A pathological counterexample for hypothesis about non-Fredholm property for algebras of weighted translation operators (Q1921699)

Let \((\Omega, \Sigma, \mu)\) be a standard measure space with atomless measure \(\mu\). Denote by \(L^p = L^p(\Omega,\Sigma, \mu)\) the Lebesgue spaces, \(1 \leq p \leq \infty\). An operator \(S : L^p \to L^p\) is called a weighted translation operator if \[ (Sf)(\omega) = a (\omega) f(h(\omega)) \;\text{for} \omega \in \Omega, \quad f \in L^p, \] for some weight function \(a : \Omega \to \mathbb R\) and some translation \(h : \Omega \to \Omega\). Denote by \(AWS(L^p)\) the subalgebra of the Banach algebra \(B(L^p)\) of all linear bounded operators generated by the weighted translation operators. The author considers the question: Is any Fredholm operator in \(AWS(L^p)\) invertible in \(B(L^p)\)? Due to \textit{L. W. Weis} [North Holland Math. Stud. 90, 95-115 (1984; Zbl 0549.47016)] the answer is yes, if \(p=1\) and \(p=\infty\). The author proves that for \(1 < p < \infty\) there is always a non-invertible Fredholm operator in \(AWS(L^p)\).