Professional Documents
Culture Documents
net/publication/320212951
CITATIONS READS
4 379
2 authors, including:
George Grätzer
University of Manitoba
359 PUBLICATIONS 8,332 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by George Grätzer on 03 November 2017.
Abstract. Let the finite distributive lattice D be isomorphic to the congruence lat-
tice of a finite lattice L. Let Q denote those elements of D that correspond to
principal congruences under this isomorphism. Then Q contains 0, 1 ∈ D and all the
join-irreducible elements of D. If Q contains exactly these elements, we say that L is
a minimal representations of D by principal congruences of the lattice L.
We characterize finite distributive lattices D with a minimal representation by
principal congruences with the property that D has at most two dual atoms.
1. Introduction
1.1. The problem. Let Princ L denote the ordered set of principal congru-
ences of the finite lattice L. Then
Let us say that a finite lattice L has a minimal set of principal congruences if
we have equality in (3), that is,
1.2. Two illustrations. We provide two examples. The first one is from
G. Grätzer and H. Lakser [17].
Illustration 1. The eight element Boolean lattice B3 = C32 has no minimal
representation.
See [17] for a proof. Basically, if the lattice L is a minimal representation,
then on any maximal chain in L, we find two adjacent prime intervals generat-
ing distinct atoms of Con L. The two intervals together form an interval that
generates the join of two atoms of Con L, contradicting minimality.
Illustration 2. The nine element distributive lattice D = C23 has a minimal
representation.
We take N6 (see the first diagram of Figure 1) as a minimal representa-
.
tion of the chain C3 . Then the glued sum N6 + N6 (see the second diagram
of Figure 1) is a congruence representation of D = C23 but it is not minimal;
indeed, con(a, b), con(b, c) < con(a, c) so con(a, b)∨con(b, c) = con(a, c) is prin-
cipal and joint-reducible. The third diagram of Figure 1 provides a minimal
representation of D = C23 .
.
Figure 1. The lattices N6 and N6 + N6
1.4. Related results. This paper continues G. Grätzer [10] (see also [9, Sec-
tion 10-6] and [12, Part VI]), whose main result is the following statement.
1.5. Notation. We use the notation as in [12]. You can find the complete
Part I. A Brief Introduction to Lattices and Glossary of Notation
of [12] at
tinyurl.com/lattices101
Proof. Let
x = c0 ≺ c1 ≺ · · · ≺ cn = y
be a (maximal) chain C in the interval [x, y] and let β i = con(ci , ci+1 ) for
0 ≤ i < n Then β i is a join-irreducible congruence of L and
_ _
( β i | 0 ≤ i < n ) = con(x, y) = A.
Proof. Assume that the finite lattice L with bounds o and i provides a minimal
representation of D, that is, Princ L = {0, 1} ∪ J(Con L) and there is an
isomorphism between D and Con L. Let Q ⊆ D correspond to Princ L under
this isomorphism.
By Theorem 8 applied to the interval [o, i], for each a ∈ A there is a b ∈
J(D), such that a ∨ b is a join-reducible element of Q. Since A has at least
3 elements, a ∨ b 6= 1, a contradiction.
Vol. 00, XX Minimal representations by principal congruences 5
Corollary 10. Let D = B3 , the eight element Boolean lattice. Then D does
not have a minimal representation.
Theorem 11. Let D be a finite distributive lattice with more than two dual
atoms. Then D does not have a minimal representation.
3.1. Preliminaries. We will need the Technical Lemma for Finite Lattices,
see G. Grätzer [11].
br
a0 = b0 a1 = b1
ar
fp,q
dp,q hp,q
ep,q
c ∗p,q
cp,q gp,q
br
a0 = b0 a1 = b1
ar
a′r
a′q a′p
a′1 = b′1
a′0 = b′0
b′q b′p
b′r
i = i′
br
a1 = b1
a0 = b0
ar
a′ a′r
b′ u′ b′r u′r
i = i′ t m s i = i′ tr mr sr
b br ur
u
a ar
o′
a′r
b′q a′p
a′1 = b′1
a′0 = b′0 u′r
a′q b′p b′r
i = i′ sr
tr mr
br
ur
a1 = b1
a0 = b0
ar
for x ∈ K,
(
u for x ≤ a;
u∨x=
b ∨ x, otherwise,
and dually.
