Iranian Journal of Fuzzy Systems Vol. 4, No. 1, (2007) pp.

75-87 75

TRIANGULAR FUZZY MATRICES

A. K. SHYAMAL AND M. PAL

Abstract. In this paper, some elementary operations on triangular fuzzy

numbers (TFNs) are defined. We also define some operations on triangu-
lar fuzzy matrices (TFMs) such as trace and triangular fuzzy determinant
(TFD). Using elementary operations, some important properties of TFMs are
presented. The concept of adjoints on TFM is discussed and some of their
properties are. Some special types of TFMs (e.g. pure and fuzzy triangular,
symmetric, pure and fuzzy skew-symmetric, singular, semi-singular, constant)
are defined and a number of properties of these TFMs are presented.

1. Introduction
After Zadeh’s work [20] uncertainty can be classified into two types - probabilistic
uncertainty and fuzzy uncertainty, though people were aware of fuzzy uncertainty
before the mathematical formulation of fuzziness by Zadeh. Fuzziness can be rep-
resented in different ways. One of the most useful representation is membership
function. Also, depending the nature or shape of membership function a fuzzy
number can be classified in different ways, such as triangular fuzzy number (TFN),
trapezoidal fuzzy number etc. Triangular fuzzy numbers (TFNs) are frequently
used in applications. It is well known that the matrix formulation of a mathemat-
ical formula gives extra facility to handle/study the problem. Due to the presence
of uncertainty in many mathematical formulations in different branches of science
and technology, we introduce triangular fuzzy matrices (TFMs). To the best of our
knowledge, no work is available on TFMs, though a lot of work on fuzzy matrices
is available in literature . A brief review on fuzzy matrices is given below.
Fuzzy matrices were introduced for the first time by Thomason [17], who dis-
cussed the convergence of powers of fuzzy matrix. Fuzzy matrices play an important
role in scientific development. Several authors have presented a number of results
on the convergence of the power sequence of fuzzy matrices [2, 4, 8]. Ragab et al.
[11] presented some properties on determinant and adjoint of square fuzzy matrix.
Kim and Roush [5] studied the canonical form of an idempotent matrix. Hashimoto
[3] studied the canonical form of a transitive matrix. Kolodiejczyk [7] presented
the canonical form of a strongly transitive matrix. Xin [18, 19] studied the con-
trollable fuzzy matrix. Ragab et al. [12] presented some properties of the min-max
composition of fuzzy matrices. Kim [6] presented some important results on deter-
minant of a square fuzzy matrices. Two new operators and some applications of

Received: January 2006; Revised: May 2006; Accepted: June 2006

Key words and phrases: Triangular fuzzy numbers, Triangular fuzzy number arithmetic, Tri-
angular fuzzy matrices, Triangular fuzzy determinant.
76 A. K. Shyamal and M. Pal

fuzzy matrices are given in [13, 14, 15, 16]. Pal [9] introduced intuitionistic fuzzy
determinant. Pal et al. [10] introduced intuitionistic fuzzy matrices.
In this article, we introduce TFN, determinant, adjoint of TFMs and discuss
some of their properties.

2. Triangular Fuzzy Number

Sometimes it may happen that some data or numbers cannot be specified pre-
cisely or accurately due to the error of the measuring technique or instruments etc.
Suppose the height of a person is recorded as 160 cm. However, it is impossible in
practice to measure the height accurately; actually this height is about 160 cm; it
may be a bit more or a bit less than 160 cm. Thus the height of that person can
be written more precisely as the triangular fuzzy number (160 − α, 160, 160 + β),
where α and β are the left and right spreads. In general, a TFN ”a” can written
as (a − α, a, a + β), where α and β are the left and right spreads of a respectively.
These type of numbers are alternately represented as ha, α, βi. The mathematical
definition of a TFN is given below.
Definition 2.1. A triangular fuzzy number denoted by M = hm, α, βi, has the
membership function
0 for x ≤ m − α

m−x



 1− α for m − α < x < m



µM (x) = 1 for x = m

 1− x − m
for m < x < m + β

β




0 for x ≥ m + β.
The point m, with membership grade of 1, is called the mean value and α, β are
the left hand and right hand spreads of M respectively.
A TFN is said to be symmetric if both its spreads are equal, i.e., if α = β and
it is sometimes denoted by M = hm, αi.
Due to the wide field of applications of TFNs, many authors have tried to de-
fine the basic arithmetic operations on TFNs. Here we introduce the definitions of
arithmetic operations due to Dubois and Prade [1].

Let M = hm, α, βi and N = hn, γ, δi be two TFNs.

