AMS Subj. Class.

:Primary54X%5,02KQ5,541)30;
Secondary 54F05,54D2cb9LX25,04K30
I ’

countabb cm contition Bairn space

SouslW conjectwe point-countable cover
caliber separable
Martin’sAxiom cardinalinvariants

Abstract: Circumstancesin which :thecountable chain condition implies sepambi3ity 8te in-
vestigated.in particular,the implication “Martin’sAxiom plus 2K0 > HI implies Souslin’s con-
jecture” is gre;itlygene&ized. Roofs rarepurely topologic& usjlltgthe strong Bairecsategoq
form of Martk’s Axiom, Extensive use is made of Sanin calilbersand other cardinal&w~uiar~o‘s.

1. Introductim

A topological space satisfies the countabl’e chain con&km (CCC for

shcrt) if eveqy collrjction of d,isjoint open sets is c0~Gable. It has long
been known that this property is equivalent to separabilitv in metxiz-
able spaces, blut is in general weake.r than it. ~~tps~&~m&xmre [33]
states that the CCC’is equivalent to separability in linearly orde.red to-
pological spaces as well. Ns conjecture has a+z~ interesting history [24]
and is now known to be consistent with and independent of theI axioms
of set theory.
In this pa.pex we investigate generalizations sf Ssuslin’s cmject
i.e. assertions that various classes af CCC spaces are separable,

* mepqu.ration of ‘&ispaper was assistedby Grant No. A-?35

ColUWil oft Iamda.
3.‘16 lm Tidlpme txwelzbk &a&l timzditin Ys.sepmW~y

b&h their consistency and independence as well. Our print;:~ ::I tool is
Itfcur&z2 Kim [ 26 ) (stated below). Zn contrast to previous work in
this area, our proofs are pureiy topologkd.
The cd&wof ,ourresearch tie in the nurmal Mo~ore-spaceproblem. In
an early version of [ 361, I proved that.if there were a locahy compact,
perfictly krmal, metacumpact space which was not paracompact, then
there wals a normal non-metrizable Moore space. At the recent Prague
Symposium, AX. Arhangel’skii informed me that he had proved ~%enon-
existence of the former spaces twelve years earlier, but tti.s. his mentor
P.S. Alexandrov had not thought it worth publishing! In fact the follow-
ing more general statement is true, which Arhangel’skii has now published
.
151.

Theorem 1.1. Every point-finite open cover of a &xh complete U$!ned

bekw) CCC space is countable.

On the other hand, Juhk [19] had shown:

‘Theorem 1.2. Martin’s Axiom p&s 2 rgcfl

> HI implies that every point-
ccuuztabJeopen cover of a locuZZy
compact CCC regulbr ,space is count-
able.

We will prove the natural conjecture= arising from these two theorems:

eissremR.3. Mar&z’s Axitm p1z.u

: T!iHo > 8, implies that every point-
count&ye open cover oja Cech complete, CCC s,w;ce is coeuztable.

Tile property that paint-countable open covers be countable (other-

wise knowrn as caliber k!a) is intermediate between the CCC and separab-
ility, and will plaly an important role in what follows. B. Sapirovskii
proved ‘Theorem ‘i +3and other results overlapping ours at approximately
the samle time. They have appeared in [28] and [29].
Our Knajorresu!ts rnaLybe summarized in the following

ecmm 1. . Let X be compk:eZy regular CCC space which is the ititer-

sectiore of X open sets ii12
some eompactifictPrtion,X C 2% Let I I, n, x,
dd, cc r;tapzdrespectively for calrdi.nality,n-weight, characters hereditary
devrsity(width) and spreua”(‘ah’defined
. irzSkcti~n 3). Martin’s Axiom
phs 2Q > 8 implies that if X satisfies any of the fol,tow~ng
1

it is separable:
(a) n(X) *< PO;
(b) &(X)li < 2”a;
(c) dd(X) < 2H0, 1x1< 22N”,x G k$);
(d) cc(X) 6 H,, x G N,.

The organization of the paper is as follows. In Section 2, Martin’s

Axiom is intro&iced and generalized. In Section 3, calibers atid a large
variety of cardinal invtiants are defined, and some new relations be-
tween the CCC, caliber K, and separability are proved. In Section 4,
these results are strengthened with the aid of Martin’s Axiom. Section 5
is devoted to Sot&n’s conjecture and related matters. .In Ssction 6, the
results of the previous sections are a.pplied to prove a few results con-
cerning paracompactness and metrization. In Section 7, a number of
examples arc exhibited.
I thank the referee for spotting a number of errors in the original
version of the paper.

2. Martin’s Axiom

Martin’s Axiom is an alto,

I+&1’.!,kc;pzto the continuum hypothesis which
_.J
is becoming increasingly ust- I l in general topology. See e.g. 1381 and
the references therein. It ha ,.HJJ~alternative formulations, for example
using partial orders, Boolear &4~as~ or toI alogical spaces. Souslin’s
conjecture also can be formulated in terms of partial order or Boolean
;ilgebras. In [ 321, Solovay a .mc!Tennenbaum derive SouslinTs conjecture
from Martin’s Axiom plu$; 21r,0 ‘;, Hr, using the partial order versions,
They then prove th.e consistency of Martin% Axiom plus 2’0 > K 1 (in-
deed 2”O =:any regular aleph) with t:he axioms of Zermelo-Fraenkel
set theory, thus establishing the conlsistency csf So&n’s conjecture (its
independence was known).

