We propose two more general methods to construct nullnorms on bounded lattices. By some illustrative examples, we demonstrate that the new method differ from the existing approaches.
The notions of triangular norms (t-norms for short) and triangular conorms (t-conorms for short) were introduced by Schweizer and Sklar [
Some constructions of nullnorms on bounded lattices were demonstrated in previously papers. Based on the existence of t-norms and t-conorms on an arbitrary bounded lattice, Karaçal et al. [
In this paper, we propose two more general construction methods of nullnorms on an arbitrary bounded lattice. The present study is organized as follows: In Section 2, we recall some basic concepts and show some existing constructions of nullnorms on an arbitrary bounded lattice. In Section 3, we introduce the notions of t-subnorm and t-subcnonorm. By using these operations, we propose new methods to obtain nullnorms on L under some additional constraints and their characteristics are examined. Finally, this summarization can be found in Section 4.
A lattice is a partially ordered set ( L , ≤ ) in which each two-element subset { x , y } has an infimum, denoted as x ∧ y , and a supremum, denoted as x ∨ y . A bounded lattice ( L , ≤ , 0 , 1 ) is a lattice that has the bottom and top elements written as 0 and 1, respectively. We denote ( L , ≤ , 0 , 1 ) simply by L in this article.
Let ( L , ≤ , 0 , 1 ) be a bounded lattice and V 1 , V 2 : L 2 → L be two binary operations on L, we can define a partial order:
V 1 ≤ V 2 ⇔ V 1 ( x , y ) ≤ V 2 ( x , y ) for all x , y ∈ L .
Given a bounded lattice ( L , ≤ , 0 , 1 ) and a , b ∈ L , a ≤ b , a subset [ a , b ] of L is defined as [ a , b ] = { x ∈ L | a ≤ x ≤ b } . Similarly, denote [ a , b ) = { a ≤ x < b } , ( a , b ] = { x ∈ L | a < x ≤ b } and ( a , b ) = { x ∈ L | a < x < b } . If a and b are incomparable, we use the notation a ∥ b . The set of all elements which are incomparable with a are denoted by I a .
Definition 2.1. ( [
Definition 2.2. ( [
Definition 2.3. ( [
Definition 2.4. ( [
Proposition 2.5. ( [
T ( x , y ) = ( F 1 ( x , y ) , if ( x , y ) ∈ ( L \ { 1 } ) 2 x ∧ y , otherwise (1)
is a t-norm on L.
Dually, if R 1 : L 2 → L is a t-subconorm on a bounded lattice L, then S : L 2 → L defined by
S ( x , y ) = ( R 1 ( x , y ) , if ( x , y ) ∈ ( L \ { 0 } ) 2 x ∨ y , otherwise (2)
is a t-conorm on L.
Definition 2.6. ( [
It is easy to see that V ( x , a ) = a for all x ∈ L , and thus a is the zero element for V [
Proposition 2.7. ( [
(i) V | [ 0, a ] 2 : [ 0, a ] 2 → [ 0, a ] is a t-conorm on [ 0, a ] ;
(ii) V | [ a ,1 ] 2 : [ a ,1 ] 2 → [ a ,1 ] is a t-norm on [ a ,1 ] .
Let ( L , ≤ , 0 , 1 ) be a bounded lattice and a ∈ L \ { 0,1 } . Let T : [ a ,1 ] 2 → [ a ,1 ] be a t-norm on [ a ,1 ] and S : [ 0, a ] 2 → [ 0, a ] be a t-conorm on [ 0, a ] . Based on the knowledge of the existence of t-norms and t-conorms on an arbitrary given bounded lattice, many construction methods of nullnorms were presented in previous papers. Generally speaking, these construction methods on an arbitrary bounded lattice under no additional constraints can be divided into two groups. One is V a ( T , S ) ( x , y ) : L 2 → L proposed by Karaçal et al. in [
V a ( T , S ) ( x , y ) = ( S ( x , y ) , if ( x , y ) ∈ [ 0, a ] 2 T ( x , y ) , if ( x , y ) ∈ [ a ,1 ] 2 a , otherwise . (3)
The structures of V a ( T , S ) is shown in
The other group is V T S and its dual, i.e., V S T : L 2 → L , which are proposed by Ümit Ertuğrul [
V T S ( x , y ) = ( S ( x , y ) , if ( x , y ) ∈ [ 0, a ] 2 T ( x , y ) , if ( x , y ) ∈ [ a ,1 ] 2 S ( x ∧ a , y ∧ a ) , if ( x , y ) ∈ [ 0, a ] × I a ∪ I a × [ 0, a ] ∪ I a × I a a , otherwise (4)
and
V S T ( x , y ) = ( S ( x , y ) , if ( x , y ) ∈ [ 0, a ] 2 T ( x , y ) , if ( x , y ) ∈ [ a ,1 ] 2 T ( x ∨ a , y ∨ a ) , if ( x , y ) ∈ [ a ,1 ] × I a ∪ I a × [ a ,1 ] ∪ I a × I a a , otherwise . (5)
The structures of V T S and V S T are shown in
In order to reduce the complexity in the proof of associativity, we introduce the following proposition.
