# Thread: Diagonalizing matrix problem with parameters

1. ## Diagonalizing matrix problem with parameters

Hi,
I have this problem I'm stuck on:
A is a 3x3 matrix
A=
a,b,c are real numbers
1) determine for what values of a,b,c the matrix A is diagnoizable. (advice: distinguish between the cases a=c and a=\=c)

My attempt at a solution has been finding det(tI-A)=0 , the characteristic polynomial is (t-c)(t-a)^2 and therefore the eigenvalues are a and c.

for t=c we get the matrix let it be C. for Cx=0 we get the system of linear equations:
(c-a)x1-x3=0
(c-a)x2 -bx3=0
likewise, for t=a we get
-1x3=0
-bx3=0
(a-c)x3=0

Now, when I tried to find the eigenspaces that correspond to a and c to find out their dimensions I got pretty lost.
I have no idea what to do. help would be very welcomed!

2. ## Re: Diagonalizing matrix problem with parameters

Case 1. $a\neq c.$ Then, $\lambda=a$ (double), $\lambda=c$ (simple). By a well known theorem, $\dim V_c=1=\mbox{ multiplic }(c).$ On the other hand:

$\dim V_a=3-\mbox{rank}\begin{bmatrix}{0}&{0}&{1}\\{0}&{0}&{b} \\{0}&{0}&{c-a}\end{bmatrix}=3-1=2=\mbox{ multiplic }(a).$ That is, $A$ is diagonalizable.

Case 2. $a=c.$ Then, $\lambda=a$ (triple). So,

$\dim V_a=3-\mbox{rank}\begin{bmatrix}{0}&{0}&{1}\\{0}&{0}&{b} \\{0}&{0}&{0}\end{bmatrix}=3-1=2\neq \mbox{ multiplic }(a).$ In this case, $A$ is not diagonalizable.

3. ## Re: Diagonalizing matrix problem with parameters

Got it. thank you!

4. ## Re: Diagonalizing matrix problem with parameters

we can try to proceed with your original approach, and "see what happens". i often do this "just out of curiosity".

it's clear that you found the characteristic equation properly, it's (t - c)(t - a)2.

now let's find some eigenvectors. first let's tackle c:

$(A - cI)x = 0 \iff \begin{bmatrix}a-c&0&1\\0&a-c&b\\0&0&0 \end{bmatrix} \begin{bmatrix}x_1\\x_2\\x_3 \end{bmatrix} = \begin{bmatrix}0\\0\\0 \end{bmatrix}$

it's clear that x3 is "free", and provided a ≠ c:

$x_2 = \frac{b}{c-a}x_3$

$x_1 = \frac{1}{c-a}x_3$

which means that any eigenvector (belonging to c) is a scalar multiple of (1,b,c-a) (we have a one-dimensional eigenspace).

(note that if we DID have a = c, we we get instead is: x3 = 0. more on this later).

next we look at a:

$\begin{bmatrix}0&0&1\\0&0&b\\0&0&c-a \end{bmatrix} \begin{bmatrix}x_1\\x_2\\x_3 \end{bmatrix} = \begin{bmatrix}0\\0\\0 \end{bmatrix}$

gives:

x3 = 0
bx3 = 0 (and thus x3 = 0, this just duplicates the first equation)
(c-a)x3 = 0 (another duplication).

thus we are free to choose x1 and x2 independently, so we can pick (1,0,0) and (0,1,0) as linearly independent eigenvectors.

so if {(1,0,0),(0,1,0),(1,b,c-a)} are linearly independent, A is diagonalizable (we have an eigenbasis). it should be clear that this will be true if and only if c -a ≠ 0, that is: if a ≠ c.

on the other hand, if a = c, we only get an eigenspace of dimension 2: namely {(x,y,0) in R3}, so there is no eigenbasis, and A is not diagonalizable.