Internat.
VOL. 21 NO. 4 (1998) 729-734
729
ATOMICITY
OFMAPPINGS
MathematicalInstitute University of
Wroctaw
Pl. Gnmwaldzki2/4,50-384Wroctaw,
POLANDMathematical Institute University of Wrodaw P1.Gnnwaldzki2/4,50-384
Wrochw, POLAND
JANUSZ J. CHARATONIK
InstitutodeMatem/ticas UniversidadNacionalAut6nomade Mxico
CircuitoExterior,CiudadUniversitaria 04510Mtxico,
D.F.,
MEXICO andWt.ODZIMIERZ J. CHARATONIK
Departamento
de Matemticas Facultadde CienciasUniversidad NacionalAut6noma deMtxico CircuitoExterior, Ciudad Universitaria
04510Mtxico,
D.F.,
MEXICO(Received December 16, 1996and in revisedform December 11,1997)
ABSTRACT. A mapping f:X--, Y between continua X and Y is said to be atomic at a subcontinuum
K
of thedomainX
provided thatf(K)
isnondegenerate andK f-I(f(K)).
Thesetofsubcontinua at which agiven mappingis atomic, considered asasubspaceof thehyperspace of all subcontinuaof
X,
isstudied. The introduced conceptisappliedtogetnew characterizationsofatomic and monotonemappings. Somerelated questionsareasked.KEY WORDS AND PHRASES: Atomic, composition factor property, continuum, contratomic, hyperspace, mapping,monotone.
1991AMSSUBJECTCLASSIFICATION CODES: Primary 54E40,54F15. Secondary54B15.
INTRODUCTION
All spaces considered inthe paperareassumedtobe metric, and a mapping means a continuous function. Acontinuum meansacompact connectedspace. Recall thatamapping
f X Y
betweeacontinua
X
andYissaidtobemonotoneifthe inverse image of eachpoifitofY
(equivalently, ofeach subcontinuumofY)
isconnected. Asurjective mappingf X Y
between continuaX
andYis said to be atomic provided that, for each subcontinuumK
ofX
such that $(K) is nondegenerate,K f-l(f(K)).
Thenotionofan atomicmappingwasintroducedby R.D.Andersonin[1]
todescribe special decompositions of continua.Soon,
atomic mappings turned out to be important tools in continuumtheoryand provedtobe interesting bythemselves, and several oftheirpropertieshavebeen discovered, e.g.in[3], [5]and[6]. The following facton atomicmappingsisknown(see [3,Theorem 1, p.49]and[6,(4.14),p.17]).
Fact. Everyatomicmapping ofa continuumis(hereditarily)monotone.
Thepaperconsistsoftwoparts. Inthefirst onethecompositionfactor propertyisdiscussed for the class ofatomicmappings. The second part dealswiththefamilyofsubcontinuaof the domain continuum
X
atwhichagiven mappingf X Y
isatomic. Inparticular,atomicmappingsaswell asmonotone ones arecharacterizedby conditions concerning thestructureofthisfamily. The paperissuppliedwith a number ofexamples;open problemsposedinboth parts of the paperindicate some directionsofafurther studyinthe area730 J.J. CHARATONIK AND W. J. CHARATONIK
Thefollowing standard notation willbe used.
N, R
andC standfor thesets of positive imegers, reals,andcomplex numbers, respectively, equippedwith theirnatural topologies,ifneeded. IntheplaneR
thesymbol(x, /)
meansapoint havingxand/asits Cartesian coordinates.1. COMPOSITION FACTOR PROPERTY
Wesaythat aclass Jdof mappings has thecomposition
factor
propertyifthe composition gohof mappingshand g is in ,MonlyifgT. Makowiakaskedin[6, (5.22),p.33]iftheclassofatomicmappingshas thecomposition factor property, and conjecturedthat itdoes.
