Вы находитесь на странице: 1из 8

Arch. Math.

72 (1999) 145 ± 152


0003-889X/99/020145-08 $ 3.10/0
 Birkhäuser Verlag, Basel, 1999 Archiv der Mathematik

Lexicographic behaviour of chains

By

JUAN C. CANDEAL and ESTEBAN INDURAÂIN *)

Abstract. This paper introduces a different approach to the study of the existence of
numerical representations of totally ordered sets (chains). We pay attention to the
properties of non-representable chains showing that, under certain conditions, those
chains must have a sort of lexicographic behaviour similar to that of the lexicographic
plane. We prove that a countably bounded connected chain …Z; † admits a
lexicographic decomposition as a subset of the lexicographic product R  Z. Then we
apply our approach to state both a sufficient and a necessary condition for the lack of
utility functions. The concept of planar chain is also introduced.

1. Introduction. A relevant problem when studying ordered sets is that of their


representability by means of a real-valued order preserving function. In the context of totally
ordered sets (chains) a first result is due to Cantor [5, 6]. Early characterizations of
representability were given by Milgram [12] and Birkhoff (see e.g., Birkhoff [1], th. 24 on p. 200).
It is well known that for a chain …Z; † to be representable it is necessary and sufficient
that Z be perfectly separable (see e.g., Ch. 1 of Bridges and Mehta [2]).
The starting-point of this note is quite different from the previous one. Let …Z; † be a
chain which is not representable by a utility function. When does it contain a subchain which
is isotonic to a non-representable subset of the lexicographic real plane ? In other words,
when is the lexicographic plane a germ of the non-representability of a chain by means of a
real-valued order preserving function ?
Of course such a non-representable chain, that we shall call planar, has to satisfy some
additional conditions because there are non-representable chains involving ordinals which
fail to have the above property.
The motivation of our work is two-fold. On the one hand the lack of examples of non-
representable chains which are meaningful especially in applications to economics and social
sciences. Ordered structures based on ordinal numbers are too sophisticated so as to find
applications in practise. By contrast many situations involving lexicographic orderings are
often encountered in the literature (see, e.g., Fishburn [10]).
On the other hand, and stressing the first point even further, in the context of algebraical
totally ordered sets the lack of representability leads to the existence of non-representable
subsets of the lexicographic plane (see, e.g., Birkhoff [1], p. 354).

Mathematics Subject Classification (1991): 06A05.


*) Correspondent author.
146 J. C. CANDEAL and E. INDURAÂIN ARCH. MATH.

An idea related to our work is that of embedding an ordered set into another one. In
relation to lexicographic orderings several authors have studied the question of embedding a
chain into a lexicographic product (see for instance, Cuesta-Dutari [7, 8], Sierpinski [13],
Fleischer [11] or Fishburn [10]). The main result in this context is:
Let …Z; † be a chain and let a be the first ordinal whose cardinality equals that of Z. Then
…Z; † is isotonic to a subset of …f0; 1ga ; L † where L stands for the lexicographic ordering.
Nevertheless no conclusion is obtained about the representability of the chain embedded
into a lexicographic structure.
The paper is organized as follows. Section 2 contains definitions and notations. The
fundamental result on lexicographic decomposition of a countably bounded and connected
chain is established in Section 3. Our approach is then applied to provide both a sufficient
condition and a necessary one for the lack of representability. We conclude in Section 4 by
discussing the concept of planar chain.

