WorksMathematics
Uninorm constructions on bounded trellises via closure and interior operators
Abstract
In this paper, we establish necessary and sufficient conditions under which six closure- or interior-based operations on bounded trellises are uninorms. These operations incorporate triangular norms, triangular conorms, or both, and form three order-dual pairs. Since the underlying pseudo-order need not be transitive and meet and join need not be associative, standard lattice arguments do not apply directly. Under a fixed ambient transitivity assumption, the first pair is characterized by image, monotonicity, and compatibility conditions for the associated maps on elements incomparable with the neutral element; an additional comparison condition used in earlier sufficient constructions is shown to be redundant. Complete characterizations are also obtained for the variants with an absorbing endpoint, including both nondegenerate and degenerate interval cases. For the pair involving both a triangular norm and a triangular conorm, we derive exact comparison conditions requiring the triangular norm to take the value~, or dually the triangular conorm to take the value~, in the relevant cases. These conditions reduce to the usual incomparability assumptions in the absence of nontrivial zero- or one-divisors. On bounded lattices, the closure and interior conditions simplify to explicit identities or image restrictions. Finite proper-trellis examples clarify the role and limitations of the ambient transitivity assumptions.
INTRODUCTION
Uninorms were introduced by Yager and Rybalov as associative and commutative aggregation operations whose neutral element need not be an endpoint [29]. Their structure on the unit interval has been studied from several viewpoints, including representability and ordinal sums [7, 8, 18]. More broadly, lattice-valued fuzzy-set frameworks have long motivated topological and logical investigations [11, 12].
The extension of uninorms to bounded lattices led to a variety of construction and characterization results [1, 2, 3, 5, 13, 15, 19, 21, 25, 31, 33]. Of particular relevance here are constructions and characterizations based on closure and interior operators [4, 20, 22, 27]. These operators are useful when defining the values of an operation on elements incomparable with its neutral element. A bounded trellis replaces the transitive order of a bounded lattice by a reflexive and antisymmetric pseudo-order while retaining a meet and a join for each pair of elements [9, 10, 23]. In this setting, meet and join need not be associative, and consecutive comparisons cannot be concatenated without a suitable transitivity hypothesis. Zedam and De Baets [30] studied triangular norms on bounded trellises. Kong and Zhao [14] subsequently introduced uninorms on bounded trellises and gave constructions using triangular norms, triangular conorms, and closure or interior operators. Further developments appear in [16, 17, 24, 26, 28, 32].
This paper examines six closure- or interior-based formulas, arranged in three order-dual pairs. Its aim is not to introduce another unrestricted construction scheme, but to determine when these particular formulas define uninorms. Kong and Zhao [14] gave several sufficient constructions on bounded trellises, distinguished by transitivity properties of the neutral element. For the first pair considered here, (C1) is retained as an ambient hypothesis, while (C2) and (C4) are shown to be necessary and sufficient; (C3) then follows from (C1) and (C2). The dual statement holds for the interior-based formula. The variants with an absorbing endpoint are treated separately according to whether the relevant strict interval is empty.
For the pair involving both a triangular norm and a triangular conorm, the usual incomparability assumption remains sufficient, but the exact requirement is the implication in (19) or its order dual (22). These conditions coincide with the incomparability assumptions when there are no nontrivial zero- or one-divisors, and a counterexample shows that the converse fails without this restriction. On bounded lattices, the closure and interior parts reduce to explicit identities or image restrictions. The related lattice constructions in [4, 20, 22, 27] use formulas or input operations that are not identical in general. The lattice results below therefore compare the criteria for the present formulas rather than claim that all of those constructions are special cases.
§ 2 introduces the notation used below. § 3 collects the hypotheses on the closure and interior data. The six operations are studied in § 4, and their bounded-lattice forms are obtained in § 5.
PRELIMINARIES
Definition 2.1. [9] A pseudo-order on a nonempty set is a reflexive and antisymmetric binary relation . A pseudo-ordered set is denoted by . We write when and , and when neither nor holds. A pseudo-ordered set is called bounded if it has a least element and a greatest element .
Definition 2.2. [10] A trellis is a pseudo-ordered set in which every pair has a greatest lower bound and a least upper bound . It is bounded if it has a least element and a greatest element . A bounded trellis is written .
The relation of a trellis is not assumed to be transitive. Accordingly, all iterated meets and joins in the trellis arguments below are explicitly parenthesized. The order dual of a bounded trellis is , where if and only if [6]. It is again a bounded trellis. Under order duality, meet and join, the least and greatest elements, right- and left-transitivity, closure and interior operators, and triangular norms and triangular conorms are interchanged, whereas middle-transitivity is preserved.
Following [14], for put and , and define and analogously. For a fixed , the incomparable set is . Thus is the disjoint union of , , , and .
Definition 2.3. [10] A pseudo-ordered set is a -semi-trellis (respectively, a -semi-trellis) if every pair of elements has a greatest lower bound (respectively, a least upper bound). A subset of a trellis is a sub-trellis if for all , with the operations and pseudo-order restricted to .
Definition 2.4. [14] An element is right-transitive if implies ; it is left-transitive if implies ; and it is middle-transitive if implies . The corresponding sets are denoted by , , and . Their intersection is denoted by .
