1
$\begingroup$

Let $-3$ be an eigenvalue of a $3\times3$ singular matrix $P$ and $$P\begin{bmatrix} 5\\ 3\\ -2 \end{bmatrix}=\begin{bmatrix} -20\\ -12\\ 8 \end{bmatrix}.$$ Then find whether $P$ is diagonalizable or not. Also find whether $P^2+3P$ is diagonalizable.

My attempt

Given $-3$ is one eigenvalue and from $$P\begin{bmatrix} 5\\ 3\\ -2 \end{bmatrix}=\begin{bmatrix} -20\\ -12\\ 8 \end{bmatrix},$$ we know $Ax=\lambda x$. So we get $-4$ is another eigenvalue . Also $$\begin{bmatrix} 5\\ 3\\ -2 \end{bmatrix}$$ is an eigenvector corresponding to the eigen value $-4$. I am struck now don't know how to find another eigenvalue and how to proceed further.

This was the question asked for NET exam for PhD entrance in India

$\endgroup$
2
$\begingroup$

"Singular" means that $0$ is a third eigenvalue. With three distinct eigenvalues $0,-3,-4$, $P$ must be diagonalizable.

$P^2 + 3P$ is also diagonalizable with respect to the same basis as $P$.

$\endgroup$
4
  • $\begingroup$ if a matrix have distict eigen values , it is diagonalizable? this is what you said right? @user165670 $\endgroup$ – srishti Jul 24 '14 at 6:15
  • $\begingroup$ @srishti Yes, an $(n \times n)$-matrix with $n$ distinct eigenvalues is always diagonalizable. $\endgroup$ – user165670 Jul 24 '14 at 6:17
  • $\begingroup$ what if it has 1,2,2 as eigen values? it may or may not diagonalize.am i right? $\endgroup$ – srishti Jul 24 '14 at 6:31
  • $\begingroup$ @srishti That's right $\endgroup$ – user165670 Jul 24 '14 at 6:34
0
$\begingroup$

A singular matrix is one which is not full rank, meaning that there is at least one row (or column) that is a linear combination of the others, since a matrix represents a linear basis-space.

If one of the eigenvalues has been given, then as shown, you can find a second eigenvalue of 4. Since the first two "discovered" eigenvalues are non-zero and the matrix is singular, thus not full-rank, then there must be a zero-valued eigenvalue.

You can also understand that a singular matrix has a determinant of 0. Since the determinant of a square matrix is equal to the product of its eigenvalues, the only way to have a zero-valued determinant is for at least one of those eigenvalues to be 0.

Once you've diagonalized a matrix, then you can now use the features of matrix calculus by applying functions of a matrix instead to its eigenmatrix.

$$f\left( \underline{\underline{A}} \right) = \underline{\underline{Q}} f\left( \underline{\underline{\Lambda}} \right) \underline{\underline{Q}}^{-1} $$

Since the function is being applied to the diagonal matrix, it only affects the diagonal values, thus the result is still a diagonalizable matrix. This is a more practical view of understanding that the operations being applied to a diagonalizable matrix are equivalent to simply operating upon the underlying space's bases.

$\endgroup$
2
  • $\begingroup$ It should be: "each of its rows and columns is a linear combinations of the others" $\endgroup$ – enzotib Jul 24 '14 at 8:12
  • $\begingroup$ @enzotib Thank you, I updated the first sentence to be clearer about that. $\endgroup$ – user164367 Jul 24 '14 at 8:15
0
$\begingroup$

To diagonalize a matrix, only "distinct eigen values" is not enough... We must consider the field over which, we are going to diagonalize. You can easily get an example of a matrix, which has distinct eigen values but "not diagonalizable over Real numbers, But diagonalizable over complex numbers".

Also there is a necessary and sufficient condition for diagonalizable, which is based on eigen vectors, not on eigen values... You can get that from any good book on linear algebra.

$\endgroup$
2
  • $\begingroup$ can you suggest me a book? i couldn't find these topics anywhere $\endgroup$ – srishti Jul 24 '14 at 8:59
  • $\begingroup$ I studied that condition in this book: Linear Algebra by Jin Ho Kwak and Sungpyo Hong $\endgroup$ – Saravanan Jul 24 '14 at 9:19

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.