3
$\begingroup$

I'm looking to diagonalize a matrix A seen below. (Find a $P$ and a $D$ such that $AP = PD$). $$ \begin{bmatrix} 1 & 2 & 3 & 4 & 5 & 6 \\ 1 & 2 & 3 & 4 & 5 & 6 \\ 1 & 2 & 3 & 4 & 5 & 6 \\ 1 & 2 & 3 & 4 & 5 & 6 \\ 1 & 2 & 3 & 4 & 5 & 6 \\ 1 & 2 & 3 & 4 & 5 & 6 \\ \end{bmatrix} $$

It doesn't seem practical to use determinants in this matrix to find the eigenvalues. I do realize that with row reduction I can limit to the matrix:

$$ \begin{bmatrix} 1 & 2 & 3 & 4 & 5 & 6 \\ 0 & 0 & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 & 0 \\ \end{bmatrix} $$

And further realize that the following vector is in the column space of A.

$$ \begin{bmatrix} 1 \\ 1 \\ 1 \\ 1 \\ 1 \\ 1 \\ \end{bmatrix} $$

By solving $Av_1=λv_1$ I get an eigenvalue of 21. But this is the only eigenvalue that I get and I believe it has a multiplicity of 1. Then is there a $λ=0$ with multiplicity 5? Further, how is it possible to find the matrix $P$ (specifically the eigenvectors that form a basis for the eigenspace associated with $λ=0$)?

$\endgroup$
1
$\begingroup$

You can immediately see that your matrix decomposes as a product $CR$ of a column matrix $C=(1~1~1~1~1~1)^\top$ and a row matrix $R=(1~2~3~4~5~6)$, so it has rank$~1$. For rank $1$ matrices, the analysis is easy: it has an eigenvalue $0$ with geometric multiplicity$~n-1$ (here $5$) and one more eigenvalue equal to the trace $t=RC$ of the matrix (here $t=21$); see this question and links from there for details. In case that $t=0$ the algebraic multiplicity of the eigenvalue$~0$ is strictly larger than its geometric multiplicity, and the matrix therefore non-diagonalisable; but if (as it is here) the trace is nonzero, then the matrix is diagonalisable.

The eigenspace for $0$ is the kernel of the matrix$~R$ (since vectors killed by $R$ are also killed by $CR$), which here is the subspace $\{(a~b~c~d~e~f)^\top\,\mid a+2b+3c+4d+5e+6f=0\,\}$. The eigenspace for the remaining eigenvalue$~t$ (if nonzero) is spanned by the column of $C$, since $CRC=Ct=tC$.

$\endgroup$
1
$\begingroup$

We need to find the eigenvectors $\mathbf{x}$ such that: $A\mathbf{x} =\mathbf{0}.$ Because the system is homogeneous, we can use the row echelon form of $A$, thus we need to solve the sytem: $$ \begin{bmatrix} 1 & 2 & 3 & 4 & 5 & 6 \\ 0 & 0 & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 & 0 \\ 0 & 0 & 0 & 0 & 0 & 0 \\ \end{bmatrix} \cdot \begin{bmatrix} x_1 \\ x_2 \\ x_3 \\ x_4 \\ x_5 \\x_ 6 \end{bmatrix} = \begin{bmatrix} 0 \\ 0 \\ 0 \\ 0 \\ 0 \\ 0\end{bmatrix},$$ Thus, we have the equation: $$x_1 + 2x_2 + 3x_3 + 4x_4 + 5x_5 + 6x_6 = 0,$$ which implies that we have $5$ free variables.

So, every eigenvector corresponding to the zero eigenvalue can be written in the form: $$x_2\cdot\begin{bmatrix} -2 \\ 1 \\ 0 \\ 0 \\0 \\0 \end{bmatrix} +x_3 \cdot \begin{bmatrix} -3 \\ 0 \\ 1 \\ 0 \\0 \\0 \end{bmatrix} +x_4\cdot \begin{bmatrix} -4 \\ 0 \\ 0 \\ 1 \\0 \\0 \end{bmatrix}+x_5\cdot\begin{bmatrix} -5 \\ 0 \\ 0 \\ 0 \\1 \\0 \end{bmatrix} +x_6\cdot \begin{bmatrix} -6 \\ 0 \\ 0 \\ 0 \\0 \\1 \end{bmatrix}.$$

The column vectors form an eigenbasis for the eigenspace that corresponds to the zero eigenvalue. So, we have the matrix $$P = \begin{bmatrix} 1 & -2 & -3 & -4 & -5 & -6\\ 1 & 1 & 0 & 0 & 0 & 0 \\ 1 & 0 & 1 & 0 & 0 & 0 \\ 1 & 0 & 0 & 1 & 0 & 0 \\ 1 & 0 & 0 & 0 & 1 & 0 \\ 1 & 0 & 0 & 0 & 0 & 1 \end{bmatrix}$$ and $D = \operatorname{diag}(21\quad 0\quad 0\quad 0\quad 0\quad 0 ).$


We can notice that matrix $A$ is a rank one matrix, so there is only one non-zero eigenvalue, which can be found by adding the diagonal elements of $A$.

$\endgroup$
  • $\begingroup$ Thank you - this does make a lot of sense. Where I seem to get most hung up here is calculating the first eigenvalue. Typically I'd find the determinant of (A-λI) and the associated characteristic polynomial which doesn't seem very doable here. This seems like a special case and I'm not sure how applicable it is for other diagonalization problems? I'm also unsure why we'd assume λ = 21 only has a multiplicity of 1 and λ = 0 has multiplicity 5? $\endgroup$ – econom May 16 '16 at 0:18
  • $\begingroup$ Yes - in a typical diagonalization problem I would find the characteristic polynomial (from determinant(A-λI)). That doesn't seem practical here because det(A-λI) would be a huge hassle to calculate by hand. I can get $λ_1 = 21$ from recognizing that $Av_1=λv_1$, but how do we know it only has multiplicity 1? And that $λ=0$ has multiplicity of 5? $\endgroup$ – econom May 16 '16 at 0:39
  • $\begingroup$ Very nice! Thanks for the trick. Again, appreciated with your help, @thanasissdr! $\endgroup$ – econom May 16 '16 at 1:20
  • $\begingroup$ @econom The matrix has rank $1$, so $0$ is an eigenvalue with multiplicity $5$, leaving the non-zero eigenvalue to have multiplicity one. Another was to see this is that the image is one-dimensional, so there’s at most a single line that’s mapped to itself by $A$, i.e., that the eigenspace corresponding to the non-zero eigenvalue is one-dimensional. $\endgroup$ – amd May 16 '16 at 3:15

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.