See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/225546079

Commutative idempotent residuated lattices

Article  in  Czechoslovak Mathematical Journal · January 2007

DOI: 10.1007/s10587-007-0055-7

CITATIONS READS

10 144

1 author:

David Stanovský
Charles University in Prague
53 PUBLICATIONS   363 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Racks and Quandles theory View project

Yang-Baxter equation View project

All content following this page was uploaded by David Stanovský on 30 May 2014.

The user has requested enhancement of the downloaded file.

 COMMUTATIVE IDEMPOTENT RESIDUATED LATTICES

DAVID STANOVSKÝ

Abstract. We investigate the variety of residuated lattices with a commuta-

tive and idempotent monoid reduct.

A residuated lattice is an algebra A = (A, ∨, ∧, ·, e, /, \) such that (A, ∨, ∧) is a

lattice, (A, ·, e) is a monoid and for every a, b, c ∈ A
ab ≤ c ⇔ a ≤ c/b ⇔ b ≤ a\c.
The last condition is equivalent to the fact that (A, ∨, ∧, ·, e) is a lattice-ordered
monoid and for every a, b ∈ A there is a greatest c such that cb ≤ a (denoted a/b)
and a greatest d such that bd ≤ a (denoted b\a). It is easy to see that the class RL
of all residuated lattices is a variety. We are concerned about the variety CIdRL
of commutative idempotent (CI) residuated lattices, i.e. the subvariety of RL given
by equations
xy ≈ yx and xx ≈ x.
In other words, residuated lattices whose semigroup reduct is a semilattice. For
example, every Heyting algebra is a CI residuated lattice, where ab = a ∧ b and
a/b = b\a = b → a for every a, b (see e.g. [3], p. 30).
Foundation of the theory of residuated lattices goes as far as to 1930’s, when
Dilworth and Ward [5] studied lattices of ring ideals. A recent introduction can be
found in [4] and [10] and commutative residuated lattices were particularly studied
in [9]. We will use the notation and terminology of these papers. We also assume
a basic familiarity with universal algebra, standard references are [3] and [12].
In CI residuated lattices, we drop the operation \, since under commutativity
x/y ≈ y\x. The lattice order will be denoted by ≤. We put a  b iff ab = a; hence
 is the semilattice order, where · is regarded as the meet; e is its top element.
When refering to an order, we mean the lattice order ≤, unless explicitly stated
otherwise. We put A+ = {a ∈ A : a ≥ e} and A− = {a ∈ A : a ≤ e} and we call
A+ the positive cone and A− the negative cone of A (regarded as lattice-ordered
monoids; indeed, they may not be closed on residuation).
The bottom element (in the lattice order) is denoted 0 and the top element is
denoted 1, if they exist; it is easy to see that, in any residuated lattice, if 0 exists,
then 1 exists, 0a = a0 = 0 and a/0 = 1/a = 1 (see also [4]); particularly, 0 is also
the bottom element of the semilattice order in any CI residuated lattice.

1991 Mathematics Subject Classification. 06F05.

Key words and phrases. residuated lattice, semilattice, finitely based variety, minimal variety.
The work is a part of the research project MSM 0021620839 financed by MŠMT. A partial
support by the Grant Agency of the Czech Republic, grants #201/02/0594 and #201/02/0148 is
acknowledged.
1
2 DAVID STANOVSKÝ

