[Link]., [Link].

, 16 (2006), Supplement
Proceedings of ICMI 45, Bacau, Sept.18-20, 2006, pp. 263-268

Properties of the Generalized Permutations

N. Shajareh-Poursalavati

Abstract

In this article, we study some properties of the generalized permutations. We

introduce a fundamental structure of all generalized permutations and by use this
concept, we introduce another generalization of permutation matrices, also we obtain
some properties of generalized permutation matrices.
Key words and phrases: generalized permutation, directed graph, semi-
inverse, generalized permutation matrices.
(2000) Mathematics Subject Classification: 20B99

1. INTRODUCTION
Suppose that X is a non empty set and let P * ( X ) be the set of all non empty
subset of X . A generalized permutation, which is a generalization of the
concept of ordinary permutation, is defined in [2], as follows. The function
σ : X → P * ( X ) is called a generalized permutation on X , if U σ ( x) = X .
x∈ X

Obviously, if we consider the singleton { f (x)} instead f (x) , for every

ordinary permutation f on X , then f is a generalized permutation. The set
of all generalized permutation on X is denoted by M X . For f1 , f 2 ∈ M X , we
say that f1 is a sub generalized permutation of f 2 , or f 2 contains f 1 , and
write f1 ⊆ f 2 , if f1 ( x) ⊆ f 2 ( x) for every x ∈ X . The mapping g for which
g ( x) = X for all x ∈ X , is called the universal generalized permutation and
contains all the element of M X . Every map f : X → P * ( X ) which contains a
generalized permutation is itself a generalized permutation. The map
i : X → P * ( X ) where i ( x) = {x} , for all x ∈ X , is called the identity
generalized permutation. The composition of two generalized permutations
f1 and f 2 is defined as follows:

263
 f1 o f 2 ( x ) = U f1 ( y ) ∀x ∈ X .
y∈ f 2 ( x )

It is easy to show that f1 o f 2 is a generalized permutation. By this

composition, we have a binary operation on the set of all generalized
permutation, i.e., f1 o f 2 ∈ M Ω .
Definition 1.1. Let f be a generalized permutation on X . The semi-inverse
of f is denoted by f ~1 , define as f ~1 ( y ) = {x ∈ X : y ∈ f ( x)} . It is easy to
show that identity element of M X , is a sub generalized permutation of
f ~1 o f and f o f ~1 , i.e., x ∈ f ~1 o f ( x) I f o f ~1 ( x) for all ∀x ∈ X .

2. GRAPH OF A GENERALIZED PERMUTATION

Let f be a generalized permutation on X . We consider X , the set of
vertices and for two vertices "x" and "y", we define a directed edge (arc) from
"x" to "y", if y ∈ f (x) . The graph of a generalized permutation is denoted by
G(f). The number of the input or output arcs to a vertex α is called input or
output degree in α and denoted by id (α ) or od (α ) , respectively. In [1], we
prove that, there is a one to one correspondence between the set of all
generalized permutation on the non empty set X and the set of all directed
graphs with vertices in X , where input and output degrees in each vertex be
positive. Also we introduced a presentation of a generalized permutation f
over a finite set X of order n. Without loss of generality, we consider X as
the set {1,2, K, n} .
Definition 2.1. Let x be a vertex of G(f) and k be the smallest positive integer
numbers such that x ∈ f k (x) . The k-cycle ( x 0 x1 x 2 K x k −1 ) where x = x 0 ,
xi ∈ f ( xi −1 ), i = 1,2,K, k − 1 and x ∈ f ( x k −1 ) , is called a closed orbit of length
k.
Note that every loop is a 1-cycle or closed orbit of length 1.
Definition 2.2. Let x be a vertex of G(f) that lies in a closed orbit or loop. The
k-symbol order [ x0 x1 x 2 K x k −1 ] where k is a positive integer number,
x = x 0 , xi ∈ f ( xi −1 ), for all i = 1,2,K , k − 1 and x k −1 lies in a closed orbit, and
for all i = 1,2, K, k − 2 , xi does not lie in any closed orbit of G(f), is called a
pure chain of length k-1.

264
The set of all strongly-connected component’s generalized permutations on a
set X is denoted by SM X . In [1], we show that, every elements of SM X is a
collection of ordinary cycle permutations over X . Also, we show that, every
elements of M X is a collection of closed orbits and pure chains over X .
Example 2.3. Let X = {1,2,3,4,5,6} and let f be generalized permutation
defined by:
 1 2 3 4 5 6 7
f :=  .
 {1,2} {3} {1,4} {5} {6} {7} {5}
We present f as the collection {(1), (1 2 3 ), ( 5 6 7 ), [ 3 4 5] } where (1) ,
( 1 2 3 ) , (5 6 7) are closed orbits of length 1, 3 and 3 respectively and
[3 4 5] is a pure chain of length 3.
In [1], we obtain an upper bound for SM X and M X . If X be a non empty set
of cardinal n, and
n(n − 1) n(n − 1)(n − 2)
m=n+ + + L + (n − 1)! , and
2 3
l = n + n(n − 1) + n(n − 1)(n − 2) + L + n! ,
then SM X ≤ 2 m and M X ≤ 2 m +l .
In the next section we define a new generalization of permutation matrices.