2. Definitions and notations. Throughout the paper we suppose that …Z; † is a chain
(totally ordered set), that is, a set Z endowed with a binary relation  which is transitive,
asymmetric and total (i.e., if x ˆj y then x  y or y  x).
 y we shall understand :…y  x†. Notice that if …Z; † is a chain then the binary
For x 
relation  is transitive, antisymmetric and complete. For every x 2 Z; L…x† (respectively
G…x†) will be the set fy 2 Z : y  x (respectively: x  y†g.
Given x; y 2 Z such that x  y, we will denote by …x; y† the subset
…x; y† ˆ L…y† \ G…x† ˆ fz 2 Z : x  z  yg, and by ‰x; yŠ the subset ‰x; yŠ ˆ …x; y† [ fx; yg.
(Observe that …x; y† could be the empty set).
On the chain …Z; † we shall consider the order topology. A subbasis for this topology is
S
given by the family fL…x† : x 2 Zg fG…y† : y 2 Zg.
The chain …Z; † is said to be without gaps if …x; y† is non-empty, for every x; y 2 Z such
that x  y. If x; y are such that x  y and …x; y† is empty, we say that fx; yg defines a gap, or
equivalently that y is consecutive to x.
A subset X 7 Z is said to be:
coinitial in Z if, for every z 2 Z, there exists some x 2 X such that x   z,
cofinal in Z if, for every z 2 Z, there exists some x 2 X such that z   x,
totally order connected if it is coinitial and cofinal in Z and the subchain …X; † is without
gaps.
The chain …Z; † is said to be:
countably bounded if there exists a countable subset X 7 Z which is coinitial and cofinal
in Z,
Dedekind complete if any non-empty bounded above subset of Z has a supremum (i.e., the
least upper bound) in Z.
It is well known that a chain …Z; † is connected with respect to the order topology iff it is
Dedekind complete and has no gaps.
The order  is called perfectly separable if there exists a countable subset D 7 Z such that
for every x; y 2 Z with x  y, it holds that D \ ‰x; yŠ is non-empty.
Two chains …Z; † and …Z0 ; 0 † are said to be isotonic (order isomorphic) if there is an
order preserving (bijective) function f : Z ! Z0 , i.e., z1  z2 () f …z1 † 0 f …z2 †.
Vol. 72, 1999 Lexicographic behaviour of chains 147

A utility function for the chain …Z; † is a real-valued function u : Z ! R such that for any
x; y 2 Z, it holds that x  y () u…x† < u…y†. If there exists a utility function defined on
…Z; †, the chain is said to be representable (equivalently, …Z; † is isotonic to a subset of the
real line).
Given two chains …Z; † and …Z0 ; 0 † we define their lexicographic product as the
Cartesian product Z  Z0 endowed with the following order L (called lexicographic):
…z1 ; z2 † L …y1 ; y2 † if and only if z1  y1 or z1 ˆ y1 and z2 0 y2 .
A chain …Z; † is said to be planar if it contains a subchain that is isotonic to a non-
representable subset of the lexicographic plane …R2 ; †. Given a subset X 7 Z, Card X will
denote the cardinality of X.

3. Lexicographic decomposition of chains. In this section we prove the main result of the
paper, namely: a countably bounded connected chain admits a lexicographic decomposition.

Theorem 1. Let …Z; † be a countably bounded connected chain with no largest and no
smallest element. Then there exist a subchain X 7 Z, and an equivalence relation r on Z, with
equivalence classes ‰zŠ ˆ fy 2 Z : yrzg, such that:
(a) X is isotonic to the real line R,
S
(b) Z is order isomorphic to the lexicographically ordered set fxg  ‰xŠ.
x2X
P r o o f. Let D ˆ …dn †n2N be a countable set that bounds Z. In other words,
S
Zˆ …dm ; dn †. Between each pair x; y 2 D such that fx; yg is a gap in …D; † we add
m;n2N
to D an element w from Z with x  w  y. (The existence of such w is guaranteed by
connectedness of Z.) Then we also add elements u and v, from Z with x  u  w  v  y,
and continue with this process indefinitely. This gives a countable extension D0 of D with the
properties:
(i) D0 has no gaps,
(ii) D0 has no largest and no smallest element.
It is known (see e.g., Birkhoff [1], p. 200) that such a set D0 is order isomorphic to the set
of rationals in …0; 1†, so that by connectedness of …Z; †, there is an extension X of D0 in Z
which is order isomorphic to …0; 1† and hence to R. Because the chain X is isotonic to R, it is
Dedekind complete and has no gaps. Moreover, it is not bounded above or below in Z. For
each x 2 X, let
 T   T 
‰xŠ ˆ L…b† \ G…a† ;
xb; b2X ax; a2X