1. Motivation
Our interest in this particular variety comes from the following observation.
1.1. Observation. Let V be a non-trivial subvariety of residuated lattices based (rel-
atively to RL) by equations in the language of monoids. Then V contains CIdRL
as a subvariety. (In other words, any monoid equation with a non-trivial residuated
lattice model is implied by commutativity and idempotency.)
Proof. Let u ≈ v be an equation in the language of monoids valid in V. In order
to prove that every CI residuated lattice is in V, it is enough to show that u ≈ v
holds in every semilattice. Indeed, this happens, iff the terms u and v contain the
same variables. Hence, suppose that a variable x occurs in the term u and does
not occur in the term v. Put all other variables equal to e and obtain an equation
xn ≈ e for some n, valid in V. However, this implies that V is trivial, because any
non-trivial lattice-ordered monoid contains an element a comparable to e and we
get a contradiction either by e < a ≤ a2 ≤ · · · ≤ an = e if a > e, or similarly if
a < e. 
Our motivation was the following result of Bahls, Cole, Galatos, Jipsen and
Tsinakis [1].
1.2. Theorem. Let V be a non-trivial subvariety of residuated lattices based (rela-
tively to RL) by equations in the language of lattices. Then V does not satisfy any
non-trivial monoid equation (precisely, for every equation ε in the language ·, e, if
V  ε, then all monoids satisfy ε).
Proof. Let L be a bounded lattice. We construct a residuated lattice L0 , whose
monoid reduct is the free monoid over the alphabet L and whose lattice reduct
satisfies the same lattice equations as L (it generates the same variety as L). We
identify words of length n over L with n-tuples of elements of L and define a lattice
structure on the free monoid to be the ordinal sum of L0 (consisting of the empty
word), L1 , L2 , L3 , . . . (with the empty word on top). One can check that the
resulting structure becomes a residuated lattice. Now, if a monoid identity holds
in V, it holds in L0 for every L satisfying the relative base of V. Hence it holds in
free monoids and thus in every monoid. See [1] for details. 
Is there a similar theorem, with the role of lattice and monoid reducts inter-
changed?
1.3. Theorem. The variety CIdRL does not satisfy any non-trivial lattice equation
(precisely, for every equation ε in the language ∨, ∧, if CIdRL  ε, then all lattices
satisfy ε).
Proof. Let L be a bounded lattice. We construct a CI residuated lattice L0 , whose
lattice reduct satisfies the same lattice equations as L (it generates the same variety
as L). Let us denote 1 the top element of L and e the bottom element of L. Let
L0 be the disjoint union of L and {0}. The lattice structure on L0 is defined so
that 0 is added to L as a new bottom element. We define the multiplication by
00 = 0a = a0 = 0 for every a ∈ L and ab = a ∨ b for every a, b ∈ L. It is easy to
check that this is a lattice-ordered CI monoid and it admits residuation as follows:
a/0 = 1, 0/a = 0, a/b = a for b ≤ a and a/b = 0 for b 6≤ a, a, b ∈ L. Now, if a
lattice identity holds in CIdRL, it holds in L0 for every bounded lattice L and thus
it holds in all lattices. 
 COMMUTATIVE IDEMPOTENT RESIDUATED LATTICES 3

1.4. Corollary. Let V be a non-trivial subvariety of residuated lattices based (rel-

atively to RL) by equations in the language of monoids. Then V does not satisfy
any non-trivial lattice equation.
Proof. According to Observation 1.1, the variety CIdRL is a subvariety of V and
thus Theorem 1.3 applies. 

2. Basic properties
2.1. Lemma. Let A be a lattice-ordered idempotent monoid and a, b ∈ A.
(1) a ∧ b ≤ ab ≤ a ∨ b.
(2) If a, b ≥ e, then ab = a ∨ b.
(3) If a, b ≤ e, then ab = a ∧ b.
(4) If a ≤ e ≤ ab, then ab = b.
(5) If ab ≤ e ≤ a, then ab = b.
Proof. (1) a∧b ≤ a, b ≤ a∨b, hence a∧b = (a∧b)(a∧b) ≤ ab ≤ (a∨b)(a∨b) = a∨b.
(2) If a ≥ e, then ab ≥ eb = b and similarly also ab ≥ a. Thus ab ≥ a ∨ b. The
other inequality was proven in (1). Similarly for (3).
(4) b = eb ≤ abb = ab ≤ eb = b. Similarly for (5). 
The following two statements about congruence lattices of CI residuated lat-
tices are immediate consequences of results in [4] and [9]. The second sentence of
Proposition 2.2 appears also in [8] (in a more general setting).
2.2. Proposition. The congruence lattice of A is isomorphic to the lattice of filters
on A− . In particular, if A is finite, then Con(A) ' (A− )∂ .
Proof. Blount and Tsinakis described in [4] a correspondence between congruences
of a residuated lattice A and convex normal submonoids of A− . We prove that
convex normal submonoids in CI residuated lattices are precisely filters.
Let M ⊆ A− . Since a ∧ b = ab for all a, b ≤ e, M is closed on meet iff it is
closed on multiplication. If e ∈ M (it indeed is, whenever M is a submonoid or a
filter), then M is convex iff it is an upper set. Hence, it remains to show that every
filter is normal. Since (ba)/b = (ab)/b ≥ a for all a, b, every conjugation mapping
γ(x) = ((bx)/b) ∧ e maps a negative element onto a greater one. Consequently,
congruences of a CI resuduated lattice correspond to filters. 
2.3. Corollary. A CI residuated lattice A is simple, iff |A− | = 2. It is subdirectly
irreducible, iff e is completely join-irreducible.
It is well-known that residuated lattices are congruence distributive and congru-
ence permutable. In particular, the negative cone of a non-trivial CI residuated
lattice is always distributive (in fact, it is a Heyting algebra) and contains at least
two elements.

