About a forty years old conjecture by P. Erdős (Q4452883)

\textit{P. Erdős} [Ann. Math. (2) 47, 1--20 (1946; Zbl 0061.07902)] proved that if a real-valued additive function \(f\) satisfies \(f(n+1)-f(n)\to 0\) as \(n\to+\infty\), then \(f(n)=c\,\log n\) with \(c=f(2)/\log 2\). He made some conjectures when the conditions are weakened, including: Conjecture E: If \(f\) is additive and \(f(n+1)-f(n)\to 0\) as \(n\to+\infty\), on a sequence of density \(1\), then \(f(n)=c\,\log n\). This has not been established or disproved. The wide ranging survey reviewed here describes various extensions of Conjecture E which depend on different interpretations of the conditions, and outlines progress made on them. The author concentrates on the modification that (1) \(f(u_{i+1})-f(u_i)\to 0\) as \(i\to+\infty\) on an increasing sequence \(U\) of positive integers with density \(1\), and considers different densities that could be used, starting with the asymptotic density. In [Atti Accad. Sci. Torino, Cl. Sci. Fis. Mat. Nat. 111, 503--510 (1977; Zbl 0379.10034)], the author deduced that if \(f\) is additive, (1) holds with \(U\) having asymptotic density \(1\), and in addition (2) \(u_{i+1}-u_i=O(1)\), then \(f(n)=c\,\log n\). If instead \(U\) is assumed to have Banach-Buck measure \(1\), then it follows that \(U\) has asymptotic density \(1\) and that (2) holds so the conclusion is the same; see the author's paper [Rend. Mat. (6) 10 (1977), 607--616 (1978; Zbl 0379.10035)] in which weaker conditions are investigated. The survey goes on to discuss necessary and sufficient conditions under which a real-valued function \(f\) defined on a suitable subset \(W\) of \(\mathbb{N}\) has the form \(f(n)=c\,\log n+O(1)\) for \(n\in W\). In the final section the author describes a variant of the problem involving an asymptotically additive function \(f\), i.e. one satisfying \(f(mn)-f(m)-f(n)\to 0\) as \(m,n\to+\infty\), such as \(f(n)=\log (n+1)\).