I)efinitisn 2.1. Let K be a cardinal. A space ia a ~-&~ire SJNNXif the inter-

section of fewer than K dense open sets is dense. Thus t
spaces are H 1-Baire spaces. i4 stm
Martin’s Axiom is the assertion that e
is a stron
318 ED. Ta& The countablechainconditionV;Rseparability

The &es of Babe spaces is much wider tkn the compact Hausdorff
ones, see e.g. [ 11; similarly, the class of strong Baire spaces, assuming
Martin’s Axiom,. is much wider than the class of compact CCC Nausdorff
spaces, It does not, however, coincide with the class of CCC Baire spies
- see Example 7.4, We &all deal with two representative classes of strong
Baire spaces, leaving modifications such as localization to the reader.

n 2.2. A collection of sets is centeM if every finite c!rbcollec-

non-empty intersection. A space is ~~~is-corn~~ct [4-11 if it has
a base 16 such that if 6?C_ c1sis centered, then

A completely regular space is an absoZ~& G, (X a cardinal) if it is the

intersection of X open sets in its Stone-tech (or any (see [ 143)) compact-
ification. Absolute GNPspaces are also shown as absoltrte G, ‘s or
tech-ccmplete spaces.

Note that X is a rc-Baire space if and oz;ilyif no open set is the union
of fewer than K nowhere dense sets. It is not hard to verify that closed
subspaces, anal subspaces which are the intersection of < X open sets,
inherit the property of being absolute G,. On. the other hand, basis-
compMnVess is not necessarily inherited by closed subsets. Basis-compact-
ness and Cech completeness coincide with the usual completeness in
metrizable spaces but diverge outside [ 351.

Theorem 2.3. Martin’s Axiom implies that every CCC space which is
either regular and basis-compact, or is atudt!te G,, X < 2#0, is a strong
Bal’respace.

The proof of the first half is an obvious modifica&;i of the proof of

[ 19, Theorem 5.23. The proof of the second haif is analogous to the
standard proo;’ that since compact Hausdorff spaces are Baire, so are
absolute G, spaces, once it is observed that x is CCC is (and only if)
x is.

subsets pf a space X is poirll)-u if each

O c4ections asre
 RD. Ea.ll~
27~ emmtabk chin ctmiitiottvs. sepambiti~ 319

called point-finite; point-H 1 collections are called point-eoufltuble.

collection E! is locaZly-uat a point p if there is an open nei&borhood of
p meeting fewer than u members elf e . E? is locti&-K if it & locally fc at
As above, WI!use Escal~~) ,f;iWe i/k&:j; ~unt~ble) fsr lo-
ky H, )<_A5qace is metacompact On&Q-Lindeliifl if every
a p&--&finite (point-countable) open refinement. A
mp(dct(par-a-lindeliif) if every open cover has a locally
finite (locally countable) open refinement. A space has ccrliber K if every
point-u collection of open sets has power < K. A space satisfies the ianin
condition if for every regular uncountable K, x has caliber K.

It is easy to see that a space has caliber K if and only if every family of
power K of open sets has 8 subfamily of power K with non-empty inter-
section. It is also routine to verify:

dtion 3.2. Separable implies the $anin condition impks caliber H1

impbes CCC

“<heproofs are routine. No implication can be reversed. See Section 7.

Calibers are interesting because of the following result of Sanin [25,
261. Contrast with separability, preserved only under products of si 2’0
fact W-s,and the countable chain condition (see Section 5).

hoposition 3.3. If K> H, is regular, caliber K is preserved by arbitrary

proc?arc ts.

In this section we assume calibers plus other properties to get separab-

ility. In the next section, by employing M, we get the calibers from the
CCC.
The locally countable version of the following proposition was proved
by I. Juhsisz in 1969 in response to a question of the author.

Proposition 3.4. Let Cu ue an open cover of a CCC space such that

{p: Cu is localZy-uat p)
is dense, where K is a cardinal of uncountable cofinality. Then % has
cardinality k’essthan K.

roof. Tht?re is a maximal collection of disjoint open sets, each inter-

secting fewer ahain K members of the cover. Every member of the coveI*
intersects a member of the collection. Therefore the cover has cardin&ty
less than K* Cl
320 . ED. Tati,*fie cxwntd!&?
chaincon&ion vs sepuraaisity

ft Is perhapsof interest to see that CC!Cspaces can be characterize!&

usin& the propositions.

{p: (u is locaUy countuble at p}

is dense is mm table.

Pmof. Only one direction requires proof. Let {&},< wI be a collection

of disjoint open sets. By hypothesis,
{PI cv,>,,,,
u {X’)is hxally countable at p)
is not dense-. Thus there is an open Vsuch that {UJa<wl u {X) is nat lo-
cally countiable at any p E K In particular, V intersects some U&. ‘Br_d
A? U&, is 31neighborhood of p E V n U’ which %ersects only t%Mo mem-
bers of the cover, a contradiction Cl

With a bit of Baire, the CCC yields calibers.

Theorem 3.6. Let X be &Baire md CCC. Let X < fc be a regular s.:urdinal.

Then X has caliber h.

Corollary 3.7. M implies that if X is a CCC space which is either regular

and basis-compact or is absolute &for X less than 2’0, then X has culi-
ber fcfor every regular u C 2”9 in &rticular

The most interesting case is Theorem 1.3. Corollary 3.7 cannot be im

proved to allow K < 2No;see Example 7.5.

roofs.Let X be K;+-Bake. Let Cu = {b/ar)ar<h be a point-h open cover of

A’. We show that
(p: 94 is locally-X at p)
is dense. By Proposition 3.4, since .Wis CCC, it follows that Cu has car-
dinality C A. Suppose on the cxx\tralry that there is an open V siuch that
‘U isinot locally-X at any point of 61:Let
 ut each ,F’@is nowhere dense,, for if p E in& F”, then ?c is locaXly4 at p,
contradicting the definition of Iv, Since X < cdand X is &Bake implies
X is A+-Baire! the theorem is proved. The corollary follows from Theorem
2.3. El

A variation. of the proof of Theorem 3.6 suffices t3 prove Arhangel’skii’s

result (our Theorem 1.l), which of course inspired our work. We state a.
slightly stronger version.