3. Finitely and non-finitely based subvarieties

3.1. Proposition. CI residuated lattices have definable principle congruences.
Proof. Principal congruences correspond to principal filters, which are, of course,
first-order definable. It can be checked easily that a congruence corresponding to
a definable convex normal submonoid is also definable (generally for residuated
lattices). 
4 DAVID STANOVSKÝ

In fact, N. Galatos proved a stronger result in [8]: principal congruences in

commutative n-potent residuated lattices are equationally definable. This result is
indeed more complicated.
3.2. Corollary. A subvariety V of CIdRL is finitely based, iff the class of subdi-
rectly irreducible algebras in V is first-order definable.
Proof. This is an immediate consequence of a theorem of K. Baker and J. Wang
[2]. 
A non-finitely based variety of lattices was found by R. McKenzie in [11]. He
constructed an infinite independent family ε1 , ε2 , . . . of lattice equations and finite
lattices B1 , B2 , . . . such that Bn 6 εn and Bn  εm for every m 6= n. We modify his
construction to get an example of a non-finitely based subvariety of CI residuated
lattices.
3.3. Proposition. Let V be a variety with a lattice reduct and assume that for every
finite lattice L there is an algebra AL ∈ V such that L and (AL , ∨, ∧) satisfy the
same lattice equations. Then the subvariety of V based (relatively to V) by ε1 , ε2 , . . .
is not finitely based.
Proof. Let us denote the subvariety W. If there were a finite base Σ of W, by the
compactness theorem, only finitely many εi ’s were necessary to prove that Σ holds
in W. Thus there is n such that CIdRL, ε1 , . . . , εn  Σ. Hence, since Σ is a base
of W, a CI residuated lattice is in W, iff it satisfies ε1 , . . . , εn . But it means that
ABm+1 ∈ W, because Bm+1 satisfies all equations ε1 , . . . , εm . On the other hand,
W  εm+1 and ABm+1 6 εm+1 . This is a contradiction. 
Proposition 3.3 applies to the variety CIdRL; we can take, for example, AL =
L0 from the proof of Theorem 1.3. It applies also to the variety of cancellative
residuated lattices, if we take AL = L0 from the proof of Theorem 1.2.