Proposition 3.1. ( [
1) H ( H ( x , y ) , z ) = H ( x , H ( y , z ) ) for all ( x , y , z ) ∈ ( A , A , A ) ∪ ( B , B , B ) ∪ ( C , C , C ) ∪ ( A , A , B ) ∪ ( A , B , B ) ∪ ( A , A , C ) ∪ ( A , C , C ) ∪ ( B , B , C ) ∪ ( B , C , C ) ;
2) H ( H ( x , y ) , z ) = H ( x , H ( y , z ) ) = H ( H ( x , z ) , y ) for all ( x , y , z ) ∈ ( A , B , C ) .
Now, we introduce two construction methods which can be regard as generalizations of existing methods.
Theorem 3.2. Let ( L , ≤ , 0 , 1 ) be a bounded lattice and a ∈ L \ { 0,1 } . Let
T : [ a ,1 ] 2 → [ a ,1 ] be a t-norm on [ a ,1 ] , S : [ 0, a ] 2 → [ 0, a ] be a t-conorm on [ 0, a ] and R : [ 0, a ] 2 → [ 0, a ] be a t-subconorm on [ 0, a ] . If S ≤ R and
S ( x , R ( y , z ) ) = R ( R ( x , y ) , z ) = R ( S ( x , y ) , z ) forall x , y , z ∈ [ 0, a ] , (6)
then V T S , R : L 2 → L is a nullnorm on L with the zero element a, where
V T S , R ( x , y ) = ( S ( x , y ) , if ( x , y ) ∈ [ 0, a ] 2 T ( x , y ) , if ( x , y ) ∈ [ a ,1 ] 2 R ( x ∧ a , y ∧ a ) , if ( x , y ) ∈ [ 0, a ] × I a ∪ I a × [ 0, a ] ∪ I a × I a a , otherwise . (7)
Proof. The commutativity of V T S , R can be proven directly based on its description. Similarly, we can express V T S , R ( x , 0 ) = S ( x , 0 ) = x for all x ∈ [ 0, a ] and V T S , R ( x , 1 ) = T ( x , 1 ) = x for all x ∈ [ a ,1 ] .
Monotonicity: Let us prove that if x ≤ y , then V T S , R ( x , z ) ≤ V T S , R ( y , z ) for all z ∈ L . If x , y ∈ [ 0, a ] , or x , y ∈ I a , or x , y ∈ ( a ,1 ] , then it is clear that U ( x , z ) ≤ U ( y , z ) because ( x , z ) and ( y , z ) are in the same piece of U and U is monotonic in each piece. Moreover, ( x , y ) ∈ ( a ,1 ] × [ 0, a ] ∪ I a × [ 0, a ] ∪ I a × ( a ,1 ] contradicts the assumption that x ≤ y . Therefore, there are only three cases left to consider, namely, ( x , y ) ∈ [ 0, a ] × ( a ,1 ] , ( x , y ) ∈ [ 0, a ] × I a , and ( x , y ) ∈ I a × ( a ,1 ] .
(I) Assume that x ∈ [ 0, a ] and y ∈ ( a ,1 ] .
(i) If z ∈ [ 0, a ] , then V T S , R ( x , z ) = S ( x , z ) and V T S , R ( y , z ) = a . As S ( x , z ) ≤ a , we have V T S , R ( x , z ) ≤ V T S , R ( y , z ) .
(ii) If z ∈ ( a ,1 ] , then V T S , R ( x , z ) = a and V T S , R ( y , z ) = T ( y , z ) . As a ≤ T ( y , z ) , we have V T S , R ( x , z ) ≤ V T S , R ( y , z ) .
(iii) If z ∈ I a , then V T S , R ( x , z ) = R ( x ∧ a , z ∧ a ) and V T S , R ( y , z ) = a . As R ( x ∧ a , z ∧ a ) ≤ a , we have V T S , R ( x , z ) ≤ V T S , R ( y , z ) .
Therefore, V T S , R ( x , z ) ≤ V T S , R ( y , z ) holds for ( x , y ) ∈ [ 0, a ] × [ a ,1 ] .