Theorem 3.8. Every point-finite open cover o,f c6CCC B&e spme is count-
able.

Proof. Making an obvious definition of point-Pa; one proves by induction,

as pointed r_ but by Arhangel’skii, that for each natural number pz,every
point-n open cover of ,a CCC space, is countable. Let Cu be an uncount-
able point4nite open cover of a CCC Baire space X. By Proposition 3.4,
there is an open V such that %Yis not locahy countable at any point of
V. Let ’

I;, = {p E V: tv is point-@ + 1) at p).

jr== II-ncwA JTn md X is Baire,, so for some .JI?

int Fi + Q).
Ijut then int F, intersects only finitely many members of tu , contra-
dicting our assumption on V. El

The following defmitions are taken +- with minor changes - from the
fundalnental tract [ 19 1. We use IX!,to Ltenote the cardinahty of a set X.

Definition 3.9. Let X be a.topological space.

weight :
w(x) = H, mint193 I: % is a basis for X];
n-weight :
n(X) = N, min{I 91: 5;Bis a ?rmbasisforX],
322 F*D. Td& The etountuble&in condition vs. sqmahilO’y

width :
dd(.X) = supid( Y c_ x);
cxzllularity:
c(X) = H, sup(le 1: 6 is a collection of disjoint open sets};

cc(X) = sup{c(Y): Y c_X}

= HOsup{lY]: Y C_X, Y discrete};
LindelGf tziwnber:
L(x) = PC,min {K : each open cover of X has a subcover oQ’
power G K);
hereditary LindelGf number:
= sup{L(Yj: Y s X};

= min {I c)[1,1: 9, ir; a neighbourhood basis at p} ;

=sup{~@,x):pEx);

= min(I E?+: eP a collection of open sets such that

n (% ==<yH;

= min (K: p E X E 2 implies that there is B C_1d

such that p E B, @I = K),
= sup{tQl, X): p E X}.

(Note that whenever t ghtncss is mentioned, X is assumed T,). ’

Arhangel’skii [ 41 has introduced a new cardinal invariant, the h@ht
of a spare. Unfortunately, it is not thle same as the height of Juhkz [ 19 3,
which is why we have used LL 5x the: latter,

. f;(X) is the least cardinal b: such that each point is (son-

pact 291 K of character 4 ‘K)i.e., such
 p2.D.Tall, i%e courttablechain conditionvs.sqwability 323

cv,>,,ia, X <
sets K, each including , such that any open set about K
includes one of them.

e next list some relations among these functions that we will be re-
e numbers in parentheses .are the theorem numbers in [ 191.

ion 3.11 [ 19, (2.l)j. t(X) \< m!in(x(X), dd(X)}.

roposition 3,12 [ 19, (2.26)]. Let Y C: X. Then d(Y) S d(y) t(P).

Pro.position 3.13 [ 19, (2.16)]. X Hausdorfi implies 1x1 G 2°((x)+Ccx)).

Proposition 3.14 [ 19, (2.7~1. X Hausdorff implies d(X) < 2cc(?

Proposition %I5 119, (2.1)]. n(X) G d(X) J&X).

Proposition 3.16 EI?, (2.3)]. X regulur and ‘7 = X imply n(S) = n(X).

Propositim 3.17 [ 19, (2. H7)]. X Hazrsdorff impkes e(X) < LL(X).

Proposition 3.18 [ 19? (2.1)]. cc(X) G LL(X).

SapiruvslG [ 27 ] has shown:

Proposition 3.19. If X is Haus&wff, t(X) G cc(X) -h(X). If X is a k-spncs

also, then t(X) < cc(X).

With lCsurdictionary well stocked, we may proceed to new results.

Tightness, caliber and density are closely related. For example:

Propositian 3.20. If d(X) = (Xl, the cofinality of d(:X) is not a cailiE& 0.~
X. Thus iOf1x1 = Hi and X has caliber H,, X is separable.

ropwition 3.21. If t(X) is less than the cofinality of d(X), then the co-
j)‘inality o.f d(X) is not a caliber qf’_K

iro~fs. The proofs are the same. The method is useld in [ 71 and f 121s
Let d(X) :=K. Let L be cofinal in K? I C 1 = X, the csfinality of IC.ILet:
(xv Q < K} be dense in X, 9n the firs- case, assume ,&Q tl~t:
l
 ‘lkis is essentia’ y proved in [7].

mma 3.23. Suppose d (X>> u. Then tdzereis Y C_X such that

IYi =d(Y) = K+.

The

hereditarily separable, rt has a non-

ble subspace Y of power Hl. d(Y) Q 3 t, which is a caliber of P,
80 by Proposition 3,2 1, y is separable. But then by 3.12, Y is;separable;
contradiction. U

Corollary 3.25. If X is a k-spaceand each closed subset of X has caliber

13I, then ;h’is hereditarily separable.

aof. Sinr-=e
each closed subset has caliber:H1, certainly every closed
subset and therefore every subset has the CW. Therefore by Pioposi-
tion 3.19, X has countable tightness, and the result follows. 13

Next we have two applications of the fact that a space o

Hd having caliber Hl is separable.

’ is hereditarily separable if md onlryif it henlditarily

[ 121. Suppose 2”O=

 RD. Toil* The eotrprtddr && ecrqditbn vs. sqw&lity x5

se, for the first, tha X is not hereditarily separable, Then

25, it ks a non-separable subspace of cardinality Hfc But
tihis is impxsible. The second foXiow& from Proposition 3.13. which im-
plies that the xdinality Q~X is c 2% e3

Observe that irom Corollary 3.22 it follows thax if thi; density of a

space with count,ible tightness satisfying the Sanin condition is less than
N,, it is countable. Csfcul&ting cardinalities to ensure this, we get: -

(2) t(X) < H, and cc(X) G N,, or

(3) cc(X) < H, a& ,Y is a k-space.