4. More examples
W
WA complete lattice L is called infinitely join distributive, if x∈X (x ∧ y) =
( x∈X x) ∧ y holds for any X ⊆ L and y ∈ L.
Example. Let D be a complete infinitely join distributive Wlattice. Then the algebra
(D, ∨, ∧, ∧, 1, /) is a CI residuated lattice, where a/b = W{c : c ∧ b ≤ a}. (Indeed,
since a/b is the greatest c such that c ∧ b ≤ a, it must be {c : c ∧ b ≤ a}. And the
big join is less than a, if D is infinitely join distributive.)
Example. Let L be a bounded lattice and D a complete infinitely join distributive
lattice, suppose L ∩ D = ∅. We construct a CI residuated lattice L t D on the set
L ∪ D. Let L, D be sublattices of L t D with all elements of L greater then any
element of D. Denote e the bottom element of L and t the top element of D, while
0, 1 refer to the top and bottom of L t D. Put ab = a ∨ b for a, b ∈ L, ab = a ∧ b
for a, b ∈ D and ab = ba = b for a ∈ L, b ∈ D. It is easy to check that this is a
lattice-ordered CI monoid and it admits residuation as follows:
• a/b = a for e ≤ b ≤ a.
• a/b = 1 for b ≤ a, b ≤ e.
• a/b = a for a ≤ e ≤ b.
• a/b = t for b 6≤ a, a, b ≥ e.
 COMMUTATIVE IDEMPOTENT RESIDUATED LATTICES 5

W
• a/b = {c ∈ D : c ∧ b ≤ a} for b 6≤ a, a, b ≤ e.
Consequently, for every bounded lattice L and complete infinitely join distribu-
tive lattice D, there is a CI residuated lattice A with (A+ , ∨, ∧) = L, (A− , ∨, ∧) =
D+{e} and all elements comparable to e. Note that the lattice LtD is subdirectly
irreducible.
In particular, there exists a simple CI residuated lattice L0 with (L0+ , ∨, ∧) =
L (take D trivial). By Lemma 2.1(2), any simple CI residuated lattice with no
elements incomparable to the unit is some L0 . Also, by Jónsson’s lemma, L0 ’s are
the only subdirectly irreducible algebras in the variety they generate, hence they
generate a proper subvariety of CIdRL. This variety is finitely based, according to
Corollary 3.2. In fact, one can use the Galatos’ algorithm [7] and find a basis: it is
based (relatively to CIdRL) by the single equation ((e/x) ∧ e) ∨ ((y/x) ∧ e) ≈ e.
It is easy to check that there is (up to isomorphism) one 2-element CIRL, two
3-element CIRLs and four 4-element CIRLs. Using a computer, on can compute
that there are twenty 5-element CIRLs; every 5-element lattice is a reduct of a
CIRL; and in any 5-element lattice, one can choose e 6= 0, 1 arbitrarily, except for
the following case:
t

t e
@@
t @t
@@
@t

We proved that every bounded lattice is a subreduct of a CI residuated lattice.

However, there is a 6-element lattice, which is not a reduct of a CI residuated
lattice.
4.1. Proposition. Let L be a lattice and Mn be the (n + 2)-element lattice with
n atoms, n ≥ 3. Then the ordinal sum L0 of L and Mn (with L on top) is not a
reduct of a CI residuated lattice.

Lm
tc
Q AAQ
te q q q At a Qt b
 Q
QQ
 
Q

Qt
0

Proof. Assume there is a CI residuated lattice A with the lattice reduct L0 . First
of all, note that the unit element must be one of the atoms — otherwise, A− is
not a non-trivial distributive lattice. Let us denote e, a, b three distinct atoms and
assume that e is the unit element. Let c = e ∨ a ∨ b be the top element of Mn . It is
well known (see [4]) and easy to prove that in any residuated lattice multiplication
distributes over joins, in symbols
x(y ∨ z) ≈ (xy) ∨ (xz).
6 DAVID STANOVSKÝ