(II) Assume that x ∈ [ 0, a ] and y ∈ I a such that x ≤ y .
(i) If z ∈ [ 0, a ] , then V T S , R ( x , z ) = S ( x , z ) and V T S , R ( y , z ) = R ( y ∧ a , z ∧ a ) . As S ( x , z ) = S ( x ∧ a , z ∧ a ) ≤ R ( x ∧ a , z ∧ a ) ≤ R ( y ∧ a , z ∧ a ) , we have V T S , R ( x , z ) ≤ V T S , R ( y , z ) .
(ii) If z ∈ [ a ,1 ] , then V T S , R ( x , z ) = a and V T S , R ( y , z ) = a , and thus V T S , R ( x , z ) ≤ V T S , R ( y , z ) .
(iii) If z ∈ I a , then V T S , R ( x , z ) = R ( x ∧ a , z ∧ a ) and V T S , R ( y , z ) = R ( y ∧ a , z ∧ a ) . As R ( x ∧ a , z ∧ a ) ≤ R ( y ∧ a , z ∧ a ) , we have V T S , R ( x , z ) ≤ V T S , R ( y , z ) .
Therefore, V T S , R ( x , z ) ≤ V T S , R ( y , z ) holds for ( x , y ) ∈ [ 0, a ] × I a .
(III) Assume that x ∈ I a and y ∈ ( a ,1 ] such that x ≤ y .
(i) If z ∈ [ 0, a ] , then V T S , R ( x , z ) = R ( x ∧ a , z ∧ a ) and V T S , R ( y , z ) = a . As R ( x ∧ a , z ∧ a ) ≤ a , we have V T S , R ( x , z ) ≤ V T S , R ( y , z ) .
(ii) If z ∈ [ a ,1 ] , then V T S , R ( x , z ) = a and V T S , R ( y , z ) = T ( y , z ) . As a ≤ T ( y , z ) , we have V T S , R ( x , z ) ≤ V T S , R ( y , z ) .
(iii) If z ∈ I a , then V T S , R ( x , z ) = R ( x ∧ a , z ∧ a ) and V T S , R ( y , z ) = a . As R ( x ∧ a , z ∧ a ) ≤ a , we have V T S , R ( x , z ) ≤ V T S , R ( y , z ) .
Therefore, V T S , R ( x , z ) ≤ V T S , R ( y , z ) holds for ( x , y ) ∈ I a × ( a ,1 ] .
Combining the above cases, we obtain that V T S , R ( x , z ) ≤ V T S , R ( y , z ) holds for x , y , z ∈ L such that x ≤ y . Therefore, V T S , R is monotonic.
Associativity: It can be shown that V T S , R ( x , V T S , R ( y , z ) ) = V T S , R ( V T S , R ( x , y ) , z ) for all x , y , z ∈ L . By Proposition 3.1, We only need to consider the following cases:
(i) If x , y , z ∈ [ 0, a ] , then sinceS is associative, we have V T S , R ( V T S , R ( x , y ) , z ) = V T S , R ( x , V T S , R ( y , z ) ) .
(ii) If x , y , z ∈ [ a ,1 ] , then since T is associative, we have V T S , R ( V T S , R ( x , y ) , z ) = V T S , R ( x , V T S , R ( y , z ) ) .
(iii) If x , y , z ∈ I a , then V T S , R ( V T S , R ( x , y ) , z ) = V T S , R ( R ( x ∧ a , y ∧ a ) , z ) = R ( R ( x ∧ a , y ∧ a ) , z ∧ a ) , V T S , R ( x , V T S , R ( y , z ) ) = V T S , R ( x , R ( y ∧ a , z ∧ a ) ) = R ( x ∧ a , R ( y ∧ a , z ∧ a ) ) . As R is an associative function on [ 0, a ] , we have V T S , R ( V T S , R ( x , y ) , z ) = V T S , R ( x , V T S , R ( y , z ) ) .
(iv) If x , y ∈ [ 0, a ] and z ∈ [ a ,1 ] , then V T S , R ( V T S , R ( x , y ) , z ) = V T S , R ( S ( x , y ) , z ) = a and V T S , R ( x , V T S , R ( y , z ) ) = V T S , R ( x , a ) = a , and thus V T S , R ( V T S , R ( x , y ) , z ) = V T S , R ( x , V T S , R ( y , z ) ) .
(v) If x ∈ [ 0, a ] and y , z ∈ [ a ,1 ] , then V T S , R ( V T S , R ( x , y ) , z ) = a and V T S , R ( x , V T S , R ( y , z ) ) = V T S , R ( x , T ( y , z ) ) = a . Thus V T S , R ( V T S , R ( x , y ) , z ) = V T S , R ( x , V T S , R ( y , z ) ) .