and note that ‰xŠ \ X ˆ fxg because if x  y then ‰xŠ 7 L…y† so that y 2j ‰xŠ, and similarly if
y  x.
We shall now show that the sets ‰xŠ; x 2 X, define a partition of Z, so that each z 2 Z lies
in some set ‰xz Š, where xz 2 X is uniquely determined by z. We shall also see that each ‰xŠ is a
closed interval in Z and this means that the collection f‰xŠ : x 2 Xg inherits in a natural way
an order from Z. It follows that the map z 7! …xz ; z† is an order preserving map of Z into the
lexicographically ordered product X  Z and this will prove Theorem 1. All these
statements follow from the next result. h
148 J. C. CANDEAL and E. INDURAÂIN ARCH. MATH.

Lemma 1. The sets ‰xŠ have the following properties:


(i) If x; y 2 X and x  y then ‰xŠ  ‰yŠ and ‰xŠ \ ‰yŠ ˆ ;. If x ˆ y then ‰xŠ ˆ ‰yŠ,
(ii) The union of all ‰xŠ, where x 2 X, is Z,
(iii) If u; v 2 ‰xŠ and u  v, then ‰u; vŠ 7 ‰xŠ,
(iv) Each ‰xŠ is of the form ‰a; bŠ where a; b 2 Z.

P r o o f. (i) Suppose that x  y, where x; y 2 X. As X has no consecutive elements, there


exists a 2 X with x  a  y. It follows that ‰xŠ 7 L…a† and ‰yŠ 7 G…a†, and this shows that
‰xŠ  ‰yŠ. Of course, if x ˆ y, then ‰xŠ ˆ ‰yŠ.
(ii) We need to show that each z 2 Z lies in some ‰xŠ, where x 2 X. Given z there are
points x1 ; x2 2 D and hence in X, with x1  z  x2 . This means that in the Dedekind
complete chain …X; †, the subset L…z† \ X is non-empty and bounded above by x2 ; thus
x ˆ sup …L…z† \ X† exists (in X). Similarly x ˆ inf …G…z† \ X† exists in X. It is easy to see
that x   x and we shall now show that x ˆ x . If not, then x  x and as X has no
gaps, there are infinitely many elements of X in …x ; x †. This means that there are points of
X either in …x ; z† or in …z; x †, which is a contradiction. Therefore x ˆ x. As there cannot
be any points of X between z and x ˆ x , it follows that z 2 ‰x Š. This proves that
S
‰xŠ ˆ Z.
x2X
(iii) Suppose now that u; v 2 ‰xŠ and let I be any L…a† or G…a†, where a 2 X. If x belongs to
I, then I contains ‰xŠ; hence u; v 2 I and so I contains ‰u; vŠ. It follows that the intersection of
all such I contains ‰u; vŠ. Therefore, by definition of ‰xŠ, ‰u; vŠ 7 ‰xŠ.
(iv) Given x 2 X there are points a; b 2 X with a  x  b so that ‰xŠ 7 …a; b†; thus
‰xŠ is bounded above and below. Since …Z; † is connected, it is Dedekind complete so that
zi ˆ inf ‰xŠ; zs ˆ sup ‰xŠ

exist in Z. In particular, ‰xŠ 7 ‰zi ; zs Š 7 …a; b†. We show that zs 2 ‰xŠ: If zs 2j ‰xŠ, then, by
definition, there is an X-interval …a; b† that contains x but not zs . As a  x   zs we see by
 zs . Now x and b are in X, that has no gaps, so that there is some g in X with
part (iii) that b 
x  g  b. Thus ‰xŠ 7 …a; g† and this means that g is an upper bound of ‰xŠ such that
g  zs ˆ sup ‰xŠ. Contradiction. Therefore zs 2 ‰xŠ. A similar argument proves that zi 2 ‰xŠ,
so that by part (iii) again, ‰zi ; zs Š 7 ‰xŠ. Since ‰xŠ 7 ‰zi ; zs Š by definition, we conclude
‰xŠ ˆ ‰zi ; zs Š. h