(1) Addition: M + N = hm + n, α + γ, β + δi.
(2) Scalar multiplication: Let λ be a scalar, λM = hλm, λα, λβi, when
λ ≥ 0.
λM = hλm, −λβ, −λαi, when λ ≤ 0. In particular, −M = h−m, β, αi.
(3) Subtraction: M − N = hm, α, βi − hn, γ, δi = hm − n, α + δ, β + γi.
For two TFNs M and N , their addition, subtraction and scalar multi-
plication, i.e., M + N, M − N and λM are all TFMs.
(4) Multiplication: It can be shown that the shape of the membership func-
tion of M.N is not necessarily a triangular, but, if the spreads of M and
 Triangular Fuzzy Matrices 77

N are small compared to their mean values m and n then the shape of
membership function is closed to a triangle. A good approximation is as
follows:
(a) When M ≥ 0 and N ≥ 0 (M ≥ 0, if m ≥ 0)
M.N = hm, α, βi.hn, γ, δi ' hmn, mγ + nα, mδ + nβi.
(b) When M ≤ 0 and N ≥ 0
M.N = hm, α, βi.hn, γ, δi ' hmn, nα − mδ, nβ − mγi.
(c) When M ≤ 0 and N ≤ 0
M.N = hm, α, βi.hn, γ, δi ' hmn, −nβ − mδ, −nα − mγi.
When spreads are not small compared with mean values, the following
is a a better approximation:
hm, α, βi.hn, γ, δi ' hmn, mγ + nα − αγ, mδ + nβ + βδi for M > 0, N > 0.
Throughout the paper we use the previous definition.
Now, we define inverse of a TFN based on the definition of multiplication.
(5) Inverse: The inverse of a TFN M = hm, α, βi, m > 0 is defined as,
M −1 = hm, α, βi−1 ' hm−1 , βm−2 , αm−2 i.
This is also an approximate value of M −1 and it is valid only a neigh-
bourhood of 1/m.
Division of M by N is given by
M
= M.N −1 .
N
Since inverse and product both are approximate, the division is also an
approximate value. The formal definition of division is given below.
M
(6) Division: = M.N −1 = hm, α, βi.hn−1 , δn−2 , γn−2 i
N 
m mδ + nα mγ + nβ
' , , .
n n2 n2
From the definition of multiplication of TFNs, the power of any TFN M
is defined in the following way.
(7) Exponentiation: Using the definition of multiplication it can be shown
that M n is given by
M n = hm, α, βin ' hmn , −nmn−1 β, −nmn−1 αi, when n is negative,
' hmn , nmn−1 α, nmn−1 βi, when n is positive.
Consider two TFN’s with a common mean value. Then subtraction produces a
TFN whose mean value is zero and the spreads are the sum of both the spreads of
computed TFN. The quotient of same TFNs is a TFN having mean value one. THE
Inverse of a TFN whose mean value is zero does not exist and we cannot divide by
such a number. The addition and multiplication of TFNs are both commutative
and associative. But the distributive law does not always hold.
For example, if A = h2, 0.5, 0.5i, B = h3, 0.8, 0.7i, C = h5, 1, 2i and D = h−5, 2, 1i,
then A.(B + C) = A.B + A.C holds but, A.(C + D) 6= A.C + A.D.
It may be remembered that
hm, α, βi.h0, 0, 0i = h0, 0, 0i.
78 A. K. Shyamal and M. Pal