(vi) If x , y ∈ [ 0, a ] and z ∈ I a , then V T S , R ( V T S , R ( x , y ) , z ) = V T S , R ( S ( x , y ) , z ) = R ( S ( x , y ) , z ∧ a ) and V T S , R ( x , V T S , R ( y , z ) ) = V T S , R ( x , R ( y ∧ a , z ∧ a ) ) = S ( x , R ( y ∧ a , z ∧ a ) ) . It follows from (6) that V T S , R ( V T S , R ( x , y ) , z ) = V T S , R ( x , V T S , R ( y , z ) ) .
(vii) If x ∈ [ 0, a ] and y , z ∈ I a , then V T S , R ( V T S , R ( x , y ) , z ) = V T S , R ( R ( x ∧ a , y ∧ a ) , z ) = R ( R ( x ∧ a , y ∧ a ) , z ∧ a ) and V T S , R ( x , V T S , R ( y , z ) ) = V T S , R ( x , R ( y ∧ a , z ∧ a ) ) = S ( x , R ( y ∧ a , z ∧ a ) ) . It follows from (6) that V T S , R ( V T S , R ( x , y ) , z ) = V T S , R ( x , V T S , R ( y , z ) ) .
(viii) If x , y ∈ [ a ,1 ] and z ∈ I a , then V T S , R ( V T S , R ( x , y ) , z ) = V T S , R ( T ( x , y ) , z ) = a and V T S , R ( x , V T S , R ( y , z ) ) = a . Thus V T S , R ( V T S , R ( x , y ) , z ) = V T S , R ( x , V T S , R ( y , z ) ) .
(ix) If x ∈ [ a ,1 ] and y , z ∈ I a , then V T S , R ( V T S , R ( x , y ) , z ) = a and V T S , R ( x , V T S , R ( y , z ) ) = V T S , R ( x , R ( y ∧ a , z ∧ a ) ) = a . Thus V T S , R ( V T S , R ( x , y ) , z ) = V T S , R ( x , V T S , R ( y , z ) ) .
(x) If x ∈ [ 0, a ] , y ∈ ( a ,1 ] , z ∈ I a , then V T S , R ( V T S , R ( x , y ) , z ) = a , V T S , R ( x , V T S , R ( y , z ) ) = a and V T S , R ( V T S , R ( x , z ) , y ) = V T S , R ( R ( x ∧ a , z ∧ a ) , y ) = a . Thus V T S , R ( V T S , R ( x , y ) , z ) = V T S , R ( x , V T S , R ( y , z ) ) = V T S , R ( V T S , R ( x , z ) , y ) .
From (i) to (x), we obtain that V T S , R ( x , V T S , R ( y , z ) ) = V T S , R ( V T S , R ( x , y ) , z ) for all x , y , z ∈ L by Proposition 3.1. Therefore, V T S , R is a nullnorm on L with the zero element a.¨
Theorem 3.3. Let ( L , ≤ , 0 , 1 ) be a bounded lattice and a ∈ L \ { 0,1 } . Let T : [ a ,1 ] 2 → [ a ,1 ] be a t-norm on [ a ,1 ] , F : [ a ,1 ] 2 → [ a ,1 ] be a t-subnorm on [ a ,1 ] and S : [ 0, a ] 2 → [ 0, a ] be a t-conorm on [ 0, a ] . If F ≤ T and T ( x , F ( y , z ) ) = F ( F ( x , y ) , z ) = F ( T ( x , y ) , z ) for all x , y , z ∈ L , then V S T , F : L 2 → L is a nullnorm on L with the zero element a, where
V S T , F ( x , y ) = ( S ( x , y ) , if ( x , y ) ∈ [ 0, a ] 2 T ( x , y ) , if ( x , y ) ∈ [ a ,1 ] 2 F ( x ∨ a , y ∨ a ) , if ( x , y ) ∈ [ a ,1 ] × I a ∪ I a × [ a ,1 ] ∪ I a × I a a , otherwise . (8)
Proof. This can be proved similarly as Theorem 3.2.¨
The structures of V T S , R and V S T , F from Formula (7) and Formula (8) are shown in
Let ( L , ≤ , 0 , 1 ) be a bounded lattice and a ∈ L \ { 0,1 } . Let T : [ a ,1 ] 2 → [ a ,1 ] be a t-norm on [ a ,1 ] , S : [ 0, a ] 2 → [ 0, a ] be a t-conorm on [ 0, a ] . Taking R ( x , y ) = S ( x , y ) in Formula (7), we obtain that
V T S , S ( x , y ) = ( S ( x , y ) , if ( x , y ) ∈ [ 0, a ] 2 T ( x , y ) , if ( x , y ) ∈ [ a ,1 ] 2 S ( x ∧ a , y ∧ a ) , if ( x , y ) ∈ [ 0, a ] × I a ∪ I a × [ 0, a ] ∪ I a × I a a , otherwise, (9)
which is equal to V T S ( x , y ) given by Formula (4).