R e m a r k 1 . (i) Theorem 1 can be extended to situations in which Z is a countably


bounded connected chain but has a smallest or a largest element (or both). In these cases,
with similar arguments to those given above, it is straightforward to construct a lexico-
S
graphic decomposition fxg  ‰xŠ of Z in which now X is order isomorphic to one of the
x2X
sets ‰0; 1† , …0; 1Š , or ‰0; 1Š of real numbers, depending on the existence of end-points in Z. The
definition of a class for the smallest element xm and the largest element xM will be
T T
respectively, ‰xm Š ˆ L…b† and ‰xM Š ˆ G…a†.
xm b; b2X axM ; a2X
(ii) In the proof of Theorem 1 the equivalence relation r and the subchain X have been
constructed, because X was not given a priori. However, with analogous arguments, similar
equivalence relations are well-defined on a general chain …Z; † provided that there exists a
Dedekind complete and totally order connected subchain X from which we start a
Vol. 72, 1999 Lexicographic behaviour of chains 149

construction similar to that of Theorem 1. Also, the lexicographic decomposition


S
fxg  ‰xŠ is order isotonic to Z. Nevertheless, the properties on X and ‰xŠ given in
x2X
Lemma 1, are not necessarily true now. In particular, X may fail to be isotonic to a real
interval, and the classes ‰xŠ may or may not be order-intervals ‰a; bŠ as in Lemma 1 (iv). h
There are two main reasons, of different nature, for which the chain Z could fail to be
representable. On the one hand if some class ‰xŠ is not representable then Z is not either. On
the other hand, we shall prove that if the cardinality of the set of non trivial classes is bigger
than the cardinal @0 of natural numbers, then Z is not representable. We shall prove this
sufficient condition for the lack of representability in the next result.

Theorem 2. Let …Z; † be a chain and let X 7 Z be a subchain so that Z admits a


S
lexicographic decomposition fxg  ‰xŠ. Then if the set A ˆ fx 2 X : Card ‰xŠ > 1g is
x2X
uncountable, the chain …Z; † is not representable.

P r o o f. Assume by way of contradiction that f is a utility function for …Z; †. Then, for
every x 2 X we define the real set S…x† ˆ ff …y† : y 2 ‰xŠg and consider the real interval
Jx ˆ …inf fS…x†g ; sup fS…x†g†. Notice that by hypothesis, fx 2 X; Jx ˆ j ff …x†gg is uncoun-
table. Moreover we have that if x1 ; x2 2 X, with x1 ˆ j x2 2 X then Jx1 \ Jx2 is empty. So we
could associate to each x 2 A a rational number q…x† in such a way that q…x1 † ˆ j q…x2 †
whenever x1 ˆ j x2 . But this leads to a contradiction since A is uncountable. h
Let us now state, based on the lexicographic decomposition given by Theorem 1, a
necessary condition for the lack of representability.

Theorem 3. Let …Z; † be a non-representable chain and let X  Z be a representable


S
subchain so that Z admits the decomposition fxg  ‰xŠ. If ‰xŠ is representable for every
x2X
x 2 X, then …Z; † is planar.
P r o o f. Let f : X ! R be a utility function for …X; †. Since …Z; † is non-representable
and by hypothesis ‰xŠ is representable for every x 2 X, it follows that
A ˆ fx 2 X : Card ‰xŠ > 1g is uncountable. To see this, notice that if A were countable then
…Z; † would be perfectly separable, hence representable, which is a contradiction. Now
choose for every x 2 A an element y…x† ˆ j x, y…x† 2 ‰xŠ and consider the set
S ˆ X [ A‡ [ Aÿ , where A‡ ˆ fx 2 A : x  y…x†g and Aÿ ˆ fx 2 A : y…x†  xg. Define
then the function  : …S; † ! …R2 ; L † as …s† ˆ …f …s†; 0† if s 2 X, …s† ˆ …f …s†; 1† if s 2 A‡
and …s† ˆ …f …s†; ÿ1† if s 2 Aÿ. Notice that  is order-preserving and clearly …S† is a non-
representable subset of the lexicographic plane …R2 ; L †. h
E x a m p l e 1 . Let Y ˆ ‰0; W†, W being the least uncountable ordinal number. Between each
ordinal a and its successor a ‡ 1 put one copy of the interval …0; 1†. The space, denoted by L,
that we get in this way is called the long line and it is ordered in the obvious way (denoted
by ). Since the sets ‰0; g†  Y are countable for all g  W and each countable subset has a
supremum in ‰0; W† (see Dugundji [9], Th. 9.1 on p. 54) it follows that L is a connected space.
Actually it is a path-connected topological space (see Steen and Seebach [14], pp. 71 ± 72).
Clearly L is not countably bounded (hence it is not representable). It is not hard to see that
150 J. C. CANDEAL and E. INDURAÂIN ARCH. MATH.

