ArticlePDF Available

Some Results Connected with a Problem of Erdos. II

Authors:

Abstract

It is shown, using the continuum hypothesis, that if $E$ is an uncountable subset of the real line, then there exist subsets $S_1$ and $S_2$ of the unit interval, such that $S_1$ has outer Lebesgue measure one and $S_2$ is of the second Baire category and such that neither $S_1$ nor $S_2$ contains a subset similar (in the sense of elementary geometry) to $E$. These results are related to a conjecture of P. Erdős.
PROCEEDINGS OF THE
AMERICAN MATHEMATICAL SOCIETY
Volume 75, Number 2, July 1979
SOME RESULTS CONNECTED WITH
A PROBLEM OF ERDÖS. Ü
HARRY I. MILLER1
Abstract. It is shown, using the continuum hypothesis, that if E is an
uncountable subset of the real line, then there exist subsets 5, and S2 of the
unit interval, such that S¡ has outer Lebesgue measure one and 52 is of the
second Baire category and such that neither Sx nor S2 contains a subset
similar (in the sense of elementary geometry) to E. These results are related
to a conjecture of P. Erdôs.
1. Introduction. P. Erdôs [2] presented the following conjecture at the
problem session of the Fifth Balkan Mathematical Congress (Belgrade, June
24-30, 1974):
Conjecture. Let E be an infinite set of real numbers. Then there exists a set
of real numbers S of positive Lebesgue measure which does not contain a set
£" similar (in the sense of elementary geometry) to E.
If E is a finite set of real numbers, then every set of real numbers S of
positive Lebesgue measure contains a subset E' similar to E. This follows
from a result of M. S. Ruziewicz [5] or as P. Xenikakis has shown (in a private
communication) from Theorem 3 in [3]. By Theorem 4 in [3], the corre-
sponding result holds for Baire sets S of the second Baire category (S is a
Baire set if it can be written in the form S = (G \ C) u D, where G is an
open set and C and D are sets of the first Baire category).
H. I. Miller and P. Xenikakis [4] have proven the following two theorems
related to the conjecture of Erdôs.
Theorem A. If A ci (the real line) possesses the Baire property and is of
the second Baire category in R and if (£„),?_ x is a convergent sequence of reals,
then A contains a set A' which is similar to the set [zn; n = 1, 2, . . . }.
Theorem B. If A c R can be written in the form A = (G \ C) U D, where
G is a nonempty open set and C and D are sets of Lebesgue measure zero and if
(z„)n°=x is a convergent sequence of reals, then A contains a set A' which is
similar to the set {z„; n = 1, 2, . . . }.
The purpose of this work is to show, using transfinite induction and the
continuum hypothesis, that if E is an uncountable subset of the real line, then
Received by the editors May 1, 1978.
AMS (A/OS) subject classifications (1970). Primary 28A05, 26A21.
'The work on this paper was supported by the Scientific Research Fund of Bosna and
Herzegovina.
© 1979 American Mathematical Society
0002-9939/79/0000-031 S/S02.00
265
License or copyright restrictions may apply to redistribution; see http://www.ams.org/journal-terms-of-use
266 H. I. MILLER
there exist subsets S, and 52 of the interval [0, 1] such that m*(Sx) = 1 and S2
is of the second Baire category in R, and such that neither 5, nor S2 contains
a subset similar to E. Here m* denotes outer Lebesgue measure.
2. Results. Our first theorem makes use of the following lemma whose
proof can be found in [1].
Lemma A. Let C be a closed subset of the real line. Let B be a subset of C
such that B has nonempty intersection with every closed subset of C of positive
Lebesgue measure. Then m*(B) = m(C).
We now proceed to prove our first theorem.
Theorem 1. If E is an uncountable set of real numbers, then there exists a
subset S of [0, 1] such that m*(S) = 1 and no subset of S is similar to E.
Proof. Let 2 - [Q; Q a closed subset of [0, 1] and m(Q) > 0}. By the
continuum hypothesis 2 can be written in the form 2 = {o„;a<fi} where
ß denotes the first uncountable ordinal. Similarly S, the family of all subsets
of R similar to E, can be written in the form $ = [Ea; a < ß}. This is true
as there are c, the cardinality of the continuum, similarity transformations;
since each similarity transformation / is of the form/(x) = ax + b. We now
proceed to construct two transfinite sequences {*a}a<a and {^a}a<ß of real
numbers.