Dually, taking F ( x , y ) = T ( x , y ) in Formula (8), we obtain that
V S T , T ( x , y ) = ( S ( x , y ) , if ( x , y ) ∈ [ 0, a ] 2 T ( x , y ) , if ( x , y ) ∈ [ a ,1 ] 2 T ( x ∨ a , y ∨ a ) , if ( x , y ) ∈ [ a ,1 ] × I a ∪ I a × [ a ,1 ] ∪ I a × I a a , otherwise (10)
which is equal to V S T ( x , y ) given by Formula (5).
Taking R ( x , y ) = a for all ( x , y ) ∈ [ 0, a ] 2 in Formula (7), then
V T S , a ( x , y ) = ( S ( x , y ) , if ( x , y ) ∈ [ 0, a ] 2 T ( x , y ) , if ( x , y ) ∈ [ a ,1 ] 2 a , otherwise, (11)
which is equal to V a ( T , S ) given by Formula (3).
Taking R ( x , y ) = a for all ( x , y ) ∈ [ 0, a ] 2 in Formula (8), then it is clear that V S T , a ( x , y ) also coincides with V a ( T , S ) , which is given by Formula (3). Therefore, the two methods proposed in this study are more generalized than the methods proposed previously by [
Example 3.4. Let ( L , ≤ , 0 , 1 ) be a bounded lattice and let a ∈ L \ { 0,1 } .
(i) Let T : [ a ,1 ] 2 → [ a ,1 ] be a t-norm on [ a ,1 ] and b , c ∈ L be such that c ≤ b ≤ a . Let S : [ 0, a ] 2 → [ 0, a ] and R : [ 0, a ] 2 → [ 0, a ] be two functions on [ 0, a ] defined by
S ( x , y ) = ( x , if y = 0 and x ∈ [ 0, a ] y , if x = 0 and y ∈ [ 0, a ] x ∨ y ∨ c , if ( x , y ) ∈ [ 0, a ) 2 (12)
and
R ( x , y ) = x ∨ y ∨ b . (13)
Then S is a t-conorm and R is a t-subconorm on [ 0, a ] . It is easy to verify S ≤ R and the condition (6) holds. Therefore,
V 1 ( x , y ) = ( x , if y = 0 and x ∈ [ 0, a ] y , if x = 0 and y ∈ [ 0, a ] x ∨ y ∨ c , if ( x , y ) ∈ ( 0, a ] 2 T ( x , y ) , if ( x , y ) ∈ [ a ,1 ] 2 ( x ∧ a ) ∨ ( y ∧ a ) ∨ b , if ( x , y ) ∈ [ 0, a ] × I a ∪ I a × [ 0, a ] ∪ I a × I a a , otherwise (14)
is a nullnorm on L with the zero element a by Theorem 3.2.
(ii) Dually, let S : [ 0, a ] 2 → [ 0, a ] be a t-conorm on [ 0, a ] and j , k ∈ L be such that a ≤ j ≤ k . Then V 2 : L 2 → L is a nullnorm on L with the zero element a by Theorem 3.3, where
V 2 ( x , y ) = ( x , if y = 1 and x ∈ [ a ,1 ] y , if x = 1 and y ∈ [ a ,1 ] x ∧ y ∧ k , if ( x , y ) ∈ [ a ,1 ) 2 S ( x , y ) , if ( x , y ) ∈ [ 0, a ] 2 ( x ∨ a ) ∧ ( y ∨ a ) ∧ j , if ( x , y ) ∈ [ a ,1 ] × I a ∪ I a × [ a ,1 ] ∪ I a × I a a , otherwise . (15)
In this study, based on the existing constructions of nullnorms on L, we continue to study construction methods of nullnorms on bounded lattices. Two methods for obtaining nullnorms on L are presented in this paper. Some examples were provided to show that the construction methods proposed in this paper generalized the methods presented in previous studies.
The authors declare no conflicts of interest regarding the publication of this paper.
Xie, J.J. and Ji, W. (2021) New Constructions of Nullnorms on Bounded Lattices. Journal of Applied Mathematics and Physics, 9, 1-10. https://doi.org/10.4236/jamp.2021.91001