A problem in the theory of subordination (Q1317295)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | A problem in the theory of subordination |
scientific article |
Statements
A problem in the theory of subordination (English)
0 references
21 November 1996
0 references
The authors obtain an estimate for the Fourier coefficients for a measurable real-valued function \(f(x)\) on \([0,2\pi]\) where \[ (1) \quad |f(x)|\leq 1 \qquad \text{ and } \qquad (2) \quad f(x)\sim {1 \over 2} a_0 + \sum^\infty_{n=1}(a_n \cos nx + b_n \sin n x), \] \[ {1\over 2} \sum_{n=2}^\infty {a_n^2 + b_n^2 \over n^2-1} \leq {4\over \pi}-1.\tag{3} \] This proves a conjecture made earlier by the authors with \textit{A. W. Weitsman} [J. Anal. Math. 56, 87-123 (1991; Zbl 0747.31004), hereinafter cited as [HHW]]; note that the sum in (3) starts at \(n = 2\). It thus gives a sharp refinement of the geometrical result of [HHW] on the isoperimetric ratio \(L^2 /4\pi A\) for a plane region in terms of its asymmetry (Theorem 1). To state the main result, we say that a sequence \(\{\lambda_n\}_{-\infty}^\infty\) is in \(\mathcal A(k)\) if the \(\lambda_n\) are non-negative, \(\lambda_0 = 0\), \(\lambda_{-n} = \lambda_n, \lambda_k \geq \lambda_{2k} \geq \dots\), and \[ \lambda_m \leq (k^2/ \pi^2) \sin^2 (\pi m/k)\sum \lambda_n/(n-m)^2 \tag{4} \] where the sum is over \(n \equiv 0 \pmod k\). The authors include an example which shows the importance of (4). The main result is that if \(f\) is as in (1) and (2), then whenever \(\{\lambda_n\} \in \mathcal A(k)\), then \[ {1\over 2} \sum_{n=1}^\infty \lambda_n (a_n^2 + b_n^2) \leq {8\over \pi^2} (\lambda_1 + {\lambda_{3k} \over 9} + {\lambda_{5k} \over 25} + \cdots).\tag{5} \] Equality holds for the function \(f(x) = \text{sgn} (\sin kx).\) This was conjectured in [HHW] for \(k =2\) and particular choices of the \(\{\lambda_n\}\), in which case (5) reduces to (3). The theorem and techniques may be useful in other problems dealing with estimates of Fourier coefficients. By choosing the \(\{\lambda_n\}\) property, it can give a new method for estimating an integral \(\int |H(\theta)|d\theta\), which can be sharp if the sign of \(H\) alternates on congruent intervals.
0 references
isoperimetric inequality
0 references
subordination
0 references
Fourier coefficients
0 references
measurable real-valued function
0 references
isoperimetric ratio
0 references