MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Let $A$ be a real symmetric matrix whose spectrum is $\lambda_1,\lambda_2,\ldots,\lambda_n$. Let $A'$ be the matrix obtained by adding a perturbation to $A$. The requirement is that only the second eigenvector of $A'$ is known and fixed in advance. Is there a way to obtain the set of all possible perturbations which will guarantee that $A'$ has the specified second eigenvector?

share|cite|improve this question
It is not too hard to characterize rank one perturbations which make a given vector an eigenvector. I assume what makes the problem difficult is to guarantee that this eigenvector is the "second" one. You did not state the criterium for ordering, however. – Michael Renardy May 31 '11 at 15:03
To clarify, can you confirm: $A'=A+\epsilon H$; $A$ and $H$ are symmetric and $H$ is rank-one, thus $H=u\otimes u$. The second eigenvector of $A'$ is prescribed. It is not necessarily the second eigenvector of $A$, though. Also, it seems implicitly assumed that the second eigenvalue (in decreasing order?) of $A'$ is simple. – Pietro Majer May 31 '11 at 18:22

The perturbation matrix $\Pi$ must have $v_2$ (your second eigenvector) as an eigenvector. The associated eigenvalue is either $0,$ or not. In the second case, $\Pi = c v_2 v_2^t$ (in other words, a multiple of the projection onto the space spanned by $v_2.$ In the first case, $\Pi = c w w^t,$ where $(w, v_2) = 0$ (in English, a multiple of a projection onto some one-dimensional subspace orthogonal to $v_2.$ As for the ordering of the eigenvalues (as mentioned by @Denis), this is very easy to check in the first case ($c < \lambda_1 - \lambda_2$), and slightly less easy in the second case (you need to compute the projection of $w$ onto $v_1$).

share|cite|improve this answer

@Pietro Majer,

A few clarifications in the question. I am interested in the case of the matrix being the Laplacian of some undirected weighted graph. Thus, A would now be real,symmetric and positive semi definite. Also A' = A + \epsilon H where H is rank-one. Importantly the second eigen-vector i meant was the one the corresponds to the second "smallest" eigen-value.


share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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