• Nie Znaleziono Wyników

Lattices with doubly replaceable decompositions1. Abstract. We study meet decompositions of an element of a complete lattice. We give

N/A
N/A
Protected

Academic year: 2021

Share "Lattices with doubly replaceable decompositions1. Abstract. We study meet decompositions of an element of a complete lattice. We give"

Copied!
8
0
0

Pełen tekst

(1)

Series I: COMMENTATIONES MATHEMATICAE XXX (1991) ROCZN1KI POLSKIEGO TOWARZYSTWA MATEMATYCZNEGO

Séria I: PRACE MATEMATYCZNE XXX (1991)

An d r z e j Wa l e n d z i a k

(Warszawa)

Lattices with doubly replaceable decompositions

1. Abstract. We study meet decompositions of an element of a complete lattice. We give a generalization of some results of [3].

2. Basic notions. Let L be a complete lattice. Lattice join, meet, inclusion and proper inclusion are denoted respectively by the symbols v , л , ^ and <.

Let 0 be the least element and 1 the greatest element of L.

An element m e L is called irreducible iff, for all x, y e L , m = x л у implies m = x or m = у. M(L) is defined to be the set of all irreducible elements of L.

If aeL, then a representation a = f \ T with T Ç M(L) is called a (meet) decomposition of a. A decomposition a = Д T is irredundant if Д (T— {t}) Ф a for all t e T.

If a = Д T is a decomposition of a, we define t i , • • •, tn. = Д T—({tl , , tn}), for each subset {tl5 ..., tn} of T.

We say that a e L has doubly replaceable irredundant decompositions iff it has at least one irredundant decomposition and whenever a = Д T = Д R are two irredundant decompositions of a, for each t0 e T there exists r0 e R such that

a = t0 A r 0 = t0 A r 0 (1).

The lattice L has doubly replaceable irredundant decompositions iff each a e L has doubly replaceable irredundant decompositions.

A complete lattice L is called upper continuous iff, for every a e L and for every chain C ^ L, a

a \ J

C

= \ J

(a

a

c: ceC). It is well known that every algebraic lattice is upper continuous.

For two elements a, b e L (a ^ b) we define b/a : = [xeL : a ^ x ^ b}.

If b/a = {a, b}, then we say that b covers a, in notation: a -<b. A lattice L is said to be strongly atomic if, for any a, b e L with a < b, there exists peb/a such that a-<p.

O Cf. [1], p. 252.

(2)

In a complete strongly atomic lattice L, for each aeL , let Pa denote the set of all elements covering a, and set ua: = \ f P0.

If, for every a eL , the sublattice u ja is modular, then we say that L is locally modular.

3. Some lemmas. Throughout this section L will denote an upper continuous, strongly atomic, locally modular lattice. In the proofs of Theorems 3.7 and 7.3 of [2] it was not used that L is an algebraic lattice but only that L is upper continuous. Therefore, from 3.7 of [2] it follows that the lattice L is semimodular, and by 7.3 of [2] we have

Le m m a

1. I f a , b , p 1,p 2eL, and if pt , p2eP a, b л {px v p2) = a and Pi v b = p2 v b, then = p2.

Now we shall prove

Le m m a 2 .

For each a e L the sublattice u ja is a point lattice, i.e., for every pair of elements x > y in ua/a, x is the join of elements covering y.

P ro o f. Let x, y e u ja with y < x. By the proof of Lemma 4.1 of [3], every element of u ja is the join of elements of Pa. Then

x = \ / ( p v y: a ^ p ^ x , p £ y ).

By semimodularity, y -< pv y whenever a-<p y. Therefore, x is the join of elements covering y and hence ua/a is a point lattice.

Le m m a

3. Let a be an element of L, and let x e l / a and y e u ja . I f x —<x v y, then x

а

у —су.

P ro o f. Since ua/a is a point lattice, we need only show that b = x л у is irreducible in the sublattice y/b. Suppose on the contrary that y/b contains two distinct atoms p 1 and p2. Obviously we have b ^ x л (pt v p2) ^ x л y = b, i.e.,

(1) x A ( p j v p 2) = b.

By semimodularity, x —c x v p ^ < x v j and x - < x v p 2 ^ x v y . Since x —<x v y we have

(2) x v p j = x v p 2 = x v y .

