11
$\begingroup$

I have been examining a problem dealing with finding the rank of a $n \times n $ matrix $M$ as follows:

\begin{bmatrix} 0&1&4&9&16&\cdots &(n-1)^2\\ 1&0&1&4&9&\cdots&(n-2)^2\\ 4&1&0&1&4&\cdots&(n-3)^2\\ 9&4&1&0&1&\cdots&(n-4)^2\\ 16&9&4&1&0&\cdots&(n-5)^2\\ \vdots&\vdots&\vdots&\vdots&\vdots&\ddots&\vdots\\ (n-1)^2&(n-2)^2&\cdots&(n-4)^2&\cdots&\cdots&0 \end{bmatrix}

That is, the matrix with $k^2$ on its $k^{th}$ super and subdiagonals. One of my colleagues claims the matrix is of rank 3. I have tried to factor this matrix somehow, but have not really gotten very far.

Can we develop a proof of the claim that the rank of $M$ is indeed 3 (assume that $n\geq3$).

$\endgroup$
  • 5
    $\begingroup$ If you down vote, at least indicate why. What do you find objectionable about the question? That way I can fix it. $\endgroup$ – Vladhagen Feb 14 '14 at 20:46
7
$\begingroup$

When $n\ge4$, let $P=\pmatrix{1&-1\\ &\ddots&\ddots\\ &&\ddots&-1\\ &&&1}^3 =\pmatrix{1&-3&3&-1\\ &\ddots&\ddots&\ddots&\ddots\\ &&\ddots&\ddots&\ddots&-1\\ &&&\ddots&\ddots&3\\ &&&&\ddots&-3\\ &&&&&1}$. Then $P^TMP=\pmatrix{0&1&1\\ 1&-6&1\\ 1&1&0}\oplus 0_{(n-3)\times(n-3)}$. Hence $M$ has rank 3.


Remark.

Knowing that $M$ has rank $3$ allows us to find the eigenvalues of $M$ relatively easily. One can readily verify that $(1-n,3-n,\ldots,n-3,n-1)^T$ is an eigenvector of $M$ corresponding to the eigenvalue $-\frac{n(n^2-1)}{6}$. Since $M$ has zero trace, if we also know $\operatorname{tr}(M^2)$, the other two nonzero eigenvalues can be found. It turns out that $$ \operatorname{tr}(M^2)=2\sum_{i=1}^{n-1} i(n-i)^4 =\frac1{30}n^2(n^2-1)(2n^2-3) $$ (see also OEIS A101089) and the other two nonzero eigenvalues of $M$ are $$ \frac{n(n^2-1)}{12}\pm\frac{n}{2}\sqrt{\frac{(n^2-1)(3n^2-7)}{60}}. $$

$\endgroup$
  • 5
    $\begingroup$ How on Earth did you come up with this? (+1 BTW) $\endgroup$ – John Feb 14 '14 at 20:50
  • 4
    $\begingroup$ @John The answer was obtained by observing that the third-order Newton difference of the sequence $0,1,4,9,16,\ldots$ is zero. $\endgroup$ – user1551 Feb 14 '14 at 20:54
  • $\begingroup$ A very elegant solution. $\endgroup$ – John Feb 14 '14 at 21:02
  • 3
    $\begingroup$ I see no way that someone could top this. Most excellent. $\endgroup$ – Vladhagen Feb 14 '14 at 21:18

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.