First let me make two statements to give my question the proper context.

  1. Consider $D$ a diagonal matrix and $u v^T$ a rank 1 matrix. From my current knowledge there exist computationally cheap algorithms to diagonalize $D+uv^T$ as discussed here: maximum eigenvalue of a diagonal plus rank-one matrix

  2. Diagonalizing a matrix of the form $Duv^T$ is straightforward, see here: Eigenvector of a diagonal matrix times a rank-1 matrix

Now to my question(s):

  1. I want to diagonalize $D^2+Duv^T$. Is $Duv^T$ still a rank 1 matrix? I considered some examples but I am not really sure if this is a general statement.

  2. This question depends on the answer of the first. If the statement is true the algorithms in statement 1 should be applicable to diagonalization of $D^2+Duv^T$, no? If it is not true are there others ways to tackle this?


Let $u=v=(1,0)^T$, and




And $Duv^T=0$.

Of course, if $D$ has only nonzero diagonal elements, then it has full rank and the product has rank $1$, see matrix rank properties on Wikipedia. And if $Duv^T$ has not rank $1$, it's necessarily the null matrix, so it's not a very interesting case.

  • $\begingroup$ Ok thanks, so its actually very easy to write a proof from these properties... let D be of full rank ($rank(D)=n$) and use sylvester rank inequality: $1 = rank(uv^T) = rank(D)+rank(uv^T)-n \le rank(Duv^T) \le min(rank(D),rank(uv^T)) =1 $ $\endgroup$ – J. Goe Apr 6 '16 at 14:02

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.