…L; † does not contain copies of non-representable subchains of the lexicographic plane
…R2 ; L †. So, …L; † does not satisfy the hypothesis of Theorem 3. Nevertheless by
considering the chain Z ˆ L [ fWg, with the obvious ordering also denoted by , we obtain
an example of a countably bounded, connected and non-representable chain which does not
contain copies of non-representable subchains of the lexicographic plane. This implies that
…Z; † does not satisfy the hypothesis of Theorem 3 either. Thus, since by Theorem 1 the
lexicographic decomposition of Z does exist, there must be at least one class ‰xŠ that is not
representable. This is indeed the case and there is just one non-representable class; namely, ‰WŠ.
A simple rephrasal of Theorem 3 states a necessary condition for a chain to be non-
representable.

Corollary 1. Let …Z; † be a non-representable chain, and let X be a representable subchain


S
X  Z so that Z admits the decomposition fxg  ‰xŠ. Then either there exists x0 2 X such
x2X
that …‰x0 Š; † is a non-representable subchain or …Z; † is planar.

4. Final remarks. In view of Corollary 1, an interesting question is the analysis of


conditions that force a non-representable chain to be planar. From now on we shall assume
by technical reasons derived from Theorem 1, that …Z; † is connected.
So, starting from a non-representable countably bounded connected chain …Z; † we are
interested in applying recurrently the approach given by Theorem 1 in order to get a planar
subchain of Z, proving in this way that Z itself is planar. Certainly, this recurrence does not
always lead to a happy end, because, as shown in Example 1, there are non-representable
countably bounded connected chains that are not planar.
Anyway, let us see what happens in the general case.
If after applying Theorem 1 the set A ˆ fx 2 X : Card ‰xŠ > 1g is uncountable, we are
done. If it is countable, then we take a connected subchain Z1 7 Z such that Z1 contains all
the non-representable subchains …‰xŠ; † with x 2 A. So proceeding we may obtain a family
of Dedekind complete, totally order connected and representable subchains …Xn †n ^ 1 (with
X1 ˆ X) and a sequence of nested non-representable subchains …Zn †n ^ 0 (with Z0 ˆ Z): If in
a step that starts with Znÿ1 , the corresponding set An ˆ fx 2 Xn : Card ‰xŠZnÿ1 > 1g is
uncountable (‰xŠZnÿ1 meaning here the class corresponding to the element x 2 Xn
that Xn defines on Znÿ1 ), we are done again by Theorem 2 and we stop the process.
Otherwise, we continue in the same way. If the process goes on indefinitely, it clearly
T
follows that Zn is a non-representable subchain of Z: Indeed, observe that
T n ^0
Z ˆ …… Zn † [ …Z0 n Z1 † [ …Z1 n Z2 † [ . . . [ …Znÿ1 n Zn † [ . . .†, and ……Znÿ1 n Zn †; † is re-
n^0
presentable, by construction, for every n ^ 1.
The next examples allow us to illustrate these observations.
E x a m p l e 2 . Let Z be the following subset of ‰0; 2ŠN :
Z ˆ f…zn †n ^ 1  ‰0; 2ŠN : if for some n0 2 N; zn0 ˆj 1; then zm ˆ 0 for every m > n0 and
zk ˆ 1, for every k < n0 g.
Endow Z with the natural lexicographic ordering L given by: …an †n ^ 1 L …bn †n ^ 1 if and
only if there exists k 2 N such that ai ˆ bi if i < k, and ak < bk . Observe that …Z; L † is a
countably bounded and connected chain. Let X1 ˆ f…zn †n ^ 1 2 Z : zk ˆ 0 for every k > 1g,
Vol. 72, 1999 Lexicographic behaviour of chains 151

