ON FOURIER RE-EXPANSIONS E. LIFLYAND Abstract. WestudyanextensiontoFouriertransformsoftheoldproblemonabsoluteconvergence ofthere-expansioninthesine(cosine)Fourierseriesofanabsolutelyconvergentcosine(sine)Fourier series. The results are obtained by revealing certain relations between the Fourier transforms and 2 1 their Hilbert transforms. 0 2 n a 1. Introduction J 0 In 50-s (see, e.g., [5] or in more detail [6, Chapters II and VI]), the following problem in Fourier 2 Analysis attracted much attention: Let {a }∞ be the sequence of the Fourier coefficients of the absolutely convergent sine (cosine) ] k k=0 A Fourier series of a function f : T = [−π,π) → C, that is |a | < ∞. Under which conditions on k C {ak} the re-expansion of f(t) (f(t)−f(0), respectively) inPthe cosine (sine) Fourier series will also . h be absolutely convergent? t a The obtained condition is quite simple and is the same in both cases: m [ ∞ (1) |a |ln(k +1) < ∞. 1 k v X k=1 7 1 In this paper we study a similar problem for Fourier transforms defined on R+ = [0,∞). Let 2 4 ∞ . 1 F (x) = f(t)cosxtdt c 0 Z0 2 be the cosine Fourier transform of f and 1 : v ∞ i X F (x) = f(x)sinxtdt s Z r 0 a be the sine Fourier transform of f, each understood in certain sense. Let ∞ (2) |F (x)|dx < ∞, c Z 0 and hence 1 ∞ (3) f(t) = F (x)costxdx, π Z c 0 or, alternatively, 1991 Mathematics Subject Classification. Primary 42A38; Secondary 42A50. Key words and phrases. Fourier transform, integrability, Hilbert transform, Hardy space. 1 2 E. LIFLYAND ∞ (4) |F (x)|dx < ∞ s Z 0 and hence 1 ∞ (5) f(t) = F (x)sintxdx. π Z s 0 1) Under which (additional) conditions on F we get (4), c or, in the alternative case, 2) under which (additional) conditions on F we get (2)? s The answer is given by the following theorem. To formulate it we turn to the Hilbert transform of an integrable function g 1 g(t) (6) Hg(x) = dt, π Z x−t R where the integral is understood in the improper (principal value) sense, as lim . It is |t−x|>δ δ→0+ not necessarily integrable, and when it is, we say that g is in the (real) Hardy spaRce H1(R). If g ∈ H1(R), then (7) g(t)dt = 0. Z R It was apparently first mentioned in [8]. Theorem 1. In order than the re-expansion F of f with the integrable cosine Fourier transform s F be integrable, it is necessary and sufficient that its Hilbert transform HF (x) be integrable. c c Similarly, in order than the re-expansion F of f with the integrable sine Fourier transform F c s be integrable, it is necessary and sufficient that its Hilbert transform HF (x) be integrable. s 2. Proof of Theorem 1 Let (2) holds true. Then we can rewrite ∞ 1 ∞ (8) F (x) = F (u)costudu sinxtdt. s Z (cid:20)π Z c (cid:21) 0 0 The right-hand side can be understood in the (C,1) sense as 1 N t ∞ lim (1− ) F (u)costudu sinxtdt. π N→∞Z N Z c 0 0 In virtue of (2) we can change the order of integration: ON FOURIER RE-EXPANSIONS 3 1 ∞ N t lim F (u) (1− )costu sinxtdtdu π N→∞Z c Z N 0 0 1 ∞ 1 N t = lim F (u) (1− )[sin(u+x)t−sin(u−x)t]dtdu. π N→∞Z c 2 Z N 0 0 We now need the next simple formula N t 1 sinNA (9) (1− )sinAtdt = − . Z N A NA2 0 Applying it yields 1 ∞ 1 sin(u+x)N F (x) = lim F (u) − du s π N→∞Z c (cid:20)u+x N(u+x)2 (cid:21) 0 1 ∞ 1 sin(u−x)N (10) − lim F (u) − du = I +I . π N→∞Z c (cid:20)u−x N(u−x)2 (cid:21) 1 2 0 Let us begin with I . Substituting u−x = t, we obtain 2 1 ∞ 1 sinNt