Definition 2.5. [30] A binary operation is increasing if
More generally, for , a map is increasing if the same implication holds whenever . A unary map , where , is increasing if in implies . A triangular norm on a bounded pseudo-ordered set is an increasing, associative, and commutative operation with neutral element 1. A triangular conorm is defined dually, with neutral element 0. A uninorm on is an increasing, associative, and commutative operation having a neutral element .
The four-variable definition of increasingness is essential. In a nontransitive pseudo-order, separate inequalities and do not by themselves imply the required joint comparison.
Definition 2.6. [20] A map is a closure operator if it is extensive, increasing, and idempotent:
An interior operator is increasing and idempotent and satisfies .
We first recall a necessary condition for the neutral element of a uninorm.
Lemma 2.7. [14] If a uninorm on a bounded trellis has neutral element , then .
Proposition 2.8. [30] Let be a bounded trellis, and let be a triangular norm on the sub-trellis with neutral element . Then for every .
CONDITIONS ON THE CLOSURE AND INTERIOR DATA
Let . For a closure operator , put
and
For the closure-based constructions, we use the following conditions.
(C1) and .
(C2) , the map is increasing on , and for every .
(C3) If , , and , then .
(C4) is increasing in the joint sense, that is,
and for all .
The corresponding interior-side conditions are stated explicitly. For an interior operator , put
and
(I1) and ;
(I2) , the map is increasing on , and for every ;
(I3) if , , and , then .
(I4) is increasing in the joint sense, namely,
and for all .
Lemma 3.1. For , one has and . If (C1)–(C4) hold, then for all . Dually, (I1)–(I4) imply .
Proof. The element is a lower bound of and , so the defining property of the meet gives . The dual argument gives .
Clearly . If , then . Right-transitivity of would give , contrary to . Here we use the hypothesis (C1) which says . Thus . The last assertion is dual. □
Lemma 3.2. Assume (C1) and let . Then
implies ;
implies ;
implies ;
implies .
The order-dual statements hold under (I1).
Proof. For the first two assertions use right-transitivity of . If and , then and right-transitivity of contradicts , proving the third assertion. For the fourth assertion, if , then item (1) would force , contrary to ; if or , items (2) and (3), respectively, give the same contradiction. Hence . The uses of antisymmetry are confined to the endpoint identifications in the interval definitions. □
CONSTRUCTION METHODS FOR UNINORMS ON BOUNDED TRELLISES
A meet-based construction and its dual
Theorem 4.1. Let be a bounded trellis, and let . Let be a closure operator, and let be a triangular conorm on the sub-trellis with neutral element . Assume (C1) and define by
Then the following statements are equivalent:
is a uninorm on with neutral element ;
conditions (C2) and (C4) hold.
Under these conditions, (C3) follows from (C1) and (C2), and is an absorbing element of .
Proof. Assume first that (C2) and (C4) hold. If , , and , then is an upper bound of and . Hence . Since and , right-transitivity gives . Thus (C3) follows, and (C1)–(C4) are available in the sufficiency proof below.
The displayed clauses assign exactly one value to every pair. The formula is symmetric, and direct substitution gives for every .
We first verify joint increasingness. Take and . Simultaneously interchanging with leaves the required comparison unchanged, so the roles of and may be interchanged whenever commutativity is used below. By Theorem 3.2, a point of can move only to , a point equal to can move only to or , and a point of can move only to or . We now inspect every possible case for .
If lies in , , or , then , so the required comparison follows from the minimality of .
Suppose and . Then and . If , then and . If , increasingness of gives . Since by (C1) and , it follows that . If , then
and by Theorem 3.1. Right-transitivity of again gives .
If , then , and the desired inequality is precisely the joint increasingness of .
Suppose next that and . Here and . If , then , so . If and , then , and right-transitivity of yields the conclusion. Finally, if , increasingness of gives . Combining this with by right-transitivity of proves the remaining comparison.
Suppose and . Then and . If , (C2) gives . If , then (C3) and increasingness of give . Because , the two displayed comparisons yield without invoking transitivity of the ambient relation.
It remains to take . Put . By Theorem 3.1, . If , (C4) directly gives . If, say, and , then . From the defining properties of the meets and repeated applications of right-transitivity of , we obtain
where each adjacent comparison is combined with the preceding one using that the fixed initial element is right-transitive. The case , is symmetric. If , the same argument first gives , while . One more application of right-transitivity of yields . The listed cases exhaust all transitions allowed by Theorem 3.2; hence is jointly increasing.
It remains to verify associativity. The element is neutral, and inspection of (5) shows that is absorbing. Thus triples containing or are immediate. For all remaining triples, the following list checks every possible choice of the inner pair, not merely one ordering for each placement of the three arguments among , , and .
If all three elements lie in , every inner product is , so every bracketing is .
If all three lie in , the assertion is the associativity of .
If all three lie in , every inner product belongs to by Theorem 3.1, and its product with the remaining incomparable element is . Hence all three choices of the inner pair give .
For two elements of and one element of , pairing the two lower elements first gives , whereas pairing a lower and an upper element first returns that lower element, whose product with the other lower element is .
For two elements of and one element of , every choice of the inner pair gives either at the first step or at the second step.
Let and . Pairing first gives . Pairing with either upper element first also gives , and pairing the result with the remaining upper element again gives .