3. Preliminaries and Definitions

Definition 3.1. Triangular fuzzy matrix (TFM). A triangular fuzzy matrix
of order m × n is defined as A = (aij )m×n , where aij = hmij , αij , βij i is the ijth
element of A, mij is the mean value of aij and αij , βij are the left and right spreads
of aij respectively.
As for classical matrices define the following operations on TFMs. Let A = (aij )
and B = (bij ) be two TFMs of same order. Then we have the following.
(i) A + B = (aij + bij )
(ii) A − B = (aij − bij ),
(iii) For A = (aij )m×n and B = (bij )n×p , A.B = (cij )m×p , where
Pn
cij = k=1 aik .bkj , i = 1, 2, · · · , m and j = 1, 2, · · · , p.
(iv) A0 = (aji ) (the transpose of A)
(v) k.A = (kaij ), where k is a scalar.
We now define special types of TFMs corresponding to special classical matrices.
However, because of fuzziness we will have more than on type of TFM corresponding
to one type of classical matrix.
Definition 3.2. Pure Null TFM. A TFM is said to be a pure null TFM if all
its entries are zero, i.e., all elements are h0, 0, 0i. This matrix is denoted by O.
Definition 3.3. Fuzzy Null TFM. A TFM is said to be a fuzzy null TFM if all
elements are of the form aij = h0, 1 , 2 i, where 1 · 2 6= 0.
Definition 3.4. Pure Unit TFM. A square TFM is said to be a pure unit TFM
if aii = h1, 0, 0i and aij = h0, 0, 0i, i 6= j, for all i, j. It is denoted by I.
Definition 3.5. Fuzzy Unit TFM. A square TFM is said to be a fuzzy unit
TFM if aii = h1, 1 , 2 i and aij = h0, 3 , 4 i for i 6= j for all i, j, where 1 · 2 6= 0,
3 · 4 6= 0.
Definition 3.6. Pure Triangular TFM. A square TFM A = (aij ) is said to be
a pure triangular TFM if either aij = h0, 0, 0i for all i > j or aij = h0, 0, 0i for all
i < j; i, j = 1, 2, · · · , n.
A pure triangular TFM A = (aij ) is said to be pure upper triangular TFM
when aij = h0, 0, 0i for all i > j and is said to be a pure lower triangular TFM if
aij = h0, 0, 0i for all i < j.
Definition 3.7. Fuzzy Triangular TFM. A square TFM A = (aij ) is said to be
a fuzzy triangular TFM if either aij = h0, 1 , 2 i for all i > j or aij = h0, 1 , 2 i for
all i < j; i, j = 1, 2, · · · , n and 1 · 2 6= 0.
Definition 3.8. Symmetric TFM. A square TFM A = (aij ) is said to be sym-
metric if A = A0 , i.e., if aij = aji for all i, j.
Definition 3.9. Pure Skew-symmetric TFM. A square TFM A = (aij ) is said
to be pure skew-symmetric if A = −A0 and aii = h0, 0, 0i, i.e., if aij = −aji for all
i, j and aii = h0, 0, 0i.
 Triangular Fuzzy Matrices 79

Definition 3.10. Fuzzy Skew-symmetric TFM. A square TFM A = (aij ) is

said to be fuzzy skew-symmetric if A = −A0 and aii = h0, 1 , 2 i, i.e., if aij = −aji
for all i, j and aii = h0, 1 , 2 i, 1 · 2 6= 0.

4. Basic Properties
In this section some basic properties of TFMs are presented. The commutative
and associative laws are valid for TFM under the operation ‘+’ only.
Property 4.1. For any three TFMs A, B and C of order m × n we have:
(i) A+B=B+A,
(ii) A+(B+C)=(A+B)+C,
(iii) A+A=2A,
(iv) A-A is a fuzzy null TFM ,
(v) A+O=A-O=A.
Property 4.2. Let A and B be two TFMs of the same order and k, l be two scalars.
Then:
(i) k(lA)=(kl)A,
(ii) k(A+B)=kA+kB,
(iii) (k+l)A=kA+lA, if k.l¿0,
(iv) k(A-B)=kA-kB.
Property 4.3. If A and B be two TFMs such that A + B and A.B are defined
then:
(i) (A0 )0 = A,
(ii) (A + B)0 = A0 + B 0 ,
0
(iii) (A.B) = B 0 .A0 .
Corollary 4.4. Let A and B be two TFMs and k, l be two scalars then:
(i) (k.A)0 = k.A0 ,
(ii) (k.A + l.B)0 = k.A0 + l.B 0 .
Property 4.5. Let A be a square TFM then
(i) A.A0 and A0 .A are both symmetric,
(ii) A + A0 is symmetric,
(iii) A − A0 is fuzzy skew-symmetric.

5. Trace of TFM
Definition 5.1. Trace of TFM. The trace of a square TFM A = (aij ),denoted
by
Pntr(A), is the sum of the principal diagonal elements.In other words,tr(A) =
i=1 aii .

Property 5.2. Let A = (aij ) and B = (bij ) be any two square TFMs of order
n × n then,
(i) tr(A) + tr(B) = tr(A + B)
80 A. K. Shyamal and M. Pal

0
(ii) tr(A) = tr(A )
(iii) tr(A.B) = tr(B.A).

Proof. (i) Let A = (aij )n×n and B = (bij )n×nPbe two TFMs where aij = hmij , αij , βij i
n
Pn bij = hnij ,P
and γij , δij i. Now, tr(A)
n Pn = i=1 aii = hM, TL , TU i, where M =
m
i=1 Pii , T L = i=1 Pα ii and T U = i=1 iiPSimilarly, tr(B) = hN, PL , PU i where
β .
n n n
N = n
i=1 ii , P L = γ
i=1 ii and PU = i=1 δii . Therefore, tr(A) + tr(B) =
hM, TL , TU i + hN, PL , PU i = hM + N, TL + PL , TU + PU i.

