Krzysztof Ciesielski Department of Mathematics, West Virginia University, , Morgantown, WV 26506-6310, USA. e-mail: KCies@wvnvms.wvnet.

edu Saharon Shelah Institute of Mathematics, the Hebrew University of , Jerusalem, 91904 Jerusalem, Israel, and Department of Mathematics, Rutgers University, New Brunswick, NJ 08854, USA

UNIFORMLY ANTISYMMETRIC FUNCTIONS WITH BOUNDED RANGE

Abstract The goal of this note is to construct a uniformly antisymmetric function f : R R with a bounded countable range. This answers Problem 1(b) of Ciesielski and Larson [6]. (See also the list of problems in Thomson [9] and Problem 2(b) from Ciesielskis survey [5].) A problem of existence of uniformly antisymmetric function f : R R with nite range remains open.

modified:1999-09-02

A function f : R R is said to be uniformly antisymmetric [6] (or nowhere weakly symmetrically continuous [9]) provided for every x R the limit limn (f (x + sn ) f (x sn )) equals 0 for no sequence {sn }n< converging to 0. Uniformly antisymmetric functions have been studied by Kostyrko [7], Ciesielski and Larson [6], Komjth and Shelah [8], and Ciesielski [1, 2]. (A a connection of some of these results to the paradoxical decompositions of the Euclidean space Rn is described in Ciesielski [3].) In particular in [6] the authors constructed a uniformly antisymmetric function f : R N and noticed that the existence of a uniformly antisymmetric function cannot be proved without an essential use of the axiom of choice.
Key Words: uniformly antisymmetric function, Hamel basis Mathematical Reviews subject classication: 26A15, 26A21 Received by the editors May 3, 1998 The rst author was partially supported by NSF Cooperative Research Grant INT9600548, with its Polish part being nanced by Polish Academy of Science PAN, and 1996/97 West Virginia University Senate Research Grant. Papers authored or co-authored by a Contributing Editor are managed by a Managing Editor or one of the other Contributing Editors. The authors like to thank the referee for many valuable comments and suggestions. This work was supported in part by a grant from Basic Research Foundation of the Israel Academy of Sciences and Humanities. Publication 680.

revision:1999-09-02

615

680

616

Krzysztof Ciesielski and Saharon Shelah

The terminology and notation used in this note is standard and follows [4]. In particular for a set X we will write |X| for its cardinality and P(X) for its power set. Also 2 will stand for the set of all functions from = {0, 1, 2, . . .} into 2 = {0, 1}. We consider 2 as ordered lexicographically. Theorem 1. There exists a function f : R R with countable bounded range such that for every x R there exists an x > 0 with the property that the set Sx = {s R : |f (x s) f (x + s)| < x } is nite. In particular f is uniformly antisymmetric. Proof. First notice that it is enough to nd a compact zerodimensional metric space T, d and a function g from R into a countable subset T0 of T such that for every x R there is a x > 0 for which the set Sx = {s R : d(g(x s), g(x + s)) < x } is nite. To see this assume that such a function g : R T exists and take a homeomorphic embedding h of T into R. We claim that f = h g : R R is as desired. Indeed, f [R] = h[g[R]] is countable, as it is a subset of a countable set h[T0 ], and it is bounded, since it is a subset of a compact set h[T ]. So take x R and x > 0 for which Sx is nite. Since h1 : h[T ] T is uniformly continuous, we can nd an x > 0 such that |y1 y2 | < x implies d(h1 (y1 ), h1 (y1 )) < x
modified:1999-09-02

for every y1 , y2 h[T ]. But for such a choice of x we have Sx = {s R : |h(g(x s)) h(g(x + s))| < x } Sx proving that Sx is nite. Thus, we proceed to construct a function g described above. The value of g(x) will be dened with help of a representation of x in a Hamel basis; i.e., a linear basis of R over Q. For this we will use the following notation. Let {y : 2 } be a one-to-one enumeration of a Hamel basis H. For every x R let 2 qx, y , with qx, Q for 2 , be the unique representation of x in basis H and let wx = { 2 : qx, = 0}. Thus wx is nite and x= qx, y .
wx

revision:1999-09-02

The denition of the space T is considerably more technical since it reects several dierent cases of the proof that the sets Sx are indeed nite. To this

