, then the

#− inverse of A2 is KA#
k K.

Proof A2 (KA# 2 2 #
k K)A = A KAk KA
2

= KAA#
k AK

= KAK

= A2 .

In the similar manner, we will prove

(KA# # #
k K)A(KAk K) = KAk K

Next, by the definition, A = A2 A# # 2

k = Ak A

Multiplying K 2 on both sides,

K 2 (A# 2 # 2
k A ) = (Ak A )K
2

that is, (KA# 2 2 #

k K)A = A (KAk K).

Hence KA# 2
k K is the group inverse of A .

Lemma 5.3.3 If A#
k is the #− inverse of an idempotent zκ matrix A, then

A2 = A#
k.

Proof By theorem 5.1.14, A2 ∈ A{1, 2} for the idempotent zκ matrix A,

we have AA# # # # # #
k A = A, Ak AAk = Ak and AAk = Ak A.

Now A = A3 = AA2 = A2 A.

Hence, A# 2
k = A .

 
0.1 0.1 0.1
Example 5.3.4 Let A = 0.2 0.4 0.2 .
0.7 0.6 0.8

A study on k−idempotent fuzzy matrices 65

 5 Some Inverses on k−Idempotent Fuzzy Matrices

    
0 0 1 0.1 0.1 0.1 0.1 0.1 0.1 0 0 1
2
We have KA K = 0 1 0 0.2 0.4 0.2 0.2 0.4 0.2 0 1 0
1 0 0 0.7 0.6 0.8 0.7 0.6 0.8 1 0 0
 
0.1 0.1 0.1
= 0.2 0.4 0.2
0.7 0.6 0.8
A is κ−idempotent.

Hence A# #
k exists and Ak = A.

Here AAT A 6= A ⇒ AT is not a γ−inverse of A.

Thus A+
k does not exists.

Hence for an idempotent fuzzy matrix A, if A# +

k exists then Ak need not

exists.

Theorem 5.3.5 If A# k #
k is the #− inverse of A ∈ Fn , then (KA)k = KA and

(AK)#
k = AK.

By Lemma 5.3.3, proof follows immediately.

Theorem 5.3.6 If A ∈ Fn and An ∈ Fnk , then (An )# n

k = A for all n > 1.

Proof By using AA# #

k = Ak A,

we easily verify that A(An )# n #

k = (A )k A.

Hence #− inverse of An is (An )#

k , n > 1.

Converse of the above theorem is not true, which is illustrated in the

following example.

 
0 1
Example 5.3.7 Consider A = and for any positive integer n > 1,
1 1
 
1 1
An = is an κ−idempotent matrix.
1 1

A study on k−idempotent fuzzy matrices 66

 5 Some Inverses on k−Idempotent Fuzzy Matrices

Here (An )# n
k = A for all n > 1.

However, A#
k does not exists.

Theorem 5.3.8 If A ∈ Fnk has a #− inverse A# n # # n

k , then (A )k = (Ak ) .

Proof By Theorem 5.3.2, (A2 )# #

k = KAk K

= (A# 2
k) .

By induction, we get (An )# # n

k = (Ak ) .

Definition 5.3.9 An idempotent zκ matrix A is said to be −matrix if, and

#
only if A+
k = Ak

Theorem 5.3.10 If A ∈ Fnk is −matrix, then KA and AK are also −matrix.

#
Proof If A is −matrix, then A+
k = Ak

A+
k = A
2

A+
k = KAK

A+
k K = KA

#
(KA)+
k = (KA)k .

#
In the similar way, we prove (AK)+
k = (AK)k .

Hence KA and AK are −matrix.

Definition 5.3.11 For an idempotent zκ matrix A, the δ− inverse of A denoted

as Adk is the solution of the following equations :

(i) For some positive integer m, Am = Am+1 .X

(ii) X = X 2A

(iii) AX = XA

A study on k−idempotent fuzzy matrices 67

 5 Some Inverses on k−Idempotent Fuzzy Matrices

The smallest positive integer m for which the above condition is satisfied is called
the κ−index of A.

Theorem 5.3.12 If Adk is the δ− inverse of idempotent zκ matrix A,

then Adk = A2 .

Proof We have A = A4 = A2 .A2 . Hence m = 1

A2 = A5 = (A2 )2 .A and

AA2 = A3 = A2 .A.

Hence Adk = A2 .

Theorem 5.3.13 If Adk is the κ−commuting δ− inverse of idempotent zκ

matrix A, then (i) (KA)dk = KA and (ii) (AK)dk = AK

Proof Since KA commutes with itself and it is a tripotent matrix, we have

KA = KA.KA.KA

= (KA)2 .KA with m = 1.

It follows that (KA)dk = KA.

The proof of (ii) is similar to (i).

Theorem 5.3.14 For an idempotent zκ matrix A, if G is the δ−inverse of A

with κ−index P, then for any s ≥ p, As has #− inverse and Gs is the #−inverse
of As .

Proof Suppose A ∈ Fnk has δ− inverse G of index p, then by definition,

Ap+1 G = Ap , G2 = G and AG = GA.

We claim that Gp is the #− inverse.

A study on k−idempotent fuzzy matrices 68

 5 Some Inverses on k−Idempotent Fuzzy Matrices

Now, Ap Gp Ap = Ap Gp−2 (G2 A)Ap−1 = Ap Gp−1 Ap−1

= . . . = Ap GA = Ap+1 G = Ap

Also, (Gp )Ap (Gp ) = (Gp )2 Ap = (G2 )p = Gp .

Thus, Gp is the #− inverse of Ap .

Hence by using Theorem 5.3.6, we get for any s ≥ p, As has #− inverse

and Gs is the #− inverse of Ap .

A study on k−idempotent fuzzy matrices 69