The paper contains two theorems relating the fine structure of differentiable functions, in one or more dimensions, to the behavior of Fourier-Stieltjes transforms on sets that are small in various ways. In this paper we prove two theorems on the transformation of certain sets, defined as follows. A set E in a metric space is an L-set if there are sequences Ek?-)>O and 6k?_0, and for each k a decomposition E= UI Ei, wherein diam(Ei) k (isi'). For each compact L-set E of real numbers there is a function h of class C'(oo, oo) with h'>O, so that h(E) is a Kronecker set ([2], [3]). The first theorem is a complement to this. THEOREM I. Let co be a monotone, positive function on (0, oc), and c(O +) = O; let C' be the set offunctions q in C1 with p'> O, I p'(a)-p'(b) I _ w)(la-bl)for all real a and b. Then there is a compact L-set E so that g(E) is an MO-set for each q9 in C ,. To prove the theorem we choose a sequence of positive numbers (cj) so that co= 1, wo(cn)<n-2, and c,,+<n -3c0. We now construct finite sets Fn and En; the peculiar construction of Fn is the main point in the argument. Fn is a sequence of n2 elements x(m) = x(O) + MCn + m2cnn-5!2, 1< m _ n2. Here x(O)= -cn-cnn-512 so that x(1)=O. En is then a union of translates of Fn, say Ui (En+aj). Then ao=O, while aj+i-aj=n2c,+n-1/2cn. In different terms, the final term in each translate becomes x(O) in its successor to the right. The number of translates is to be [c,,1lc1n -1316] for n_1. (a) In Fn we have the inequalities Cn < x(m + 1) x(m) < (n2 +? n312)cn < 2n2cn. Thus En has diameter <2n-1/6cn-l. The vector sum E= n En iS then an L-set. (It is somewhat easier to verify that, for large enough r, the subset Received by the editors January 15, 1971. AMS 1969 subject classifications. Primary 4272, 4615.
Read full abstract