Let \(\delta \) be a linear mapping from a von Neumann algebra \(\mathcal {A}\) into itself. For \(a, b\in \mathcal {A},\) the following statements are equivalent: (1) \(ab=0\) implies \(b\delta (a)+\delta (b)a=0\) .
(2) \(ab^{*}=0\) implies \(b^{*}\delta (a)+\delta (b)^{*}a=0\) , and \(\delta (P)=0,\) where \(P\in \mathcal {A}\) is the central projection such that \(P\mathcal {A}\) contains no commutative direct summands.
(3) \(\delta (a)=\xi a\) , where \(\xi \) belongs to the type \(\textrm{I}_{1}\) direct summand of \(\mathcal {A}\) .
If \(\delta \) is continuous, a similar conclusion holds for a \(C^{*}\) -algebra \(\mathcal {A}\) mapping into \(\mathcal {A}^{**}\) , and for a locally compact group \(\mathcal {G}\) , it remains valid for the group algebra \(L^{1}(\mathcal {G})\) mapping into the measure convolution algebra \( M(\mathcal {G})\) .