Simultaneous approximation of a pair of conjugate functions (Q791797)

Let us choose an arbitrary but fixed real number \(r\geq 0\) and a modulus of continuity \(\omega\) (\(\delta)\) (\(\omega\) (\(\delta)\to 0\) with \(\delta\to 0)\). If \(r=0\), then \(\omega\) (\(\delta)\)/\(\delta\) is supposed to be integrable in a neighbourhood of 0. With each real \(\alpha\) we associate a class \({\mathfrak M}_{\alpha}\) of functions \[ f(x)=\sum^{\infty}_{k=1}(1/\pi k^ r)\int^{\pi}_{- \pi}\phi(t+x)\cos(kt+\alpha \pi /2)dt\tag{1} \] where \(\phi\) (t) are 2\(\pi\)- periodic functions satisfying \(\int^{\pi}_{-\pi}\phi(t)dt=0\), \(| \phi(t)-\phi(t')| \leq \omega(| t-t'|)\) and in case \(r>0\) also \(| \phi(t)| \leq 1\). Let us note that (1) is a uniformly convergent trigonometric series. Hereafter \(f\in {\mathfrak M}_{\alpha}\) and \(f_{\beta}\in {\mathfrak M}_{\alpha +\beta}\) (\(\beta\) real) stand always for a pair of functions which are defined by the same \(\phi\) in (1). Clearly, f and \(f_ 1=\tilde f\) are conjugate functions. That is why f and \(f_{\beta}\) are called \(\beta\)-conjugate functions if \(\beta\) is arbitrary. We shall denote for each natural n \(U_ n(f;x;\Lambda)=\lambda^{(n)}_ 0\quad a_ 0(f)/2+\sum^{n}_{k=1}\lambda^{(n)}_ k\quad(a_ k(f)\cos kx+b_ k(f)\sin kx)\) where \(a_ k(f)\) and \(b_ k(f)\) are Fourier coefficients of f and \(\lambda^{(n)}_ k\) correspond to a summation method \(\Lambda\). \(U_ n({\mathfrak M}_{\alpha};\Lambda)=\sup_{f\in {\mathfrak M}_{\alpha}}| f(x)-U_ n(f;x;\Lambda)|\) and for any \(p\geq 0\), \(q\geq 0 {\mathfrak E}^{p,q}_{n,\beta}({\mathfrak M}_{\alpha};\Lambda)=\sup_{f\in {\mathfrak M}_{\alpha}}(p| f(x)- U_ n(f;x;\Lambda)| +q| f_{\beta}(x)-U_ n(f_{\beta};x;\Lambda)|)\) are values independent on x, the latter characterizing the simultaneous approximation of \(\beta\)-conjugate functions by a \(\Lambda\)-method. The main theorem of this paper says that \({\mathfrak E}^{p,q}_{n,\beta}({\mathfrak M}_{\alpha};\Lambda)=\max(A U_ n({\mathfrak M}_{\gamma};\Lambda);B U_ n({\mathfrak M}_{\gamma_ 1};\Lambda))\) where A,B,\(\gamma\) and \(\gamma_ 1\) are given by explicit formulas in terms of p,q,\(\alpha\) and \(\beta\). Thus earlier results concerning the behaviour of \(U_ n({\mathfrak M}_{\alpha};\Lambda)\) with \(n\to \infty\) for various \(\alpha\) and r can be used. Especially for conjugate functions (i.e. for \(\beta =1) A=B=(p^ 2+q^ 2)^{1/2}, \gamma =\alpha +2\epsilon\) and \(\gamma_ 1=\alpha -2\epsilon\) where \(\epsilon =arctg(p/q)\). Moreover, in this case \(U_ n({\mathfrak M}_{\gamma};\Lambda)=U_ n({\mathfrak M}_{\gamma_ 1};\Lambda)\) provided that \(\alpha\) is an integer.