Vol. 00, XX Minimal representations by principal congruences 9
We apply Lemma 13 five times for each r ∈ R, that is, we add tr , ur , u0r , sr ,
and mr successively, to conclude that the lattice L obtained for Theorem 4(ii)
in the previous section is a lattice.
Lemma 14. The relation <F is a strict order relation on F = Fuse(P, A) and
ψA : P → Fuse(P, A) is an isotone map.
We are finally left with the case where x ≮P y and y ≮P x. Then there are
a1 , a2 , a3 , a4 ∈ A with x <P a1 , a2 <P y, y <P a3 , and a4 <P x. Then for
instance, a4 <P x <P a1 , again contradicting the convexity of A.
Consequently, <F is antisymmetric.
Finally, we establish transitivity. So let x, y, z ∈ F with x <F y <F z.
We first consider the case x = ι. Then y, z ∈ / A and there is an a ∈ A with
a <P y. By the convexity of A, y <F z cannot follow from the second case
in (7). Then y <P z and so a <P z, that is, x = ι <F z.
If z = ι, we use the dual argument.
If y = ι, then by (5), (6), and the second case in (7), we get x <F z.
If one of x, y, z is ι, we then have transitivity. So let x, y, z all differ from
ι. If x <P y and y <P z, then x <P z, and so x <F z. On the other hand,
if, say, x ≮P y, then the second case of (7) holds, and so x <F ι <F y <F z.
By transitivity whenever one of the three entries is ι, we get first ι <F z, and
then x <F z. Similarly, x <F z if y ≮P z.
Thus transitivity has been established, concluding the proof that <F is an
order relation.
It is immediate that ψA is isotone. Indeed, let x, y ∈ P with x <P y.
If x, y ∈ A, then ψA (x) = ι = ψA (y). If x ∈ A and y ∈ / A, then by (5),
ψA (x) = ι <F y = ψA (y). Similarly, ψA (x) <F ψA (y) if y ∈ A and x ∈ / A. If
x, y ∈/ A, then by the first case of (7), ψA (x) = x <F y = ψA (y).
The ordered set Fuse(P, A) is the ”freest” ordered set with the convex subset
A fusing to a single element. This is formalized by the following Universal
Mapping Property.
Lemma 15. Let P, Q be ordered sets, let A be a nonempty convex subset of P ,
and let ϕ : P → Q be an isotone map with ϕ(a1 ) = ϕ(a2 ) for all a1 a2 ∈ A.
Then there is an isotone map ϕ0 : Fuse(P, A) → Q with
ϕ0 ψA = ϕ, (8)
0
and ϕ is determined uniquely by (8).
Proof. Let a ∈ A be arbitrary. Then we must have ϕ0 (ι) = ϕ(a). For x ∈/ A,
we must have ϕ0 (x) = ϕ(x). So the isotone property for ϕ0 follows from (5),
(6) and (7).
Lemma 16. Let P and Q be ordered sets, and let A be a nonempty convex
subset of P . Let ϕ : P → Q be a surjective isotone map with ϕ(a1 ) = ϕ(a2 )
for all a1 , a2 ∈ A. Assume that, for all x, y ∈ P with x P y and with
ϕ(x) ≤ ϕ(y), there are a1 , a2 ∈ A with x ≤P a1 and a2 ≤P y. Then the
isotone map ϕ0 : Fuse(P, A) → Q determined by the condition ϕ0 ψA = ϕ is an
isomorphism.
Proof. As before, we set F = Fuse(P, A). The map ϕ0 is surjective since the
map ϕ is. Thus we need only show that, for x, y ∈ F , whenever ϕ0 (x) ≤ ϕ0 (y),
then x ≤F y. We may assume that x, y are distinct.
Vol. 00, XX Minimal representations by principal congruences 11
Note that admissibility is a self-dual concept: (iii) is the dual of (i), and
(iv) is the dual of (ii).