Pick je, G Qx andyx G Ex such that xx ¥=yx- Suppose that w is an ordinal
number u < ß and that {*a}a<w and {ya)a<a have been selected such that:
(a) xa E Qa,ya E Ea for each a < w, and
(b) [xa; a< ß)n{ya;a< ß) =0, for every ß, ß < co.
Then we can find xu E Qa and yu G Ea such that {xa; a < w} n [ya;
a < w} is the empty set, since Qu and Ea are uncountable sets and u < ß
implies that w is a countable ordinal. Therefore by transfinite sequences
{xa}a<a aQd 0„L<a such to-t xa E Qa, ya E Ea for each a,a<Ü, and
such that {xa; a < ß) n [ya; a < ß) =0 for every ß, ß < ß. Let S denote
the set [xa; a < Q). Then S c [0, 1] and by Lemma A we have m*(S) = 1.
Furthermore, no subset of S is similar to E. For if some subset, say S',of S is
similar to E we have S' = Ey for some y < fi. This in turn implies yy G S'
and hence yy G S. Therefore there exists 5 < ß such that yy = xs or [xa;
a < ß) n [ya; a < ß] ¥=0, where ß = max(y, 5), which is a contradiction.
We need the following lemma in the proof of Theorem 2.
Lemma B. Let 2 denote the collection of subsets of[0, 1] given by the formula
2 = {(/? \ U" i F¡) n [0, 1] where each F¡ is a closed and nowhere dense
subset of R). If B is a subset of [0, 1] and has the property that Q n B =£0 for
every g G 2, then B is a set of the second Baire category in R.
Proof. If B is a set of the first Baire category, then we have B = U"i X¡,
with each X¡ nowhere dense in R. This implies that fie U"i Cl( X¡), where
License or copyright restrictions may apply to redistribution; see http://www.ams.org/journal-terms-of-use
SOME RESULTS CONNECTED WITH A PROBLEM OF ERDÖS. II 267
Cl denotes the closure operator. From this it follows that fln(Ä\
U," i C1(A",)) =0, contradicting the assumption that Q n B ¥=0 for every
0<ES.
Theorem 2. If E is an uncountable set of real numbers, then there exists a
subset S of [0, 1] with the property that no subset of S is similar to E and such
that S is of the second Baire category in R.
Proof. The proof of Theorem 2 is essentially the same as that of Theorem
1. Here we make use of Lemma B and the fact that the family of sets 2. given
in Lemma B can, by assuming the continuum hypothesis, be written in the
form 2 = {g,;a<Q}, where as before, ß denotes the first uncountable
ordinal.
3. Remark. Professor John C. Oxtoby has observed that Theorems 1 and 2
can be obtained by applying a theorem of Sierpiñski [6] which reads:
Theorem C. Given a set X, a family $ of subsets of X, and a group G of 1-1
mappings of X onto itself, such that card X = card Í» = card G = 3C, and such
that X \ U¡fi(H¡) is uncountable for each sequence {/,} c G and {//,} c $,
then there exists an uncountable set S C X with the properties:
H E $ implies H n S is countable, and
f E G implies f(S) \ S is countable.
We will sketch Oxtoby's proof. In the following, let £ = {L; L an uncount-
able subset of R that contains no uncountable meager subset} and S = (S;
S an uncountable subset of R that contains no uncountable subset of measure
zero}. Furthermore let X = [0, 1], ¥, denote all Gs nullsets contained in X,
^2 denote all F0 meager subsets of X, Gx denote all 1-1 Borel measurable and
nullset-preserving transformations of X, and G2 denote all 1-1 Borel
measurable and category-preserving transformations of X. Let S denote all
sets similar to a given fixed uncountable subset of R. Define
$, = *, U [E E & ; E c X and E E S }
and
$2 = *2 u [E E & ; E c X and E E £}.
It is easy to verify, by assuming the continuum hypothesis, that the hypothe-
ses of Sierpiñski's theorem are satisfied when we take G = Gx and $ = $,.