Using this identity, we get for every atom x 6= e in L0 that xc = x(e∨x) = x∨x = x.
Another use of this identity yields a = ac = a(e ∨ b) = a ∨ (ab) and similarly
b = b ∨ (ab), so ab ≤ a and ab ≤ b and thus ab = 0. Now, choose d ∈ L. We have
(da) ∨ (db) = d(a ∨ b) = dc = d (because multiplication coincides with the join on
positive elements). Hence, at least one of da, db must be greater than c; assume it
is da. Then c(db) ≤ (da)(db) = d(ab) = d0 = 0. However, this is possible iff db = 0,
because cx ≥ c for every x positive and we proved above that cx = x for every
atom x 6= e. But db ≥ eb = b, a contradiction. 
A different argument shows examples of infinite lattices which are not reducts
of any CI residuated lattice. Let L be an arbitrary simple atomless lattice (e.g.
the dual of the lattice of subspaces of an infinite-dimensional vector space) and let
A be a CI residuated lattice with the lattice reduct L. By adding operations to a
simple algebra, one gets again a simple algebra. Hence A is simple, but A− cannot
have two elements, because there are no atoms in A, which contradicts Corollary
2.3.
The following propositions describe all totally ordered CI residuated lattices (i.e.
those, where the lattice reduct is a chain).
4.2. Proposition. Let A = (A, ∨, ∧, ·, e) be a structure such that (A, ∨, ∧) is a
chain and (A, ·, e) is a semilattice with a unit. Then the following are equivalent.
(1) A is a lattice-ordered monoid.
(2) ab = a ∨ b for every a, b ∈ A+ , ab = a ∧ b for every a, b ∈ A− and the
semilattice reduct is a chain.
Proof. (1) ⇒ (2) follows from Lemma 2.1. If a, b are both positive or both negative,
2.1(2) or 2.1(3) applies. Otherwise, since ≤ is a chain, we may assume that a ≤
e ≤ b. In this case, either e ≤ ab and 2.1(4) applies, or ab ≤ e and 2.1(5) applies.
(2) ⇒ (1). Note that on the positive cone, a ≤ b iff b  a, and on the negative
cone, a ≤ b iff a  b. Let a ≤ b. We need to prove that ac ≤ bc for every c ∈ A.
Since (A, ) is a chain, ac ∈ {a, c} and bc ∈ {b, c}. Hence the only bad situation is
either (a) ac = a, bc = c and a > c, or (b) ac = c, bc = b and c > b. We prove that
none of them is actually possible. In (a), we have c < a < b and a ≺ c ≺ b. The
element a can’t be positive, because in this case b is also positive and a < b implies
b ≺ a. On the other hand, a can’t be negative, because then c is also negative
and c < a implies c ≺ a. This is a contradiction. In (b), we have a < b < c and
b ≺ c ≺ a and a similar argument works. 
4.3. Corollary. Let A = (A, ∨, ∧, ·, e) be a structure such that (A, ∨, ∧) is a chain
and (A, ·, e) is a semilattice with a unit. Then the following are equivalent.
(1) (A, ∨, ∧, ·, e, /) is a residuated lattice for some /.
(2) ab = a∨b for every a, b ∈ A+ , ab = a∧b for every a, b ∈ A− , the semilattice
reduct is a chain and for every a, b there is the greatest c such that ac ≤ b.
In particular, for A finite, the conditions are equivalent to
(3) ab = a ∨ b for every a, b ≥ e, ab = a ∧ b for every a, b ≤ e and the semilattice
reduct is a chain with 0 in bottom.
Proof. (1) ⇔ (2) follows obviously from the previous proposition. If (1),(2) are
true, then (3) follows from the fact that 0 exists and 0a = a0 = 0 for all a in any
residuated lattice with 0. And if (3) holds, then there is always some c, namely
 COMMUTATIVE IDEMPOTENT RESIDUATED LATTICES 7

c = 0, such that ac ≤ b, and thus there is also the greatest such c. (Note that it is
enough to assume that the dual of (A, ∨, ∧) is well-ordered with a top element, not
necessarily finite.) 

