Internat. J. Math. & Math. Scl. VOL. 15 NO. 3 (1992) 441-448 441 A MAJORANT PROBLEM RONEN PERETZ Department of Mathematics University of Michigan Ann Arbor, MI 48109 (Received September 25, 1990) Let ABSTRACT. f(z) k=O e Inflnlte complex vector k induces a 0,1,2 akzk a o be analytlc in the unlt disc. 0 (eo,et,e2 ) function re(Z) such that k=O akekZk lekl Any 1, whlch is still analytic In the unit disc. In this paper we study the problem of maximizing the p-means: e over all possible vectors and for values of close to r 0 p<2. It is proved that a maxlmlzlng function Is and that r -laoi f,{z} + and for all . laklZ k=l k could be taken to be any positive number which Is smaller than the radius of the largest disc centered at the orlgln which can be Inscrlbed in the zero sets of f Thls problem is originated by a well known maJorant problem for Fourier coefficients that was studied by Hardy and Llttlewood. One consequence of our paper is that for for the Hardy-Llttlewood problem should be p < 2 -[ao[ the extremal function r. laklz + k=1 We also give some appllcatlons to derive some sharp Inequalltles for the classes of Schllcht functions and of functions of positive real part. Key Words and Phrases: Fourier coefficients, analytic continuation, Schllcht functions, functions of positive real part 1980 AMS SUBJECT CLASSIFICATION CODES: 1. 42A05, 30BlO, 30D55 DEFINITIONS AND NOTATIONS. {z e C We will denote D{a,r) Iz-al < r} and H{U} is the set of all functions analytic In U. If f(z} kffiO akzk complex vector such that e H(U) lekl and If I, k e (eo,el,e2 0,1,2 U ) D(0,1). Is an infinite then we will define 442 R. PERETZ fO (z) For fixed p n e Z If (Usually f R by the formuIa: J (0) r then we wilI denote: p and z e U. p and {fe} functional on laklOk zk, E k:O such that (f) min(1,min((lzl Akzk H(U) F(z) If(retO)lP e -i nodO. will be fixed so we will shorten f z, and satisfies k=O then we will say that we will define a II. fs(reiO),PdO. cf, p,n (r) The radius of the zero-free region of If 0 s r < 1, and c (r) c_ (r)). I ,p0n n will be defined by o, fc-) lak[ -< Ak. 0,1,2 k F majorizes f and write f F. Some familiar families of analytic functions in U are: S f(O) O, f’ (0) (f H(U) 1, f is univalent in H(U) f(O) P Re{f (z)} > 0 V z e U}. 1, {f 2. U} THE MAIN RESULTS. f(z) If k=m Let k s f=(z) k’ akzk e H(U) -I + k=l k a m Denote 0 k zk, f(z) -lam Jzm+k=l lam+kl m+k f (z} -I + k=l k zk H(U) satisfy 1,2,3 The following theorems will be proved in the paper. For every p < 2 and every 0 s r < 6(f ), we have o.f(rei’)IPd# If1(rei’)lPd,. s The inequality is usually not valid for For every p < 2 For every p < 0 n =I,2,3 THEOREM D. n 1,2,3 we and every 0 s r < and 0 s r < 6(fl), (fB), we have and for all haveloJf(rei#)JPe-in#d#l }f,(rei,)jPe-in#d#. For every p < 0 and every 0 s r < we have o 3. r > f,(re i#)’ Pe-in#d# s o. f,(rei#). 8(fB) and for all Pe- in#d#. A MAJORANT PROBLEM TO IIARI}Y AND L ITTLEWOOD. In this paper we solve the following extremal problem: max J (8) for every p < 2 and every 0 p,r We also find the extremal directions e’ for which J (8’) p,r (The compactness of {re} in the topology of uniform max J O p.r (e]. convergence on compact MAJORANT PROBLEM U subsets of 443 A majorant guarantees the existence of maximizing functions). problem for Fourier coefficients, that was studied by Hardy and Littlewood, goes as follows: p If P P Are there any other values of ffl F e L and p, f e L then p and, FII f in fact, F f is an