Proof. The first again follows from Proposition 3.13. The third fOrlows
from the second by Proposition 3.19. The second follaiws from Proposi-
tion 3.14. Cl

4. Martin’s Axiom, the CCC md separability

We now explore the implications of Martin’s Axiom fur the question

of when CCC spaces are separable, Theorems 4.1. an3 4.3 are the funda-
second and third pressings
mental results; the rest is comprised of %Eze
of thz grapes.

Theorem 4.1. Suppc;lse M..Let X be an absolute GAspacle, for mm

X < PO (or a regub space with all closed subsets basis-compact), with
countable tightness *andsuch that every closed subset of A’ AUS
~c.
Then X is hereditarily separable.

Corollary 4.3,. Suppose M. Then every tech-complete space with closed

subsets CCC is her&itariJy sepnmble.

Proofs. To prove the theorem, by Theorem 3.24 it suffices to aihow that

each closed set has (caliber )Ilt1. But Corollary IL’7assures this. ,l[‘hecoral-
lary is a special case: of the theorem, since &Ned sets inherit C&h c
pleteness, closed se-is 3 imply cc(X) =: H
k-spaces [2], and so t
326 i%D.Talil,The countablecl&n conditionvs.sepmb#ity

ewern 4-3. Suppose Let X be absolute q, X < 2Nflfor regular and

is-compact), and CCC. If a(X) < 2N9 then X isseparablc

. Suppose M. Let X be absolute GA, X < ?2!{Q(w regular and

basis-compact)p and CCC. If there is a dense set-? s X, 1YI < %*O,such
that x(y, X) < 2”o foreach y E Y, then X is sepcrable. \

5 Suppose M. Let X be ab.w&e G,, h < 2H* (or regular

with closed subsets basis-compact], and CCC. If (x(X))’ < Z?o, then
X ii separable.

The proofs are all adapted from the proofs for the basis-compact
case in [ 161, which used the partial order version of Marti,r 2 Axiom. We
state three lemmas, the first of which wiil be discussed in Section 5, and
then prove the theorem.

~Lemma4.6 [ 19, Theorem 5.5j. M it-vlpliesthat any product c f CCC

spaces is CCC.

Lemma 4.7. [ 14, Theorem 2.91. A countable product of absolute G,

spaces is absolute G, .

Lemma 4.8 [ 4 1, Theorem 2.3 11. Any product of basis-compact spaces

is basis-compact.

hoofs. Let c10be .he usual basis for the product topology on X? Let
ri be the zth proje&ion map. Let 9 be a n-basis for X, (91 < 2% For
eachi% P let

Then LIP is dense open in Xw, for, given. any basic open set B, change
its value from X to P at some coordinate. The result is included in
DP (I 8. Since X” is CCC and either absolute G,, X < 2K*, or regular
and basis-compact, by M there is a p E N (I$: P E 5p}. Claim
S = {QJI~: t < ti} is dense in X. For let U be nonempty open in X.
Then there is P E F such that P & U. Since p E DP, there is a B E g
and an i *< W such that p E B arid w#3)= P. ut then vi(p) E P and
therefore Ril$) E C . CI

The: proof actually establishes someth

 327

ore Let X be absolute G,, X < 2*S kw regular apzd

basis-compa~~t)and CCC. Let @ be QcolIection of fewer than2”o open
subsets of X. Then there is IIcountable D C_X such that euch waemberof
63 contains c member ofD.

This result is due to Sapirovskii [ 291, who probably used it for MS

proof
. of Theorem 1.3 since that follows easily We prefer our proof since
i~ works for all strong Baire spaces, not just those classes closed under
c>unta,ble products.
Corollary 4.4 follows immediately, using Propositions 3.15 and 3.16
and:

Lemma 4.10 f21, Corollary 3.21. Martin’s Axiom implies that 2%~ re-
gular.

To prove Corollary 4.5, we need:

Lemma 4aI.1 [ 19, Theorem 2.251 vIf x(X).< d(X), then there is Y C_X,,
such that i YI = d(Yj = (x(X))’ and c(Y) < c(X).

) < x(X). If so, n(X) < 2”0, and we are done. If not, bye
Lemma 4%1,I there is CCC Y -c X such that
= d(X)
IYI = (x(X))” < 2%
Consider y. It has the CCC and is a&s*oluteG, (or basis-compact);
n(P) < 2”0, so it is separable. But by Proposition 3.12,
d(Y) G d(Y) t(P) G x(X),
contradiction. i3

In f 151 the question is raised whether in Corollary 4.5 the conditiolvr

x’ < 2No can be weakened to x < 2”o or even x 6 2% If there is a
model of set theory in which Martin% Axiom holds and 2”4,is inacces-
sible, then in that model x < 2”Q will d-2. However this isnot a very sa
tisfactory answer. In approaching the problem, it may be useful to not e
that bgr Theorem 4.9, if x < 2H0,then each non-isolated p,oint is in the
closure of a countable subset of its) complement. If x = 2”o, the ques-
tion h:as a negative answer - see Example 7.5.
It would be interesting to know whether character could be replaced
by tightness in 4.5. In fact the foregoing statement applies to a number
338 lW. Tull, The countablechin condWonus,separability

of problems in the area of cardinal jtnvariants. However, we do have a

number of results in which tightness replaces or supplements cha
with additional ad hoc hypotheses. There is no claim that these results
are best possible. The first one counts as another corollary of Theorem
4.3 and combines in its proof the characteristic lemmas associated with
character and tightness, Lemma 4.11 and Proposition 3.12.

Corollary 4.12. Suypose M. Let A’ be absolute G,, X < ZHo for be regular
and Izaveclosed seis basis-compact), and CCC. If there is a dense set
Y c X such that (x(Y))’ < 2”o and t(X) < x(Y), then X is separable.

