We give sufficient conditions (in terms of differentiability and growth properties) for a radial function to be an L!(R2) Fourier multiplier. These conditions are in the nature of best possible.The well-known multiplier theorem of Hormander (ref. 1; see also p. 96 in ref.2) may be stated as follows. Let 4 be a fixed nonnegative smooth bump function supported in [1,2], and let D,, a > 0, denote the fractional derivative defined initially on f(R") (the space of rapidly decreasing C' functions) by (Dng)A(4) = 11agA(e) and then extended to L1(R').For n = 1 we write D' = Di . Suppose m is a bounded function on R' that satisfies SU t2a-n faDn' [4(u) m(1)1 2 = A2 < 00 [1] for some real number a > n/2. Then m is a Fourier multiplier of LP(R") [ Condition 1 for cr > n/2 actually implies that m is bounded, while for a s n/2, the space of functions satisfying 1 contains unbounded functions: in this sense, Hormander's result is sharp; but when additional assumptions about m and p are made, the differentiability conditions required of m may be relaxed (see, for example, refs. 3 and 4).On the other hand, let us consider the Bochner-Riesz multipliers ms(f) = (1 -If12)5 for 8 > 0. Carleson and Sjolin (5) [see also the work of Fefferman (6) and the contributions of C6rdoba (7) 10 equals R(q, a) with equivalent norms when a > -, 1 < q q <0. The case q = 2 and Sobolev-type embeddings give the case q < 2. Interpolation of the case q = 2 with M2(R2) = L'(R2) gives the case q > 2. These matters will be discussed in a forthcoming paper. It should be noted that while the multiplier theorems are most naturally proved in the R(q, a) context, the WBVq,a conditions are easier to verify.(ii) Theorem 2 is best possible in the sense that the "criti- n -i1 await a better understanding of the multipliers m6 for n . 3. However, Dappa (12) has obtained the optimal result in the case q = 1, a > (3n + 1)/(2n + 2) by using the Stein-Tomas restriction theorem (13).(iv) Gasper and Trebels proved the Hankel multiplier version of Theorem 2 in ref. 9 and gave necessary conditions for Hankel multipliers (and hence for radial Fourier multipliers) in terms of WBV-spaces in ref. 14.The proof of Theorem 1 uses square functions in the spirit of the Igari-Kuratsubo multiplier theorem (3) and Stein's proof of Hormander's theorem. Prior to the proof, let us make the following definitions in which all Fourier transforms are in R2, with functions on (0, 00) extended to be radial where appropriate.
3254The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. §1734 solely to indicate this fact.