From (1) and (2) we conclude, by Lemma 1, that pA = p2. This contradiction shows that b is irreducible in y/b. Thus the proof is complete.

Le m m a

4. Let a e L and let p t , p2 be two distinct elements covering a. I f m eM (L), m ^ a and m ^ pt , p2, then a —cm л (p2 v p2).

P ro o f. Apply Lemma 3 and the proof of Lemma 4.2 of [3].

De f in it io n

(cf. [3], p. 10). If a = Д T is a decomposition of a, let

H T: = {peP a: p < t for some te T } and hT: = \J H T.

(3)

Lemma 5.

Let a = Д

Г

be an irredundant decomposition of a. Then (i) For each p e H T there exists a unique t(p)eT such that p < t(p).

(ii) For each t e T there is a unique t e P a with t ^ t.

(iii) H T is an independent subset of Pa.

P ro o f, (i) Let p e H T and suppose that there are two distinct tt , t2e T such * that p < Fx and p ^ F2. Since ф t2, t2 ^ t1 and therefore p ^

a

= a, contrary to the assumption that p e H T. Thus t t = t2 and (i) holds.

(ii) Let te T . Since the decomposition a = Д T is irredundant and the lattice L is strongly atomic there is p e P a with p ^ t. Suppose that there are two distinct such elements p1, p2. Hence p l5 p2 ^ t, and therefore, by Lemma 4, a -<t л (pl v p2). Since this contradicts t л (pt v p2) ^ t

a

t = a we must have

Pi = Pi

and (ü) follows.

(iii) Now, let p e H T and q e H T — {p}. Then p ^ t(p) and q < t(q) < tip).

Therefore, \ / (H T — {p}) ^ t(p). Since p ^ t(p) it follows that p ^ \ J (HT — {p}).

Thus H T is independent and (iii) holds.

The next lemma is a generalization of Lemma 4.4 of [3].

Lemma 6.

Let a = Д

T

be an irredundant decomposition of aeL. I f T = (tl5 ..., tn} is a finite subset of T, then p < Д (Г — T') implies p ^ \ / ( t : te T ') for each p e P e.

P ro o f. We use induction cn n. If n = 1, the lemma follows from Lemma

5

(ii). Assume the lemma holds for n = к — 1 and.let a —<p ^ tx, ..., tk. We shall consider two cases.

C ase 1. Let p < tk. Then p ^ t lt ..., tk- t and hence by the induction hypothesis

p ^ v ... v tk- ! ^ t± v ... v tk.

C ase 2. Let p £ t k, and set tk : = tk л (p v tk). Obviously, tk ^ tk and hence from Lemma 4 it follows that tke P a. Since p ^ ^ , . . . , ^ and tk ^ t k

^ tl t ..., tk we have

Ik

^

t-1

> • • • »

tk

Л

~ t i f

• • • »

tk- i .

Now by the induction hypothesis we obtain

(3) t'k < v ... v fk_!.

Observe that

(4) P v tk = £* v tk.

Indeed, £k v tk = [£k

a

(p v tk)] v < p v tk and since tk ~<tk v tk and tk —< p v tk we conclude that (4) holds. From (3) and (4) we obtain

p ^ p v t k = t'k v t k ^ t l v . . . v tk.

(4)

Thus the lemma follows by induction.

Lemma 7.

I f a = f \ T is an irredundant decomposition of a eL , then hT = \J(t: te T )

a

ua.

P ro o f. Apply Lemma 6 and the proof of Theorem 4.4 of [3].

Lemma 8.

I f A is an independent set of atoms of the lattice L, and St (i e I) is a collection of subsets of A such that (r')f6jS. =

0 ,

then / \ ( \ / Sp. ie l) = 0.

P ro o f. As the proof is similar to that of Theorem 6.6 in [2] it will be omitted.

Lemma 9.

Let a = f \ T be an irredundant decomposition of a sL . I f b, c e L are such that b v c = ua and b л c = a, then b v (tip)

a

c ) ^ u a for each p e H T.

P ro o f. Let p e H T. Since p ^ t{p) we conclude that p £ t(p). Hence, by semimodularity of the lattice L, t(p) -^ p v tip). Since ua/a is a point lattice (cf.