Proof. If 1 is separable, so is X. Assume Y is not separable. As a dense

subset of a CCC space, Y is CCC, so by Lemma 4.11 there is 2 C_Y,
2 CCC, such that d(Z) = 121 = (x(Y))’ . By Corollary 4.4, z is separable.
Also, t(z) < t(X) < a~(Y) and d(Z) < d(Z) t(Z), so d(Z) < x(Y), contra-
diction. 0

There is a proposition asserting the existence of dense sets of low

character under certain circumstances:

Broposition 4.13 [ 19, Theorem 2.24 1.If X is absolute G, and 1x1 *< 2K,
then (p: x(p, X) < K) is dense.

(The referee points out that the remark after 2.22 in [ 191 that 2.24
can be extended to absolute G, spaces, X < ic, is incorrect.)
In particular, if X is absolute G, and 1x1C 22”, e.g. if 1x1
(p: x(p, X) < 2”p) is dense.
y we do not have x+ < 2No(unless 2NQis inaccessi’ble), so
Corollasy 4.12 seems to be a dead end. However, Proposition 4.13 does
have applications:

m 4.14. Suppose M, Let X be CCC absolute G,, I;YI< 22**, and

d(X) t(X) < 2% Then X is separable.

roof. The set Y = {p: X(J),X) fW2No) is dense, and d(Y) < d(X) t(X) C 2%
Let Z%be dense in Y, and IZI < 2% Then 2 is dense in X, and by Corollary
4.4, X is separable.

. Suppose CC absolute

_ 7 ,
I
,/ ’ : ’ <
.
’
I .I
L’ ’

’ :
/-
 F.D. TuA?,The ctxmtcvble
chain conditionvs.separability 329

n 5291 it is stated that IK i: plies that any CCC absolute GHI space
with countable tightness is separable. I do not.:know how to prove this.
ogue of Lemma 4.11 plus the following theorem would
stiffice.

Theorem 4.16. Suppose M. Let X be absolute G,, X < 2*0 (01’regular

and basis-comlract.),and CCC. If d(X) < 2*O,then X satisfies i’hejanin
condition, If in addition t(X) = H, and d(X) < H,, then X is separable.

Proof. X is a strong l3aire space, and so has as calibers all regular ca.rainals
less than 2% On the other hand, clearly each regular K > d(X) 4s a caliber
of X. Thus X satisfies the Sank1 condition. The second clause follows
from Corollary 3.22. KFI

A strong &tire space of cardinality < 2H0is easily seen to have a dense
set of isolated points, and hence if CCC, is separable. M may be dispensed
with in the absolute G, case in favor of Proposition 4,13 to yield:

Theorem 4.17. Let X absolute G, and CC , 1x14<2% Then X is separable.

We have shown that, assumrng M, certain classes of non-separable CCC

spaces do not exist. Other classes can be shown pot to exist by showing
that the inclusion order on the topology of a member would include 3
Souslin tree. See [39].
A Souslin space (see Section 5) is an example of a +ereditarily Lindelof
regular space which is not separable. It is an old problem whether a ‘kea1”
example exists. One naturally wonders what are the implications of 14
for this problem.

Theorem 4.18. Suppose M. Let X be herditarilyLindeEf and absolwl’e

GA, A+ < 2% Then X is separable.

Corollary 4.19. Suppose M. Let X be hereditarily LindeEf and tech

complete. Then X is hereditarily separable.
.‘..
For the compact case, this is Juhk [ 181.
333 ED. Tutt,The countable chin con&on vs mptwubiiify

ofs, It is well-known [ 19, Theo~i~ 0.81 that in a compact Hausdorff

space, charac er and pseudocharacter are equal. By Proposition 3.17, a
hereditarily Lindelii ausdorff space has countabk pseudocharacter,
zily) CCC. By consid-Tring a compactifkation,
m is seen to have character G X, and hence
by Corollary 4.5 is separable.
The corollary follows from Theorem 4.1.0

finitkn 4.20. A space X is perfect if every closed set Is G6 u

It is well known that a regular space is hereditarily Lind&f if and

only if it is Lindeliif and perfect.

Corollary 4.21. Suppose M Let X be CCC, absolute C,, X+< ZHo.,

meta-lindeltif and perfect. Then X is separableand hereditarily
LindelG$

Proof. j/(X) < N,, hence x(X) < X+, so X is separable by Theorem 4.18.
A separ&le meta=Lindelof space is Lindelof, so X is hereditarily
Lindeliii ‘, 0

Corollas-1 4.22. Suppose M Let X be CCC, tech complete, meta-

Lindel6i and perfect. Then X is hereditarilyLindeliif and hereditarily
separabltI

:roof. By Corollarv 4.2 1, X is hereditarily Lindeliif. Ey Corollary 4.19,

X is hereditarily seii?arable.

Corollary 4.23. Suppose M. Let X be CCC, tech complete and here-

ditarily meta-LindeltiJ Then X is hereditarily Lindeliif and hereditarily
separable.

roof. It suffices to show that every open subset of X is Lindeliif, for

then X is heredi;:arily Lindeltif, and hence perfect, so Corollary 4.22
$an be applied. By Corollary 3.7, X has caliber H,. Calibex H 1 is inherited
‘by open sets, so open subsets of X are Lindelof. U
 usk ‘s conjecture in the inboduction
used, but rather one well-known to
equivalent to it. Souslin actually conjectured that if X is a linearly
ordered set with no first or last element, such that every collection of
disjoint open intervals is counta le and X is connected in the induced
interval topology, then X is sep le and (hence) homeomorphic to She
real line. A counterexample to either form of the conjecture is know3.s
as a SousEinspace.
It is also well known (and easily shown by using the Dedekind com-
pletion, see e.g. [ 19,O. 171) that t&e conjecture is equivalent to asserting
that all CCC compact linearly o ered topological spaces are separable.
Let X be a CCC compact linearly-ordered topological space which is not
separable. From [ 19, Theorems 2.5, 2, lo] it follows that X is hereditar-
ily Lindeliif. and
c(X) = cc(X) = t(X) = x(X) = J/(X) = H,,
d(X) = dd(X) = n(X) = tE,,
IX]G 2No. /
.’
1x1 = 2Hofollows from our Theorem 4.17. Thus every coasequerce of M
in Section 4 can be used to establish Souslin’s conjecture. Startmg from
scratch, we prefer to prove first that if X is a CCC linearly oru’.eredspace,
then
cc(X) = t(X) = H,;