The resulting set Sx c [0, 1] has the following properties:
(i) Sx G S and hence is nonmeasurable and of the first category on every
perfect set;
(ii) Sx contains no member of & ;
(iii)/ G Gx implies f(Sx)ASx is countable, and
(iv) m*(Sx) = 1.
Similarly, taking G = G2 and $ = i»2 and by assuming the continuum
hypothesis we obtain a set S2 c [0, 1] with the following properties:
License or copyright restrictions may apply to redistribution; see http://www.ams.org/journal-terms-of-use
268 H. I. MILLER
(i) S2 G £ and hence is of measure zero and does not possess the Baire
property;
(ii) S2 contains no member of S ;
(iii)/ G G2 implies f(S2)AS2 is countable, and
(iv) S2 is of the second category at each point of [0, 1].
References
1. A. Abian, Partition of nondenumerable closed sets of reals, Czechoslovak Math. J. 26(101)
(1976), no. 2, 207-210. MR 53 #5820.
2. P. Erdôs, Problems, Math. Balkánica (Papers presented at The Fifth Balkan Mathematical
Congress), 4 (1974), 203-204.
3. H. I. Miller, Relationships between various gauges of the size of sets of real numbers. II, Akad.
Nauka i Umjet. Bosne i Hercegov. Rad. Odjelj. Prirod. Mat. Nauka 59 (1976), 37-48.
4. H. I. Miller and P. I. Xenikakis, Some results connected with a problem of Erdôs. I, Akad.
Nauka i Umjet. Bosne i Hercegov. Rad. Odjelj. Prirod.' Mat. Nauka (to appear).
5. M. S. Ruziewicz, Contribution à l'étude des ensembles de distances de points, Fund. Math. 7
(1925), 141-143.
6. W. Sierpiñski, Un théorème de la théorie générale des ensembles et ses applications, C. R. Soc.
Sei. Varsovie 28 (1936), 131-135. Zbl. 15, 103.
Department of Mathematics, University of Sarajevo, Sarajevo 71000, Yugoslavia
License or copyright restrictions may apply to redistribution; see http://www.ams.org/journal-terms-of-use
... How big must a set in R" be in order to contain a subset similar (in the sense of elementary geometry) to a given finite or countable set? Some results concerning this question have been given in [6,10,111, in the case of the real line R. ...
... However, we can prove, using the same methods as in Theorems 3 and 4, that if JV is a a-ideal of subsets of l%" closed under translations with R" $ N, then the last two theorems hold for sets of the form (G\P) u Q, where G is open and P, Q E N. For example, the set of all nullsets of R" is such a a-ideal. So Theorems 3 and 4 in [6] are also true in R". ...
... If (b,: n E 6~) is similar to (c,: n E CO) and meets Y in an infinite set, then it meets some Y, in an infinite set. This follows from the lemma and the continuity of the Next we improve on a theorem of H. Miller [33]. Erdijs conjectured that for every convergent sequence there is a set of reals of positive measure which contains no subset similar to the sequence. ...
... H. Miller and Xenikakis [32] have proved that the set of reals cannot have full measure in any interval. It is also easy to see that for every finite set of reals E and positive measure set X, X contains a set similar to E (see [33]). It is impossible to partition the reals into two (or even finitely many) sets neither of which contains a set similar to a given finite set E. This is the one-dimensional case of Gallai's Theorem [13, p. 381. ...
Article
The topic of this paper is Borel versions of infinite combinatorial theorems. For example it is shown that there cannot be a Borel subset of [ω]ω which is a maximal independent family. A Borel version of the delta systems lemma is proved. We prove a parameterized version of the Galvin-Prikry Theorem. We show that it is consistent that any ω2 cover of reals by Borel sets has an ω1 subcover. We show that if V \= L, then there are π11 Hamel bases, maximal almost disjoint families, and maximal independent families.
... H. I. Miller and P. Xenikakis [4] have proven the following two theorems related to the conjecture of Erdôs. ...
Article
Full-text available
It is shown, using the continuum hypothesis, that if E is an uncountable subset of the real line, then there exist subsets S1 and S2 of the unit interval, such that S1 has outer Lebesgue measure one and S2 is of the second Baire category and such that neither S1 nor S2 contains a subset similar (in the sense of elementary geometry) to E. These results are related to a conjecture of P. Erdôs.
