Complex Eigenvalues

Today we consider how to deal with complex eigenvalues in a linear homogeneous system of rst order equations. We will also look back briey at how what we have done with systems recapitulates what we did with second order equations. 1. Complex Eigenvalues 2. Second Order Equations as Systems

Complex Eigenvalues

We know that to solve a system of n equations (written in matrix form as x = Ax), we must nd n linearly independent solutions x1 , . . . , xn . In the case where A has n real and distinct eigenvalues, we have already solved the system by using the solutions ei t vi , where i and vi are the eigenvalues and eigenvectors of A. We now consider the case where A has complex eigenvalues. We will assume that A has only real entries. Then the characteristic polynomial |A rI | has real coecients, and therefore any eigenvalues occur in conjugate pairs: r = a + bi and r = a bi Only slightly more surprising is the fact that the eigenvectors also occur in conjugate pairs. For example, suppose we have eigenvalue r with eigenvector v. Then they satisfy the equation (A rI )v = 0 Now if we take the complex conjugate of both sides, and note that both A and I have only real entries, we get (A rI )v = 0

Therefore, an eigenvector associated with r is v! If we have a solution ert v, we also have its conjugate ert v, and this means that we also have its real and imaginary parts, since i 1 Re(x) = (x + x) and Im(x) = (x x) 2 2

Now let us write the eigenvector split into real and imaginary parts, as v = a + bi (Note that a and b are real vectors.) If we also write our eigenvalues with real and imaginary parts as r = + i, then one solution can be rewritten as follows: (a + bi)e(+i)t = (a + bi)et (cos(t) + i sin(t)) = et (a cos(t) b sin(t)) + iet (a sin(t) + b cos(t)) Of course, we also have the complex conjugate of this solution. Therefore, we can get both the real and imaginary parts as solutions. So we have found two real solutions: u(t) = et (a cos(t) b sin(t)) and v(t) = et (a sin(t) + b cos(t)) Example: Solve the system x = 6 13 1 0 x.

First we nd the eigenvalues of the matrix A in x = Ax:

6 13 = (6 )() + 13 = 2 6 + 13 = 0 1 Solving for yields = 6 36 52 6 4i = = 3 2i. 2 2

We only need to nd the eigenvector associated with one of these eigenvalues. Lets nd the eigenvector for = 3 + 2i by solving (A I )v = 0. We row-reduce the augmented matrix 0 6 (3 + 2i) 13 1 (3 + 2i) 0 2 or 0 3 2i 13 1 3 2i 0 .

A useful trick to convert a complex value into a real value is to multiply by the complex conjugate, so to get rid of the complex number in the rst column of row one, let us multiply by the conjugate 3 + 2i. Then (3 2i)(3 + 2i) = 9 4i2 = 9 + 4 = 13, and we get 3 2i 13 0 1 3 2i 0 13 39 26i 0 1 3 2i 0 1 3 2i 0 1 3 2i 0

after also dividing through by 13 on row one. Then we can subtract row one from row two, and we end the row reduction with: 1 3 2i 0 0 0 0 We note that we now have v1 + (3 2i)v2 = 0 in the rst row, and nothing in the second row. (Note that as expected, we have eliminated at least one row in solving for our eigenvectors.) So we have v1 = (3 + 2i)v2 , and v2 is a free variable. Lets assign v2 = 1, and then we have the eigenvalue/eigenvector pair = 3 + 2i, So we get a solution of the form e(3+2i)t 3 + 2i 1 = e3t e2it 3 + 2i 1 = e3t (cos(2t) + i sin(2t)) 3 + 2i 1 and 3 + 2i 1 .

(Remember: eit = ) Multiplying through and separating into real and imaginary parts yields 3e3t cos(2t) 2e3t sin(2t) + i [3e3t sin(2t) + 2e3t cos(2t)] e3t cos(2t) + ie3t sin(2t) 3e3t cos(2t) 2e3t sin(2t) e3t cos(2t)

+i

3e3t sin(2t) + 2e3t cos(2t) e3t sin(2t)

We know that the real and imaginary parts are both solutions, so our general solution is 3e3t cos(2t) 2e3t sin(2t) 3e3t sin(2t) + 2e3t cos(2t) c1 + c2 3t e cos(2t) e3t sin(2t) 3

+ i

If we wish to set an initial condition, such as x(0) = c2 : c1 3e0 cos(0) 2e0 sin(0) e0 cos(0) + c2

5 3

, we can solve for c1 and

3e0 sin(0) + 2e0 cos(0) e0 sin(0) c1 3 1 + c2 2 0

= = 5 3

which gives us the following augmented matrix: 3 2 5 1 0 3 Row reduction leads to 1 0 3 0 1 2 .

so c1 = 3 and c2 = 2 are the required constants. Example: Solve the system x = Eigenvalues: 1 2 1 3 x, x(0) = 2 1

Eigenvectors:

General Solution: x= Solving the initial condition: (x(0) = 2 1 )

Solution: 4

x= or x1 (t) = x2 (t) =

Second Order Equations as Systems

We know that any order n equation can be converted to a system of n rst order equations. Lets see what happens when we use this approach to solve a second order equation. Example: Solve y + 2y 3y = 0. We know the characteristic equation is r 2 + 2r 3 = 0, which has roots r = 3 and r = 1. Thus we know the general solution is y (t) = c1 e3t + c2 et If we rst convert to a system, we set x1 = y , x2 = y , and get the following: x1 (t) = x2 x2 (t) = 3x1 2x2 or x = We nd our eigenvalues: 1 = (2 ) + 2 = 2 + 2 3 = 0 3 2 Thus = 3 and = 1 are the eigenvalues. 5 0 1 3 2 x

We nd our eigenvectors: 3 1 3 1 x=0

We have the single relation 3x1 = x2 , so we can use (1, 3)T . For = 1, we solve 1 1 3 3 x=0

Here we get x1 = x2 , so we use (1, 1)T . Thus, our general solution is c1 or x1 (t) = c1 e3t + c2 et x2 (t) = 3c1 e3t + c2 et Since x1 = y , we see that we have obtained the same solution as we did before. Example: Solve y + 4y = 0. We know that the characteristic equation is r 2 + 4 = 0, so r = 2i. Thus our general solution is y (t) = c1 cos(2t) + c2 sin(2t) If we convert this to a system, we let x1 = y and x2 = y to get x1 = x2 x2 = 4x1 or x = We get our eigenvalues: 1 = 2 + 4 = 0 4 So our eigenvalues are = 2i. We can then nd eigenvectors: If = 2i, we get 2i 1 4 2i 6 x=0 0 1 4 0 x 1 3 e3t + c2 1 1 et

Solving this, we get eigenvector (i, 2)T . The eigenvector for = 2i is then the conjugate, (i, 2)T . So expanding the solution corresponding to = 2i and (i, 2)T into real and imaginary parts yields (cos(2t) + i sin(2t)) So our solution is c1 i 2 = sin(2t) cos(2t) +i cos(2t) sin(2t) .

sin(2t) 2 cos(2t)

+ c2

cos(2t) 2 sin(2t)

The rst row gives c1 sin(2t) c2 cos(2t). Since c2 could be any value, this is equivalent to the answer we would get from solving the second order equation previously.