Eigenvalues with constraints?

115 Views Asked by At

Note: This is a short version of https://math.stackexchange.com/questions/979834/

For a $n$-dimensional symmetric matrix A, orthogonal matrix C exists so that the relationship $$AC = CD$$ is fulfilled, where $D$ is a diagonal matrix. For $D=\delta_{ij}d_j$ and $C=\begin{pmatrix}c_1 & \cdots & c_n\end{pmatrix}$ (i.e. $c_j$ are the column vectors of $C$ the eigenvalue equation $$Ac_j = d_jc_j \Rightarrow (A-d_jE)c_j = 0$$ results, which are $n$ equations ($E$ is the identity matrix).

Now, if a constraint is given to $D$ (i.e. the eigenvalues) such as $$D = \begin{pmatrix}m_1 d_1 & & \\ & \ddots & \\ & & m_n d_n\end{pmatrix}$$ it should result in $n$ equations with $$(A-m_1d_1E)c_1 = 0$$ $$ \vdots$$ $$(A-m_n d_n E)c_n = 0$$

Instead, my textbook suggests $$(A-d_j M)c_j = 0$$ where $M$ is a diagonal matrix with: $$ M = \begin{pmatrix}m_1 & & \\ & \ddots & \\ & & m_n \end{pmatrix}$$

Which one is correct and why?