For one element of and two elements of , pairing the two incomparable elements first gives an element of and hence after the second multiplication; pairing the lower element with either incomparable element first gives immediately. Thus all bracketings are .
Let and . Pairing first gives . The other two choices of inner pair give, respectively, and , where (C4) and the symmetry of are used.
Let and . Pairing first gives . Pairing with either upper element first gives , and the second multiplication gives by (C2).
Let , , and . Pairing first gives and then ; pairing first gives ; and pairing first gives , whose product with is . Hence all three bracketings are $0.
These are the ten possible placements of three arguments in , , and , and every possible inner pair has been covered. Therefore is associative and hence is a uninorm with neutral element .
Conversely, suppose that is a uninorm. We first record two facts that do not use (C2)–(C4). For , Theorem 3.1 gives . Moreover, for one has . Equality would give , and right-transitivity of would imply , contrary to . Thus
Fix . If , then right-transitivity of , applied to , would give , a contradiction. Consequently, the decomposition shows that belongs either to or to .
Assume temporarily that . Associativity at the triple gives
Since a triangular conorm on satisfies , the two sides are and , respectively. Indeed, increasingness and neutrality give , so antisymmetry yields . Hence . Associativity at would then give
contradicting (6). Therefore
If and , increasingness of , applied directly to , yields . Thus is increasing on . Using (7), associativity at now reads
This proves (C2).
On the operation agrees with . Hence the joint increasingness of implies the joint increasingness of . Finally, for , associativity at and (6) give
Thus (C4) also holds.
Kong and Zhao [14] introduced uninorms on bounded trellises and gave four construction methods based on closure or interior operators together with suitable transitivity assumptions. Theorem 4.1 has a more specific purpose: after fixing the formula (5) and the ambient hypothesis (C1), it identifies (C2) and (C4) as the exact remaining conditions and shows that (C3) is redundant. In particular, the necessity proof derives the image, increasingness, idempotence, and compatibility requirements from the uninorm axioms instead of imposing them only as hypotheses. This is the precise sense in which the result is sharper for the fixed formula (5); the necessity is part of the conclusion rather than a restatement of the assumptions. It is not a characterization of all closure-based uninorms on bounded trellises, and the same distinction applies to Theorem 4.3 by order duality. The following example shows both why (C1) is used in Theorem 4.1 and why it need not hold for every individual operation defined by (5).
Example 4.2. Let . Besides reflexivity and the boundary comparisons , impose only and . Thus . Comparable pairs have their usual meet and join, and , . This gives a bounded trellis, which is proper because but .
Let be the identity map. Then , , , , and . In particular, the lattice-style equality holds. On take the unique triangular conorm. Formula (5), in the order ,
becomes
Table 1.
The table is symmetric and has neutral element . Associativity follows from the absorbing role of , the neutrality of , and the fact that the product of two elements of is . Direct inspection against the listed comparisons shows joint increasingness. Hence is a uninorm. Nevertheless, is not right-transitive, since but ; therefore (C1) fails. The same example shows that the lattice-style equality alone does not imply all of (C1)–(C4) on a proper trellis.
Theorem 4.3. Let be a bounded trellis, let , and let be an interior operator. Let be a triangular norm on the sub-trellis with neutral element . Assume (I1) and define
Then is a uninorm with neutral element if and only if (I2) and (I4) hold. Under these conditions, (I3) follows from (I1) and (I2), and is an absorbing element of .
Proof. Consider the order-dual bounded trellis . With respect to , the interior operator is a closure operator, the interval in becomes in , and becomes a triangular conorm with neutral element . Conditions (I1)–(I4) become, respectively, (C1)–(C4), while and correspond to and . Formula (8) becomes (5). The assertion therefore follows from Theorem 4.1 by order duality. □
Variants with an absorbing endpoint
Theorem 4.4. Let be a bounded trellis, let , let be a closure operator, and let be a triangular conorm on with neutral element . Assume (C1)–(C4) and , and define by
Then is a uninorm on with neutral element if and only if
Proof. We first prove necessity. Suppose for some . Then
so associativity fails. Hence (10) is necessary.
Conversely, assume (10). Formula (9) is symmetric and unambiguous, and direct substitution gives for every .
To prove joint increasingness, let and . If or , then , so the desired comparison follows from the maximality of . Otherwise . Antisymmetry then implies : for example, together with would force . On , the operations and agree pointwise. Consequently, the joint increasingness proved in Theorem 4.1 gives
We next show that is closed under . For inputs in , condition (10) applies when both inputs belong to ; when one input equals , neutral-element identity gives a value different from . A mixed incomparable–upper clause has value by (C2), an incomparable–incomparable clause has value in by Theorem 3.1, the mixed clauses return one of the inputs, and the remaining clause returns . Hence is closed under .
If at least one of equals , both bracketings of their product are , because is absorbing for . If none equals , the closure just proved ensures that both intermediate products also differ from . Since on , associativity of yields
Therefore is a uninorm with neutral element .
Proposition 4.5. Let be a bounded trellis, let , let be a closure operator, and let be a triangular conorm on with neutral element . Assume (C1), and define by (9).
If , then the following statements are equivalent:
(a) is a uninorm on with neutral element ;
(b) conditions (C2) and (C4) hold, and
Under these equivalent conditions, (C3) follows from (C1) and (C2) and is therefore redundant.