680

Uniformly antisymmetric functions with bounded range

617

end let {qj : j < } be a one-to-one enumeration of Q with q0 = 0. For i < let Pi = P({qj : j < i}) and put Pi = P(2i {0, 1} Pi Pi ). Note that each Pi is nite, so T = i< Pi , considered as the standard product of discrete spaces, is compact zerodimensional. We equip T with a distance function d dened between dierent s, t T by d(s, t) = 2 min{i< : s(i)=t(i)} and let T0 = {t T : (n < )(i n) t(i) = } . Clearly T0 is countable. Now we are ready to dene g : R T0 T . For this, however, we will need few more denitions. For x R, q Q, i < , and 2i such that {( i) : wx } we dene: p(i) {0, 1} as the parity of i, i.e., p(i) = i mod 2; ki (q) = {qj Q : qj < q & j < i} Pi ; (x, ) to be the minimum of { wx : } (in the lexicographical order); (x, ) to be the minimum of { wx : } \ {(x, )} provided |{ wx : }| = 1; otherwise we put (x, ) = (x, ); nx < to be the smallest number n > 0 such that (i) n= n for any dierent , wx , and

(ii) qx, {qj : j < n} for every wx .

modified:1999-09-02

Consider the function g : R T0 dened as follows. For every x R and i < we dene g(x)(i) Pi as , p(|{ wx : }|), ki (qx,(x,) ), ki (qx,(x,) ) : {( i) : wx }

provided i nx and we put g(x)(i) = for nx < i < . In the argument below the key role will be played by the function ki in general, and the coordinate ki (qx,(x,) ) in particular. The key step in the proof that g has the desired property is that for every x R and s = 0 if nx max{nxs , nx+s } then g(x s)(nx ) = g(x + s)(nx ). (1)

revision:1999-09-02

To see (1) assume that nx nx+s . If nxs < nx then g(x s)(nx ) = = g(x + s)(nx ), where g(x + s)(nx ) = since wx+s = as nxs < nx nx+s implies x + s = 0. Thus, we can assume that nx min{nxs , nx+s }. Take an

680

618

Krzysztof Ciesielski and Saharon Shelah

wxs wx+s such that qxs, = qx+s, and let = nx . Note that, by the denition of nx , the set S = { wx : } has at most one element. If S = then { wxs : } = { wx+s : } = and so (xs, ) = (x+s, ) wx while qxs,(xs,) +qx+s,(x+s,) = 0. Thus q0 = / 0 separates qxs,(xs,) and qx+s,(x+s,) implying that knx (qxs,(xs,) ) = knx (qx+s,(x+s,) ). Therefore g(x s)(nx ) = g(x + s)(nx ). So, assume that S = and let be the only element of S. Then wxs wx+s . If belongs to precisely one of the sets wx+s and wxs , say wx+s , then { wx+s : } = { wxs : } { }. In particular, p(|{ wx+s : }|) = p(|{ wxs : }|) implying that g(xs)(nx ) = g(x + s)(nx ). So, we can assume that wxs wx+s . Then { wxs : } = { wx+s : } and (x s, ) = (x + s, ). We will consider three cases. Case 1: = (x s, ) = (x + s, ). Then qxs,(xs,) + qx+s,(x+s,) = 0, so q0 = 0 separates qxs,(xs,) and qx+s,(x+s,) . Thus knx (qxs,(xs,) ) = knx (qx+s,(x+s,) ) and g(x s)(nx ) = g(x + s)(nx ). Case 2: = (x s, ) = (x + s, ) and qxs,(xs,) = qx+s,(x+s,) . Then qxs,(xs,) + qx+s,(x+s,) = 2qx, and, by the denition of nx , qx, {qj : j < nx }. Since qx, separates qxs,(xs,) and qx+s,(x+s,) we conclude that knx (qxs,(xs,) ) = knx (qx+s,(x+s,) ) and g(x s)(nx ) = g(x + s)(nx ). Case 3: = (x s, ) = (x + s, ) and qxs,(xs,) = qx+s,(x+s,) . Then Z = { wxs : }\{(xs, )} = { wx+s : }\{(x+s, )} is non-empty, since it contains , and so (x s, ) = (x + s, ) wx . / Therefore, as in Case 1, qxs,(xs,) + qx+s,(x+s,) = 0, so q0 = 0 separates qxs,(xs,) and qx+s,(x+s,) . Thus knx (qxs,(xs,) ) = knx (qx+s,(x+s,) ) and g(x s)(nx ) = g(x + s)(nx ). This nishes the proof of (1). Next, for every x R put x = 2nx . To nish the proof of the theorem it is enough to show that every Sx dened for such a choice of x is a subset of a nite set Zx = {s R : wx+s wx & nx+s < nx } =
wx