I = − lim F (x+t) − dt. 2 π N→∞Z c (cid:20)t Nt2 (cid:21) −x For I , we first substitute u = −v. Thus 1 1 0 1 sin(−v +x)N I = lim F (−v) − dv 1 π N→∞Z c (cid:20)−v +x N(x−v)2 (cid:21) −∞ 1 0 1 sin(v −x)N = − lim F (−v) − dv π N→∞Z c (cid:20)v −x N(v −x)2 (cid:21) −∞ 1 0 1 sin(v −x)N = − lim F (v) − dv. π N→∞Z c (cid:20)v−x N(v −x)2 (cid:21) −∞ The last equality follows from the evenness of F . Substituting v −x = t, we obtain c 1 −x 1 sinNt I = − lim F (x+t) − dt. 1 π N→∞Z c (cid:20)t Nt2 (cid:21) −∞ Therefore, 1 ∞ 1 sinNt F (x) = − lim F (x+t) − dt. s π N→∞Z c (cid:20)t Nt2 (cid:21) −∞ We are now in a position to apply the following result (see [17, Vol.II, Ch.XVI, Th. 1.22]; even more general result can be found in [15, Th.107]). |f(t)| Theorem A. If is integrable on R, then the (C,1) means 1+|t| 1 ∞ 1 sinNt − f(x+t) − dt π Z (cid:20)t Nt2 (cid:21) −∞ 4 E. LIFLYAND converge to the Hilbert transform Hf(x) almost everywhere as N → ∞. It follows from Theorem A that for almost all x) (11) F (x) = HF (x). s c We remark that any integrable function satisfies the assumption of Theorem A. Now, let (4) holds true. Then we can rewrite ∞ 1 ∞ (12) F (x) = F (u)sintudu cosxtdt. c Z (cid:20)π Z s (cid:21) 0 0 The right-hand side can be understood in the (C,1) sense as 1 N t ∞ lim (1− ) F (u)sintudu cosxtdt. π N→∞Z N Z s 0 0 In virtue of (4) we can change the order of integration: 1 ∞ N t lim F (u) (1− )sintu cosxtdtdu π N→∞Z s Z N 0 0 1 ∞ 1 N t = lim F (u) (1− )[sin(u+x)t+sin(u−x)t]dtdu. π N→∞Z s 2 Z N 0 0 Applying (refces), we get 1 ∞ 1 sin(u+x)N F (x) = lim F (u) − du s π N→∞Z s (cid:20)u+x N(u+x)2 (cid:21) 0 1 ∞ 1 sin(u−x)N (13) + lim F (u) − du = J +J . π N→∞Z s (cid:20)u−x N(u−x)2 (cid:21) 1 2 0 Let us begin with J . Substituting u−x = t, we obtain 2 1 ∞ 1 sinNt J = lim F (x+t) − dt. 2 π N→∞Z s (cid:20)t Nt2 (cid:21) −x Treating J as I above, we get 1 1 1 0 1 sin(v −x)N J = − lim F (−v) − dv 1 π N→∞Z s (cid:20)v −x N(v −x)2 (cid:21) −∞ 1 0 1 sin(v −x)N = lim F (v) − dv. π N→∞Z s (cid:20)v −x N(v −x)2 (cid:21) −∞ The last equality follows from the oddness of F . Substituting v −x = t, we obtain s 1 −x 1 sinNt J = lim F (x+t) − dt. 1 π N→∞Z s (cid:20)t Nt2 (cid:21) −∞ Therefore, ON FOURIER RE-EXPANSIONS 5 1 ∞ 1 sinNt F (x) = lim F (x+t) − dt. c π N→∞Z s (cid:20)t Nt2 (cid:21) −∞ Finally, it follows from Theorem A that for almost all x) (14) F (x) = −HF (x). c s (cid:3) This completes the proof. Let us comment on the obtained results. In fact, the proof of Theorem 1 shows that more general results than stated are obtained. Indeed, formulas (11) and (14) are more informative than the assertion of Theorem 1. To be precise, such formulas are known, see [7, (5.42) and (5.43)]. However, the situation is much more delicate. These formulas are proved in [7] for square integrable functions by applying the Riemann-Lebesgue lemma in an appropriate place (5.44). But in [7, §6.19] more details are given (see also [2]) and it is shown that the possibility to apply the Riemann-Lebesgue lemma in that argument is equivalent to (Carleson’s solution of) Lusin’s conjecture. In our L1 setting this is by no means applicable. And, indeed, our proof is different and rests on less restrictive Theorem A. This is well agrees with what E.M. Dyn’kin wrote in his well-known survey on singular integrals [2]: ”In fact, the theory of singular integrals is a technical subject where ideas cannot be separated from the techniques.” 