Although we make no use of it in this paper, observe that the Tab Lemma
(G. Czédli, G. Grätzer, and H. Lakser [8, Lemma 12]) easily follows.
Corollary 23 (Tab Lemma). Let L be any finite lattice and let u be a tab
of L in the covering multidiamond [a, b]. Set K = L − {u}, a sublattice of L.
Let α be a congruence relation on K, and set β = conL (α).
If a 6≡ b (mod β), then βeK = α.
If a ≡ b (mod β), then βeK = α ∨ conK (a, b).
16 G. Grätzer and H. Lakser Algebra univers.
Proof. Let c and c0 be two other atoms of the multidiamond [a, b]. The lattice
L then is obtained by adding a relative complement of c in [a, b]K . Thus
Corollary 22 applies.
The element a is meet-reducible in K and b is join-reducible because of the
element c0 . By Corollary 22, if α is not admissible, then βeK = α∨conK (a, b).
Furthermore, a ≡ b (mod β). So if α is not admissible, then a ≡ b (mod β)
and βeK = α ∨ conK (a, b).
On the other hand, if α is admissible, then βeK = α which is α∨conK (a, b)
exactly when a ≡ b (mod α) exactly when a ≡ b (mod β).
Summarizing, if a 6≡ b (mod β), then α is admissible, and βeK = α, and if
a ≡ b (mod β), then βeK = α∨conK (a, b) whether or not α is admissible.
We present three results that serve as the foundation of the bridge construc-
tion.
or
conK (b, c) ≤ α and conL (α)eK = α ∨ conK (a, c).
Proof. If a 6≡ c (mod α) and b 6≡ c (mod α), then 20(ii) and (iv) hold.
Now a ≡ c0 (mod α) iff b ≡ c (mod α) and b ≡ c0 (mod α) iff a ≡ c
(mod α). Since c0 and c are the only upper covers in K of a and the only
lower covers in K of b, 20(i) and (iii) also hold. Thus if a 6≡ c (mod α) and
b 6≡ c (mod α), then α is admissible.
Vol. 00, XX Minimal representations by principal congruences 17
Lemma 26. Let the lattice K0 be a sublattice of the lattice K1 , and let
x, y, z, w ∈ K0 . If conK0 (x, y) ≤ conK0 (z, w), then conK1 (x, y) ≤ conK1 (z, w).
Now there are three possibilities for conL (x1 , y1 )eK. Either conK (x1 , y1 ) is
admissible, and so we have
conL (x1 , y1 )eK = conK (x1 , y1 ), (15)
or either
conK (a, c) ≤ conK (x1 , y1 ) and conL (x1 , y1 )eK = conK (x1 , y1 ) ∨ conK (b, c)
or
conK (b, c) ≤ conK (x1 , y1 ) and conL (x1 , y1 )eK = conK (x1 , y1 ) ∨ conK (a, c)
holds by Lemma 25. Thus by (14), either
conK (a, c) ≤ conK (x1 , y1 ) and conK (x0 , y0 ) ≤ conK (x1 , y1 ) ∨ conK (b, c) (16)
or
conK (b, c) ≤ conK (x1 , y1 ) and conK (x0 , y0 ) ≤ conK (x1 , y1 )∨conK (a, c). (17)
Since x0 ≺ y0 , conK (x0 , y0 ) is a join-irreducible congruence relation. By (13),
(16) implies
conK (b, c) ≤ conK (x1 , y1 ) and conK (x0 , y0 ) ≤ conK (a, c),
that is, implies (b).
Similarly, (17) implies (c).
Thus if (a) fails, then (12) implies that either (b) or (c) holds. Consequently,
(12) implies that at least one of (a), (b), (c) holds, concluding the proof.
We start with the Bridge Theorem, which examines how the congruences
of a bridge extension behave.
or
iff either
conK (x0 , y0 ) ≤ conK (x1 , y1 )
or
conK (x0 , y0 ) ≤ conK (a, b) and conK (a0 , b0 ) ≤ conK (x1 , y1 )
or
conK (x0 , y0 ) ≤ conK (a0 , b0 ) and conK (a, b) ≤ conK (x1 , y1 ),
thereby verifying (vi) and concluding the proof of the theorem.
