6 Systems of Linear Differential Equations
This chapter treats systems of first-order linear ODEs with constant coefficients and their solution by means of the eigenvalues and eigenvectors of the coefficient matrix. Table 6.1 summarizes the symbols used.
| Symbol | Meaning | Type | Defined in |
|---|---|---|---|
| \(a_{ik}\) | Element of the coefficient matrix \(A\) of an ODE system | Scalar | Definition 6.1 |
| \(\mathrm{e}\) | Euler’s number | Constant | Example 2.1 |
| \(g(x)\), \(g_i(x)\) | Forcing function of a linear ODE; as a vector \(g(x)\) or \(g(t)\) for systems | Function | Definition 6.1 |
| \(i\) | Running index (component, row, eigenvalue number, spring number) | Index | Definition 6.1 |
| \(\mathrm{j}\) | Imaginary unit, \(\mathrm{j}^2 = -1\) | Constant | Equation 2.5 |
| \(k\) | Running index (time/iteration step, mode, eigenvalue number); here column index in \(a_{ik}\) | Index | Definition 6.1 |
| \(k\), \(k_0\) | Multiplicity of an eigenvalue (“root of multiplicity \(k\)”); \(k_0\) smallest exponent with \(\operatorname{rank} B^{k_0} = n-k\) | Index | alg. 6.1 |
| \(\ell\) | Running index (component of an eigenvector, solution number) | Index | Proposition 6.2 |
| \(n\) | Order of an ODE; dimension of a matrix or system; number of steps | Index | Definition 6.2 |
| \(p(\lambda)\) | Characteristic polynomial \(\det(A - \lambda E)\) | Function | Equation 6.8 |
| \(v\), \(v_i\) | Eigenvector; \(v_i\) eigenvector for the eigenvalue \(\lambda_i\) or \(\mu_k\) | Vector | Definition 5.1 |
| \(x\) | Independent variable (Ch. 02 §1–4, 03, 06, 08 eigenfunctions) | Scalar | Definition 6.1 |
| \(y\) | Unknown function \(y(x)\) or vector of unknown functions; \(y', y'', y^{(n)}\) derivatives | Function | Definition 6.1 |
| \(y_1, \dots, y_n\) | Components of the state vector \(S\) in the reduction of order (\(y_1 = y\), \(y_2 = y'\), …) or of the solution vector \(y\) of a system | Function | Definition 6.1 |
| \(A\) | Matrix (coefficient, system, or difference matrix, \(e = Au\)) | Matrix | Definition 6.1 |
| \(B\) | Further matrix (similar matrix \(A = P^{-1}BP\); \(B = A - \lambda E\) in the brute-force (“Hau-Ruck”) method) | Matrix | alg. 6.1 |
| \(C_1, \dots, C_n\) | Constant of integration; \(C_1, \dots, C_n\) constants of integration or coefficients of the general solution | Constant | Theorem 6.2 |
| \(E\) | Identity matrix | Matrix | Definition 5.1 |
| \(Y_i\), \(Y_\ell\) | Basis solution (vector) of a homogeneous ODE system, \(Y_i = \mathrm{e}^{\lambda_i x} v_i\) | Vector | Equation 6.9 |
| \(\mathbb{R}\), \(\mathbb{R}^n\) | Set of real numbers or of \(n\)-tuples | Set | Section 2.1 |
| \(\alpha\) | Real part of a complex eigenvalue, \(\lambda_{1,2} = \alpha \pm \mathrm{j}\omega\) | Scalar | Theorem 6.1 |
| \(\lambda\) | Eigenvalue or root of the characteristic equation (exponent in the ansatz \(\mathrm{e}^{\lambda t}\)); \(\lambda_i\), \(\lambda_k\), \(\lambda_{1,2}\), \(\lambda_{k\pm} = \pm\mathrm{j}\omega_k\) | Scalar | Definition 5.1 |
| \(\xi_1, \xi_2\) | Components of an unknown eigenvector | Scalar | Example 6.2 |
| \(\omega\) | Angular frequency (excitation angular frequency; imaginary part of an eigenvalue; frequency of an eigenfunction) | Scalar | Theorem 6.1 |
| \(y', y'', y^{(n)}\) | First, second, \(n\)-th derivative with respect to the independent variable | Operator | Example 2.1 |
| \(\det\) | Determinant | Operator | Definition 5.1 |
| \(\operatorname{tr}\) | Trace of a matrix | Operator | Equation 6.2 |
| \(\operatorname{Re}, \operatorname{Im}\) | Real and imaginary parts | Operator | Equation 2.19 |
| \(\overline{\lambda}\) | Complex conjugate | Operator | Proposition 6.2 |
6.1 Systems of First-Order Linear Differential Equations with Constant Coefficients
We first consider two coupled first-order linear ODEs with constant coefficients. Systems of more than two such equations are treated in Section 6.1.3.
