Correction up to a function with sparse spectrum and uniformly convergent Fourier series (Q549427)

Let \(G\) be a compact Abelian group and \(\Gamma\) its dual group. Let \(l\) be a positive integer and \(\gamma=(\gamma_1,\dots,\gamma_l)\in\Gamma^l\) be an ordered collection of \(l\) characters of \(G\). For \(m=(m_1,\dots,m_l)\in\mathbb{Z}^l\), denote by \(\langle m,\gamma\rangle\) the character \(m_1\gamma_1+\dots+m_l\gamma_l\), i.\,e., the function \(x\mapsto\gamma_1(x)^{m_1}\dots\gamma_l(x)^{m_l}\) on \(G\). A pattern (more precisely, \(l\)-pattern) means an arbitrary finite set \(M=\{m^{(1)},\dots,m^{(k)}\}\subset\mathbb{Z}^l\) satisfying that \(m^{(j)}\neq0\) for all \(j\), but there exist nonnegative integers \(n_1,\dots,n_k\) not all equal to zero such that \(n_1m^{(1)}+\dots+n_km^{(k)}=0\). A collection \((R_1,\dots,R_k)\) of \(k\) subsets of \(\Gamma\) is said to be sufficient with the pattern \(M\) (or \(M\)-sufficient) if for every finite set \(K\subset\Gamma\), there exists a collection \(\gamma=(\gamma_1,\dots,\gamma_l)\) of characters of \(G\) such that \(\langle m^{(j)},\gamma\rangle+K\subset R_j\) for \(j\in\{1,\dots,k\}\). Denote by \(\mathcal{P}=\mathcal{P}(G)\) the set of all trigonometric polynomials on \(G\). Also, fix the notation for the Fourier transform: \[ \widehat{f}(\gamma)=\int_Gf(x)\overline{\gamma(x)}\,dm(x),\quad f\in L^1(G),\,\gamma\in\Gamma. \] For every finite set \(E\subset\Gamma\), the corresponding operator of partial Fourier sums, \[ P_Ef=\sum_{\gamma\in E}\widehat{f}(\gamma)\gamma, \] is well defined at least on \(\mathcal{P}\). Let \(\mathcal{B}\) be a system of finite subsets of \(\Gamma\) with the property that for every finite set \(K\subset \Gamma\), there exists \(E\in\mathcal{B}\) with \(E\supset K\). Such a \(\mathcal{B}\) is called a summation basis. Denote by \(u(G,\mathcal{B})\) the completion of \(\mathcal{P}\) in the norm \[ \|f\|=\sup_{E\in\mathcal{B}}\|P_Ef\|_{\infty}. \] Also, denote by \(K_{\mathcal{B}}\) the union of all sets \(E\in\mathcal{B}\) such that both \(K\cap E\) and \(K\setminus E\) are nonempty. A summation basis \(\mathcal{B}\) and an \(M\)-sufficient collection \((R_1,\dots,R_k)\) of subsets of \(\Gamma\) are said to be (mutually) coordinated if for every finite set \(K\subset\Gamma\), the collection \((R_1\setminus K_{\mathcal{B}},\dots,R_k\setminus{\mathcal{B}})\) is sufficient with the same \(l\)-pattern \(M\). Denote also by \(u(G,\mathcal{B},R_j)\) the set of \(f\in u(G,\mathcal{B})\) such that \(\widehat{f}(\gamma)=0\) for \(\gamma\notin R_j.\) Finally, let \(U=U(G,\mathcal{B},R)\) be the set of all sums of the form \(h=f_1+\dots+f_k\), where \(f_j\in u(G,\mathcal{B},R_j),\,j=1,\dots,k\). Endow \(U\) with the norm \[ \|h\|_U=\inf\sum_{1\leq j\leq k}\|f_j\|_{u(G,\mathcal{B},R_j)}, \] where the infimum is taken over all representations of \(h\) as above. In this paper, the authors prove the following result: Assume that the system \(R\) is coordinated with the summation basis \(\mathcal{B}\). Then for every \(\varepsilon>0\) and every \(f\in L^{\infty}(G)\), there exists \(g\in U\) such that \(m\{f\neq g\}\leq\varepsilon\) and \(\|g\|_U\leq C\varepsilon^{-D}\|F\|_{\infty}\), where \(m\) is a normalized Haar measure. The constants \(C\) and \(D\) are independent of \(f\).