Let K and L be as in Theorem 28, and let conK (a, b) k conK (a0 , b0 ). Then
we can fuse the set A = {conK (a, b), conK (a0 b0 )} and get the ordered set
Fuse(J(Con K), A). Applying Lemma 16, we get an immediate corollary:
Corollary 29. Assume that conK (a, b) k conK (a0 , b0 ) and set
and
ϕ0 : conK (x, y) 7→ conL (x, y),
otherwise, is an order isomorphism.
Theorem 30. Let the finite lattice K be as in Theorem 28 and let K further-
more satisfy the following five conditions:
(i) If x < i, then con(x, i) = con(b, i).
(ii) If x > i, then con(x, i) = con(b0 , i).
(iii) If x < y, then either con(x, y) = 1 or con(x, y) is a join-irreducible
congruence.
(iv) If x < b and x 6= a, then con(x, b) = con(b, i).
(v) If x > b0 and x 6= a0 , then con(x, b0 ) = con(b0 , i).
Let L be obtained by attaching the bridge Bridge depicted in Figure 6 to K as
in Theorem 28. Then L satisfies conditions (i), (ii), and (iii).
Vol. 00, XX Minimal representations by principal congruences 21
Proof. Let x ∈ L with x < i. Then x ∈ K and so, by (i) for K, conK (x, i) =
conK (b, i). Then conL (x, i) = conL (b, i), verifying (i) for L.
(ii) is just the dual of (i).
Now let us denote by o the zero of K, and so of L, and by o0 the unit of K,
and so of L. By (i) and (ii), which were assumed for K and verified for L, we
note that if x < i < y, then, whether we are referring to K or L, con(x, y) = 1.
Indeed,
con(x, y) = con(x, i) ∨ con(y, i) = con(b, i) ∨ con(b0 , i)
= con(o, i) ∨ con(o0 , i) = con(o, o0 ) = 1.
We now verify (iii) for L.
Assume first that x, y ∈ K. By (iii) for K, either
conK (x, y) = 1K = conK (o, o0 )
and so
conL (x, y) = conL (o, o0 ) = 1L ,
or there are v, w ∈ K with v ≺ w in K and with conK (x, y) = conK (v, w).
Then conL (x, y) = conL (v, w) and, by Theorem 28(i), v ≺ w in L, that is,
conL (x, y) is join-irreducible.
Otherwise, by duality, we can assume that y ∈
/ K. If x ≺ y in L, we are
done. So we may assume that y does not cover x in L Then x ∈ K, and so
x ≤ b.
Assume, first, that x = b. Then y = u0 . So
conL (x, y) = conL (b, t) ∨ conL (t, u0 ) = conL (i, b0 ) ∨ conL (a0 , b0 ) = conL (i, b0 ),
since
conL (a0 , b0 ) ≤ conL (a0 , i) = conL (i, b0 )
by (ii). Thus if x = b, then conL (x, y) is join-irreducible.
Thus we are left with the case when x < b.
Assume further that x = a. Then y, not covering x, must be one of
t, m, s, u0 . Then conL (u, y) = conL (a0 , b0 ), and so
conL (x, y) = conL (a, u) ∨ conL (u, y) = conL (i, b0 ) ∨ conL (a0 , b0 ) = conL (i, b0 )
since conL (a0 , b0 ) ≤ con(i, b0 ). Thus again, conL (x, y) is join-irreducible.
We are then left with the case x < b and x 6= a.
By (iv) for K, conK (x, b) = conK (b, i). Therefore, conL (x, b) = conL (b, i).
Since y ∈
/ K, y ≥ t. Then
1L = conL (b, b0 ) = conL (b, i)∨conL (b0 , i) = conL (x, b)∨conL (b, t) ≤ conL (x, y),
since b < i < b0 . So in this final case, we have conL (x, y) = 1L .
We have thus verified (iii) for L, concluding the proof.
Lemma 31. Let the finite lattice K be as in Theorem 28, and let L be obtained
by attaching the bridge Bridge depicted in Figure 6. Assume further that there
are elements a0 , b0 ∈ K, with a0 6= a, b0 6= b, with a0 ≺ b0 ≺ i in K, such that
for any x ∈ K with x 6= a0 and x < b0 , we have conK (x, b0 ) = conK (b0 , i).