Again, let A + B = D = (dij ), where dij = hmdij , αdij , βdij i,

mdij = mij + nPij , αdij = P
αij + γij and βdij = βij + δij .
n n
Now, tr(D) = i=1 dii = i=1 hmij + nij , αij + γij , βij + δij i
= hM + N, TL + PL , TU + PU i.

Hence, tr(A) + tr(B) = tr(A + B).

(ii) the proof is trivial.
(iii) Let A = (aij ) and B = (bij ) be two TFMs of order n × n, where aij =
hm
Pnij , αij , βij i and bij = hnij ,P
γij , δij i. Also
Pn let PnC = A.B = (cij ), then cij =
n
k=1 aik .bkj . Now tr(A.B) = i=1 cii = i=1 k=1 aik .bki .
Pn
Again, let B.A = D = (dij ) where dij = k=1 bik .akj .
Pn Pn
Now tr(B.A)
Pn =Pni=1 k=1 bik .aki
= k=1 i=1 aik .bki (interchanging the indices i and k).
Hence, tr(A.B) = tr(B.A). 

Property 5.3. The product of two pure upper triangular TFMs of order n × n is
a pure upper triangular TFM.

Proof. Let A = (aij ) and B = (bij ) be two upper triangular TFMs where aij =
hmij , αij , βij i and bij = hnij , γij , δij i. Since A and B are pure upper triangular
TFMs then aij = h0, 0, 0i and P bij = h0, 0, 0i for
Pnall i > j; i, j = 1, 2, · · · , n. Let
n
A.B = C = (cij ), where cij = k=1 aik .bkj = k=1 hmik , αik , βik i.hnkj , γkj , δkj i.
We shall now show that cij = h0, 0, 0i if i > j; i, j = 1, 2, · · · , n.
For i > j we have aik = h0, 0, 0i for k = 1, 2, · · · , i−1 and similarly bkj = h0, 0, 0i
for k = i, i + 1, · · · , n.
Pn Pi−1 Pn
Therefore, cij = k=1 aik .bkj = k=1 aik .bkj + k=i aik .bkj = h0, 0, 0i.
Pn Pi−1 Pn
Now, cii = k=1 aik .bki = k=1 aik .bki + aii .bii + k=i+1 aik .bki
= h0, 0, 0i = aii .bii ,
since aik = h0, 0, 0i for k = 1, 2, · · · , i−1 and bki = h0, 0, 0i for k = i+1, i+2, · · · , n.
Hence the result follows. 

Property 5.4. The product of two pure lower triangular TFMs of order n × n is
also a pure lower triangular TFM.
 Triangular Fuzzy Matrices 81

6. Determinant of TFM
The triangular fuzzy determinant (TFD) of a TFM, minor and cofactor are
defined as in classical matrices. But, TFD has some special properties due to the
sub-distributive property of TFNs.
Definition 6.1. Determinant of TFM. The triangular fuzzy determinant of a
TFM A of order n × n is denoted by | A | or det(A) and is defined as,
X
|A| = Sgn σ hm1σ(1) , α1σ(1) , β1σ(1) i · · · hmnσ(n) , αnσ(n) , βnσ(n) i
σ∈Sn
X n
Y
(1) = Sgn σ aiσ(i) ,
σ∈Sn i=1

where aiσ(i) = hmiσ(i) , αiσ(i) , βiσ(i) i are TFNs and Sn denotes the symmetric group
of all permutations ofthe indices {1, 2, ..., n} and  Sgn σ = 1 or −1 according as
1 2 ... n
the permutation σ = is even or odd respectively.
σ(1) σ(2) . . . σ(n)
The computation of det(A) involves several product of TFNs. Since the prod-
uct of two or more TFNs is an approximate TFN, the value of det(A) is also an
approximate TFN.
Definition 6.2. Minor. Let A = (aij ) be a square TFM of order n × n. The
minor of an element aij in det(A) is a determinant of order (n − 1) × (n − 1), which
is obtained by deleting the ith row and the jth column from A and is denoted by
Mij .
Definition 6.3. Cofactor. Let A = (aij ) be a square TFM of order n × n.
The cofactor of an element aij in A is denoted by Aij and is defined as, Aij =
(−1)i+j Mij .
Definition 6.4. Adjoint. Let A = (aij ) be a square TFM and B = (Aij ) be a
square TFM whose elements are the cofactors of the corresponding elements in |A|
then the transpose of B is called the adjoint or adjugate of A and it is equal to
(Aji ). The adjoint of A is denoted by adj(A).
Here |A| contains n! terms out of which n2 ! are positive terms and the same
number of terms are negative. All these n! terms contain n quantities at a time in
product form, subject to the condition that from the n quantities in the product
exactly one is taken from each row and exactly one from each column. P
n
Alternatively, a TFD of a TFM A = (aij ) may be expanded in the form j=1 aij .
Aij , i ∈ {1, 2, ..., n}, where Aij is the cofactor of aij . Thus the TFD is the sum
of the products of the elements of any row (or column) and the cofactors of the
corresponding elements of the same row (or column). We refer to this method as
the alternative method.
In classical mathematics, the value of a determinant is computed by any one of
the aforesaid two processes and both yield same result. But, due to the failure of
distributive laws of triangular fuzzy numbers, the value of a TFD, computed by
82 A. K. Shyamal and M. Pal

