Approximation of differentiable functions by positive summability methods for Fourier series (Q801501)

Let \(W^ rK\) denote the class of \(2\pi\)-periodic function f for which \(f^{(r-1)}(x)\) is an absolute continuous function on [-\(\pi\),\(\pi\) ] and \(| f^{(r)}(x)| \leq K\) a.e. If \(f\in W^ rK\), denote by \(a_ k,b_ k\) the coefficients of f and consider the positive operator \(U_ n(f,x,\lambda)=a_ 0/2+\sum^{n-1}_{k=1}\lambda_ k^{(n)}(a_ k\cos kx+b_ k\sin kx)\) where \(\lambda =\{\lambda_ k^{(n)}\}\) satisfy the condition \(+\sum^{n-1}_{k=1}\lambda_ k^{(n)}\cos kt\geq 0\) (0\(\leq t\leq \pi)\). Define \(E(r,\lambda)=\sup_{f\in W^ rK}\| U_ n(f,\lambda)-f\|_ c,\quad {\mathcal E}(r)=\min_{\lambda}E_ n(r,\lambda).\) The author proves \({\mathcal E}(r)=kN_ rn^{-2}+o(n^{-2}),\) where \(N_{2m}=(\pi^{2m}/(2^{2m-1}(2m-2)!))| C_{2m-2}|,N_{2m- 1}=((2^{2m}-1)n^{2m+1}/(2m)!)| B_{2m}|,\) \(m=1,2,...\), \(B_ r,C_ r\) are the Bernoulli numbers and Euler numbers respectively. Furthermore, let \(X^ r=\| x^ r_{ik}\|^ n\) be the square matrix with elements \(x^ r_{i_ k}=(-1)^{(| i-k| -1)(r- 1)/2}| i-k|^{-r-1},\) if \(| i-k|\) is odd, \(x^ r_{i_ k}=0\), if \(| i-k|\) is even. Then the coefficients \(\lambda =\{\lambda_ k^{(n)}\}\) attaining to \(E(r,\lambda)={\mathcal E}(r)\) are \(\lambda_ k^{(n)}=\sum^{n-k}_{m=1}u_ mu_{m+k})/(\sum^{n}_{m=1}u^ 2_ m)\) \((k=1,2,...,n-1)\) where \(\| u_ 1,...,u_ n\| '=u\) is the eigenvector of the matrix \(X^ r\) corresponding to the greatest eigenvalue.