If , then is a uninorm if and only if is jointly increasing on . In this degenerate case no condition involving is necessary.
Proof. Assume first that and that is a uninorm. Condition (11) is necessary. Indeed, if and , then
contrary to associativity. Hence (11) holds.
Condition (C1) also implies
Indeed, . If , then , and right-transitivity of would give , contrary to .
Fix and . Since , if , right-transitivity of would imply , a contradiction. Thus . Associativity at the triple gives
If , then the last member of (13) equals , whereas its first member belongs to by (12). This is impossible. Consequently,
If and , increasingness of , applied to , gives . Hence is increasing on .
Increasingness of and neutrality of yield . Since , antisymmetry shows that . Together with (11), this gives .
Associativity at and (14) now yield
Thus (C2) holds. The restriction of to is . Therefore the joint increasingness of implies that is jointly increasing on . Finally, for , associativity at gives
where (12) is used in the last equality. Hence (C4) holds.
Conversely, suppose that (C2), (C4), and (11) hold. As in Theorem 4.1, conditions (C1) and (C2) imply (C3). Therefore (C1)–(C4) hold, and Theorem 4.4 shows that is a uninorm.
It remains to consider . In this case the definition of reduces to
In particular, neither nor a nontrivial value of occurs in the formula.
If is a uninorm, its restriction to shows immediately that is jointly increasing. Conversely, assume that is jointly increasing. Commutativity and neutrality of follow directly from the displayed formula.
For increasingness, let and . If or , the required comparison follows from the maximality of 1. Otherwise, Theorem 3.2 and imply that forces , and forces ; the same statements hold for and . If or , neutrality reduces the desired inequality to the comparison involving the other variable. If , the assertion is the joint increasingness of . In every remaining case . Thus is jointly increasing.
For associativity, a triple containing 1 gives the value 1 under both bracketings, and a triple containing reduces by neutrality. Every remaining entry belongs to . If all three entries belong to , each inner product belongs to by (12), and the second product is 0. If exactly two entries belong to , pairing them first gives an element of and hence gives 0 at the second step, while pairing an incomparable and a lower element first gives 0 immediately. If at most one entry belongs to , both bracketings are again 0. Therefore is associative, completing the proof. □
Theorem 4.6. Let be a bounded trellis, let , let be an interior operator, and let be a triangular norm on with neutral element . Assume (I1)–(I4) and , and define
Then is a uninorm if and only if
Proof. We first prove necessity. If for some , then
Thus associativity implies (16).
Conversely, assume (16). The formula is the order dual of the definition in Theorem 4.4, so it is well defined and symmetric, and is neutral. We give the remaining dual argument explicitly.
Let and . If or , then . Otherwise , and antisymmetry implies ; for example, would force . On , the operations and agree pointwise. The joint increasingness of therefore gives
The set is closed under . Indeed, for inputs in , condition (16) treats the case in which both inputs belong to , while treats the case in which one input is . The mixed incomparable–lower clauses take values in by (I2), the incomparable–incomparable clause takes values in by Theorem 3.1, the mixed clauses return a nonzero input, and the remaining clause returns 1.
If one of is 0, both bracketings are 0, because 0 is absorbing for . If none is 0, closure keeps both intermediate values in , where . Hence
by associativity of . Thus is a uninorm. □
Combining the preceding result with the order-dual necessity argument gives the criterion for the interior construction with absorbing element 0.
Corollary 4.7. Let be a bounded trellis, , be an interior operator, and be a triangular norm on with neutral element . Assume (I1), and define by (15).
If , then is a uninorm if and only if (I2), (I4), and for all hold. Under these conditions, (I3) is redundant.
If , then is a uninorm if and only if is jointly increasing on ; no condition involving is necessary.
Proof. Apply Theorem 4.5 to the order-dual bounded trellis. Under this duality, (C1)–(C4), , , and condition (11) become (I1)–(I4), , , and condition (16), respectively. The nondegenerate and degenerate criteria therefore dualize exactly as stated. □
Constructions using both a triangular norm and a triangular conorm
Let be a bounded trellis, let , let be a closure operator, let be a triangular norm on with neutral element , and let be a triangular conorm on with neutral element . We shall use the following sufficient condition:
Define
The following result gives the necessary and sufficient condition under (C1).
Theorem 4.8. Let be a bounded trellis, let , let be a closure operator, let be a triangular norm on with neutral element , and let be a triangular conorm on with neutral element . Assume (C1), and define by (18). Then the following statements are equivalent:
is a uninorm on with neutral element ;
condition (C2) holds and
Under these assumptions, (C3) follows from (C2), whereas (C4) is not required for this construction.
Proof. Suppose first that is a uninorm. For , put . By Theorem 3.1, . If , then right-transitivity of would imply , contrary to . Hence .
Assume that . Since is absorbing for the triangular conorm , associativity at gives
Thus . Associativity at would then give
a contradiction. Consequently . If in , joint increasingness applied to gives . Moreover, associativity at , now using , yields
This proves (C2).
Let and satisfy . Then componentwise, by the definition of , for and , and joint increasingness gives . Since is the least element, (19) follows.
Conversely, assume (C2) and (19). We first recover (C3). If , , and , then is an upper bound of both and , and hence . Condition (C2) gives , so (C1) makes right-transitive. Therefore , which is (C3).