and Z0 ˆ Z. Let now X2 ˆ f…zn †n ^ 1 2 Z : z1 ˆ 1 ; zk ˆ 0 for every k > 2g, and


Z1 ˆ f…zn †n ^ 1 2 Z : z1 ˆ 1g. In the general step Xn ˆ f…zn †n ^ 1 2 Z : z1 ˆ . . . ˆ zn ˆ 1;
zk ˆ 0 for every k > n ‡ 1g, and Znÿ1 ˆ f…zn †n ^ 1 2 Z : z1 ˆ . . . ˆ zn ˆ 1g. It is not difficult
T
to check that An ˆ fx 2 Xn : Card ‰xŠZnÿ1 > 1g is always countable. Moreover Zn is the
n^0
subset f…zn †n ^ 1 2 Z : zn ˆ 1 for every n 2 Ng which has a unique element and is obviously
representable. Thus we are induced to believe that …Z; L † is representable. This occurs
indeed because …Z; L † is perfectly separable. To see that, let Dm ˆ f…zn †n ^ 1 2
S
Xm : zn 2 Q \ ‰0; 2Š; for every n 2 Ng. The set D ˆ Dm is a countable order dense
m2N
subset of Z.
The generalization of the above situation to a continuum case is quite different, leading to
a non-representable chain.

E x a m p l e 3 . Let Z ˆ fc‰0;r† ‡ a  cfrg : a; r 2 ‰0; 1Šg (cB stands for characteristic function
of B). Consider on Z the usual lexicographic ordering of functions L and observe that
…Z; L ) is a countably bounded, connected chain. However …Z; L † is not representable
because from the above construction, setting X ˆ fc‰0;rŠ : r 2 ‰0; 1Šg 7 Z, it follows that X is
isotonic to ‰0; 1Š, hence representable, and the set A ˆ fx 2 X : Card ‰xŠ > 1g is exactly X
which is uncountable.
Let us explain which facts can be obtained and which difficulties appear. Suppose that
…Z; † is a countably bounded and connected chain. Assume in addition that …Z; † is first
countable. Then …Z; † can be embedded into another connected and first countable chain
having both a largest and a smallest element. Thus without loss of generality we start with a
connected, first countable and non-representable chain of the form Z ˆ ‰a; bŠ. Consider the
family of all non-representable closed intervals in Z:
D ˆ f‰z1 ; z2 Š 7 Z : ‰z1 ; z2 Š is non-representableg:

D is nonempty since Z 2 D. Define the following relation R on D: Given Z1 , Z2 2 D we say


that Z2 R Z1 if and only if Z2 7 Z1 and Z2 contains all the non-representable classes with
respect to the relation r given in Theorem 1 or Remark 1 (ii), over Z1 for some totally order
connected and Dedekind complete representable subchain X1  Z1 .
It is easy to see that R is reflexive and antisymmetric. Moreover if R fails to be transitive,
then …Z; † is planar. Suppose then that R is transitive. In other words, R is a partial order
on D. By using the first axiom of countability and Zorns lemma it holds that if there is no
R -minimal element in D, then …Z; † is planar. Otherwise let Zm be a R -minimal element
in D. The first difficulty arises here since Zm could not be planar. Notice in addition that Zm
could even be Z. This is another problem.
Finally, observe that the main two examples of connected and countably bounded chains
not being representable are the lexicographic plane and the closed long line L [ fWg. But
the latter is not planar. Since L [ fWg is not first countable, we could conjecture that every
connected countably bounded and first countable chain is planar. This conjecture leads to
deep set-theoretical concepts, namely the concept of countable chain condition (shortly,
c.c.c.) and the Souslin Hypothesis (SH). A chain …Z; † satisfies c.c.c. if every family of
pairwise disjoint open intervals in Z is countable. The Souslin Hypothesis says that every
connected and compact chain satisfying c.c.c. is representable. It is known that SH is
152 J. C. CANDEAL and E. INDURAÂIN ARCH. MATH.

