Negative property of shape-preserving finite-dimensional linear operators (Q1433649)

The paper deals with a linear operator \(L_n:C^k[0,1]\mapsto C^k[0,1]\), of finite rank \(n+1\), which satisfies \(D^kL_ne_k=D^ke_k\), where \(e_i(x):=x^i\), \(i\geq0\), and \(D^k\) denotes the \(k\)th derivative. If the operator preserves a certain cone \(C_{h,k}(\sigma)\) (described below), then the main result claims that not all the \(k\)th derivatives of \(L_ne_i\), \(i=h\dots,k+2\), \(i\neq k\), approximate well the \(k\)th derivatives of \(e_i\), \(i=h\dots,k+2\), \(i\neq k\). The cone is defined for a sequence \(\sigma:=\{\sigma_i\}\), with \(\sigma_i\in\{-1,0,1\}\), and \(0\leq h<k\) satisfying \(\sigma_h\sigma_k\neq0\). Let denote \[ C_{h,k}(\sigma):=\{f\in C^k[0,1]\mid \sigma_iD^if\geq0,\,h\leq i\leq k\}. \] Unfortunately, the reviewer cannot follow the proof of the main lemma, where it seems that the author confuses the index \(i_{j+1}\) with the index \(i_j+1\), where the indices \(i_j\) are those indices \(h\leq i_j\leq k\), for which \(\sigma_{i_j}\neq0\). Therefore the proof of the lemma is valid only when \(\sigma_i\neq0\) for all \(h\leq i\leq k\), and in particular the example given in the paper is correct.