Later,
in [7, Chapter l, Example, p. 7] he has answered his questionin thenegative. Another answerwasgiven byE.E. GraceandE. J.Voughtin[4,Section 4,p 140],who have shown that for the naturalprojectionf
of the circle ofpseudo-arcsX
ontothe circleZ
(which is clearly an atomic mapping) there exist a continuumY
and two mappingsh:X--,Y
andg:Y--, Z,
such thatf
can be factored as the composition goh and g is not atomic. Both the conjecture ofMakowiakanditsnegative solution by himself andby GraceandVoughtshow that the composition factor property for the class oftheatomicmappings should bestudied in a more detailed way,andthat there are interesting problems aroundthispropertyworthwhileclarif3dng.Ingeneral,thefollowingproblemcanbeposed.
Problem 1.1. Let
X, Y
andZ
be continua, and let h:X--,Y
andg:Y- Z
be surjective mappings. Determine conditionsconcerning(a)
thecontinuumX, (b)
the continuumY, (c)
the mapping h, underwhichthe implication holdsif goh isatomic, then g is atomic (1 2)
Tobe more precise,imroduce thefollowingdefinition.
Definition 1.3.
A
class ofC
ofcontinua is said tohave thecompositionfactor
propertyfor
aclass.M of
mappings provided that for eachcontinuumX
ECif the composition gohdefinedonX
isin,M,
then g is inAcontinuumis said tobedecomposableif it istheunionoftwoitspropersubcontinua. Otherwise it is said to be indecomposable. A continuum is said to be hereditarily decomposable (herechtarily indecomposable)providedthateach ofitsnondegneratesubcominua isdecomposable(indecomposable, respectively). Finally recall thataspace
X
issaidtobehomogeneousprovided that for everytwopoints pandqofX
thereis ahomeomorphismf X
---,X
such thatf(p) q.Itis shownin
[2]
that thecircleofpseudo-arcs(thathasbeenusedin[4]
asmentionedabove) is constructedinthe Euclideanplane,isdecomposable,andishomogeneous. Therefore the result ofGrace andVoughtcanbeformulatedeven in astrongerform,asfollows.Theorem 1.4. The following classes ofcontinuadonothave the composition factorpropertyfor the class ofatomicmappings: planecontinua, decomposable continua, homogeneous cominua,aswell as the intersectionofanyof these classes.
Onthe otherhand,it isknownthat each atomic mappingdefined on an arcwiseconnectedcontinuum isahomeomorphism provided that the imagecontinuum isnondegenerate(see [6,
(6.3),
p.51]).
Since the class of homeomorphisms obviously has the composition factor property [6, (5.14), p. 32], the following resultis immediate.Statement1.5. Theclass ofarcwiseconnectedcontinua
(of
locallyconnectedones,inparticular) has thecomposition factor property for the class ofatomicmappings.Thusthefollowing problemisnatural.
Problem1.6. Determinethe classes ofcontinua whichhave the composition factor property for the classofatomic mappings.
Twoparticular questions relatedtothisproblemareofaspecialinterest.
731 Question 1.7. Does the class of hereditarilydecomposable continua havethe composition factor propertyfor theclassof atomic mappings?
Question 1.8. Doestheclass of hereditarily indecomposablecontinuahavethe composition factor property forthe classofatomicmappings?
Asurjective mapping h
X
---}YbetweencontinuaXandYis saidtobeweaklyconfluent
providedthatforeach subcontinuumQof
Y
thereis a subcontinuumCofX
suchthath(C)
Q. Inconnection withProblem 1.1, part(c),recallthefollowing result(see [6, (5.29),p.35]).Proposition 1.9. If the mapping h
X Y
isweakly confluent,then implication(1.2)issatisfied.Notethatthe conversetoProposition1.9is not true. Namelywehavethe following example Example1.10. Therearemappings h
X
--}Yandg Y-
Zsuch that thecompositiongohand the secondmappingg areatomic,whilehis notweakly confluent.Proof. TakeasXthe wellknown
sin(1/x)-curve
SdefinedbyS
{(0,y) e R2:
ye [- 1,1]}
U{(x, sin(1/x)) e R2:x e (0,1]}, (1.11)
and let
L
be thelimit segment ofS. Identify thetwoend points ofL
and denotebyhX Y
the identification mapping. ThusYisthe unionofahalflineand thecircleh(L).
