Digital sums and divide-and-conquer recurrences: Fourier expansions and absolute convergence (Q1772233)

From MaRDI portal
scientific article
Language Label Description Also known as
English
Digital sums and divide-and-conquer recurrences: Fourier expansions and absolute convergence
scientific article

    Statements

    Digital sums and divide-and-conquer recurrences: Fourier expansions and absolute convergence (English)
    0 references
    0 references
    0 references
    15 April 2005
    0 references
    Let \(k\) be a natural number satisfying \(1\leq k\leq n\) and set \(n^{-1}S(n)=n^{-1} \sum_{1\leq k\leq n}\nu(k),\nu(k)\) being the number of 1's in the binary expansion of \(k\). The paper deals with \(n^{-1}S(n)\). As a result of Trollope and Delange it is known that (*) \(n^{2}S(n)=2^{-1}\log_2n+F(\log_2n)\), \(F(x)\) being an absolutely convergent Fourier series representing a nowhere differentiable function. If we chose at random one of the numbers \(1,2,\dots,n\) each with probability \(n^{-1}, n^{-1}S(n)\) is the expectation of the random variable \(\nu (k)\); since the \(\nu(k)\)'s are sums of mutually independent random variables, standard methods of probability theory can be applied. Different generalizations of the result of Trollope and Delange are given. Sometimes only the proofs are new, sometimes also the results. Instead of only (*) the corresponding higher moments are calculated, together with the to \(F(x)\) in (*) corresponding function. Reviewer's remark: The authors consider one of the main results of the paper a simple construction of an absolutely convergent Fourier series representing a nowhere differentiable function, However, the construction, based on the following statement is in the opinion of the rewiever at least as simple. Let \(n_1,n_2,\dots\) be a lacunary sequence of integers, \(a_1,a_2,\dots\) a sequence of real numbers satisfying \(\sum^\infty_{k=1}a_k<\infty\). Then the Fourier series \(f(x)=\sum a_k\cos N_kx\) represents a nowhere differentiable function if the \(n_k\)'s increase rapidly enough; the speed of increasing ensuring the non-differentiability depends on the coefficients \(a_k\) and can be calculated effectively.
    0 references

    Identifiers

    0 references
    0 references
    0 references
    0 references
    0 references
    0 references
    0 references