We next prove joint increasingness. If , monotonicity and neutrality of give
and . Similarly, for one has and . Let and . By commutativity, we may simultaneously interchange with and with in the cases below.
Suppose first that and put . If , the assertion follows from joint increasingness of . If one of lies in and the other in , or if both lie in , then or has an element of on the right. Condition (19), together with commutativity of , gives . If, say, and , then and ; right-transitivity of completes the comparison. If , then , and middle-transitivity of applies. If with , right-transitivity of gives . If one of is and the other lies in , the same argument gives .
Suppose next that and . Then and Theorem 3.2 gives . If , then and . If , then , and middle-transitivity of applies. If , then is when and is when ; the required comparison follows from and right-transitivity of . If and , then . If , then and joint increasingness of applies. Suppose next that and . If , then is either or , and together with right-transitivity of gives the result. If , then is either or an -value above , and the same argument applies.
Finally, suppose that and . If , increasingness of applies. If , condition (C3) gives , after a possible simultaneous interchange of the coordinates; since is right-transitive, the comparison follows. If , then . These cases prove joint increasingness of without using (17) or (C4).
It remains to prove associativity. Triples containing reduce by neutrality. On and , associativity follows from that of and , respectively, while every pairing on is 0. For two lower and one upper input, every pairing gives their -value; for two lower and one incomparable input, every pairing gives 0. For one lower and two upper inputs, every pairing gives the lower input, and for one lower and two incomparable inputs every pairing gives 0, using . For one upper and two incomparable inputs, (C2) places the two possible -values in , so all pairings are 0. For one incomparable and two upper inputs, all pairings give by the idempotence in (C2). For one input from each of , , and , every pairing is 0. Thus is associative. Commutativity and neutrality of are immediate from (18), completing the proof. □
Corollary 4.9. Assume (C1)–(C3) and (17). Then the operation in (18) is a uninorm with neutral element .
Proof. Let and with . Condition (17) excludes , so and . Hence (19) holds, and the result follows from Theorem 4.8. □
The incomparability restrictions used in the closure-based trellis constructions of [14] are sufficient for controlling pairs consisting of a lower and an incomparable element. For the fixed formula (18), Theorem 4.9 gives the corresponding sufficient statement, whereas Theorem 4.8 shows that the exact requirement is (19). The two requirements are equivalent when has no zero divisors by Theorem 4.11; Theorem 4.12 shows that they can differ when zero divisors are present. Thus the improvement concerns the criterion for (18), rather than an inclusion between all classes constructed in the two papers.
The construction in Theorem 4.9 occurs on the following small proper bounded trellis.
Example 4.10. Let . Besides reflexivity and , impose , , and , with no further comparisons. Comparable pairs have the indicated lower and upper elements, whereas every incomparable pair has meet 0 and join 1. Hence is a bounded trellis. It is proper because but .
Define and let fix . This map is extensive and idempotent. It is also increasing: the only nonboundary comparisons not fixed by have an element of on the right, and all three of these elements have image 1. Thus is a closure operator. Moreover, , , and . The element is middle- and right-transitive, while 0 and are right-transitive, so (C1) holds. Moreover, and ; since is a singleton, conditions
(C2)–(C4) follow immediately. Condition (17) is vacuous because ; hence (19) holds.
On take the unique triangular norm. On define and put whenever . The operation is commutative and has neutral element . If at least two of three inputs differ from , both associative products are ; otherwise associativity follows from neutrality. For increasingness, a value different from can occur only when one input is , in which case the required comparison is one of the assumed coordinate comparisons. Hence is a triangular conorm, and Theorem 4.8 yields a uninorm on this proper bounded trellis.
Corollary 4.11. Under the hypotheses of Theorem 4.8, suppose that has no zero divisors, that is,
Then (19) is equivalent to (17). Consequently, under (C1) and the no-zero-divisor assumption, condition (C2) together with (17) is necessary and sufficient for to be a uninorm.
Proof. Assume first (17). If and , then cannot belong to , so ; hence . Thus (19) holds. Conversely, assume (19). If and , taking gives , which contradicts the absence of zero divisors. The reverse comparison is also impossible: since , the chain would imply . Hence , as required. □
The converse relation between (19) and (17) is clarified by the following counterexample.
Remark 4.12. Condition (17) always implies (19), but the converse may fail when has zero divisors. Indeed, let be the bounded lattice whose nontrivial covering relations are
Take to be the identity map. Then and , so (C1) and (C2) hold. On the three-element chain define the drastic triangular norm
and take the triangular conorm on determined by and otherwise. Since , one has for all , so (19) holds. Hence Theorem 4.8 shows that is a uninorm. Nevertheless, , and therefore (17) fails. Thus condition (17) is sufficient but is not necessary in general.
Theorem 4.13. Let be an interior operator, be a triangular norm on , and be a triangular conorm on , both with neutral element . Assume (I1)–(I3) and the condition
Define
Then is a uninorm with neutral element .
Proof. Consider the order-dual bounded trellis . The incomparable set of is unchanged.
With respect to , the map is a closure operator. Indeed, is extensive in the dual order because in the original order implies . Conditions (I1)–(I3) are exactly conditions (C1)–(C3) in the dual pseudo-order, and the expression corresponding to is . The triangular conorm on the original interval is a triangular norm on the dual interval from to , whereas is the corresponding triangular conorm on the dual interval from to . Moreover, (20) is precisely (17) in the dual pseudo-order.