Lemma 2) and c e u ja there exists a subset Q ^ P a such that c -= \J Q. For each qsQ , if q ^ t(p), then t(p) - ^ q v t(p), by semimodularity. Since t(p)eM(L) and hence is covered by a unique element, p v tip) = q v tip) for each q e Q such that q £ t(p ). Then

c v

t ( p )

= V ô v

t ( p ) = p v t ( p )

>-

t { p ) .

By Lemma 3, we obtain t{p)

a

c -<c. Since ua/a is modular,

C

A

(b V (tip)

A

с)) = (Ь Л с) V (tip)

A

c) = tip)

A

C.

Now b v (tip)

а

с) ф ua, since otherwise с = ua

a

c = tip)

a c

, contrary to tip)

а с

—(c. Thus

ua — b y c >— b V (tip)

A

c).

Now we prove the following

Lemma

10. I f a, b e L and a ^ b , then there exists meM(L) such that m ^ a and m^-b.

P ro o f. Since L is strongly atomic and a < a v b, there exists p e L such that a ^.p < a v b. Let

S : = {x e L : = x ^ a , x ^ p } .

S is nonempty, since a e S. Let C be a chain in S. Then upper continuity yields p

a

\J C = \J ip

a

c: ceC ) — a.

Thus \J C e S, and S contains a maximal element m, by Zorn’s lemma. Clearly, meM(L), m ^ a and m ^ b .

The following lemma is a generalization of Theorem 4.3 of [3].

(5)

Lemma 11.

Let a — f \ T be an irredundant decomposition of aeL. Then if J is an independent subset of Pa such that J 3 H T, there exists an irredundant decomposition a = / \ R such that HR = J.

P ro o f. Since the lattice L is upper continuous and J is an independent subset of Pa, J is contained in a maximal independent subset M of Pa. Let := J — H T and M x := M — J t , and let b : = \ f and c : = \ J M x. By 6.5 of [2], b v c = \J M = ua and by Lemma 8, b л c = a. Finally, we define sp: = b v (t(p)

a

c) for each p e H T and sp: = c v V (Л — {p}) f°r each р е Д . From Lemma 8 we conclude that

(5) Д (\f(Ji~{p}y- p e J i) = a.

Since Mfl/a is modular, by the Isomorphism Theorem (cf. [4]) the lattices b/a = b/b

a

c and ua/c = b v c/c are isomorphic. Therefore, (5) implies that / \ ( s p: p e J 1) = c. Hence

/ \ ( s p: p e J) = /\{ s p: р е Д ) л Д (sp: p e H T) = с л Д (sp: p e H T)

= A (c

a

(b v (t(p) л с)): p e H T) = Д ({b л c) v (t{p) л c): p e H T)

= Д (t(p) л c: p e H T) = с л f\(t(p): p e H T) = с л Д T= a.

Thus we have

(6) a = / \ ( s p: peJ).

Observe that

(7)

P < / \ ( Sq: Qe J ~{p})»

for each p e J. Indeed, if p e H T, then

sp = b v (t{p) л c) = V Д v (t(p) л c) ^ \ / J i v V (я г “ {р}) = V ( J ~ {P})»

and if р е Д , then

Sp = c V \ / ( Д - {р} )

> \ / H T

v V ( J r W ) = V ( J - { p } ) -

Therefore, for each p e J , sp ^ \ / { J — {p})- Then for each p, q s J with p Ф q we have p ^ \ / ( J — {<?}) ^ sq> and hence we obtain (7).

By Lemma 9, sp-<ua for each p e H T. Moreover, if р е Д , then from the semimodularity of L and the maximality of M, sp -<ua. Thus sp-^ua for each p e J. From Lemma 10 it follows that for every p e J there exists rpe M (L ) such that rp ^ sp and rp > ua, i.e., ua

a

rp = sp.

Let Я = {rp: p e J } and suppose that a < Д (rp: peJ). Since L is strongly atomic, there is q e P a such that q ^ f \ ( r p: peJ). Therefore,

a < q = q A u a ^ Д (rp

a

ua\ peJ) = f\{ s p. peJ),

contrary to (6). Therefore,

(6)

(8) a = f\{ r p: peJ).

Now we shall prove that the decomposition (8) of a is irredundant. Suppose on the contrary that there is p e J such that a = Д (rq: q e J — {p}). Hence using (7) we have

a = ua

a

a — / \ ( r q

a

ua: q e J - { p } ) = / \ ( s q: q e J - { p } ) ^ p .