next, that M implies that every compac&Hausdorff CCC space has caliber
H,, and, finally, that if the clozed subsets of a space with countable
tightness all have caliber H19 the space is (hcreditanfy) separable.
Stated in this form, the proaf raises the questim of the &I-W or”W
hypotheshs we shall calI H :
I-k Ev~y compact Hausdorff CXC space has caliber H, .
H has many of the consequences of INin the CCC versus separ-abiYty
are; ‘.ut is not obviously incompatible with the continuum hypoth.esis.
Indeed we conjectured in [ 341 and in the original version of this paper
that the two were consistent, his question is also raised as [6, I?rob
S.S.l].) This conjunc been an interesting axiom., im
xample that every cornpa ausdorff CCC space of ~ardi~~~lit~
332

of power and caliber M, . Example

< 2Lc0is separable, since it is a s~pacxb
7S (due to P. Erdos and the author) establishes, howevcx, that H implies
2Ho> H, Complete details will appear in [40 j. It is not known whether
H implie; 2N0= 2% In [4O] it is shown that the slightly stronger assump-
tion that.eech-complete CCC spaces have caliber k+_does imply 2’0 t=2%
This sheds some light on the questi XI of why one cannot seem to prove
that tech-complete CCC spaces has*ecaliber H, merely by assuming their
compactifications do. Surprisingly, however, it follows from [ 19, Thee-
rem 5.41 that H implies that regular basis-compact CCC spaces have ca-
liber. k$ q
H has a number of interesting equivalents. We shall mention two here,
more are in [4S].

Definition 5.1. A space has prectlliber H, if every uncountable collection

of open sets has an uncountable cf:ntered subcollection.

Remark 5.2. Clearly caliber N, implies precaIiber H, implies CCC. W’t:

have previously noted thait a subspacI: X is CCC if and only if x is. X
has precaliber H, if x has caliber 13,. Also, if X is completely regular
and x is a compactifkation of X, then, if X lhas precaliber H19 x has
caliber HI.

We now introduce:

Every (completely re,gtilar)CCC spuce has preculiber N,.

It follows from Remark 5.2 that ’ is equivalent to H. The hypothesis

oj‘complete regularity can be removed with some; effort, e.g. by shutt-
ling back and forth between the original space and the Stone space of
its Boolean algebra of regular open sets. Or see [ 19, Ch. 5 ] for a differ=
ent approach.
Another equivalent of H pinpoints how it represents a
weakening of M.

H” : ~fwAY<,,
is a collection of derase open sets in a compact
Hausdorff CCC spwe, such that Q < fl C ul implies Uo,2 UP,
then cucof Z& is dense.

to prove the iother ditectie:l, observe that

(Cor #F-V X7), we really only used
 Examming the proufs in Sectio 4, and ignoring basis-compa&.nessfor
simplicity, WGsee that:

usdurff and has elmed sub-

ereditarily se~~~arable.
If’ X is comp~!ct Hausd’orff
CCC, t(X) = HO9aradd(X) < H,, then X is sqvarable.

By what should now be familiar techniques, from the latter h(alfone

can prove :

Corollary 5.4 [ 19,5.6]. H implies that every compact first-countable

CCC Hausdorff space is separable.

None of our other results appear to follow from H instead c3fM, There
are two other propositions intermediate between H and Souslin’s conje*
ture.

Theorem 5.5. Each of the folluwir’ngpropositk~~ implies the next one

(a) W
(W H;
(c) given an uncountable collection i? of open sets in a CCC space,,
there is an uncountable subcollection @’ such that any twa members of
E!’ intersect;
(d) every product of CCC spaces is CCC;
(e) Sotrsli~~
3 conjecture.

Proof. That (c) implies (d) is due to K. Kunen. There is a pr~f on f 19,
p. 621. (d) implies (e) is due to Kurepa [ 2Q’Je ff

Jensen has shown that the continuum hypo h&s is consistent with
Souslin’s conjecture. Example ‘7.5satisfies (c), but if the contiliuum
hypothesis holds, it does not have caliber HI, as remarked earli(3r Other
relations among these propositions rmain unknown. See [40] for
further discussion,

6. Paracompactness an iza:tio

As mentioned in the introduction, we came to the subject of t?Cspa-

per via a round bout route from questions of paracompact
334

metrizxttion. Vk therefore list the implications of what has gone before

for these questions. Once one knows how, the proofs are trivial and
therefore are omitted. The last four need results of Smirnov [ 3 11 quoted
below.

m 6.1 (Juhgsz) . ,Every CCC pnra-Lindeliif space is Lindeiiif.

Thweum 6.2, Every CCC (hereditarily 1 metacompact Baire space is

(hereditarily) Lindel6.f

Theorem 6.3. Every locally CCC regular para-Lindeliif space is pizracoIn-

pi;ct.

Theorem 6.4. Every ZocalZyCCC regular space with Qo-locally count-

able base is metrizable.

Theorem 6.5. Every locally CCC re~@ar metucompact Bai.rc space is

paracompact. (Hence every locally con. 0ac& perfectly normal, metu-
compact space is paracompact. )

Theorem 6.6. Every !ocdlly CCC regukr Bake *cpacewith a u-point-finite

base is metrizabk.

Liefinilion 6.7. A coHec ion of sets is star-countable if each intersects at

most counfably many others.