Applying Theorem 4.9 to these dual data gives exactly the cases displayed in (21): the lower and upper intervals are interchanged, and are interchanged, and is replaced by . Hence the joint increasingness and all three associativity pairings proved for dualize without any additional assumption. Therefore is a uninorm with neutral element . □
We next state the exact criterion for the two-sided interior construction.
Corollary 4.14. Let be a bounded trellis, , be an interior operator, be a triangular norm on with neutral element , and be a triangular conorm on with neutral element . Assume (I1), and define by (21). Then is a uninorm if and only if (I2) holds and
Under these assumptions, (I3) follows from (I2), whereas (I4) is not required. If has no nontrivial one-divisors, in the sense that
then (22) is equivalent to (20). Consequently, under (I1) and the absence of nontrivial one-divisors, condition (I2) together with (20) is necessary and sufficient for to be a uninorm.
Proof. Apply Theorem 4.8 and Theorem 4.11 to the order-dual bounded trellis. Under this duality, (C1)–(C3), condition (19), and the absence of zero divisors become (I1)–(I3), condition (22), and the absence of nontrivial one-divisors, respectively. This gives all the stated equivalences. □
BOUNDED-LATTICE REDUCTIONS AND SCOPE
Conditions (C1)–(C4) and (I1)–(I4) are formulated for trellises because ordinary lattice manipulations cannot be used without transitivity and associativity. On a bounded lattice, these conditions reduce to explicit identities and restrictions on the operator images. Combined with the characterizations in § 4, these reductions yield exact lattice criteria for all six constructions.
Proposition 5.1. Let be a bounded lattice and let .
If is a closure operator, then (C1)–(C4) hold if and only if for all .
If is an interior operator, then (I1)–(I4) hold if and only if for all .
If is a closure operator, then the following statements are equivalent:
(a) conditions (C1)–(C3) hold;
(b) for every ;
(c) for every .
Dually, for an interior operator , one has
Proof. Every element of a lattice is right-, left-, and middle-transitive, and lattice meet and join are associative and jointly increasing. Thus (C1) and its order dual hold automatically.
For (1), let and put . Since , the inequality would imply . If , then
which would imply . Both alternatives contradict , so . Monotonicity of follows from that of , meet, and join. Moreover, and , whence and . Therefore , proving (C2). Joint increasingness of follows from that of meet, and
so (C4) holds. Finally, if and , then , which is (C3). This proves the sufficiency direction in (1).
Conversely, suppose that (C1)–(C4) hold. For put . Condition (C2) gives , so (C4) may be applied with . Since , lattice absorption and associativity of meet give . On the other hand,
because . Hence , proving the necessity direction. Assertion (2), including both directions, is its order dual.
For (3), fix . Extensivity gives , so would imply , which is impossible. Consequently, can be comparable with only in the direction . Since every element is below , this proves the equivalence .
Suppose that for every , and put . As before, , so is impossible. If , then would give , again a contradiction. Hence . The map is increasing because closure, meet, and join are increasing on a lattice. Moreover, and , so closure idempotence and lattice absorption give and . It follows that , proving (C2). If with , then is an upper bound of both and ; thus , which is (C3). Condition (C1) is automatic on a lattice, and hence (C1)–(C3) hold.
Conversely, suppose that (C1)–(C3) hold. By (C2), . If , then is a lower bound of both and , whence , contrary to . Therefore , and the first paragraph shows that , or equivalently . This completes the closure part of (3). The interior statement follows by order duality. □
Corollary 5.2. Let be a bounded lattice, let , let be a triangular conorm on with neutral element , and let be a closure operator on . Define by (5). Then the following statements are equivalent:
for every ;
conditions (C1)–(C4) hold;
is a uninorm on with neutral element ;
is associative.
Thus the closure equality is necessary and sufficient for this bounded-lattice construction.
Proof. The equivalence of (1) and (2) is Theorem 5.1, and (2) implies (3) by Theorem 4.1. Clearly (3) implies (4). It remains to show that (4) implies (1).
Fix and write . Since , one cannot have . Suppose that . Associativity at gives , and hence . Associativity at then gives
which is impossible because . Thus . Applying associativity to gives
where was used in the last equality. Finally,
Therefore , as required. □
Remark 5.3. Ouyang and Zhang [20, Theorem 4.1] gave a sufficient construction on a bounded lattice from a closure operator and a triangular norm. Çaylı [4, Theorems 3.1 and 3.11] studied two further lattice formulas and obtained necessary and sufficient conditions. Qi and Xiu [22] instead start from a uninorm on a subinterval and extend it by means of a closure operator and a triangular superconorm, while Xiu and Si [27] consider closure operators together with triangular conorms. These papers are close in method, but their input data and defining clauses are not identical in general.
The preceding corollary concerns only the operation in (5). For this formula, it shows that the displayed closure equality is not merely sufficient but also necessary. It neither classifies all uninorms on a bounded lattice nor subsumes all the constructions cited above. On a proper trellis, moreover, the ambient hypothesis in Theorem 4.1 cannot in general be recovered from the lattice equality, as shown by the five-element example.