undecidable in ZFC, i.e., it is consistent with the axioms of Zermelo-Fraenkel set theory, plus
Choice axiom, that SH be either true or false (see Todorcevic [15], p. 272). On the other
hand it is not difficult to prove that if a chain satisfies c.c.c. then it is first countable. So, at
least from the point of view of set theory and assuming that SH does not hold, there is a
countably bounded connected first countable and non-representable chain which is not
planar, and the conjecture is false.
We leave as an open problem to find a whole characterization of planar chains.
A c k n o w l e d g e m e n t s. Thanks are given to Professors Alan F. Beardon (Cambridge,
U.K.), Gerhard Herden (Essen, Germany) , Ghanshyam B. Mehta (Brisbane, Australia) and
Jose Carlos RodrgÂuez-Alcantud (Salamanca, Spain) for helpful comments and valuable
suggestions.
This work has been partially supported by the Government of Navarre, Spain, through the
research project ªAnaÂlisis MatemaÂtico de la Preferenciaº. (Dec. 1996).

References
[1] G. BIRKHOFF, Lattice theory. Amer. Math. Soc. 1967.
[2] D. S. BRIDGES and G. B. MEHTA, Representations of preference orderings. Berlin 1995.
[3] J. C. CANDEAL and E. INDURAÂIN, RepresentacioÂn numeÂrica de oÂrdenes totales. Rev. Real. Acad.
Cienc. Exact. Fís. Natur. Madrid 84, 415 ± 428 (1990).
[4] J. C. CANDEAL and E. INDURAÂIN, Utility functions on chains. J. Math. Econom. 22, 161 ± 168 (1993).
[5] G. CANTOR, Beiträge zur Begründung der transfinite Mengenlehre (I). Math. Ann. 46, 481 ± 512
(1895).
[6] G. CANTOR, Beiträge zur Begründung der transfinite Mengenlehre (II). Math. Ann. 49, 207 ± 246
(1897).
[7] N. CUESTA-DUTARI, Teoría decimal de los tipos de orden. Revista MatemaÂtica Hispano-Americana
3, 186 ± 205, 242 ± 268 (1943).
[8] N. CUESTA-DUTARI, Notas sobre unos trabajos de Sierpinski. Rev. Mat. Hispano-Americana 7,
128 ± 131 (1947).
[9] J. DUGUNDJI, Topology. Boston 1966.
[10] P. C. FISHBURN, Lexicographic orders, utilities and decision rules: a survey. Management Sci. 20
(11), 1442 ± 1471 (1974).
[11] I. FLEISCHER, Embedding linearly ordered sets in real lexicographic products. Fund. Math. 49, 147 ±
150 (1961).
[12] A. N. MILGRAM, Partially ordered sets, separating systems and inductiveness. In: Reports of a
mathematical colloquium, K. Menger, ed., University of Notre Dame. Notre Dame, IN. 1939.
[13] W. SIERPINSKI, Sur une proprieÂte des ensembles ordonneÂs. Fund. Math. 36, 56 ± 67 (1949).
[14] L. A. STEEN and J. A. SEEBACH JR., Counterexamples in topology. New York 1970.
[15] S. TODORCEVICÂ, Trees and linearly ordered sets. In: Handbook of Set Theoretic Topology, 235 ± 293
(1984).

Eingegangen am 10. 10. 1996


Anschrift der Autoren:
Juan C. Candeal Esteban InduraÂin
Universidad de Zaragoza Universidad PuÂblica de Navarra
Facultad de Ciencias EconoÂmicas y Empresariales Departamento de MatemaÂtica e InformaÂtica
Departamento de AnaÂlisis EconoÂmico Campus Arrosadía s.n
c/ Doctor Cerrada 1 ± 3 E-31006-Pamplona
E-50005-Zaragoza Spain
Spain

Вам также может понравиться