... Then there exists a set of real numbers S of positive Lebesgue measure which does not contain a set E' similar to£. H. I. Miller [3] proved, using the continuum hypothesis, that if E is an uncountable subset of the real line, then there exist subsets Sx and S2 of the unit interval, such that Sx has outer Lebesgue measure one and S2 is of the second Baire category and such that neither S, nor S2 contains a subset similar to E. Here we prove without the continuum hypothesis the following theorem. ...
Article
It is shown that if E E is a subset with more than two points of the real line, then there exists a subset S S of the unit interval, such that S S has outer Lebesgue measure one and S S is of the second Baire category and such that S S does not contain a subset similar (in the sense of elementary geometry) to E E . This result is related to a conjecture of P. Erdös.
Article
It is shown that if E is a subset with more than two points of the real line, then there exists a subset S of the unit interval, such that S has outer Lebesgue measure one and S is of the second Baire category and such that S does not contain a subset similar (in the sense of elementary geometry) to E. This result is related to a conjecture of P. Erdos.
Article
Full-text available
Erdős asked if for every infinite set, A, of real numbers there exists a measurable subset of the reals having positive measure that does not contain a subset similar to A. In this note we transform this question to a finite combinatorial problem. Using this translation we extend some results of Eigen and Falconer concerning slow sequences for which the answer to Erdős’ question is positive.
Article
Erdős posed the following problem. "Let $E$ be an infinite set of real numbers. Prove that there is a set of real numbers $S$ of positive measure which does not contain a set $E^\prime$ similar (in the sense of elementary geometry) to $E$.'' The proof is known for only a few special cases; and not included among these is the geometric sequence $\{2^{-n}\}_{n=1}^\infty$. In this paper we examine the known literature, present some new results, and ask a few related questions.
Article
Erdös has conjectured that given a decreasing sequence of real numbers convergent to 0 there always exists a measurable set of positive measure that contains no similar copy of the sequence. We prove this conjecture if the sequence does not converge too rapidly.
Article
In this paper, assuming a certain set-theoretic hypothesis, a positive answer is given to a question of H. Kraljevi, namely it is shown that there exists a Lebesgue measurable subsetA of the real line such that the set {c R: A + cA contains an interval} is nonmeasurable. Here the setA + cA = {a + ca: a, a A}. Two other results about sets of the formA + cA are presented.
Article
In un lavoro classico Steinhaus ha dimostrato undici teoremi interessanti nei quali si considerano insiemi distanziali di sottoinsiemi della linea reale,aventi misura di Lebesgue positiva. Majumder ha dimostrato che sono validi teoremi analoghi a quelli di Steinhaus per insiemi del tipoR(E)={x/y: x,y∈E} doveE∪R; 0∉E. In questo lavoro si considerano funzioni generali e si investiga sui quali degli undici teoremi di Steinhaus possono essere generalizzati. Nell'articolo, inoltre, usando una funzione generalef, vengono dati risultati analoghi a quelli di Steinhaus per insiemi di Baire.
Un théorème de la théorie générale des ensembles et ses applications
  • W Sierpiñski
W. Sierpiñski, Un théorème de la théorie générale des ensembles et ses applications, C. R. Soc.
Yugoslavia License or copyright restrictions may apply to redistribution
  • Sei
Sei. Varsovie 28 (1936), 131-135. Zbl. 15, 103. Department of Mathematics, University of Sarajevo, Sarajevo 71000, Yugoslavia License or copyright restrictions may apply to redistribution; see http://www.ams.org/journal-terms-of-use
Balkánica (Papers presented at The Fifth Balkan Mathematical Congress
  • P Erdôs
  • Problems
  • Math
P. Erdôs, Problems, Math. Balkánica (Papers presented at The Fifth Balkan Mathematical Congress), 4 (1974), 203-204.
  • Umjet Nauka I
Nauka i Umjet. Bosne i Hercegov. Rad. Odjelj. Prirod. Mat. Nauka 59 (1976), 37-48.
Contribution à l'étude des ensembles de distances de points
  • M S Ruziewicz
M. S. Ruziewicz, Contribution à l'étude des ensembles de distances de points, Fund. Math. 7 (1925), 141-143.