This contradiction shows that the decomposition a = Д R of a is irredundant.

Moreover, HR = J, and the proof is complete.

4. Main results.

Theorem

1. I f a is an element of an upper continuous, strongly atomic, locally modular lattice L and a = f \ T = Д R are two irredundant decom­

positions of a with hT ^ hR, then there exists a subset S Я R and a one-to-one mapping f of T onto S such that, for each te T , a = t

a

f (t).

P ro o f. For each t e T we define Rt = {reR: a = г л Г}. Let {tl f ..., tn} be a finite subset of T. Suppose that jRtl vj ... и Rtn contains m < n elements, say rt , . . . , rm. Then, for each r e R — {r1, ..., rm}, r

a f >

a (i= 1, . . . , и). Hence, since tt contains only one element of Pa, tt < r for each i = 1, . . . , n, r e R —{/*!,..., rm}. Thus

h v ... v t n ^ r i, . . . , r m.

From Lemma 6 we conclude that t x v ... v tn v ... v fm. But this is impossible, since the dimension m of the sublattice f x v ... v r j a is less than the dimension n of the sublattice tx v ... v tn/a. Therefore,

(9) |Rt l u . . . u R J ? n ( 2).

Let te T . Since hT < hR, t ^ \ / H R, and hence there exists a finite subset {Pi> •••, Pk} ^ HR such that t ^ px v ... v pk. By Lemma 5 (i), pt < r(pt) for each i = l , . . . , k . Then t ^ г ( р х) , ..., r(pk), and therefore JRt s {r(px) , ..., r{pff}. Hence for each te T , Rt is finite. From this and (9) we obtain the theorem, by Theorem 1 of [5] on representatives of subsets.

Our second theorem is a generalization of Theorem 4.2 of [3].

Theorem 2.

I f an element a of an upper continuous, strongly atomic, locally modular lattice L has two decompositions a = Д T = Д R, then for each t0e T

there exists r0e R such that a = r0

a

t0. Moreover, the resulting decomposition is irredundant if so is a = Д T.

P ro o f. Let t0eT. Set S := { s: s = T0

a

r, reR }. Now we prove that a = t0

a

t0 is irreducible in t0/a. Suppose on the contrary that Tja contains two distinct atoms px and p2. By semimodularity, t0 -<t0 v px and t0 -<t0 v p2.

(2) For every set X , |X| denotes the cardinality of X.

(7)

Hence, since t0 is irreducible we have t0 v p l = t0 v p 2. Moreover,

*o л (Pi v Pi) = a- Then, from Lemma 1 it follows that pt = p2. This contradic­

tion shows that a is irreducible in t ja . Therefore, since L is strongly atomic, a is completely irreducible in t j a , i.e., for every subset U £ t j a , a = / \ U implies aeU. Consequently, from a = / \ R = ï0 л / \ R = f \ S and S Ç t j a we get a e S, i.e., a = T0

a

r0 with r0 e R.

By the second part of the proof of Theorem 4.2 of [3] we conclude that if the decomposition a = Д T is irredundant, then so is a = r0 л F0. Therefore the proof is complete.

R em ark 1. Theorem 2 does not yield that |T| = \R\ if a = Д T = Д R are two irredundant decompositions of a. In [3] (p.15) an example of an algebraic modular strongly atomic lattice is constructed in which there is an element with two irredundant decompositions of different cardinalities.

As an immediate consequence of Theorem 2 we get

Theorem 3.

Let L be an upper continuous, strongly atomic, locally modular lattice, and let a be an element of L. I f a = / \ Т = Д R are two irredundant decompositions of a such that hT = hR, then \T\ — |R| and for each t0e T there exists r0ER such that a = r0 л t0 is an irredundant decomposition of a.

R e m ark 2. Since every algebraic lattice is upper continuous, this theorem implies the Corollary of [3] (p. 15).

The next theorem is a generalization of Theorem 4.6 of [3].

Theorem 4.

Let L be an upper continuous, strongly atomic, locally modular lattice, and let a = \ J T — f \ R be two irredundant decompositions of a e L with hT ^ h R. Then for each t0e T there exists r0e R such that

(10) a = r0 A t 0 = t0 A r0.

P ro o f. The proof is similar to that of Theorem 4.6 of [3]. Here we apply Lemmas 5 and 6.