Then for any x ∈ L with x 6= a0 and x < b0 , we have conL (x, b0 ) = conL (b0 , i).
8.1. Some technical results. We first summarize the properties of the lat-
tice FrameW P of G. Grätzer [16].
Lemma 32. Let P be a finite ordered set with a greatest element p0 and let
L = FrameW P . There is an order isomorphism ζP : P → J(Con L) such that
ζP (p0 ) = 1 and such that the following five statements hold.
(i) α ∈ J(Con L) for all α ∈ Princ L.
(ii) conL (x, i) = 1 = ζP (p0 ) for all x ∈ L distinct from i.
(iii) For each p ∈ P distinct from p0 , there are ap , bp ∈ L with ap ≺ bp ≺ i
such that conL (ap , bp ) = ζP (p).
(iv) ap and bp are meet-irreducible for each p ∈ P distinct from p0 .
(v) For each p ∈ P distinct from p0 and each x ∈ P distinct from ap , if
x < bp , then conL (x, bp ) = 1 = ζP (p0 ).
Let P be a finite ordered set with exactly two maximal elements, p0 and p1 .
To prove Theorem 4(ii), we proceed by mathematical induction on the size of
the subset ↓ p0 ∩ ↓ p1 . First, the result that provides the base of the induction.
Lemma 33. Let P be a finite ordered set with exactly two maximal elements
p0 and p1 , and let ↓ p0 ∩ ↓ p1 = ∅. Then there is a finite lattice L with
zero o, unit o0 , and element i distinct from o and o0 , and there is an order
isomorphism ζP : P → J(Con L) such that the following nine statements hold.
(i) If α ∈ Princ L and α 6= 1, then α ∈ J(Con L).
(ii) conL (x, i) = ζP (p0 ) for all x ∈ L with x < i.
(iii) conL (x, i) = ζP (p1 ) for all x ∈ L with x > i.
(iv) For each p ∈ P with p < p0 , there are ap , bp ∈ L with ap ≺ bp ≺ i such
that conL (ap , bp ) = ζP (p).
(v) For each p ∈ P with p < p1 , there are a0p , b0p ∈ L with i ≺ b0p ≺ a0p such
that conL (a0p , b0p ) = ζP (p).
(vi) For each p ∈ P − {p0 , p1 }, if p p1 , that is, if p < p0 , then ap and bp
are meet-irreducible in L.
(vii) For each p ∈ P − {p0 , p1 }, if p p0 , that is, if p < p1 , then a0p and b0p
are join-irreducible in L.
Vol. 00, XX Minimal representations by principal congruences 23
a′q1
aq0
since conL (aq0 , bq0 ) = conL (a0q1 , b0q1 ). By Lemma 15, we get
ϕ0 : Fuse(J(Con K), A) → J(Con L),
where
ϕ0 : ι 7→ conL (aq0 , bq0 ) = conL (a0q1 , b0q1 )
and
ϕ0 : conK (x, y) 7→ conL (x, y),
otherwise.
By Theorem 28(v), ϕ is surjective, and so
ϕ0 : Fuse(J(Con K), A) → J(Con L)
is an order isomorphism, by Theorem 28(vi) and Lemma 16. The order iso-
morphism
ζQ : Q → J(Con K)
yields an order isomorphism
ζ 0 : Fuse(Q, {q0 , q1 }) → Fuse(J(Con K), A)
with ζ 0 : ι{q0 ,q1 } 7→ ιA and ζ 0 : p 7→ ζQ (p) otherwise. We then have the order
isomorphism
ϕ0 ζ 0 : Fuse(Q, {q0 , q1 }) → J(Con L)
with
ι{q0 ,q1 } 7→ conL (aq0 , bq0 ) = conL (a0q1 , b0q1 )
and p 7→ ϕζQ (p) otherwise. By Lemma 18, there is an order isomorphism
P → Fuse(Q, {q0 , q1 }) with q 7→ ι{ q0 , q1 } and p 7→ p otherwise. We then get
the desired order isomorphism
ζP : P → J(Con L)
with
ζP : p 7→ ϕ(ζQ (p)) = conL (ζQ (p)) (21)
for p 6= q and
ζP : q 7→ conL (aq0 , bq0 ) = conL (a0q1 , b0q1 ).
and, by (21) and statements (iv) and (v) for Q and K, if p 6= q,
(
ϕ(conK (ap , bp )) = conL (ap , bp ) for p ≤ p0 ,
ζP : p 7→ 0 0 0 0
ϕ(conK (ap , bp )) = conL (ap , bp ) for p ≤ p1 .