Nowletusshrinkh(L)
toapdint,andlet g Y
-
Zbethequotient mapping. ThusZ
is anarc,bothg and gohareatomic,while h is notweakly confluent.Asurjective mapping h
X Y
betweencontinuaX
andY
is saidtobeconfluent
provided that for eachsubcontinuumQofY
and foreverycomponentCof theinverseimageh-(Q)
wehave h(C)
Q.Since a continuum
Y
is hereditarily indecomposableifandonlyifeach mapping froma continuumontoY isconfluent(compare [6, (6.11),p.53]),and sinceeachconfluent mapping obviouslyisweaklyconfluent, wegetacorollarytoProposition 1.9,which isrelatedtopart(b)of Problem1.1.Corollary1.12. If thecontinuum
Y
ishereditarily indecomposable, then implication(1.2)is satisfied.2. ATOMICITY
Given a continuum X with a metric d, we let2x denotethe hyperspace ofallnonempty closed subsets of
X
equippedwiththe HausdorffmetricH
definedbyH(A,B) max{sup{d(a,B)
aEA},sup{d(b,A)
bB}}
(equivalently: withtheVietoristopology,see e.g. [8, (0.1),p. and(0.12),p.
10].
Further,wedenote byC(X)
the hyperspace of allsubcontinuaofX,
i.e., of allconnected elementsof2x,
and byF1 (X)
thehyperspace of singletons. The readerisreferredtoNadler’s book[8] for,needed information on the structureofhyperspaces. Inparticular, the following is well known
(see
[8,Theorem(1.13), p.65]).
Fact2.1. Foreach continuum
X
thehyperspaceC(X)
isa subcontinuumof the hyperspace2x.
Given amapping
f X
-}Y
betweencontinuaX
andY,
we considermappings(calledthe inducedones)
2f:
2x 2Y andC(f): C(X)
---*C(Y)
definedby
2f
(A) f(A)
foreveryA e
2x andC(f)(A) f(A)
foreveryAe C(X).
Thus,byFact2.1, thefollowingisobvious.
Fact2.2. Foreverycontinua
X
andY
and for each mappingf X - Y
wehave2IIC(X) C(f)
A proofof thenextfactisstraightfoward.
Fact 2.3. Let a mapping f:X---, Y between continua X and
Y
be given ThenC(f)(F(X))
CFI(Y).
732 J.J. CHARATONIK AND W. J. CHARATONIK
Foranarbitrary surjection
f" X
-,Y
betweencontinua we considersubcominua ofX
atwhichthe mapping satisfies the atomicity condition. More precisely, given a surjective mappingf" X Y
betweencontinuaX
andY,
wedenote by.,4(X, f)
the family of all subcontinuaKofXsuchthatf(K)
isnondegenerateand the equality
K f-1 (f(K))
holds, i.e.,t(X,f) {K C(X)\(C(f))-(FI(Y)) K f-(f(K))}.
(2 4) Thus thefollowing resultis aconsequence of thisdefinition.Statement2.5. Letamapping
X Y
betweencontinuaX
andY
be given. ThenX
4(X, f),
so,4(X, f)
isnonempty; (2 6)t(X, f) c
(2.7)Further,weput
B(X, f) A(X, /) (C(/)) - (F (Y)). (28)
Proposition 2.9. For every decreasing sequenceofcontinuabelonging
to/3(X, f)
the limit of the sequence alsois inB(X, f).
Proof. Foreachn Nassume
K, B(X, f)
andK+
CKn.
PutK
LimK
and notethatK
f{K,.,’n N}.
Considertwocases. First, if
K,., (C(f))-I(F1 (Y))
for almost allnN,
thenK (C(f))-I(F1 (Y)),
too, because
Fi(Y)
is compact, and so is its preimage underC(.f).
ThusK B(X, f).
Second, ifK A(X, f)
for almostall nN,
thenf-l(.f(K,)) K,
for theseindices n, and wehavef-(f(g)) f-l(f(Limg,.,)) f-l(Lim
f(g,))f-l(N{f(g,.,)
nN})
N{S-(f(K.)) , e r} N(K. e }
g.Thus either g
A(X,f) (if f(K)
is not a singleton), orK (C(f))-I(F(Y)) (if
f(g) is degenerate). Consequently,K B(X, f)
by(2.8).