3. Sufficient conditions Analyzing the proof in [5], one can see that in fact their results are similar to ours, that is, can also be given in terms of the (discrete) Hilbert transform. In that case (1) is simply a sufficient condition for the summability of the discrete Hilbert transform. An analog of (1) for functions cannot be the only sufficient condition for the integrability of the Hilbert transform. Indeed, a known counter-example of the indicator function of an interval works here as well: of course, it stands up to the multiplication by logarithm. Thus, we are going to give sufficient conditions for the integrability of the Hilbert transforms in the spirit of those in [5] for sequences supplied by some additional properties. 3.1. General conditions. First of all, examining integrability of the Hilbert transform, one can test the integral over, say, |t| ≤ 3|x|. Indeed, for x > 0, we have 2 ∞ ∞ g(t) dt dx Z (cid:12)Z3 x−t (cid:12) 0 (cid:12) 2x (cid:12) (cid:12) (cid:12) ∞(cid:12) 2t/3 (cid:12)dx ∞ ≤ |g(t)| dt = ln3 |g(t)|dt. Z Z x−t Z 0 0 0 The rest is estimated in a similar manner. Further, since a dt (15) = 0 Z x−t −a for any a > 0 when the integral is understood in the principal value sense, we can always consider 6 E. LIFLYAND a g(t)−g(x) dt Z x−t −a instead of the Hilbert transform truncated to [−a,a]. When in the definition of the Hilbert transform (6) the function g is odd, we will denote this transform by H , and it is equal to o 2 ∞ tg(t) (16) H g(x) = dt. o π Z x2 −t2 0 When g is even its Hilbert transform H can be rewritten as e 2 ∞ xg(t) (17) H g(x) = dt. e π Z x2 −t2 0 Of course, both integrals should be understood in the principal value sense (see, e.g., [7, Ch.4, §4.2]). Since the functions F and F are even and odd, respectively, their Hilbert transforms can be c s represented as 2 ∞ xF (t) (18) H F (x) = c dt. e c π Z x2 −t2 0 and 2 ∞ tF (t) (19) H F (x) = s dt. o s π Z x2 −t2 0 These may be useful in some applications. Infact, themostknownconditionisthefollowing. Ifg isofcompact support, aclassical Zygmund LlogL condition (see, e.g., [17]) ensures the integrability of the Hilbert transform. More precisely, the condition is the integrability of glog+|g|, where the log+|g| notation means log|g| when |g| > 1 and 0 otherwise. As E.M. Stein has shown in [13], this condition is necessary on the intervals where the function is positive. However, this condition looks quite restrictive in our case. We will prove the following result. Theorem 2. Let g be an integrable function on R which satisfies conditions (7), (20) |g(x)|log3|x|dx Z |x|≥1/2 and 1 2min(|x|,1) g(x+t)−g(x) (21) dtdx. ZR Z−21min(|x|,1)(cid:12)(cid:12) t (cid:12)(cid:12) (cid:12) (cid:12) Then g ∈ H1(R). (cid:12) (cid:12) Since each function can be represented as the sum of its even and odd parts, we will prove Theorem 2 separately for odd and even functions. Thus, from now on we can consider g to be defined on R and analyze either (16) or (17) rather than the general Hilbert transform. + ON FOURIER RE-EXPANSIONS 7 3.2. Odd functions. Though an odd function always satisfies (7), not every odd integrable func- tion belongs to H1(R), for a counterexample see, e.g., [10]. Paley-Wiener’s theorem (see [12]; for alternative proof and discussion, see