Lemma 6.8 (Smimov). Let X he a regrrkurspace, If every open cover of

X has a star-r?ountabBeopen refinement, then X is paracompact. If X has
a star-countable base, then X is met,vWble.

eplacing “point-finite” by “point-ccruntable” and “Bake” by

‘% 2-Baire”, we get :

. k?ery CCC meta-LintFelof f%@aire space is LindeZtif

ewe . Every locally WC regular meta-Lindeliif $?3ire space

is paracompact,

c9un t6rMebase is me trizable.

 .Suppose et X be absolute c;,, h < 2No(or regulcjr
crnd basis-compact). If
X is ahereditarily Lintlcitif space with a point-
countable base, X is metrim

It is not known if M implies that every hereditarily Lindeliif regui!ar

space with a point-countable base hs met&able. Ponomarev [23] ant!
Bennett [s] have established that if Souslin’s conjecture fails, there is:a
non-met&able such space. See also [ 371. Sapirovskii [ 291 has proved
that CCC tech-complete spaces with point-countable bases are met&-
able.
In contrast to the consequences of M in Section 4, I do not know
if the one in Corollary 6.12 and the ones implicit in Lemma 6.8 and
Theorems 6.10 and 6.11 are independent of the axioms of set theory.

7. Exampies

Example 7.1. A compact Hausdorff space which satisfies the jhrtin curt-
dition, but is not separable. Let X be the product of 22”’ copies of the
two-point discrete space- Then d(X) = 2’0 (set [ 19, Theorem 4.11). Ca-
liber K, for regular
Y uncountable K, is preserved by arbitrary products, so
X satisfies the Sanin condition.

7.2. A space with caliber HI which does not hmc caliber N,,
and hence does not satisfy the janin conditiort. Let X be the set of or-
dinals less than m2, with all sets having complements of cardinahty < tf,
being open.
There are better examplcis, this space not being flausdorff. See [30] q
where Hausdorff spaces having pred;.etermini:;ds&s of calibers are con-
structed:

ExampJle7.3. A CCC space which does mt haven&ber M,, Pixley and

Rx>y1223 construct a completely regula: CCB’J non-separable Moore
space .X with a o-point-finite base CiB.Sir+,ce,
X is not met&able, c1Bis
unl:ountable. Not only does any uncountable subcollection of g have
void intersection, there is also an uncountable iB’ C_q such that any
infinite subcollection of g’ has ~id intersec:tion. This shows the in-
consistency of extending ‘ from finite to ctiuntable,
336

hangeS’&ii’s p;rpaces [33 comprise a natural extension of the class

of Cech-compkte spaces. X is
a p-space f 111. Id follows that tin’s Axiom doe
spaces, since X is CCC but do

.A CSC hire sp m forwhich

“work”. I am grateful to
number of other interesting p&
space. Then d(L%) = K,, dd@ *-1 r=H, (se
be the subspac : of DHf consisting of those functions which are 0 except
at countably many places, It is not hard to see that X is not separable,
thus d(X) = H,. We simultaneo ly show X is not &&ire and does not
have caliber H , by constructi dense open sets {Uo)crKwl such that for
any uncountable A !Z,ol.

Partition C+ into H 1 disjoint uncountable sets (B,), < w1. Let

U. = CfE .X: for some C’jER,, fTp) = 1).
U’ is dense open in X, but the intersection of any uncountable collec-
tion of U,‘s is empty, else it would contain a function having uncount-
ably many l’s in its range. .
X is dense in D” 1, so ..Yis CCC; indeed X has precaliber H,. By the
same method Booth f9] uses to plrove the consistency of D*l being se-
quentially compact, ave proevethat X is se .juentially compact and hence
Baire, Let uR},, Wbl=:a sequence in X. Le; S =
(cu: for some yt,fn(a)= l},,
LGt 0 = sup S. Let I map w one-one onto 0. Let

PA:,= {n : frdl(i(k))
E = E}.
Let
if Po,o is Minite,
Q0 otherwise.
 e countuble chuin co 337

There is an infinite such that for ea is infinite.

f~ Dnl by

ace ing to whether Pk 0 or Pk 1 was chosen aI.bove.Define f to be 0

everywhere else. Then ii the nakal ordering on Q,

since ’ Ifor each 01<<oI, Ii(cu) agrees with j’(a) for almost all n.
R. Engeking and E. van Douwen have pointed out that it follows
from hat X is normal.

Example 7.5. A compact Hausdorffspace

. X such that
( 1) every uncountable collection of open sets has an uncountable sub’-
coilection such that each pair has rton-empty intersection.
(2) x(X) = 2%
(3) X is not separable.
(4) Martin’s Axiom implies that X does not have caliber Z*O,,hewe
the continuum hypothesis implies that X does not have caliber H1.

X is the dual space

1 of the Boolean algebra c10of tebesgue measurable:
subsets of the unit interval, module sets of measure 0, (1) follows from
the analogous statement for C’k9, proved for arbitrary measure algebras in
[ 17 3. (3) is an exercise on [ 15, p. 108]. (4) is proved in [40]. (2) follows
from results of Efimov [ 131: since X is compact CCC and extremally dis-
connected, by [ 13, Corollary 81, LIzHo is a continuous image of X. Then
by [ 13, Corollary 111, x(X> 3 2% On the other hand, tcls I = 2’0 since
every r-4easurable set is equivalent to a Bore1 set, so vv(X) G 2’0, and
therefore x([X; -= 2’0.