modified:1999-09-02

p y : p {qj : j < nx } .

680

revision:1999-09-02

Indeed, take an s Sx . Then, by (1) and the denition of the distance function d, we have max{nxs , nx+s } < nx . Notice also that if nxs = nx+s , say nxs < nx+s , then g(x s)(nx+s ) = = g(x + s)(nx+s ) implying that d(g(x + s), g(x s)) 2nx+s > 2nx = x , which contradicts s Sx . So, we have nxs = nx+s . To prove that s Zx it is enough to show that

Uniformly antisymmetric functions with bounded range

619

wx+s wx . But if it is not the case then there exists an wx+s \ wx . Moreover, qx+s, = qxs, = 0 and = (x + s, ) = (x s, ), where = nx+s . In particular, q0 = 0 separates qx+s,(x+s,) and qxs,(xs,) . Therefore knx+s (qxs,(xs,) ) = knx+s (qx+s,(x+s,) ) and g(x s)(nx+s ) = g(x + s)(nx+s ). So d(g(x + s), g(x s)) 2nx+s > 2nx = x again contra dicting s Sx . Thus, wx+s wx and s Zx .

References
[1] K. Ciesielski, On range of uniformly antisymmetric functions, Real Analysis Exch. 19(2) (199394), 616619. [2] K. Ciesielski, Uniformly antisymmetric functions and K5 , Real Analysis Exch. 21(2) (199596), 147153. (Preprint available.1 ) [3] K. Ciesielski, Sum and dierence free partitions of vector spaces, Colloq. Math. 71 (1996), 263-271. (Preprint available.) [4] K. Ciesielski, Set Theory for the Working Mathematician, London Math. Soc. Student Texts 39, Cambridge Univ. Press 1997. [5] K. Ciesielski, Set theoretic real analysis, J. Appl. Anal. 3(2) (1997), 143 190. (Preprint available.) [6] K. Ciesielski, L. Larson, Uniformly antisymmetric functions, Real Analysis Exch. 19(1) (199394), 226235.
modified:1999-09-02

[7] P. Kostyrko, There is no strongly locally antisymmetric set, Real Analysis Exch. 17 (199192), 423425. [8] P. Komjth, S. Shelah, On uniformly antisymmetric functions, Real Anala ysis Exch. 19(1) (199394), 218225. [9] Brian Thomson, Symmetric Properties of Real Functions, Marcel Dekker, 1994.

revision:1999-09-02

1 Preprints marked by are available in electronic form from Set Theoretic Analysis Web Page: http://www.math.wvu.edu/homepages/kcies/STA/STA.html

680