A Constant Coefficient System of Homogeneous ODE can be written as follows:
$$ \vec{y}'(x) = A \vec{y}(x) $$
The solution to the constant coefficient equation is of the form \( \vec{y}(x) = \vec{ζ} e^{λx}\).
Substitution gives
$$ \vec{ζ} λ e^{λx} = A \vec{ζ} e^{λx} \to A \vec{ζ} = λ \vec{ζ} $$
which is an eigenvalue problem. It has a nontrivial solution if
$$ |A - I λ| = 0 $$
If the eigenvalues are distint then the eigenvalues are linearly independent
then we have \(n\) linearly independent solutions
$$ \vec{y_1}(x)=\vec{ζ_1} e^{λ_1 x}, \; \vec{y_2}(x)=\vec{ζ_2} e^{λ_2 x}, \; \cdots \vec{y_n}(x)=\vec{ζ_n} e^{λ_n x}$$
The general solution to the homogeneous problems is
$$ \vec{y}(x) = c_1 \vec{ζ_1} e^{λ_1 x} + c_2 \vec{ζ_2} e^{λ_2 x} + \cdots + c_n \vec{ζ_n} e^{λ_n x}$$
Example 1
Write the following ODE as a first order system and find the general solution
$$ y'' + 9y = 0 $$
solution:
$$ y_1 = y,\;\;\; y_2 = y' $$
The equations
$$ y_1' = y_2 $$
$$ y_2' = -9 y_1 $$
As a system
$$
\begin{bmatrix} y_1 \\ y_2 \end{bmatrix}' =
\begin{bmatrix} 0 && 1
\\-9 && 0
\end{bmatrix}
\begin{bmatrix} y_1 \\ y_2 \end{bmatrix}
$$
Find the eigenvalues
$$
\begin{vmatrix} -λ && 1
\\-9 &&-λ
\end{vmatrix} = 0 \to λ^2 + 9 = 0 \to λ = \pm 3i
$$
Find the eigenvectors
$$
λ_1 = 3i \to
\begin{bmatrix} -3i && 1
\\-9 && -3i
\end{bmatrix} \vec{ζ_1} = 0 \to
\vec{ζ_1} = \begin{bmatrix} 1 \\ 3i \end{bmatrix}
$$
$$
λ_2 = -3i \to
\begin{bmatrix} 3i && 1
\\-9 && 3i
\end{bmatrix} \vec{ζ_2} = 0 \to
\vec{ζ_2} = \begin{bmatrix} 1 \\ -3i \end{bmatrix}
$$
$$ \begin{bmatrix} y_1 \\ y_2 \end{bmatrix} = c_1 \begin{bmatrix} 1 \\ 3i \end{bmatrix} e^{3i x} + c_2 \begin{bmatrix} 1 \\ -3i \end{bmatrix} e^{-3i x} $$
Using Euler's identity it can be written as follows
$$ \begin{bmatrix} y_1 \\ y_2 \end{bmatrix} = k_1 \begin{bmatrix} \cos{3x} \\ -3\sin{3x} \end{bmatrix} + k_2 \begin{bmatrix} \sin{3x} \\ 3\cos{3x} \end{bmatrix} $$
where
$$ k_3 = c_1+c_2, k_4 = i(c_1-c_2)$$
or
$$ y(x) = k_1 \cos{3x} + k_2 \sin{3x} $$
Repeating Eigenvalues
If some of the eigenvalues are repeating, the corresponding eigenvectors can still be linearly independent.
Then it is still like the distinct eigenvalues case.
But if some of the eigenvectors are not linearly independent then we have to use a method similar to scalar case.
There we had to multiply with \(x\) to find a linearly indepent solution. Here it is not sufficient to multiply with \(x\) alone.
If \(λ\) is the eigenvalue repeating twice and there is only one independent eigenvector \(\vec{ζ}\)
than we have to look for a solution of the form
$$ \vec{y_2}(x) = (x \vec{ζ} + \vec{η}) e^{λx} $$
in addition to \( \vec{y_1}(x) = \vec{ζ} e^{λx} \).
Substitute in the equation
$$ \vec{y_2}' = \vec{ζ} e^{λx} + (x \vec{ζ} + \vec{η}) λ e^{λx} = A (x \vec{ζ} + \vec{η}) e^{λx} $$
Since \( \vec{ζ} e^{λx} \) is a solution this can be simplified
$$ \vec{ζ} e^{λx} + \vec{η} λ e^{λx} = A \vec{η} e^{λx} $$
Further simplification
$$ (A - λ I) \vec{η} = \vec{ζ} $$
This is the generalized eigenvalue problem and \(\vec{η}\) is the generalized eigenvector.