the aforesaid two processes will differ from each other. For this reason the value of
a TFD should be determined according to the definition, i.e., using the following
rule only
X
|A|= Sgn σ hm1σ(1) , α1σ(1) , β1σ(1) ii · · · hmnσ(n) , αnσ(n) , βnσ(n) i.
σ∈Sn

On the other hand the value of a TFD computed by the alternative process yields
a different and less desirable result.
Property 6.5. Let A = (aij ) be a TFM of order n × n.

(i) If all the elements of a row (column) of A are h0, 0, 0i,

then | A |= h0, 0, 0i.
(ii) If a row (column) be multiplied by a scalar k,
then | A | is multiplied by k.
(iii) If A is triangular
Qn TFM,
then | A |= i=1 hmii , αii , βii i.
Proof. (i) Let A = (aij ) be a square TFM of order n × n where aij = hmij , αij , βij i.
We definePthe determinants PnE1 , E2 , . . . , En as follows:
n
Ei (A) = j=1 aij .Aij = j=1 hmij , αij , βij i.Aij ,
where Aij is the cofactor of aij in the determinant A. Obviously, E1 (A) = E2 (A) =
· · · = En (A) =| A |. Let all the elements of rth row, 1 ≤ r ≤ n, be h0, 0, 0i. Then
Er (A) = h0, 0, 0i since arj = h0, 0, 0i for all j = 1, 2, · · · , n.
Therefore, | A |= Er (A) = h0, 0, 0i.
(ii) If k = 0, then the result is obviously true since |A| = h0, 0, 0i when A has a zero
row.
Let B = (bij )n×n where bij = hnij , γij , δij i be obtained from an n × n TFM
A = (aij ) by multiplying its rth row by a scalar k 6= 0. Obviously hnij , γij , δij i =
hmij , αij , βij i, for all i 6= 0 and hnrj , γrj , δrj i = hkmrj , kαrj , kβrj i, when k is
positive and hnrj , γrj , δrj i = hkmrj , kβrj , kαrj i, when k is negative. Then, by
definition,
X
|B| = Sgn σ hn1σ(1) , γ1σ(1) , δ1σ(1) ihn2σ(2) , γ2σ(2) , δ2σ(2)i i
σ∈Sn
· · · hnrσ(r) , γrσ(r) , δrσ(r) i · · · hnnσ(n) , γnσ(n) , δnσ(n) i.
When k is positive scalar,
X
|B| = Sgn σ hm1σ(1) , α1σ(1) , β1σ(1) ihm2σ(2) , α2σ(2) , β2σ(2) i
σ∈Sn
· · · hkmrσ(r) , kαrσ(r) , kβrσ(r) i · · · hmnσ(n) , αnσ(n) , βnσ(n) i
X n
Y
= k Sgn σ hmiσ(i) , αiσ(i) , βiσ(i) i
σ∈Sn i=1
= k|A|.
 Triangular Fuzzy Matrices 83

