The adjoint alternative for matrix operators (Q1129523)

Consider \(n\times n\) matrices and let \(\text{adj}(A)\) denote the classical adjoint of \(A\). When \(B\) is given, the author classifies those \(A\) which satisfy \(\text{adj}(A)= B\). If \(\text{rank}(B)= n\), there are \(n-1\) solutions. If \(1< \text{rank}(B)< n\), there are no solutions. When \(B= 0\), if \(n\leq 2\), only \(A= 0\) is a solution, and if \(n>2\) then any \(A\) with \(\text{rank}(A)\leq n-2\) is a solution. The interesting case is when \(\text{rank}(B)= 1\). Although the author's Lemma 1 is false -- he doesn't need it -- the subsequent construction is still correct if \(c\) is taken to be a nonzero column of \(B\). The construction shows that any invertible \((n-1)\times (n-1)\) matrix \(A_1\) can be completed to an \(n\times n\) matrix \(A\) such that \(\text{adj}(A)= B\) is true. The remark in the last line that \(\text{adj}(A)\) could be \(0\) is also false.