Transactions of the American Mathematical Society · 1934 · 59 citations · 9 references
EngineeringMechanical EngineeringStochastic ProcessesTime Series AnalysisVibrationCharacteristic FunctionsFourier AnalysisFourier Series Analysis
I. On the character of oscillation of the PARTIAL SUMS of Fourier series 1.The fundamental theorem.Completing a well known result of Kolmogoroff [12], Marcinkiewicz [15] has recently constructed a function integrable L, whose Fourier series possesses partial sums oscillating finitely almost everywhere.It is, therefore, natural to ask what may be said about the relative position of the interval of oscillation of s"(x) and the value f(x), beyond the well known fact that the said interval contains/(x).The result proved in this note is a first attempt in this direction.Theorem.If the partial sums sn(x) of the Fourier series of a function fix) integrable L, 1 (1) f(x) ~ -a0 + 22 (a* cos nx + bn sin nx), 2 n-X satisfy an inequality(2) sn(x) £ -<¡>(x) (0 g x á 2t; ft -0,1, 2, • • • ), with <p(x) ^0 integrable L, then, for almost every x,(3) fix) = -lim sup s"(x) + lim inf sn(x) .2 |_ n-*» n-.oo _| 2. Statement of lemmas.The proof of our theorem is based on three lemmas which will be stated in this section and the proof of which will be given in the next section.Let o-"(x), S7(x), sn(x) denote, respectively, the first arithmetic mean's of the series (1), of the conjugate series oo 22 (an sin nx -bn cos nx), n-l and the partial sums of the latter series.We have [20,24,18,30] * Presented to the Society, March 31, 1934; received by the editors November 5, 1933.The six notes constituting this paper are independent of one another, although they treat related topics.The numbers in square brackets refer to the Bibliography at the end of the paper.Ln(t) ¿ J Kn(u)Kn(u -t)du.Since the integral is an even function of t, we may restrict ourselves to the values O^igir.Break up the integral into four, extended over the intervals (-7T, -ir/2), i-w/2, 0), (0, t/2), (t/2, it), and denoted respectively by UnM(t), Unw(t), Un™(t), Unw(t)-Since Qn(u) is decreasing in the interval 0 ¿ u ^ 37r/2, it is readily seen thatJ 119 Hi Adding these inequalities together we obtain (6).To obtain (7), it is sufficient to replace in the integral (5) the function Kn(u-t) by its upper bound (which is 0(n)) and to notice that the remaining integral represents the Fejér means, at x=0, of the characteristic function of the set H, and so, by Lebesgue's well known criterion, tends to 0 with 1/».To prove Lemma 2, replace, in the right-hand side of (4), s"(w) by sn(u)+<t>(u) -d>(u).Then in view of (2), the first term in (8) does not exceed -I bn(u) + <p(u)]Kn(u -x)du + -j <b(u)Kn(u -x)du \¡/(u)Kn(u -x)du t •/_, where <¿(w) = 2/(w) +4c/>(»).We pass on to the proof of Lemma 3. First, we have, for almost every x in the interval (-it, t), the relation (11) -llîL -(' [sn(U)-0-n(u) + M] COS (n+l)(u-x)Kn(u-x)du + o(l).
9
Some properties of fractional integrals. I.
G. H. Hardy, J. E. Littlewood · Mathematische Zeitschrift · 1928 · 908 citations
Some new properties of fourier constants
G. H. Hardy, J. E. Littlewood · Mathematische Annalen · 1927 · 233 citations
E. C. Titchmarsh · Proceedings of the London Mathematical Society · 1929 · 149 citations