5. Minimal varieties
Minimal subvarieties of residuated lattices were investigated by several authors,
particularly by N. Galatos in [6]. He found also minimal subvarieties of CIdRL —
they are just two. We shortly reprove his result.
A residuated lattice is called integral, if all its elements are negative. Let C2 be
the two-element CI residuated lattice, C2 = {0, 1}, e = 1. Let C3 be the three-
element non-integral CI residuated lattice, C3 = {0, e, 1}, 0 < e < 1. (Note that, in
fact, C2 is the only two-element residuated lattice and C3 is the only non-integral
three-element residuated lattice.) Let V2 , V3 be the varieties generated by C2 , C3 ,
respectively. It is clear from Jónsson’s lemma that V2 and V3 are minimal varieties.
5.1. Theorem. V2 and V3 are the only minimal subvarieties of CIdRL.
Proof. We show that every non-trivial subvariety V of CIdRL contains C2 or C3 .
According to the well known Magari’s theorem, V contains a (non-trivial) simple
algebra A. Indeed, |A− | = 2, so A has the bottom and thus also the top element.
We show that B = {0, e, 1} is a subalgebra of A — then it is isomorphic to one of
C2 , C3 , depending on whether e = 1 or not. The set B is indeed closed on join,
meet and multiplication. In any bounded residuated lattice the equations x/0 ≈ 1,
x/e ≈ x and 1/x ≈ 1 hold and 0/1 ≤ e/1 < e. Hence in a simple CI residuated
lattice 0/1 = e/1 = 0 and we are done. 
V2 is known as the variety of generalized Boolean algebras and it is based (rel-
atively to CIdRL) by x ≤ e and y/(y/x) ≈ x ∨ y. A finite base for the variety V3
can be found in [6] (or computed by the Galatos’ algorithm).
In fact, N. Galatos proved in [6] that C2 or C3 is a subalgebra of any idempotent
residuated lattice A satisfying e/x ≈ x\e. If A is integral, then {a, e} is a subalge-
bra isomorphic to C2 for every a 6= e and if A is not integral, then {e/a, e, e/(e/a)}
is a subalgebra isomorphic to C3 for every a > e. Consequently, every subvariety of
CIdRL is either integral, or contains C3 (in other words, C3 is a splitting algebra).

Aknowledgement. The paper was written during the author’s stay at Vanderbilt
University in Nashville, TN, in fall 2003 and 2004. The author is indebted to R.
McKenzie for the invitation to Vanderbilt and also to N. Galatos, M. Kozik and C.
Tsinakis for their interest, valuable comments and corrections.

References
[1] P. Bahls, J. Cole, N. Galatos, P. Jipsen, C. Tsinakis, Cancellative residuated lattices, Algebra
Universalis 50/1 (2003), 83–106.
[2] K. Baker, J. Wang, Definable principal subcongruences, Algebra Universalis 47/2 (2002),
145–151. Zbl 1063.08005
[3] S. Burris, H.P. Sankappanavar, A course in universal algebra, GTM 78, Springer, 1981. Zbl
0478.08001
[4] K. Blount, C. Tsinakis, The structure of residuated lattices, Internat. J. Algebra Comput.
13/4 (2003), 437–461. Zbl 1048.06010
[5] R.P. Dilworth, M. Ward, Residuated lattices, Trans. Amer. Math. Soc. 45 (1939), 335–354.
Zbl 0021.10801
 8 DAVID STANOVSKÝ

[6] N. Galatos, Minimal varieties of residuated-lattices, Algebra Universalis 52/2-3 (2004), 215–
239.
[7] N. Galatos, Equational bases for joins of residuated-lattice varieties, Studia Logica 76 (2004),
227–240. Zbl 1068.06007
[8] N. Galatos, Varieties of residuated lattices, PhD Thesis, Vanderbilt University, 2003.
[9] J. Hart, L. Rafter, C. Tsinakis, The structure of commutative residuated lattices, Internat. J.
of Algebra Comput. 12 (2002), 509–524. Zbl 1011.06006
[10] P. Jipsen, C. Tsinakis, A survey of residuated lattices, Ordered Algebraic Structures (J.
Martinez, ed.), Kluwer Academic Publishers, Dordrecht, 2002, 19–56. Zbl pre02208113
[11] R. McKenzie, Equational bases for lattice theories, Math. Scand. 27 (1970), 24–38. Zbl
0307.08001
[12] R. McKenzie, G. McNulty and W. Taylor, Algebras, Lattices, Varieties, Volume I. Wadsworth
& Brooks/Cole, Monterey, CA, 1987. Zbl 0611.08001

David Stanovský, Charles University in Prague, Czech Republic

[email protected]

View publication stats