# Find all Complex 2x2 matrices A

1. Oct 22, 2009

### beetle2

1. The problem statement, all variables and given/known data

Find all Complex 2x2 matrices A such that A2 = A

2. Relevant equations
$$\left( \begin{array}{cc} a & b \\ c & d \\ \end{array} \right)^2$ = A$
3. The attempt at a solution

A2=A

I need to solve the equations

$a^2+bc=a$
$(a+d)b =b$
$(a+d)c = c\$/extract_itex] $d^2 + bc = d\\$ What is the best way to do this? Should I use some kind of rearrangement like using equation 1 and 4 we get $a^2-d^2=a-d$ any help greatly appreciated 2. Oct 22, 2009 ### LCKurtz I would think about two cases, that where A-1 exists and where it doesn't exist. 3. Oct 23, 2009 ### beetle2 Are you talking about when A is an orthogonal matrix so A-1 = AT? Last edited: Oct 23, 2009 4. Oct 23, 2009 ### Office_Shredder Staff Emeritus The question is: Is A invertible? If so, then just take A-1 of both sides and what do you get? 5. Oct 23, 2009 ### beetle2 A-1 exist when $ad-cb \neq 0$ so case 1 $ad-cb \neq 0$ case 2 $ad-cb = 0$ 6. Oct 23, 2009 ### Office_Shredder Staff Emeritus Stop looking at the series of equations you have If A2=A And A-1 exists In one step you can find A 7. Oct 23, 2009 ### beetle2 If A is invertable and A2=A AA=A A-1AA=A-1A IA=I So A = I 8. Oct 23, 2009 ### Office_Shredder Staff Emeritus Now if A is not invertible, what can you say about how A acts on im(A), its image? 9. Oct 23, 2009 ### HallsofIvy Think about the fact that if $A^2= A$ then $A^2- A= A(A- I)= 0$ 10. Oct 25, 2009 ### beetle2 Would that mean that the only way that $A^2- A= A(A- I)= 0$ is that either A = the trivial solution. $\[ \left( \begin{array}{cc}0 & 0 \\0 & 0 \\ \end{array} \right)$ = A$
or

$$\left( \begin{array}{cc}-1 & 0 \\0 & -1 \\ \end{array} \right)$ = A$

regards

Last edited: Oct 25, 2009
11. Oct 25, 2009

### Dick

That's completely wrong thinking. If A and B are matrices AB=0 does NOT mean A=0 or B=0. And [[-1,0],[0,-1]] doesn't work at all. Here's some other ones that do [[1,0],[0,0]], [[1,1],[0,0]], [[1/2,1/2],[1/2,1/2]]. There's LOTS of them. You already followed up on the excellent hint that if A is invertible, then A=I is the only solution. Now suppose A is not invertible. Then det(A)=0, so you can add the condition ad-bc=0 to your original list of equations. That makes it easier. Now go to those equations and start working on cases. You should be able to write down a two parameter family of solutions that satisfy A^2=A besides the zero matrix.

12. Oct 25, 2009

### Hurkyl

Staff Emeritus
For the record, responders have been suggesting two different methods of approach. Office_Shredder's is a geometric one, and offers more insight into why idempotent matrices are interesting. The other was more algebraic.

I confess I would have suggested yet a different approach, one that starts by looking at the equation (a+d)b=b.

13. Oct 25, 2009

### Dick

Why is that different from the algebraic approach? I'm just saying that the equations the OP originally posted are a lot easier to solve if you restrict to the case of noninvertible matrices, in which case you can put ad=bc. LCKurtz already suggested this, and it's a very good suggestion.

Last edited: Oct 25, 2009
14. Oct 25, 2009

### beetle2

As idempotent matrices must eigenvalues of 1 and 0 would it be all the Matrices that can produce the diagonal matrix

$$\left( \begin{array}{cc}1 & 0 \\0 & \\ \end{array} \right)$ = A$

15. Oct 25, 2009

### Dick

No, I already pointed out that [[1/2,1/2],[1/2,1/2]] works.

16. Oct 25, 2009

### Hurkyl

Staff Emeritus
Figures you'd go for yet another solution! There are (at least) three things you need to do to turn this into a complete solution and a proof.

(2) For the diagonalizable solutions, you haven't covered all possible diagonal forms.

(3) You still need to simplify your characterization of the solutions into a form that would be considered "found".

17. Oct 30, 2009

### beetle2

So I have the first case where the matreix is invertable

If A is invertable and A^2=A
AA=A
A-1AA=A-1A
IA=I

So A = I

The other case is when A is not invertable so then the det(A) = 0

A^2 = A
AA = A
AA - A = A - A
AA - A = 0
A(I-A) = 0

SO what complex matrix will do this?

18. Oct 30, 2009

### Dick

I already told you in post 10 that that factorization isn't particularly useful. Go back to your original set of equation and add the condition det(A)=0=ad-bc. You'll find they are pretty easy to solve.

19. Nov 4, 2009

### beetle2

For det(A)= 1 = invertable

$$\left( \begin{array}{cc}1 & 0 \\0 & 1 \\ \end{array} \right)$$

for det(A) = 0 = non invertable

b or c can be any value

$$\left( \begin{array}{cc}1 & b \\0 & 0 \\ \end{array} \right)$$

$$\left( \begin{array}{cc}1 & 0 \\c & 0 \\ \end{array} \right)$$

$$\left( \begin{array}{cc}0 & b \\0 & 1 \\ \end{array} \right)$$

$$\left( \begin{array}{cc}0 & 0 \\c & 0 \\ \end{array} \right)$$

$$\left( \begin{array}{cc}0 & 0 \\0 & 0 \\ \end{array} \right)$$

also you have
for all square matrices nXn

$$\left( \begin{array}{cc}\frac{1}{n} & \frac{1}{n} \\\frac{1}{n} & \frac{1}{n} \\ \end{array} \right)$$

I'm not sure how to get that last one. I just noticed that DIck posted that one and by trying different options noticed that it worked for all nXn matrices

20. Nov 4, 2009

### Dick

Those are all one parameter special cases. At least the ones that actually work are. The really general solution has two free parameters b and c, both in the same matrix. Can you find it? It includes the [[1/2,1/2],[1/2,1/2]] case. I keep telling you to go back to your original equations and add ad=bc.