When k is negative scalar,

X
|B| = Sgn σ hm1σ(1) , α1σ(1) , β1σ(1) ihm2σ(2) , α2σ(2) , β2σ(2) i
σ∈Sn
· · · hkmrσ(r) , kβrσ(r) , kαrσ(r) i · · · hmnσ(n) , αnσ(n) , βnσ(n) i
X
= k Sgn σ hm1σ(1) , α1σ(1) , β1σ(1) ihm2σ(2) , α2σ(2) , β2σ(2) i
σ∈Sn
· · · hmrσ(r) , αrσ(r) , βrσ(r) i · · · hmnσ(n) , αnσ(n) , βnσ(n) i
X Yn
= k Sgn σ hmiσ(i) , αiσ(i) , βiσ(i) i
σ∈Sn i=1
= k|A|.
Hence the result follows.
(iii) Let A = (aij )n×n be a square triangular TFM for i < j, i.e., aij = h0, 0, 0i for
i < j. Take a term t of | A |,
t = hm1σ(1) , α1σ(1) , β1σ(1) ihm2σ(2) , α2σ(2) , β2σ(2) i · · · hmnσ(n) , αnσ(n) , βnσ(n) i.
Let σ(1) 6= 1, i.e., 1 < σ(1) and so that, m1σ(1) = 0, α1σ(1) = 0 and β1σ(1) = 0.
Consequently, t = h0, 0, 0i. Now, let σ(1) = 1 but σ(2) 6= 2. Then it is obvious that
σ(2) > 2 therefore, m2σ(2) = 0, α2σ(2) = 0 and β2σ(2) = 0 and hence t = h0, 0, 0i.
This means that each term of |A| is h0, 0, 0i if σ(1) 6= 1 or σ(2) 6= 2. Briefly we can
say that miσ(i) = 0, αiσ(i) = 0 and βiσ(i) = 0 for σ(i) 6= i. Therefore,
| A |=Qhm 11 , α11 , β11 ihm22 , α22 , β22 i · · · hmnn , αnn , βnn i
n
= i=1 hmii , αii , βii i. 
Property 6.6. If any two rows (or columns) of a square TFM A are interchanged
then determinant |A| of A changes the sign of |A|.
Proof. Let A = (aij ) be a TFM of order n × n. If B = (bij )n×n is obtained
from A by interchanging the rth and sth row (r < s) of A,then it is clear that
bij = aij , i 6= r, i 6= s and brj = asj , bsj = arj .
Now,
X
|B| = Sgn σ b1σ(1) b2σ(2) · · · brσ(r) · · · bsσ(s) · · · bnσ(n)
σ∈Sn
X
= Sgn σ a1σ(1) a2σ(2) · · · asσ(r) · · · arσ(s) · · · anσ(n)
σ∈Sn
X
= Sgn σ hm1σ(1) , α1σ(1) , β1σ(1) ihmsσ(r) , αsσ(r) , βsσ(r) i
σ∈Sn
· · · hmrσ(s) , αrσ(s) , βrσ(s) i · · · hmnσ(n) , αnσ(n) , βnσ(n) i.
 
1 2 ... r ... s ... n
Let λ = .
1 2 ... s ... r ... n
Then λ is a transposition interchanging r and s and Sgn λ = −1.
Let σλ = φ. As σ runs through all permutations on {1, 2, · · · , n}, φ also runs over
the same permutations, because σ1 λ = σ2 λ or, σ1 = σ2 .
84 A. K. Shyamal and M. Pal

Now,
φ = σλ
 
1 2 ... r ... s ... n
=
σ(1) σ(2) ... σ(r) ... σ(s) ... σ(n)
 
1 2 ... r ... s ... n
×
1 2 ... s ... r ... n
 
1 2 ... r ... s ... n
= .
σ(1) σ(2) ... σ(s) ... σ(r) ... σ(n)

Therefore φ(i) = σ(i), i 6= r, i 6= s and φ(r) = σ(s), φ(s) = σ(r).

Since λ is odd, φ is even or odd according as φ is odd or even. Therefore,
Sgn φ = −Sgn σ.
Then,
X
|B| = Sgn σ hm1σ(1) , α1σ(1) , β1σ(1) ihm2σ(2) , α2σ(2) , β2σ(2) i
σ∈Sn
· · · hmsσ(r) , αsσ(r) , βsσ(r) i · · · hmrσ(s) , αrσ(s) , βrσ(s) i
· · · hmnσ(n) , αnσ(n) , βnσ(n) i
X
= − Sgn φ hm1φ(1) , α1φ(1) , β1φ(1) ihm2φ(2) , α2φ(2) , β2φ(2) i
φ∈Sn
· · · hmsφ(r) , αsφ(r) , βsφ(r) i · · · hmrφ(s) , αrφ(s) , βrφ(s) i
· · · hmnφ(n) , αnφ(n) , βnφ(n) i
= −|A|.
0

Property 6.7. If A is a square TFM then | A |=| A | .
0
Proof. Let A = (aij )n×n be a square TFM and A = B = (bij )n×n . Then,
X
|B| = Sgn σ b1σ(1) , b2σ(2) , · · · bnσ(n)
σ∈Sn
X
= Sgn σ aσ(1)1 , aσ(2)2 , · · · aσ(n)n .
σ∈Sn