The proofis then complete.Theexamplebelow shows that theconclusionof Proposition 2.9is not treefor arbitrary sequences of continua
In
particular, the assumption "decreasing" cannotbereplaced by"increasing" in Proposition 2.9.Example 2.10. There is a continuum
X,
anincreasing sequence ofsubcontinuaK,,
inX
and a monotonemappingf"
XY
such thatK, B(X, f)
for eachnN,
whil9
LimK, B(X, f)
Proof. Let
S
be thesin(1/x)-curve
definexi by(1.11).
PutA {(0,y) R
"y[1,2]},
and defineX S
t3A.
LetY=[O, 1],
and letf X Y
be theprojectiondefinedbyf (x,
y) x. Foreachn N let
K f-([1/(n + 1), 1]).
ThenK, .A(X, f)
CB(X, f),
andLimK
S. Sincef(S) Y,
we havef-(f(S))= X,
and thus LimK,B(X,f),
as claimed. The argument is complete.Theorem 2.11. Foreach surjective mapping
f" X
--,Y
betweencontinuaXandYthefollowingassertionsareequivalent:
/
isatomic; (2 12)(X, f) C(X)\(C(f))-’(F (Y));
(2.13)B(X, f) C(X);
(2 14)A(X, f)
is anopensubsetofC(X).
(2 15)73 3
Proof. Equivalence of
(2.12)
and(2.13)
is evidentfrom the definitions, and(2.14)
is anotherform of (2 13)by(2.8). Thus(2.12), (2.13)and(2.14)areequivalent. The implication from(2.13)to(2 15)is obvious. Wewillshowthat(2.15)
implies(2.14).
Tothis aimrecall that an order arc inthehyperspaceU(X)
is afamily ofsubcontinuaofXsuchthatfor everytwomembersA
andB
of wehaveeither AC BorBC A. Letp be anarbitrary point ofXandlet denote an order arc inC(X)
from{p}toX. By (2.15)theintersection
.A(X, f)
is anopensubsetofE. DenotebyCthe componentofthe imersection that comainsX,
and by K the (only) element ofthe boundary of C in E. Then KE
fA(X, f).
Takeadecreasingsequence{K,
EC rEN}
ofsubcontinuaofXconvergingtoK. Then K
{K
nEN),
and sinceK
ECC.A(X, f) C/3(X, f)
by (2 8), we infer from Proposition 2.9 thatK B(X,f).
SinceK A(X,f),
we haveK (C(f))-:(F(Y))
by (2.8), whence itfollowsthatf(K)
EF (Y),
andtherefore,bythe definitionofanorderarc, thesubarc of from{p)
toKis contained in(C(f))-(F:(Y)),
thusinB(X, f),
whiletherestof the orderarc,
i.e.,C,
is contained in.A(X, f)
byits definition.So,
we conclude by(2.8)
thatthe whole orderarc is comainedinB(X, f). Now,
sincep waschosen asanarbitrary point ofX,
weinferfrom CB(X, f)
that
C(X)
CB(X, f),
whence(2.14)
follows. Theproofiscomplete.Thenexttheorem is a characterizationofmonotonemappingsinthe introduced terms
Theorem2.16. Foreachsurjective mapping
f:X
YbetweencominuaXandYthefollowingassertions areequivalent:
f
is monotone; (2 17)c(f)((x, f)) c(r).
(2.1s)Proof. Assume
f
is monotone. Since oneinclusionof equality(2.18)
is obvious,wehavetoshow the other one. LetL
be a nondegenerate subcominuum ofY,
i.e. L_ U(Y)\F(Y)
PuttingK
f-l(L)
weseethatKis a continuumby monotoneity off,and we have-](f(K))= f-](L)
KThusKE
.A(X, f)
C13(X, f)
by(2.4)
and(2.8),whenceL U(f)(13(X, f)),
and(2.18)follows Assume equality(2.18)
holds. Takea poim 0EY
and, to show thatf
is monotone, i.e., thatf-l(t0)
is connected, consider for each positive integer rz the componentL,
containing 0 of a closed 1/r-neighborhood about F0 in Y. ThusL
is a nondegenerate subcontinuum ofY,
e,L U(Y)\F(Y).