even integer and if for which an inequality of the type p B F F ? holds whenever f P P P In his paper [2] from 1975, Harold Shapiro characterizes the class of all smooth functions 2 o l(elldo n p function unless @ R on u + Jo @..f(eit}.. I) dtCI for which x he ddu h x p o uoh is an even integer. One consequence of our paper in connection with this problem is that for p < 2 new the majorant function of Hardy and Littlewood should be replaced by a maJorant function: -1o{ 4. PRELIMINARY (a} If and every / k=l instead of k=O {aklzk 0). (a REM/ugKS. H(u) and f f,g r < 0 {akizk g, then for every p 2k, 0,1,2 k we have o lf(rei)IPd o lg(rei)lPd. Even more is true: .f(reiO}.Pe-inbd@ l.g(rei}.Pe-inbdO, (In view of the fact that for every z k(z} f If{n)(rei@}12kd Ih(n)(rei@)12kd n,k (b) If 0, I,2 and mln and in particular (c) J {e’} 6(f) p v O, If max J e p,r f1(rei)l {e) 0 fS hP r < I.) 0 H(U), a 0 k=O then (by the triangle inequality} we have o-2 we have l+z 1-z t(z} oIk{n)(rei@}12kd@, olt(n)(rei@)12kd@, akzk f(z) h e P and every h (l_z)2 we can deduce that for all f e S O, 1,2 n -< rain o--2 and if f(rei)l -[ao[ fl(z} 0 r < + . k=l (f) -< Sff). r < and e’= satisfies 2 and if we have the Fourier expansions: $(fx) (8o,8,e 444 R. PERETZ k=O p-2 Ife,(Zll then for all 0,1,2 k bkzk fe,(z) arg(bk) we have {arg(k), arg(Ok) 8J see this we use the fact that for all Al AltCLYTIC EXTIOI FOR S. Let PIOPOSlTIO 1. (a) If Cf,p,n (b} (r) If (a) If gn . atzt . It is enough to consider the case f(z then )k t=O homogeneous polynomlal of degree l i--[P6[ 1/6= I- Co{r) extension of (b) Let 8Ch) H{U) h 8(f) [a611/6. k r < 8(f). D(0,8(f)). 0 F. 6=0 R Hence (h(zei@)) But II---J JdtJ2J point on Izl - t2 1- ]d61 1/6 8(f) (since at)z6 where is an P6 coefficients such E 6=0 a6)]2z26. lP6(a0 that zmh(z), . 6--0 IdtlZz26, h(zei@)h(ei@)) is an (h(ze I@)h(e I@)) P/2d@ d6(zei@}t, Izl Izl Co{r) < (f) for then by < ,scf:. ,h(zei@)] p/2 --(- because Since p/2 and clearly p/2 U. z o =-rmpgo{r) go{Z} Parseval’s equality we have go (z) 0. f(z) and D(0,8{f)) Finally, if n with positive integer we have for each fixed is an analytic function in for every gn H(D(O,@}). p6(ao go(Z) h{0) satisfy O, z By Parseval’s equality the desired analytic is given by analytic function in 0 gn(r) such that rmpgn{r} Cf,p,n(r} such that t=O that H(U) } and if f has a zero of order m at gn H(D(0,8{f))) f(z) =0.) k Z. n and 0 <- r < 8(f}. }4oreover for each } > 8(f} we have Proof. 80 n ) then there Is a {2klk=0,1,2 then there is a p,r 0 -< r < 1. for every p we have c (r). H(U) f (2klk=O,1,2 p 0,1,2 k (To }. / has a singular Is not a nonnegatlve integer). We remark that in (b) we have proved the representatlon 6. o )[ .f(rel@-.Pd@ r mp E 6=0 A SOLUTION FOR THE PROBLEN max J Let f(z} e 1 + F. k zk k=l H(U) Idtl z2t, p,r (O) satisfy 0 r < 8(f). FOR p < 2, 0 0 f. Let O r < 8(f). (10 O2 ’ ..) MAJORANT PROBLEM lekl be an infinite complex vector such that < (fe(z))q and LEIIA 2. dek zk, Izl ldekl Denote Proof. < e then for every ldkl, we have then (fe(z))q (l+x) q, and this implies Hence we have a representation of the form: det where ’j e R = n +... +n and depends onIy on j(lel )nl in (iz monomlals of the form (-! )l-n et ntJ j e). (not on q zz z the coefficient of 3z .)n 3 Yn,t If then + i + et n I1 (2) n (1 is is a sum of Y t. t} n But (qn) sgn and by (1) we have (-1)n q<O nl Cqn) Yn’ d, Hence using the representatlon (2) we have (3) +nj n lj+... and the conclusion therefore follows. PROOF