Example 1
Write the following ODEs as a first order system and find the general solution
$$ y_1' = y_1 + y_2 $$
$$ y_2' = y_2 $$
solution:
As a system
$$
\begin{bmatrix} y_1 \\ y_2 \end{bmatrix}' =
\begin{bmatrix} 1 && 1
\\ 0 && 1
\end{bmatrix}
\begin{bmatrix} y_1 \\ y_2 \end{bmatrix}
$$
Find the eigenvalues
$$
\begin{vmatrix} 1-λ && 1
\\ 0 && 1-λ
\end{vmatrix} = 0 \to (λ-1)^2 = 0 \to λ_{1,2} = λ =1
$$
The eigenvector
$$
λ = 1 \to
\begin{bmatrix} 0 && 1
\\0 && 0
\end{bmatrix} \vec{ζ} = 0 \to
\vec{ζ} = \begin{bmatrix} 1 \\ 0 \end{bmatrix}
$$
There is no other independent eigenvector for this matrix. We have to find the generalized eigenvector
$$
\begin{bmatrix} 0 && 1
\\0 && 0
\end{bmatrix} \vec{η} = \vec{ζ} = \begin{bmatrix} 1 \\ 0 \end{bmatrix} \to
\vec{η} = \begin{bmatrix} 0 \\ 1 \end{bmatrix}
$$
$$ \begin{bmatrix} y_1 \\ y_2 \end{bmatrix} = c_1 \begin{bmatrix} 1 \\ 0 \end{bmatrix} e^{x}
+ c_2 \left(x \begin{bmatrix} 1 \\ 0 \end{bmatrix} + \begin{bmatrix} 0 \\ 1 \end{bmatrix}\right) e^{x} $$
$$ y_1(x) = c_1 e^x + c_2 x e^x $$
$$ y_2(x) = c_2 e^x $$
This could also be obtained in a simpler way:
by solving the second equation first and using the solution in the first equation.
Example 2
Solve the initial value problem
$$
\begin{bmatrix} y_1 \\ y_2 \end{bmatrix}' =
\begin{bmatrix} 4 && 4
\\-1 && 8
\end{bmatrix}
\begin{bmatrix} y_1 \\ y_2 \end{bmatrix},
\begin{bmatrix} y_1(0) \\ y_2(0) \end{bmatrix} = \begin{bmatrix} 1 \\ 0 \end{bmatrix}
$$
solution:
Find the eigenvalues
$$
\begin{vmatrix} 4-λ && 4
\\ -1 && 8-λ
\end{vmatrix} = 0 \to (λ-6)^2 = 0 \to λ_{1,2} = λ = 6
$$
The eigenvector
$$
λ = 6 \to
\begin{bmatrix} -2 && 4
\\-1 && 2
\end{bmatrix} \vec{ζ_1} = 0 \to
\vec{ζ_1} = \begin{bmatrix} 2 \\ 1 \end{bmatrix}
$$
There is no other independent eigenvector for this matrix. We have to find the generalized eigenvector
$$
\begin{bmatrix} -2 && 4
\\-1 && 2
\end{bmatrix} \vec{ζ_2} = \vec{ζ_1} = \begin{bmatrix} 2 \\ 1 \end{bmatrix} \to
\vec{ζ_2} = \begin{bmatrix} 1 \\ 1 \end{bmatrix}
$$
$$ \begin{bmatrix} y_1 \\ y_2 \end{bmatrix} = c_1 \begin{bmatrix} 2 \\ 1 \end{bmatrix} e^{6x}
+ c_2 \left(x \begin{bmatrix} 2 \\ 1 \end{bmatrix} + \begin{bmatrix} 1 \\ 1 \end{bmatrix}\right) e^{6x} $$
Imposing the initial conditions
$$ \begin{bmatrix} y_1(0) \\ y_2(0) \end{bmatrix} = c_1 \begin{bmatrix} 2 \\ 1 \end{bmatrix}
+ c_2 \begin{bmatrix} 1 \\ 1 \end{bmatrix} = \begin{bmatrix} 1 \\ 0 \end{bmatrix} \to c_1 = 1, c_2 = -1 $$
The solution to the IVP
$$ \begin{bmatrix} y_1 \\ y_2 \end{bmatrix} = \begin{bmatrix} 2 \\ 1 \end{bmatrix} e^{6x}
- \left(x \begin{bmatrix} 2 \\ 1 \end{bmatrix} + \begin{bmatrix} 1 \\ 1 \end{bmatrix}\right) e^{6x} $$
Simplifying
$$ \begin{bmatrix} y_1(x) \\ y_2(x) \end{bmatrix} = \begin{bmatrix} 1-2x \\ -x \end{bmatrix} e^{6x} $$