Corollary 5.4. Let be a bounded lattice, let , let be a triangular norm on with neutral element , and let be an interior operator on . Define by (8). Then the following statements are equivalent:
for every ;
conditions (I1)–(I4) hold;
is a uninorm on with neutral element ;
is associative.
Proof. Apply Theorem 5.2 to the order-dual lattice. Closure, meet, triangular conorm, , , and (5) become, respectively, interior, join, triangular norm, , , and (8). This gives all four equivalences. □
The constructions with an absorbing endpoint have the following corresponding lattice criteria.
Corollary 5.5. Let be a bounded lattice and let .
Let be a closure operator, let be a triangular conorm on with neutral element , and define by (9). If , then is a uninorm on with neutral element if and only if for every and for all . If , then is a uninorm for every closure operator .
Dually, let be an interior operator, let be a triangular norm on with neutral element , and define by (15). If , then is a uninorm on with neutral element if and only if for every and for all . If , then is a uninorm for every interior operator .
Proof. Every element of satisfies (C1), and is jointly increasing. Suppose first that . By Theorem 4.5, the uninorm property of is equivalent to (C2), (C4), and condition (11). Under (C1), condition (C3) follows from (C2), while Theorem 5.1(1) identifies the conjunction of (C2) and (C4) with the displayed closure equality. This proves the first equivalence. If , Theorem 4.5 reduces the criterion to joint increasingness of , which is automatic on a lattice. The second assertion is the order dual, using Theorem 4.7 and Theorem 5.1[2]. □
The following example treats the case in which the upper interval is degenerate.
Example 5.6. Let be ordered by inclusion, with , , and . Then and . Define the closure operator by for every . For each one has , so the image requirement in (C2) fails. Nevertheless, Theorem 5.5 shows that is a uninorm. Thus the nonemptiness of is essential when the uninorm property of is used to recover conditions (C1)–(C4); its order dual gives the corresponding boundary for .
The two-sided constructions reduce to the following lattice criteria.
Corollary 5.7. Let be a bounded lattice, let , let be a triangular norm on , and let be a triangular conorm on , both with neutral element .
Let be a closure operator and define by (18). Then is a uninorm on with neutral element if and only if for every and condition (19) holds; explicitly,
Equivalently, the first displayed condition may be replaced by for every . If has no zero divisors, then condition (19) may equivalently be replaced by (17).
Dually, let be an interior operator and define by (21). Then is a uninorm on with neutral element if and only if for every and condition (22) holds; explicitly,
Equivalently, the first displayed condition may be replaced by for every . If has no nontrivial one-divisors, then condition (22) may equivalently be replaced by (20).
Proof. Condition (C1) is automatic on . By Theorem 4.8, is a uninorm if and only if (C2) and (19) hold. Under (C1), condition (C2) implies (C3); hence (C2) is equivalent here to (C1)–(C3). Part (3) of Theorem 5.1 identifies these conditions with , equivalently , on . This proves the first assertion. The second assertion is its order dual and follows from Theorem 4.14. The final equivalences follow from Theorem 4.11 and its stated order-dual counterpart. □
CONCLUSIONS
We have obtained necessary and sufficient conditions for six constructions on bounded trellises. Under the ambient transitivity condition, the first closure-based construction is characterized by (C2) and (C4), while (C3) follows from the other assumptions; the interior-based result is its order dual. For the variants with an absorbing endpoint, the nondegenerate cases require (10) or (16), whereas in the degenerate cases the closure or interior map no longer occurs in the effective formula.
For the constructions involving both a triangular norm and a triangular conorm, the necessary and sufficient requirements are (19) and its order dual (22). The stronger incomparability assumptions (17) and (20) are equivalent to these requirements when the triangular norm has no zero divisors or the triangular conorm has no nontrivial one-divisors. On bounded lattices, the closure and interior parts of the criteria reduce to explicit identities or image restrictions.
The examples also show the limitation of these reductions: a lattice identity need not imply the ambient transitivity condition on a proper trellis, and that condition need not be necessary for a particular operation to be a uninorm. Further work may determine weaker ambient hypotheses and compare the resulting classes with other uninorm constructions on bounded trellises.
DECLARATION OF COMPETING INTEREST
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
DATA AVAILABILITY
No datasets were generated or analysed in this theoretical study.
ACKNOWLEDGEMENT
This work was supported by the National Natural Science Foundation of China (Nos. 12671553 and 12271036).
Zhenyu Xiu, Corresponding author. School of Mathematical Science, Guangxi Minzu University, Nanning 530006, Guangxi, China
e-mail: [email protected]
References
- [1]S. Bodjanova, M. Kalina, Construction of uninorms on bounded lattices, in: IEEE 12th International Symposium on Intelligent Systems and Informatics, SISY 2014, 2014, pp. 61–66.DOI
- [2]G.D. Çaylı, F. Karaçal, R. Mesiar, On a new class of uninorms on bounded lattices, Inf. Sci. 367–368 (2016) 221–231.DOI
- [3]G.D. Çaylı, New methods to construct uninorms on bounded lattices, Int. J. Approx. Reason. 115 (2019) 254–264.DOI
- [4]G.D. Çaylı, A characterization of uninorms on bounded lattices via closure and interior operators, Kybernetika 59 (2023) 768–790. doi:10.14736/kyb-2023-5-0768.DOI
- [5]Y. Dan, B.Q. Hu, J. Qiao, New constructions of uninorms on bounded lattices, Int. J. Approx. Reason. 110 (2019) 185–209.DOI
- [6]B.A. Davey, H.A. Priestley, Introduction to Lattices and Order, 2nd ed., Cambridge University Press, Cambridge, 2002. doi:10.1017/CBO9780511809088.DOI
- [7]P. Drygás, D. Ruiz-Aguilera, J. Torrens, A characterization of a class of uninorms with continuous underlying operators, Fuzzy Sets Syst. 287 (2016) 137–153.DOI
- [8]J.C. Fodor, R.R. Yager, A. Rybakov, Structure of uninorms, Int. J. Uncertain. Fuzziness Knowl.-Based Syst. 5 (1997) 411–427.DOI
- [9]E. Fried, Tournaments and non-associative lattices, Ann. Univ. Sci. Bp. Rolando Eötvös Nomin., Sect. Math. 13 (1970) 151–164.
