It follows that we can define a new matrix \hat{A}\hat{Q} = \hat{Q}\hat{U}\hat{Q}, and multiply from the left with \hat{Q}^{-1} we get \hat{Q}^{-1}\hat{A}\hat{Q} = \hat{B}=\hat{U}\hat{Q}, where the matrix \hat{B} is a similarity transformation of \hat{A} and has the same eigenvalues as \hat{B} .