Let φ be the permutation of {1, 2, . . . , n} such that φσ = I, the identity permu-

tation. Then φ = σ −1 . Since σ runs over the whole set of permutations, φ also runs
over the same set of permutations. Let σ(i) = j then i = σ −1 (j) and aσ(i)i = ajφ(j)
for all i, j. Therefore,
X
|B| = Sgn σ hmσ(1)1 , ασ(1)1 , βσ(1)1 ihmσ(2)2 , ασ(2)2 , βσ(2)2 i · · ·
σ∈Sn
hmσ(n)n , ασ(n)n , βσ(n)n i.
X
= Sgn φ hm1φ(1) , α1φ(1) , β1φ(1) ihm2φ(2) , α2φ(2) , β2φ(2) i · · ·
φ∈Sn
hmnφ(n) , αnφ(n) , βnφ(n) i
= |A|.
 Triangular Fuzzy Matrices 85

0
Hence, | A |=| A | 
Property 6.8. For a square TFM A of order n × n,

(i) if A is symmetric then adj(A) is symmetric,

(ii) if A is a fuzzy null TFM then adj(A) is a fuzzy null TFM,
(iii) if A is a pure unit TFM then adj(A) is a pure unit TFM,
(iv) if A is a fuzzy unit TFM then adj(A) is a fuzzy unit TFM.
Property 6.9. For a square TFM A of order n × n,

(i) if A contains a zero row then adj(A).A is a fuzzy null TFM.

0 0
(ii) adj(A ) = (adj(A)) .
Proof. (i) Let A = (aij ) be a square TFM of order n×n, where aij = hmij , αij , βij i,
and B = adj(A). Then by the definition of adjoint matrix, the ijth element bij of
B is |Aji |, where Aij is the sub-matrix obtained from A by suppressing the ith row
and jth column. That is, |Aij | is the cofactor of aij in A. Without loss of generality
we assume that the lth row of A be the zero row. Therefore, the elements of the
lth row are of the form alj = h0, αij , βij i, for all j. Then all the elements of adj(A)
∗ ∗
are of the form |Aij | = h0, αij , βij i except j 6= l.
Pn
P Let C = adj(A).A. Then the ijth element cij of C is cij = k=1 |Aik |akj =
k6=l |Aik |akj + |Ail |alj .
∗ ∗
Now, all |Aik |, k 6= l are of the form h0, αij , βij i and alj = h0, αij , βij i. Hence cij
is of the form h0, γij , δij i for all i, j = 1, 2, . . . , n. Thus, C, i.e., adj(A).A is a fuzzy
null TFM.
(ii) The proof is straight forward. 

7. Singular and Constant TFMs

Definition 7.1. Singular TFM. A square TFM A is said to be singular if |A| =
h0, 0, 0i.
Definition 7.2. Semi-singular TFM. A square TFM A is said to be semi-
singular if the value of |A| is of the form h0, α, βi.
Definition 7.3. Constant TFM. An n × n TFM A = (aij ) is said to be constant
if all its rows are equal to each other, i.e., if hmik , αik βik i = hmjk , αjk βjk i for all
i, j, k.
Property 7.4. If A = (aij ) is a constant TFM of order n × n and B is a TFM of
the same order, then A.B is a constant TFM.
Proof. Suppose A = (aij ) is a constant TFM of order n × n and B = (bij ) is
a TFM of the same order, where aij = hmij , αij , βij i and bij = hnij , γij , δij i.
Pn aik = hmik , αik , βik i, are the same for all i; i, k ∈ {1, 2, . . . , n}. Let A.B =
Then
k=1 aik .bkj = D = (dij ). ThenP
n Pn
hpij , µij , νij i, where pij = k=1 mik nkj , µij = k=1 {mik γkj + nik αkj } and
dij = P
n
νij = k=1 {mik δkj + nik βkj }. A is a constant TFM, i.e., mik = mjk for all i, j, k.
86 A. K. Shyamal and M. Pal

This implies that p11 = p21 = · · · = pn1 , p12 = p22 = · · · = pn2 , · · · , p1n = p2n =
· · · = pnn , i.e., pij is independent of i, i ∈ {1, 2, · · · , n}.
Again, µ11 = (m11 γ11 +n11 α11 )+(m12 γ21 +n21 α12 )+· · ·+(m1n γn1 +nn1 α1n ) =
µ21 = · · · = µn1 , µ12 = µ22 = · · · = µn2 , · · · , µ1n = µ2n = · · · = µnn . Similarly,
ν11 = ν21 = · · · = νn1 , ν12 = ν22 = · · · = νn2 , · · · , ν1n = ν2n = · · · = νnn . There-
fore, dij is independent of i, i ∈ {1, 2, · · · , n} since hmik , αik , βik i = hmjk , αjk , βjk i
for all i, j, k.
Hence, A.B is constant. 
Property 7.5. If A is a square constant TFM then,
(i) A.(adjA) is constant,
0
(ii) (adjA) is constant,
0
(iii) A.(adjA ) is constant,
0 0
(iv) (A .(adjA)) is constant,
0
(v) (adjA) .A is constant,
(vi) |A| is of the form h0, α, βi.