- [10]K. Gladstien, A characterization of complete trellises of finite length, Algebra Univers. 3 (1973) 341–344.DOI
- [11]J.J.A. Goguen, L-fuzzy sets, J. Math. Anal. Appl. 18 (1967) 145–174.DOI
- [12]J.J.A. Goguen, The fuzzy Tychonoff theorem, J. Math. Anal. Appl. 43 (1973) 734–742.DOI
- [13]F. Karaçal, R. Mesiar, Uninorms on bounded lattices, Fuzzy Sets and Systems 261 (2015) 33–43.DOI
- [14]Y. Kong, B. Zhao, Uninorms on bounded trellises, Fuzzy Sets and Systems 481 (2024) 108898. doi:10.1016/j.fss.2024.108898.DOI
- [15]G. Li, H.W. Liu, On a characterization of representable uninorms, Fuzzy Sets Syst. 408 (2021) 57–64.DOI
- [16]Y. Luo, B. Pang, J.-C. Yao, Constructions of uninorms on bounded trellises, Fuzzy Sets and Systems 521 (2025) 109576. doi:10.1016/j.fss.2025.109576.DOI
- [17]A. Mehenni, L. Zedam, Pseudo triangular norms on bounded trellises, arXiv:2212.03605 (2022).arxiv.org/abs/2212.03605
- [18]A. Mesiarová-Zemánková, Ordinal sums of representable uninorms, Fuzzy Sets Syst. 308 (2017) 42–53.
- [19]A. Mesiarová-Zemánková, Uninorms internal on one or more non-trivial cuts, Inf. Sci. 653 (2024) 119793.
- [20]Y. Ouyang, H.-P. Zhang, Constructing uninorms via closure operators on a bounded lattice, Fuzzy Sets and Systems 395 (2020) 93–106. doi:10.1016/j.fss.2019.05.006.
- [21]F. Qin, L. Fu, A characterization of uninorms not internal on the boundary, Fuzzy Sets Syst. 469 (2023) 108641.DOI
- [22]J. Qi, Z.-Y. Xiu, New methods to construct uninorms by extending uninorms with closure operators and t-superconorms, Kybernetika 61 (2025) 855–871. doi:10.14736/kyb-2025-6-0855.DOI
- [23]H.L. Skala, Trellis theory, Algebra Universalis 1 (1971) 218–233. doi:10.1007/BF02944982.DOI
- [24]H. Song, B. Zhao, New constructions of uninorms on bounded trellises, Fuzzy Sets and Systems 537 (2026) 109902. doi:10.1016/j.fss.2026.109902.DOI
- [25]X.R. Sun, H.W. Liu, Further characterization of uninorms on bounded lattices, Fuzzy Sets Syst. 427 (2022) 96–108.DOI
- [26]Y.-M. Wang, Y. Dan, B. De Baets, On the classes of uninorms U_min and U_max on bounded trellises, International Journal of Approximate Reasoning 191 (2026) 109627. doi:10.1016/j.ijar.2026.109627.
- [27]Z.-Y. Xiu, Z.-Y. Si, New methods to construct uninorms on bounded lattices via closure operators and triangular conorms, Computational and Applied Mathematics 45 (2026) 12. doi:10.1007/s40314-025-03375-9.DOI
- [28]Z. Xiu, X. Zheng, Nullnorms on bounded trellises, Fuzzy Sets Syst. 521 (2025) 109599.arxiv.org/abs/2408.09321
- [29]R.R. Yager, A. Rybalov, Uninorm aggregation operators, Fuzzy Sets and Systems 80 (1996) 111–120.DOI
- [30]L. Zedam, B. De Baets, Triangular norms on bounded trellises, Fuzzy Sets and Systems 462 (2023) 108468. doi:10.1016/j.fss.2023.01.003.DOI
- [31]H.P. Zhang, M. Wu, Z. Wang, Y. Ouyang, B. De Baets, A characterization of the classes U_min and U_max of uninorms on a bounded lattice, Fuzzy Sets Syst. 423 (2021) 107–121.
- [32]Y.F. Zhao, H.W. Liu, New constructions of t-norms and t-conorms on bounded lattices and trellises, Int. J. Gen. Syst. 55 (2026) 70–90. doi:10.1080/03081079.2025.2457559.
- [33]B. Zhao, T. Wu, Some further results about uninorms on bounded lattices, Int. J. Approx. Reason. 130 (2021) 22–49.