Zygmund’s paper [16]) asserts that if g ∈ L1(R) is an odd and monotone decreasing on R function, then Hg ∈ L1, i.e., g is in H1(R). Recently, in [11, Thm.6.1], + this theorem has been extended to a class of functions more general than monotone ones. However, it is doubtful that these results are really practical in our situation. Back to Theorem 2, we can consider 2 3x/2 tg(t) dt π Z x2 −t2 x/2 instead of (16). Indeed, the possibility of restricting to that upper limit has been justified above. Similarly, ∞ x/2 tg(t) ∞ ∞ dx dt dx ≤ |g(t)|t dt Z (cid:12)Z x2 −t2 (cid:12) Z Z x2 −t2 0 (cid:12) 0 (cid:12) 0 2t (cid:12) (cid:12) 2 ∞ (22) (cid:12) (cid:12) ≤ |g(t)|dt. 3 Z 0 Now, like in (15), we have 1 3x/2 tg(t) 1 3x/2 t[g(t)−g(x)] ∞ dt dx ≤ dt dx+O( |g(t)|dt) Z (cid:12)Z x2 −t2 (cid:12) Z (cid:12)Z x2 −t2 (cid:12) Z 0 (cid:12) x/2 (cid:12) 0 (cid:12) x/2 (cid:12) 0 (cid:12) (cid:12) (cid:12) (cid:12) (cid:12) (cid:12) 1(cid:12) x/2 |g(x+t)−g(x)|(cid:12) ∞ (23) ≤ dtdx+O( |g(t)|dt). Z Z |t| Z 0 −x/2 0 When x ≥ 1 we first estimate ∞ x−1/2 tg(t) ∞ 2t dx dt dx ≤ t|g(t)| dt Z (cid:12)Z x2 −t2 (cid:12) Z Z x2 −t2 1 (cid:12) x/2 (cid:12) 1/2 t+1/2 (cid:12) (cid:12) ∞ (cid:12) (cid:12) ≤ C |g(t)|ln3tdt Z 1/2 and ∞ 3x/2 tg(t) ∞ t−1/2 dx dt dx ≤ t|g(t)| dt Z (cid:12)Z x2 −t2 (cid:12) Z Z t2 −x2 1 (cid:12) x+1/2 (cid:12) 3/2 2t/3 (cid:12) (cid:12) ∞ (cid:12) (cid:12) ≤ C |g(t)|ln3tdt. Z 1/2 These two bounds lead to the logarithmic condition (20). The remained integral ∞ x−1/2 tg(t) dt dx Z (cid:12)Z x2 −t2 (cid:12) 1 (cid:12) x+1/2 (cid:12) (cid:12) (cid:12) is estimated exactly like that in (23). A(cid:12)pplying (15), we(cid:12)obtain 8 E. LIFLYAND ∞ x+1/2 tg(t) ∞ x+1/2 t[g(t)−g(x)] ∞ dt dx ≤ dt dx+O( |g(t)|dt) Z (cid:12)Z x2 −t2 (cid:12) Z (cid:12)Z x2 −t2 (cid:12) Z 1 (cid:12) x−1/2 (cid:12) 1 (cid:12) x−1/2 (cid:12) 0 (cid:12) (cid:12) (cid:12) (cid:12) (cid:12) (cid:12) ∞(cid:12) 1/2 |g(x+t)−g(x)| (cid:12) ∞ (24) ≤ dtdx+O( |g(t)|dt). Z Z |t| Z 1 −1/2 0 Combining all the obtained estimates, we arrive at the required result. 3.3. Even functions. While an odd function always satisfies (7), in the case of even functions the situation is more delicate: the function must satisfy (7) already on the half-axis. With this in hand, the proof goes along the same lines as that for odd functions. The only problem is that an estimate like (22) does not follow immediately from the formula (17). However, using the above remark on the cancelation property for g on the half-axis, we can rewrite (17) as 2 ∞ x 2 (25) H g(x) = g(t) − dt. e π Z (cid:20)x2 −t2 x(cid:21) 0 Now, ∞ x/2 t2g(t) ∞ ∞ dx dt dx ≤ |g(t)|t2 dt Z (cid:12)Z x(x2 −t2) (cid:12) Z Z x(x2 −t2) 0 (cid:12) 0 (cid:12) 0 2t (cid:12) (cid:12) 1 ∞ (26) (cid:12) (cid:12) ≤ |g(t)|dt. 6 Z 0 This additional term 2 does not affect the other estimates of the previous subsection. x The proof is complete. 