Thus statements (iv) and (v) hold for Q and L.
We now verify the other seven statements for Q and L.
By statements (ii), (iii), (i), (viii), and (ix) for Q, K, and Theorem 30, it
follows that statements (i), (ii), and (iii) hold for P , L.
Now let p ∈ P − (↓ p1 )P be distinct from p0 . Then p ∈ Q − (↓ p1 )Q , and is
distinct from q0 as well. Then by statement (vi) for Q, K, the elements ap , bp
are meet-irreducible in K. Then by Theorem 28(ii), ap , bp are meet-irreducible
26 G. Grätzer and H. Lakser Algebra univers.
Theorem 35. Let D be a finite distributive lattice with exactly two dual atoms.
Then D has a minimal representation L.
References
[1] Czédli, G.: Representing a monotone map by principal lattice congruences. Acta
Mathematica Hungarica 147, 12–18 (2015)
[2] Czédli, G.: An independence theorem for ordered sets of principal congruences and
automorphism groups of bounded lattices. Acta Sci. Math (Szeged) 82, 3–18 (2016)
[3] Czédli, G.: The ordered set of principal congruences of a countable lattice. Algebra
Universalis 75, 351–380 (2016)
[4] Czédli, G.: Representing some families of monotone maps by principal lattice
congruences. Algebra Universalis 77, 51–77 (2017)
[5] Czédli, G.: Cometic functors and representing order-preserving maps by principal
lattice congruences. Algebra Universalis (2017, in press)
[6] Czédli, G.: Characterizing fully principal congruence representable distributive
lattices. Algebra Universalis (2017, in press)
[7] Czédli, G.: On the set of principal congruences in a distributive congruence lattice of
an algebra. Acta Sci. Math (Szeged), submitted.
[8] Czédli, G., Grätzer, G., Lakser, H.: Congruence structure of planar semimodular
lattices: The General Swing Lemma. Algebra Universalis (2017, in press)
[9] Grätzer, G.: Lattice Theory: Foundation. Birkhäuser Verlag, Basel (2011)
[10] Grätzer, G.: The order of principal congruences of a bounded lattice. Algebra
Universalis 70, 95–105 (2013)
[11] Grätzer, G.: A technical lemma for congruences of finite lattices. Algebra Universalis
72, 53 (2014)
[12] Grätzer, G.: The Congruences of a Finite Lattice. A ”Proof-by-Picture” Approach.
Second edition. Birkhäuser Verlag, Basel (2016)
[13] Grätzer, G.: Homomorphisms and principal congruences of bounded lattices. I.
Isotone maps of principal congruences. Acta Sci. Math. (Szeged) 82, 353–360 (2016)
[14] Grätzer, G.: Homomorphisms and principal congruences of bounded lattices. II.
Sketching the proof for sublattices. Algebra Universalis (2017, in press)
[15] Grätzer, G.: Homomorphisms and principal congruences of bounded lattices. III. The
Independence Theorem. Algebra Universalis (2017, in press)
[16] Grätzer, G.: Characterizing representability by principal congruences for finite
distributive lattices with a join-irreducible unit element. Acta Math Szeged (2017, in
press)
[17] Grätzer, G., Lakser, H.: Some preliminary results on the set of principal congruences
of a finite lattice. Algebra Universalis (2017, in press)
Vol. 00, XX Minimal representations by principal congruences 27
[18] Grätzer, G. and Wehrung, F. eds., Lattice Theory: Special Topics and Applications.
Volume 1. Birkhäuser Verlag, Basel (2014)