Suppose $A=uv^T$ where $u$ and $v$ are non-zero column vectors in $\mathbb{R}^n, n\geq 3$. $\lambda=0$ is an eigenvalue of $A$ since $A$ is not of full rank. $\lambda=v^Tu$ is also an eigenvalue of $A$ since $Au=(uv^T)u=u(v^Tu)=(v^Tu)u$. Here is my questions:

(1) Are there any other eigenvalues of $A$ ?

(2) If I have another matrix $B_n$ such that $B_n = F^{*}_n\Lambda_nF_n$, and $\Lambda_n$ are diagonal matrices holding the eigenvalues of $B_n$, $F_n$ is the unitary matrix (such as the discrete Fourier-matrix). Then I want to ask:

Can sum matrix $C = B_n + A$ be diagonalized by the specified unitary matrix (In fact, $B_n$ can be the $n\times n$ circulant matrix)?


On your first question: Of course, $A$ has other eigenvalues. It has $n$ eigenvalues, to be specific, counting multiplicity. You have identified one. For a rank-one matrix, the other eigenvalues are all zero.

On your second question: I don't know whether it is possible in general, but it is possible if $u$ is also an eigenvector of $B$ since then the eigenvectors of $C$ are the same as those of $B$ and only the eigenvalues change. If $u$ lies in a subspace spanned by only a few of the columns of the matrix $F$ (i.e., of only a few of the eigenvectors), then the diagonalization procedure will be fairly simple since you only need to re-diagonalize those eigenvectors in this subspace, not all of them.

  • $\begingroup$ In fact, one of both $u$ and $v$ has only last non-zero entry, such as $v=(0,...,0,x)^T,\ x\neq 0$, $x$ is the given vector. If I hold this condition, what can we obtain ? $\endgroup$ – Hsien-Ming Ku Oct 6 '14 at 21:10
  • $\begingroup$ Sorry, $x$ is the given number. @Wolfgang $\endgroup$ – Hsien-Ming Ku Oct 6 '14 at 21:18
  • 2
    $\begingroup$ The eigenvector matrix $F_n$ is unchanged iff $u$ and $v$ are both eigenvectors of $B_n$ with the same eigenvalue. Indeed, if $C=F_n^*\Lambda F_n+uv^*$, then $C=F_n^*(D+\hat{u}\hat{v}^*)F_n$ for suitable $\hat{u}$ and $\hat{v}$, so we are asking essentially when $D+\hat{u}\hat{v}^*$ is diagonal. This happens iff $\hat{u}\hat{v}^*$ has a single nonzero entry on the diagonal, that is, when both $\hat{u}$ and $\hat{v}$ are multiple of the same vector of the canonical basis. $\endgroup$ – Federico Poloni Oct 6 '14 at 22:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.