3. GENERALIZED PERMUTATION MATRICES

Let R be a field with identity 1R and characteristic of R be zero. By
definition of generalized permutation, we can define another generalization of
permutation matrix as follows:
Definition 3.1. Let X = {1,2, K , n} and σ be a generalized permutation on
X . The generalized permutation matrix Pσ with n elements is defined as:
 eσ (1) 
 
 eσ ( 2 ) 
Pσ =  ,
M 
 
e 
 σ ( n )  n×n

265
with for non empty subset S of X , e S := ( x1 x 2 L x n )1×n , where
1 if i ∈ S
xi :=  R .
0 if i ∉ S
In other words, a generalized permutation matrix is an square matrix which its
arrays are 0 or 1R , and for each column and row there are at least one array
non zero.
Example 3.2. Let X = {1,2,3,4,5,6} and let σ be a generalized permutation
defined as follows:
 1 2 3 4 5 6 
σ =  
{2,3} {3} {4} {5} {3} {1,6}
i.e., σ (1) = {2,3}, σ (2) = {3}, σ (3) = {4}, σ (4) = {5}, σ (5) = {3}, σ (6) = {1,6}.
Then
0 1 1 0 0 0
 
0 0 1 0 0 0
0 0 0 1 0 0
Pσ :=  
0 0 0 0 1 0
0 0 1 0 0 0 

1 0 0 0 0 1  6×6

The transpose of a matrix A = (aij )n×m is denoted by AT := (b ji )m×n , which
b ji = aij for all 1 ≤ i ≤ n and 1 ≤ j ≤ m .

Theorem 3.3. If α = (α 1 α 2 L α n ) is a vector in R n and Pσ is a

T

generalized permutation matrix over {1,2, K , n} then

T
 
Pσ α =  ∑ α i , ∑ α i , K , ∑ α i  .
 i∈σ (1) i∈σ ( 2) i∈σ ( n ) 

Proof. It is easy to show that ( Pσ α ) i1 = eσ ( i )α = ∑α

j∈σ ( i )
j for all 1 ≤ i ≤ n .

□
Theorem 3.4. If σ and π be two generalized permutation over {1,2, K , n} ,
then for all 1 ≤ i, j ≤ n , ( Pσ Pπ ) ij = 0 if and only if ( Pπ oσ ) ij = 0 .

266
Proof. Note that ( Pπ oσ ) ij = 0 if and only if j ∉ π o σ (i ) . Now we calculate
the ( Pσ Pπ ) ij , we have:
n
( Pσ Pπ ) ij = ∑ (eσ ( i ) ) k (eπ ( k ) ) j
k =1

= | {k ∈ σ (i ) : j ∈ π (k )} |
= | {k ∈ σ (i ) : k ∈ π ~1 ( j )} |=| σ (i ) I π ~1 ( j ) | .
Hence,
( Pσ Pπ ) ij = 0 ⇔ σ (i ) I π ~1 ( j ) = { }
⇔ {k ∈ σ (i ) : j ∈ π (k )} = { }
⇔ j ∉ π o σ (i )
⇔ ( Pπ oσ ) ij = 0
□
Remark 3.5. In general, Pσ Pπ is not a generalized permutation matrix. As an
example, which Pσ Pπ not be generalized permutation matrix, consider
 1 2 1 2 
σ :=   and π :=   generalized permutations on the set
 {1,2}{2}  {1} {1,2}
1 1  1 0  2 ⋅ 1R 1R 
{1, 2}. Then Pσ =  R R  , Pπ =  R  and Pσ Pπ =   not a
 0 1R   1R 1 R   1R 1R 
generalized permutation matrix. In each case, if Pσ Pπ is a generalized
permutation matrix, in Theorem 3.6, we show that Pσ Pπ = Pπ oσ .
In [1], we defined the directed graph and in Definition 1.1 the semi-inverse of
a generalized permutation was introduced. Hence, naturally, we can define the
semi-inverse of a generalized permutation matrix, i.e., Pσ~1 := Pσ ~1 . But in the
following theorem we show that this is not a new definition.
Theorem 3.6. Let σ be a generalized permutation. Then Pσ~1 := Pσ ~1 = PσT .
Proof. We know that the arrays of a generalized permutation matrix, are 0 or
1R . Now we have:

267
 (Pσ )ij = 1R ⇔ (eσ (i ) )1 j = 1R ⇔ j ∈ σ (i )

(
⇔ i ∈ σ ~1 ( j ) ⇔ eσ ~1 ( j ) )
1i
= 1R ⇔ (Pσ ~1 ) ji = 1R .

Therefore (Pσ )ij = (Pσ ~1 ) ji for all i and j. This completes proof.

□
Theorem 3.7. The generalized permutation matrices Pσ oσ ~1 and Pσ ~1 oσ are
symmetric matrices which the diagonal arrays are 1R .
Proof. For every 1 ≤ i, j ≤ n , ( Pσ oσ ~1 ) ij = (eσ oσ ~1 ( i ) ) j . Now by use the graph of
σ and semi-inverse of σ we show that j ∈ σ o σ ~1 (i ) ⇔ i ∈ σ o σ ~1 ( j ) . Let
j ∈ σ o σ ~1 (i ) , hence there is a k in σ ~1 (i ) such that j ∈ σ (k ) and i ∈ σ (k ) ,
then i ∈ σ o σ ~1 ( j ) . Therefore ( Pσ oσ ) ij = ( Pσ oσ ) ji and similarly
~1 ~1

( Pσ ~1 oσ ) ij = ( Pσ ~1 oσ ) ji . It is clearly that i ∈ σ o σ ~1 (i ) and i ∈ σ ~1 o σ (i ) ,

therefore ( Pσ ~1 oσ ) ii = 1R = ( Pσ ~1 oσ ) ii and the proof complet.
□

REFERENCES
[1] Shajareh-Poursalavati, N. On the generalized permutations. Journal of
Discrete Mathematical Sciences & Cryptography, 8, 3 (Dec. 2005), 365-
672.
[2] Vougiouklis, T. Hyperstructures and their representations. Hadronic Press
Monographs in Mathematics, Palm Harbor, FL, 1994.

Department of Mathematics, Shahid Bahonar University of Kerman

mail address: salavati@[Link]

268