Approximation of functions in \(L_2\)-metric with the Jacobi weight (Q1589035)

The paper studies the characterization of functions \(f\in L_{2,\lambda,\delta}[-1,1]^2\) (where \[ \|f\|_{L_{2,\lambda,\delta}[-1,1]^2}=\left(\int_{-1}^{1}\int_{-1}^1f^2(x,y) \lambda(x)\delta(y) dx dy\right)^{\frac{1}{2}}, \] is the norm) such that the errors of best approximation \(\{E_{n,m}(f)\}_{n,m\geq 0}\) given by \[ E_{n,m}(f)= \min_{k\leq n,l\leq m} \Bigl\|f-\sum_{a_{k,l}}P_k^{a_1,b_1}(x)P_l^{a_2,b_2}(y) \Bigr\|_{L_{2,\lambda,\delta}[-1,1]^2}, \] (being \(P_k^{a,b}(t)=(1-t)^a(1+t)^b\) the classical Jacobi polynomials), satisfy the rate of convergence \[ E_{n,m}(f)\leq M(1/n^{r_1+s_1}+1/m^{r_2+s_2}) \] These functions are precisely those that belong to a certain Nikolskij class of functions, \(H^{r_1,r_2}\). The results are certainly interesting, but the author only gives a draft of the proof.