Againby(2.18)
we infer thatL, C(f)(.A(X,f)),
whence(for
each n)there existsanondegeneratesubcontinuumK,
ofXsuch thatf (K,)
L,,andK,f- (f(K,)) f- (L,)
Observe that for each n we haveL,+
CL,,,
i.e., that the sequence{L.,}
is decreasing, and that{/0} t’{L,
nEN}.
Thusitfollows fromthe equalityK, f-(L)
that thesequence{K}
isdecreasing,too, and wehave
Thus
f- (/0)
isacominuum astheintersectionofadecreasingsequenceofcontinuaK,
Theproof isthencomplete.Theorems2.11and (indirectly)2.16 motivatethe following question.
Question 2.19. What is the Borel class of the set
A(X,f)
considered as a subspace ofthe hyperspaceU(X)?
In connection with
(2.6)
of Statement 2.5 observe that the mapping h of the unit circle S{z
EC[z] 1}
(where Cstandsfor thecomplex plane)omo
itself definedbyh(z)
z has the property that the whole S is theonlyelement of.A(X,f),
i.e.,.A(X,f) {S}.
Generalizingthis phenomenon,consider thefollowing class of mappings.Definition 2.20. Asurjectiv mapping
f X Y
between continuaX
andY
iscalledontratomic provided that.A(X, f) {X).
Inotherwords, a nonconstant mappingf
is contratomic ifthe only734 J.CHARATONIK AND W. J. CHARATONIK
subcontinuum
K
ofX having nondegenerate image and satisfying theequalityKf-l(f(K))
isX itself.The above consideredmappinghof
S
ontoitselfis anexample ofa contratomicmapping. Notethat the classof contratomic mappings doesnot containhomeomorphisms(moreover,any homeomorphism, being atomic,is notcontratomic). Inconnectionwith thisobservethat ifagainh S --,S is definedbyh(z)=z
2 and g:S1--,S is the identity, then the compositions hog=h and goh=h are contratomic,while g isnot. Thisleadstothefollowingobservation.Observation 2.21. The class ofcontratomic mappings does not have the composition factor property.
Proposition 2.22. Leth
X Y
and g:Y
---,Z
be surjcctive mappings between continuaX, Y
andZ,
respectively. If either horgiscontratomic, then the composition gohiscontratomic,tooProof. Forany
K
CXwehaveK
Ch-(h(K))
andh(K)C g-(g(h(K)))
Let
K _ C(X)\{X}.
If h is contratomic, thenKh-(h(K));
and if g is contratomic, then h(g)g-(g(h(g))),
whichleadstogC
h-(h(g))h-(g-(g(h(g))))
sincehissurjective. Putting
f
goh,wegetK h-(g-(g(h(K)))) f-(f(K))
in eitherofthetwoconsidered cases The proofis thencomplete.
REFERENCES
ANDERSON,
R.D.,
Atomicdecompositions of continua,DukeMath.J.
24(1956),507-514[2] BING,
R.H. andJONES, FB.,
Anotherhomogeneous planecontinuum, Trans. Amer.Math. Soc.90(1959),171-192.
[3] EMERYK, A.andHORBANOWICZ,
Z.,
Onatomicmappings,Colloq.Math. 27(1973),49-55.[4] GRACE, E.E. and VOUGHT,
E.J.,
Four mapping problems of Madkowiak, Colloq. Math. 69 (1995),133-141.[5] HOSOKAWA,
H.,
Some remarks on the atomic mappings, Bull. Tokyo Gakugei Univ. (4), 40 (1988),31-37.[6] MAtKOWIAK, T.,
Continuousmappingsof continua,DissertationesMath. (RozprawyMat.)158 (1979),1-91.[7] MAKOWIAK, T.,
Singulararc-likecontinua,DissertattonesMath. (RozprawyMat.)
257(1986),
1-35.