1
$\begingroup$

Problem: Let \begin{align*} A = \begin{pmatrix} 1 & 1 & 1 & 1 \\ 1 & 1 & 1 & 1 \\ 1 & 1 & 1 & 1 \\ 1 & 1 & 1 & 1 \end{pmatrix}. \end{align*} Compute all the eigenvalues and eigenvectors of $A$.

Attempt at solution: I found the eigenvalues by computing the characteristic polynomial. This gives me \begin{align*} \det(A - x \mathbb{I}_4) = \det \begin{pmatrix} 1-x & 1 & 1 & 1 \\ 1 & 1-x & 1 & 1 \\ 1 & 1 & 1-x & 1 \\ 1 & 1 & 1 & 1-x \end{pmatrix} = -x^3 (x-4) = 0 \end{align*} after many steps. So the eigenvalues are $\lambda_1 = 0$ with multiplicity $3$ and $ \lambda_2 = 4$ with multiplicity $1$.

Now I was trying to figure out what the eigenvectors are corresponding to these eigenvalues. Per definition we have $Av = \lambda v$, where $v$ is an eigenvector with the corresponding eigenvalue. So I did for $\lambda_2 = 4$: \begin{align*} \begin{pmatrix} 1 & 1 & 1 & 1 \\ 1 & 1 & 1 & 1 \\ 1 & 1 & 1 & 1 \\ 1 & 1 & 1 & 1 \end{pmatrix} \begin{pmatrix} x_1 \\ x_2 \\ x_3 \\ x_4 \end{pmatrix} = 4 \begin{pmatrix} x_1 \\ x_2 \\ x_3 \\ x_4 \end{pmatrix} \end{align*} I think this is only possible when $x_1 = x_2 = x_3 = x_4 = 1$. So am I right in stating that all the eigenvectors corresponding to $\lambda_2$ are of the form $t \begin{pmatrix} 1 \\ 1 \\ 1 \\ 1 \end{pmatrix}$ with $t$ some number $\neq 0$? For $\lambda_1 = 0$, I'm not sure how to find the eigenvectors. The zero vector is never an eigenvector. This means $x_1, x_2, x_3$ and $x_4$ can be anything aslong as they add to zero?

$\endgroup$
  • $\begingroup$ you are absolutely right. $\endgroup$ – Zhanxiong Jul 13 '15 at 16:41
  • $\begingroup$ the eigenspace which is corresponding to the eigenvalue 0 is also called the kernel of the matrix - and you are right $\endgroup$ – user190080 Jul 13 '15 at 16:41
  • $\begingroup$ related $\endgroup$ – Omnomnomnom Jul 13 '15 at 18:01
1
$\begingroup$

You are right, and to show how this applies more generally, look at the eigenspace corresponding to the other eigenvalue, $\lambda = 0$. This has multiplicity 3, so we expect exactly 3 linearly independent vectors in this space. The main equation looks like $A \vec{x} = 0 \vec{x} = \vec{0}$, in other words, $$ \begin{pmatrix} 1 & 1 & 1 & 1 \\ 1 & 1 & 1 & 1 \\ 1 & 1 & 1 & 1 \\ 1 & 1 & 1 & 1 \end{pmatrix} \begin{pmatrix} x \\ y \\ z \\ w \end{pmatrix} = \begin{pmatrix} 0 \\ 0 \\ 0 \\ 0 \end{pmatrix} $$ It's not hard to see that we have 4 identical equations, so the only constraint is $x + y + z + w = 0$, so we define $w = -x-y-z$ and any vector in the eigenspace now looks like $$ \begin{pmatrix} x \\ y \\ z \\ w \end{pmatrix} = \begin{pmatrix} x \\ y \\ z \\ -x-y-z \end{pmatrix} = x \begin{pmatrix} 1 \\ 0 \\ 0 \\ -1 \end{pmatrix} + y \begin{pmatrix} 0 \\ 1 \\ 0 \\ -1 \end{pmatrix} + z \begin{pmatrix} 0 \\ 0 \\ 1 \\ -1 \end{pmatrix} $$ and the three vectors which form the basis of the eigenspace are now specified.

$\endgroup$
  • $\begingroup$ Thank you, that was very clear. So all the eigenvectors corresponding to the eigenvalue zero are of the form $ x \begin{pmatrix} 1 \\ 0 \\ 0 \\ -1 \end{pmatrix} + y \begin{pmatrix} 0 \\ 1 \\ 0 \\ -1 \end{pmatrix} + z \begin{pmatrix} 0 \\ 0 \\ 1 \\ -1 \end{pmatrix}$, with the $x,y$ and $z$ numbers $\neq 0$? $\endgroup$ – Kamil Jul 13 '15 at 17:09
  • $\begingroup$ @Kamil exactly :) $\endgroup$ – gt6989b Jul 13 '15 at 17:11
  • $\begingroup$ Worth noting that these are not the only three vectors which work. You can choose any three linearly independent such vectors. $\endgroup$ – Steven Gubkin Jul 13 '15 at 17:11
  • $\begingroup$ @StevenGubkin of course, once you are in a span, any 3 lin.indep. members will form a basis, but I wanted to show an easy way to construct one, which is untuitive and direct... $\endgroup$ – gt6989b Jul 13 '15 at 17:12
2
$\begingroup$

You are exactly right! $x_1,x_2,x_3$ and $x_4$ can be anything as long as they sum to zero. This is already a complete solution in some sense (you have "found" all of the eigenvectors). If you would like, you could find 3 linearly independent such vectors, and then you would be sure that these three span the whole space.

$\endgroup$
  • $\begingroup$ So there are infinite eigenvectors corresponding to $\lambda = 0$. How would I express this mathematically, I mean in set notation? The only condition is that they add to zero. $\endgroup$ – Kamil Jul 13 '15 at 16:47
  • $\begingroup$ @Kamil yes, there are infinite solutions (over an infinite field, of course). They are the solution set of the homogeneous system $\;\det(A-xI)=0\;$ , and in this case this space has dimension three. $\endgroup$ – Timbuc Jul 13 '15 at 16:53
1
$\begingroup$

the matrix you have has rank one and can be written as $$A = uu^\top \text{ where } u = \pmatrix{1,1,1,1}^\top.$$ now, $$Av = uu^\top v=(u^\top v)u.$$ note that $u^\top v$ is a scalar, therefore $A$ has eigenvalue $0$ of multiplicity $3$ corresponding eigenvector any vector orthogonal to $u$ which has dimension $3.$ $A$ has also the nonzero eigenvalue $u^\top u = 4$ and the corresponding eigenvector $u.$

the same ideas can be used to find the eigenvalues and eigenvectors of any rank one matrix $ab^\top.$

$\endgroup$

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.