OF THEOREN A. (fez})p/2= We can assume E dkzk, I1 < Cf) By the representation In Proposition 2 y Le..a - < th 2 O, a o > O, a (ritZ))p/2= k=O Let O. dzk, I1 < (f}. we have Z k=O lf1(reillPd m and k=O lf(rei)lPd E Idk 12 r2k, 0 IdOl r2k, Idkl Idl, r < 6(f} 0 s r < k=O e hve k 0,,- nd e obtain the desired inequality. For the final part of the theorem we consider the following example: Fix 0 < We have < and define gO(z] (g) (V- I). Denote Gl(r Then and o z d@ +(R). + 2z e iezz, g1(z] o Ge{r g (reiq) G(6(g)) Ge(a(ge) lim 6(ge) 6(gI), we e_)o q 1,2,3 k kek zk, Izl < 8(fe ), x Suppose 1,2,3 1, k (fe ). (1,1,-1,1,-1,1,-1 @ If k--O 445 Since 6(g) < can do the following: 6(ge) 2 + 2z z d@ ge(rei) for all 0 e 2 446 R. PERETZ For every r So Gl(r o) < > (g) G while (r eo eo tlcular we can choose dz k, Izl < Let {f (z) )p/2 < {ge- GOo (to) k=O to. o dk zk’ By by formula 0,1,2 idkl EIjI.11j... we have < G1(ro kj done. it suffices to prove (2) In the proof of Lemma 2 EIjIIj... idol f ,(f (z) )p/2 < z By representation in Proposition (f). Idl, k Idkl o) (gl) 0 such that o could be made as large as we want and in par- so that PROOF OF IIIEORE B. k=O that eo > we can choose a kj and we are o The combination of Theorems A and B gives us, for example, the following: COROLLARY 3. modulus (a) Let p < 2 and let be an Inflnlte complex vector of 8 components. f e P For every and every r < 0 we have 3 d. l-{re i@) Moreover for every n there is a 0 n > 0 such that for all 0 r < n we have }/{n)(rei@) {n)(rei@)IPd@ (b} For every f e S and every r < 0 o" 2.n! lPd@ (t(z) vr) (2 we have (rel’) If is the minimal modulus of the zeros of every 0 s r < i P 2(1-z} + z 3 then for we have o Flnally for every there Is a n on l+(re i@) > 0 P 2 such that for all 0 r < we have Je (reilJPd Let COROLLARY 4. Jk (n)(reiqb) 2n.n! ak zk H(U) f(z) . k=O fCz) -IoI + (0) lakl zk, f (z)’) then for every p 0 IPdq a o (k(z) 0 f(rel’) jpd’ s and and for all we have flcO)flcrei’) o" 2f-(rel,)_fl (0) p z (l-z) 0 s r < mln n MAJORANT PROBLEM PROOF. Since f{z) 0 "- r < rain (h(z)) usual let . p/2 :Imk z I=kl -Ikl, and zmh(z) f(z) k=O dkzk, Izl where < (f). lklZk)_ + f (Z) fl{rei) H(U) h and we obtain 1,2,3 k fl(O f(ret 2 (5(fl)’ 5(f (0) fl (-’" Let PROOF OF THEOREM C. k D(O,8(fl) both are analytic in for every + a 447 -Pdo O. h(O) and As Then, just as in Proposition 1, we have y [f(rei)[Pe-ind For dk+nkr2k)rmP, 2( 0 -< r < (f). k=O small enough we have, as in Lemma 2: ]z] (1- z- z - 3 z 3 )q + 2]z [-q2 + q (q-l) 2! z (22) q(q-1)q(q-2)= 3! (-ql)z + f-q=3 + q(q-1) 2! + [_q=4 + q(q-1) 2 + q(q-1)q(q-2) (21+) 2a 3 (3 + q(q-1)q(q-2)(q-3) 4! and if for all d k z is the coefficient of k 0,1,2 en consideri if we choose Ire(re max e k then maxldkl arg(=j) )I a,4]z 4 + is solved simultaneously J O, 12.) 1,2,3 when q < I. we solve simultaneously aXldkl let us check 0 q < then then sgn(dk) choice arg(j) follows, and arg(dk+nk " arg(dk+nk} sgn(do) +I, O, k +1 for the choice while 0, I,2 J const. 1,2,3 sgn(dk) O, 1,2 k J arg(j)= O, -1, 1,2,3,’ k 1,2,3 If If q < 0 q < O, the This proves that when solves (I) and the conclusion u The proof of Theorem D goes along these lines in analogy to the proof of Theorem B. [11 [2] G.H. Hardy and J.E. Llttlewood, "Notes on the theory of series (XIX): a problem concernlng majorants of Fourler serles", Quart. J. Hath. (Oxford) 6 (1935), 304-315. H.S. Shapiro, "Majorant problems for Fourier coefficients", Quart. J. Math. (Oxford) {2}, 26 (1975), 9-18.