Download as pdf
Download as pdf
You are on page 1of 10
STUDIA MATHEMATICA, T, LXII. (1978) icm A wuultilinear interpolation theorem by MISHA ZAFRAN (Stanford, Calif.) Abstract, We prove a general multilinear interpolation theorem for the “espaces de moyenne” of Lions and Peetre. This result is used to obtain a multilinear Marein- kiewiez interpolation theorem. Let (B', B') be an interpolation pair of Banach spaces, and denote by (B°,B'),, (or more simply B,,), 0<8<1,10 108 M. Zafran where, as usual, we identify functions agreeing almost everywhere. If X = complex numbers, we write Li(X) = Ly. If (B°, B’) is an interpolation pair, we define the real interpolation space (B°, B), , (or B,,,) to be the collection of all elements «cB? +B* for which there exists a strongly measurable function w with values in BoB" such that : Po. dt (a) 2 { uy and (2) u(t) eT2(B) and #-*u(t) e L2(B). It follows by condition (2) that w0. kat Then UL (es «+4 My) ho0,0hh,, < Ma ME a] Plat n for all (4) ..+5%q) € @ BYaBL. In particular, if pp < co, 1 ap nttee rt and 1 3 Since w, is B2nBi-continuous, it follows by (2) that the integral defining of converges in the norm of B2n Bi, 1 98 6-0, TREE} (5) p— Bilin, 20 a8 60. Moreover, by (4) we have : WH cypaat 0 a8 a, PO. By an argument similar to that leading to inequality (14) (see, in particular (13), where now the ith term is treated separately from the other terms in the product), and by (15), we see that for 0 < e <1and 0 0 such that lays, (ny, a8, «5 ABT? ay) as a-+0, the convergence in C,,, Note that (19) is valid for any fixed, bub arbitrarily chosen &, with 0< a, <1,2 2 in (18), fix 0<5,<1 for 2< : ben, k #4, and fix 00 in (18), we see by (21) that (22) WD (ey, 22,5 --+, Map, «5 21, Malo, 0 was arbitrary, and by recalling that B, », = (Bi, Bis py Comment. The preceding proof could be simplified if we knew the validity of the following result: The elements « (see equation (3) in the proof of 2.4) satisfy leh,» = EDEL "Olney Hepa the infimum taken over all B27Bj-valued functions » with compact sup- port in (0, co), which are continuous on their support, and which satisty oat i . : the additional two properties #f = [ o(f)—- and tty 12*(Bi),j =0,1. a (Essentilly this requires the “smooth elements” a to attain their norm over “smooth functions”.) We do not know whether this, in fact, occurs, and anyway, this result does not seem to be of great intrinsic interest. ‘We remark that Theorem 2.4 plays a crucial role in the construction of multiplier transformations which are of weak type (p,p), but which are not bounded on L,. Here 1

0, depending only om the pe, qy%, and 2, such that for all (0, «+-5 9) € OBOE Lue have UD, 2 «+s kor chiyg < EME MY Tl Wllee ob 16 M. Zafran Proof, Let 1/r= Y1/p,—n-+1. The result follows by Theorem < 2.4, since for r ij pp Mas Spe till—n. Then if Te (My, My)q,, there exists 4 constant K > 0, depending only on the p,,4,1,n, and s, such that le% pl)? Le (Ll) (D(a, .., Bq) (09,0, 4S Ej, (Moti en [ lel ket . for all (a, ... 7 i ij ‘ ( z 1%) © BBE In particular, if py < co for all k, then T as a unique extension (again denoted by T) so that (1) remains valid for AW (0g, «+5 Bq) © @ (BE, Bide yp kewl is 4 PPOOL Let Lia, ty ++ ty) 2) = Do, «+ 04) for all (tr, ..., 4, 2) ane say 0, depending only on gq and K4, so that (4) NPs +4 Le log oe) < Ky My MY fa Utmlly, ao) for all H, ¢ 3, with finite, positive, jz,-measure. ‘We show that (4) yields the general result by an argument which is well known in the case of linear operators (see [3], Theorem 7). Let f, ¢ 7, ¥ with f, > 0 on 0,. We write f, = Y'dnz,,, where A,,€2, has finite, non- rit x zeT0 pymeasure, a, >0,10 on Q,. By the argument leadi ) : > . ing to (8 now using (5) in place of (4), we obtain » (8) WL (Fas Fas tgp +9 Xa lng ted = SE MMM, omy lely, 0 [ | liteylp, aps ies ee B,€ 2 with 0 <4,(B,) < 0,30,11, H? is isometric and isomorphic to {f¢L,| f(n) 0 for % < 0}. The basic facts concerning these spaces‘may be found in Chapter 7 of [13]. Norarion 3.1. Let (2,2, ) denote a c-tinite measure space, and let denote the measurable functions on 2. Denote by # the class of all complex polynomials, restricted to the unit dise. For all real numbers a> 0, we let [a] denote the greatest integer less than or equal to a. ‘THEOREM 3.2. Assume the notations of 3.1. Let 0 0, there exists an operator M; acting continuously on L°(T), given by M, f(a) = Sa(3) ane", where a, is the nth Fourier coefficient of f. Assume that the operator norms Mj are uniformly bounded. Then m defines a bounded multiplier oper- ator M on L?(R) ([3], p. 264). We wish to generalize this result by replacing the torus, which may be identified with SO(2), with the non-abelian group SO(3), or with its universal covering group SU(2), which is naturally identifiable with the unit sphere in two-dimensional complex space. By a limiting process analagous to the classical passage from the circle to the line, the group 80(3) can be shown to tend to a non-compact non-abelian group: the group of rigid motions of the Euclidean plane, denoted by M{(2). In this paper, we shall show how Fourier analysis on M(2) is closely * The author wishes to thank Professor Guido Weiss and Professor Ronald Coifman for their many helpful discussions concerning: this work.

You might also like