4. Concluding remarks First of all, the relations (11) and (14) are of interest by their own. The assertions of Theorem 1 can be reformulated in terms of Hardy spaces: belonging of F (F ) c s to the real Hardy space H1(R) ensures the integrability of F (F ). s c The problem of sharpness is simple in this case: any known counterexample of an integrable function with non-integrable Hilbert transform works perfectly. For example, let F (x) = 1 , the c 1+x2 Fourier transform of f(t) = e−|t|. Surely, F ∈ L1(R). However, f cannot be re-expanded in the c integrable sine Fourier transform, since HF (x) = x 6∈ L1(R). That this is true, one can see from c 1+x2 the fact that the odd extension of this F from the right half-axis to the whole R is not continuous c at zero. More can be said about odd functions. Certain convenient conditions for belonging of such functions to H1(R) are known for quite a long time. They are functions (Fourier transform) analogs of important sufficient sequence conditions for the integrability of trigonometric series (see, e.g., [14] and [3]) and can be found, for example, in [9] and in [4]. In fact, many of these subspaces first appeared in [1]. For 1 < q ≤ ∞, set ∞ 1 1/q kgk = |g(t)|qdt du, Aq Z (cid:18)u Z (cid:19) 0 u≤|t|≤2u ON FOURIER RE-EXPANSIONS 9 with a standard modification when q = ∞. In other words, belonging of g to one of the spaces A q ensures the integrability of the odd Hilbert transform of g. 5. Acknowledgements The author’s attention to this problem was brought by R.M. Trigub. Thanks to him and to J.-P. Kahane for stimulating discussions. References [1] D. Borwein, Linear functionals connected with strong Cesa´ro summability, J. London Math. Soc. 40 (1965), 628–634. [2] E.M. Dyn’kin, Methods of the Theory of Singular Integrals: Hilbert Transform and Caldero´n-Zygmund Theory, Commutative Harmonic Analysis I (V.P. Khavin and N.K. Nikol’skij - eds.), Springer, 1991, 167–259. [3] G.A. Fomin, A class of trigonometric series, Mat. Zametki 23 (1978), 213–222 (Russian). - English transl.: Math. Notes 23 (1978), 117–123. [4] D.V. Giang and F. Mo´ricz, Lebesgue integrability of double Fourier transforms, Acta Sci. Math. (Szeged) 58(1993), 299–328. [5] S.I. Izumi and T. Tsuchikura, Absolute convergence of trigonometric expansions, Tˆohoku Math. J. 7 (1955), 243–251. [6] J.-P. Kahane, S´eries de Fourier absolument convergentes, Springer, Berlin, 1970. [7] F.W. King, Hilbert transforms, Vol.1, Enc. Math/ Appl., Cambridge Univ. Press, Cambridge, 2009. [8] H. Kober, A note on Hilberts operator, Bull. Amer. Math. Soc., 48:1 (1942), 421–426. [9] E. Liflyand, Fourier transform of functions from certain classes, Anal. Math. 19(1993), 151–168. [10] E. Liflyand and S. Tikhonov, Weighted Paley-Wiener theorem on the Hilbert transform, C.R. Acad. Sci. Paris, Ser.I 348 (2010), 1253–1258. [11] E. Liflyand and S. Tikhonov, A concept of general monotonicity and applications, Math. Nachrichten 284 (2011), 1083–1098. [12] R.E.A.C.PaleyandN.Wiener,Notes on thetheory and application of Fourier transform,NoteII,Trans.Amer. Math. Soc. 35 (1933), 354–355. [13] E.M. Stein, Note on the class LlogL, Studia Math., XXXII (1969), 305–310. [14] S.A.Telyakovskii,OnaSufficientConditionofSidonfortheIntegrabilityofTrigonometricSeries,Mat.Zametki 14 (1973), 317–328(Russian). - English transl.: Math. Notes 14 (1973), 742–748. [15] E.C. Titchmarsh, Introduction to the Theory of Fourier Integrals, Clarendon Press, Oxford, 1937. [16] A. Zygmund, Some Points in the Theory of Trigonometric and Power Series, Trans. Amer. Math. Soc. 36 (1934), 586–617. [17] A. Zygmund, Trigonometric series: Vols. I, II, Second edition, reprinted with corrections and some additions, Cambridge University Press, London, 1968. Department of Mathematics, Bar-Ilan University, 52900 Ramat-Gan, Israel E-mail address: [email protected]