The main result of our paper is

Theorem 5.

Let L be an upper continuous, strongly atomic, locally modular lattice. An element a e L has doubly replaceable irredundant decompositions iff for every irredundant decomposition a = / \ T of a, ua ^ \ J ( t : teT).

P ro o f. Sufficiency. We know that each element of an upper continuous strongly atomic lattice has an irredundant decomposition (cf. [6], Theorem 10).

Let a = Д T = Д R be two such decompositions. By assumption,

ua ^ \J (T : t e T ) and ua ^ \ J (f: reR). Therefore, from Lemma 7 it follows that

ua = hT = hR. Hence, by Theorem 4, for every t0e T there exists r0 such that

the decompositions (10) hold. Thus a has doubly replaceable irredundant

decompositions.

(8)

Necessity. Let a = f \ T be an irredundant decomposition of a. Suppose that uaf , \ J (t : te T ). Hence, from Lemma 7 we conclude that hT < ua. Then, by Lemma 6.5 of [2], H T is not a maximal independent subset of Pa, and hence there exists an independent subset J ç Pa such that J 3 H T. By Lemma 11, there exists an irredundant decomposition a = f \ R such that HR = J. Now observe that r(p) ^ hT for each p e H R — H T. Indeed, if q e H T, then q e H R and q ф p. Hence q ^ r(q) ^ r(p), since r(p) Ф r(q). Therefore, hT = \ f H T ^ r(p).

Suppose that there exists t0 e T such that r(p) л F0 = a for some p e H R — H T. Then a ^ hT л t0 ^ hT л t0 = t0, a contradiction. Consequently, r(p) л t > a for all te T, and hence a does not have doubly replaceable irredundant decompositions. Thus ua ^ \/ (T : teT ).

Co r o l l a r y 1.

Let L be an upper continuous, strongly atomic, locally modular lattice. L has doubly replaceable irredundant decompositions iff for each a e L and for every irredundant decomposition a = f \ T of a, ua ^ \ J (t: teT ).

References

[11 P. C r a w le y , Decomposition theory o f nonsemimodular lattices, Trans. Amer. Math. Soc. 99 (1961), 246-254.

[2] P. C r a w le y and R. P. D ilw o r t h , Algebraic Theory o f Lattices, Prentice-Hall, Englewood Cliffs, N. J. 1973.

[3] R. P. D il w o r th and P. C r a w le y , Decomposition theory for lattices without chain conditions, Trans. Amer. Math. Soc. 96 (1960), 1-22.

[4] G. G r â tz e r , General Lattice Theory, Birkhâuser, Basel 1978.

[5] M. H a ll, Distinct representatives of subsets, Bull. Amer. Math. Soc. 54 (1948), 922-926.

[6] G. R ic h te r , The Kuros-O re theorem, finite and infinite decompositions, Studia Sci. Math.

Hungar. 17 (1982), 243-250.

INSTITUTE OF MATHEMATICS, PEDAGOGICAL UNIVERSITY OLESKA 48, 45-052 OPOLE, POLAND

Cytaty

Powiązane dokumenty

Thus eigenfunctions of the Fourier transform defined by the negative definite form −x 2 in one variable are the same as eigenfunctions of the classical in- verse Fourier

Thus, since Handelman proved that the ultrasimplicial property is preserved under formation of quotients by order-ideals [Han83, Theorem 3(ii)], it follows that every abelian

In the proof of this theorem, the key role is played by an effective interpretation of the well-known fact that an irreducible polynomial which is reducible over the algebraic

We study the existence and nonexistence of positive solutions of nonlinear elliptic systems in an annulus with Dirichlet boundary conditions.. Mitidieri [2] con- sidered the

The first important observation concerning the whole group of automorphisms of a hyperbolic, compact Riemann surface is contained in the Hurwitz theorem (see Theorem 1.1).. The

In [1] and [2] we introduced a modification of classical Bernstein oper- ators in C([0, 1]) which we used to approximate the solutions of suitable parabolic problems.. In this paper

Our purpose in this article is to give a brief and simple proof of this theorem in the general case of (not necessarily bounded) hyperbolic convex domains in C n.. In order for M to

An interesting feature of Example 1 is that the set {T n } ∞ n=1 is discrete.. It is similar in the spirit to Example 3.7 in [2], ours is however