8. Conclusion
In this article some elementary operations on triangular fuzzy numbers are de-
fined. Like classical matrices we also define some operations on TFMs. Using the
elementary operations, some important properties of TFMs are presented. The
concept of adjoint of TFM is discussed and some properties on it are also pre-
sented. The definition and some properties of determinant of TFM are presented
in this article. It is well known that the determinant is a very important tool in
mathematics, so an efficient method is required to evaluate a TFD. Presently, we
are trying to develop an efficient method to evaluate a TFD of large size. Some
special types of TFMs, i.e. pure and fuzzy triangular, symmetric, pure and fuzzy
skew-symmetric, singular, semi-singular and constant TFMs are defined here. Some
properties of these TFMs are also presented.

References
[1] D. Dubois and H. Prade, Fuzzy Sets and Systems: Theory and Applications, Academic Press
London, 1980.
[2] H. Hashimoto, Convergence of powers of a fuzzy transitive matrix, Fuzzy Sets and Systems,
9 (1983), 153-160.
[3] H. Hashimoto, Canonical form of a transitive fuzzy matrix, Fuzzy Sets and Systems, 11
(1983), 157-162.
[4] A. Kandel, Fuzzy mathematical Techniques with Applications, Addison-Wesley, Tokyo, 1996.
[5] K. H. Kim and F. W. Roush, Generalised fuzzy matrices, Fuzzy Sets and Systems, 4 (1980),
293-315.
[6] J. B. Kim, Determinant theory for fuzzy and boolean matrices, Congressus Numerantium,
(1988), 273-276.
[7] W. Kolodziejczyk, Canonical form of a strongly transitive fuzzy matrix, Fuzzy Sets and
Systems, 22 (1987), 292-302.
[8] W. Kolodziejczyk, Convergence of powers of s-transitive fuzzy matrices, Fuzzy Sets and
Systems, 26 (1988), 127-130.
 Triangular Fuzzy Matrices 87

[9] M. Pal, Intuitionistic fuzzy determinant, V.U.J.Physical Sciences, 7 (2001), 87-93.

[10] M. Pal, S. K. Khan and A. K. Shyamal, Intuitionistic fuzzy matrices, Notes on Intuitionistic
Fuzzy Sets, 8(2) (2002), 51-62.
[11] M. Z. Ragab and E. G. Emam, The determinant and adjoint of a square fuzzy matrix, Fuzzy
Sets and Systems, 61 (1994), 297-307.
[12] M. Z. Ragab and E. G. Emam, On the min-max composition of fuzzy matrices, Fuzzy Sets
and Systems, 75 (1995), 83-92.
[13] A. K. Shyamal and M. Pal, Two new operators on fuzzy matrices, J. Applied Mathematics
and Computing, 15 (2004), 91-107.
[14] A. K. Shyamal and M. Pal, Distance between fuzzy matrices and its applications, Acta Siencia
Indica, XXXI.M (1) (2005), 199-204.
[15] A. K. Shyamal and M. Pal, Distance between fuzzy matrices and its applications-I, J. Natural
and Physical Sciences, 19(1)(2005), 39-58.
[16] A. K. Shyamal and M. Pal, Distances between intuitionistics fuzzy matrices, V. U. J. Physical
Sciences, 8 (2002), 81-91.
[17] M. G. Thomason,Convergence of powers of a fuzzy matrix, J. Math Anal. Appl., 57 (1977),
476-480.
[18] L. J. Xin,Controllable fuzzy matrices, Fuzzy Sets and Systems, 45 (1992), 313-319.
[19] L. J. Xin,Convergence of powers of controllable fuzzy matrices, Fuzzy Sets and Systems, 62
(1994), 83-88.
[20] L. A. Zadeh, Fuzzy Sets, Information and Control, 8 (1965), 338-353.

Amiya Kumar Shyamal and Madhumangal Pal∗ , Department of Applied Mathemat-

ics with Oceanology and Computer Programming, Vidyasagar University, Midnapore -
721102, West Bengal, India
E-mail address: [email protected]

∗Corresponding author