References

111J.M. Aarts and D.J. Lutzer, Completeness propertiesdesigned for recogr&ingBairespaces,

Dissert3tionesMath. 116 (1974) l-48.
PI AV. A&angel’&& Bicompact sets and the topology of spaces, Tram!!.Moscow Math. Soe.
13 (1965) l-62.
[31A.V.Arhangel’skii,On a class of spaces containing all metric spaces wd aUlocally bicome
pact spaces, Am. Math. Sot. Transl. (2) 92 (1970) l-39,
141AX ArhangeWci~,The Suslin r‘umberand cardinality. Characte!;sof:’points in sequent%
bicompacta, Soviet Math. Dokl, 11 (1970) 597-601.
338 FL?. TuZl,Themm&ablechainconditionvg separability

(51 AX. Arhangel’ski~,The property of paracompactnessin the class of perfectly normal, lo-
cally bicompact spaces, Soviet Math. D&i. 12 (1971) 1253-1257.
A.V. A&ang&sk& On cardinalinvariants,in: J. Novak, ed., General topology and its &a-
tions to modern analysis and ytigebraIII, Rot. 3rd Prague Topological Symp., 197 I (Aca-
demia, Prague, 1972) 37-46.
A.V. ArhangeIWi and V.I. Ponomarev, On dyadic bicompacta, Soviet Math. Dokl. 9
(1968) 1220-1223.
H.R. Bennett, Quasidevelopzrblespaces9Ph.D. Thesis, Arizona State University, Tempe,
Aria (1968).
D.D. Booth, Countably indexed ultrafilters,Ph.D. Thesis, Universityof Wisconsin, Madison,
(1969 j.
‘I!!iisc.
H.H. Carson, Normality in S’tdbsets of product spaces, Am. J. Math. 81 (1959) 785-796.
G.D. Creede,Concerningsernistratifiiblespaces, Pacific J. Math.32 (1970) 47-54.
B. Efimov, Solution of some problemson dyadic bicompa&, Soviet Math. DokI. 10 (1969)
776-749.
B. Efiiov, ExtremaIIy disconnected compa :t spaces and absolutes, Trans. Moscow Math.
Sot. 23 (1970) 243-285.
Z. Fro&c, Generalizations of the Grproperty of complete metric spaces, Czech. Math. J.10
(1960) 359-379.
P.R. HahuospLectureson :tlooleanAlgebras(Van Nostrand, Princeton, N.J., 1963).
A. Hajnalandd. Juh&sz,A consequence of Martin’sAxiom, Res. PaperNo. 110, Dept. of
Mathematics,Univ. of 0lary, Calgary,Alberta.
A. Horn ar,b & Tarski,l%asures in boolean algebras, Trans Am. Math.Sot. 64 (1948)
467-497.
I. Juhaisz, Martin’s Axiom’solves Ponomarev’sproblem, Bull, Acad. Polon. Sci. S&. Sci.
Math. Astr. Phys. 18 (1970) 71-74.
1. Juh&z, Cardinal Functions in Topology (Mathematical Centre, Amsterdam, 1971).
G. Kurepa, !,a condition de Suslin et une prop&! caract&istiquedes nombres r6els,C.R.
Acad. Sci. Paris~31 (1950) 1113-1114.
D. Martin and R.M. SoWay, Internai Cohen extensions, Ann. Math. Logic 2 (1970) 143-
178.
C. PixIey and P. Roy, Uncompletable Moorespaces, in: Proc. AuburnTopology Conf.,
%rch 1969 (Auburn, Ala., 1969) 75-85.
V.I. Ponomarev, Met&ability of a finally compact pspace with a point countable base,
Soviet Math. DokI. 8 (1967) 765-768.
M.E. Rudin, Soushn’s s:onjecture,, Am. Math. Monthly 76 (1969) 1113-l Il9.
N.A. !&in, On intersection of open subsets in the product of topological spaces, Doki.
Akad. Nauk SSSR N.S. 53 (1946) 499-501.
N.A. &nin, 0n the product of topological spaces,,Tr. Mat. Inst. Akad. NalukSSSR 24
(1948) 112 pp. (in Russian].
B. &pirovskii, On discrete subspaces of topological spaces; weight, tightness and Suslin
number, Soviet Math. Dokl. 13 (1.97:2!:2115-219.
B. &pirovskii, On tile density of topological spaces, Soviet Math. DokL 13 (1972) 1271-
1275.
B. &.Grovskii,On separabilityand met&ability of spaces with Souslin’s oond$tion, Soviet
Math. DOW.13 (1972) 1633-5f1i8.
S. Sk&h, Remarksou cardinalinvarYi& in topology, preprint.
Ju. MI.Smimov, On strong@ paracompact spaces, Izv. Akad. Nauk SSSR Ser. Math. 20
(1956) 253-274 (in Russian),
R&I. Solovay and S. Tcnnenb;*um, Iterated Cohen extensions and SousIin”s prablem, Ann.
hlath. 94 (1971) 201-245.
(331 M. Souslin, Problame 3, FUnd. Math. 1 (1920) 223.
[ 341 F.D. Tall, Souslin’s conjecture Ireuisited,in: Colloq. Math. Sot. J;ink,sBolyai 8. ‘Il‘opius
in
Topology szthely (Hun?- ry]), 1972) 609-615.
[ 351 F.D. Tall, A counterexampit arbthe theories of compactness and of met&z&ion, KoninkL
Nederl. Akad. van Wetensch. Proc. Ser. A 75 (1973) 471-474.
[36 j F.D. Tall, On the existence of non-met&able metacompact normal Moore spaces, Can.
J. Math. 26 (1974) l-6.
[ 371 F.D. Tall, On the existence of non-met&able hereditarily Underif spaces with point-
counmble bases, Duke Math. J. 41 (1974) 299-304.
[38] F.D. Tall, 1L1alternative to the continuum hypothesis and its uses in genexal topology,
preprint.
[ 391 F.D. Tti, Stalking the Souslin tree - a topolo Cal guide, submitted.
[40] F.D. Tall and K. Kunen, Between Martin’s Axiom and Sot&n’s hypothesis, in preparation.
[41] J. van der Slot, Some Properties Related to Compactness (Mathematical C&&e, Amster-
dam, 1968).