Maybe context would help. We're trying to find a-priori estimates for the Dirichlet problem. For $u \in C_c^\infty(\Omega)$ we have
$$|\Delta u|^2 = \sum_{i,j} (\partial_i(\partial_i u \partial_j^2 u) - \partial_i u \partial_j^2 \partial_i u)= \sum_{i,j} (\partial_i(\partial_i u \partial_j^2 u) - \partial_j(\partial_i u \partial_i \partial_j u)) + \sum_{i,j} |\partial_i \partial_j u|^2,$$
but the last term is equal to $|\nabla^2 u|^2$. If this was equal to the LHS, then the first term on the RHS would be zero but I don't think that is necessarily the case? The lecture notes call that term t…