What is the necessary and sufficient condition (if there is any) that $n$ orthonormal vectors $v_1,v_2,\cdots,v_n$ are eigenvectors of a Euclidean distance matrix. When $n=2$, the orthonormal vectors are easily charcterized, i.e., $(1/\sqrt{2}, 1/\sqrt{2})$ and $(1/\sqrt{2}, -1/\sqrt{2})$.


This isn't an answer, but it's too long for a comment. As you're maybe aware, a real $n \times n$ matrix $M$ is a Euclidean distance matrix if and only if the following conditions hold:

  • $M_{ij} \geq 0$ for all $i, j$
  • $M_{ii} = 0$ for all $i$
  • $M$ is symmetric
  • $M$ is conditionally negative definite, that is, $$ x^t M x \leq 0 $$ whenever $x \in \mathbb{R}^n$ with $\sum_i x_i = 0$.

This was shown in: I. J. Schoenberg, Metric spaces and positive definite functions, Transactions of the AMS 44 (1938), 522-536.

  • $\begingroup$ +1, Thanks, I know this fact. Do you think if it is possible to characterize the orthogonal matrix $P$ such that $P^TMP$ is diagonal? $\endgroup$
    – Sunni
    Aug 8 '11 at 22:56
  • $\begingroup$ OK, sorry not to have added anything new. (Personally I find this fact of Schoenberg's rather surprising, in that the triangle inequality doesn't have to be mentioned.) I'm afraid I don't have any further thoughts. $\endgroup$ Aug 9 '11 at 0:22

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.