6.1.1 Two First-Order Linear Differential Equations with Constant Coefficients
The presentation in this section follows Chapter 7 (“Systems of First-Order Linear Differential Equations”) in Papula (2015).
Definition 6.1 (System of Two First-Order Linear ODEs) Two coupled first-order linear ODEs with constant coefficients of the type
\[\begin{align} y_1' = a_{11} y_1 + a_{12} y_2 + g_1(x), \\ y_2' = a_{21} y_1 + a_{22} y_2 + g_2(x) \end{align}\]
with real, constant coefficients \(a_{ik}\) (\(i, k = 1, 2\)) and forcing functions \(g_i(x)\) form a linear system of differential equations of order 2 (the order of the system is the number of its equations). In matrix form it reads
\[ \begin{pmatrix} y_1' \\ y_2' \end{pmatrix} = \begin{pmatrix} a_{11} & a_{12} \\ a_{21} & a_{22} \end{pmatrix} \begin{pmatrix} y_1 \\ y_2 \end{pmatrix} + \begin{pmatrix} g_1(x) \\ g_2(x) \end{pmatrix} \] or \[ y' = A y + g(x) \] with the vector \(y = (y_1, y_2)^T\) of the unknown functions, the vector \(y' = (y_1', y_2')^T\) of the derivatives, the coefficient matrix \(A = (a_{ik})\), and the vector of forcing functions \(g(x) = (g_1(x), g_2(x))^T\).
The system is called homogeneous if \(g(x) = 0\), that is, \[ y' = Ay. \]
An initial value problem is present when an initial value is given for each of the two unknown functions \(y_1\) and \(y_2\).
The general solution \(y_1 = y_1(x)\), \(y_2 = y_2(x)\) contains two constants, which are determined by the initial conditions.
6.1.2 Solution of the Homogeneous System for \(n=2\)
The homogeneous system \[ y' = Ay \tag{6.1}\] can be solved by means of the eigenvalues \(\lambda_i\) and eigenvectors \(v_i\) of the coefficient matrix \(A\). To this end we use the exponential ansatz \(y = v\,\mathrm{e}^{\lambda x}\) with a constant vector \(v\).
Proposition 6.1 (Exponential Ansatz for Homogeneous Systems) Let \(A\) be a constant matrix and \(v \neq 0\) a constant vector. The function \(y = v\,\mathrm{e}^{\lambda x}\) solves Equation 6.1 if and only if \[ (A - \lambda E) v = 0 \] holds, i.e., if \(\lambda\) is an eigenvalue of \(A\) and \(v\) an associated eigenvector. Here \(E\) is the identity matrix.
Proof. With \(y' = \lambda v\,\mathrm{e}^{\lambda x}\), substitution into Equation 6.1 gives \[ \lambda v\,\mathrm{e}^{\lambda x} = A v\,\mathrm{e}^{\lambda x}. \] Division by \(\mathrm{e}^{\lambda x} \neq 0\) yields the eigenvalue problem \(\lambda v = A v\), i.e., \((A - \lambda E) v = 0\).
The eigenvalues \(\lambda_i\) are the roots of the characteristic equation \(\det(A - \lambda E) = 0\). For \(n = 2\) this is the quadratic equation \[ \lambda^2 - (a_{11} + a_{22}) \lambda + (a_{11} a_{22} - a_{12} a_{21}) = 0 \] or, in short, \[ \lambda^2 - \operatorname{tr}(A)\, \lambda + \det(A) = 0. \tag{6.2}\] The solutions of this equation are the eigenvalues of the coefficient matrix \(A\). As for the homogeneous second-order linear ODE with constant coefficients (Theorem 2.4), three cases have to be distinguished.
Theorem 6.1 (General Solution of the Homogeneous System for \(n = 2\)) Let \(a_{12} \neq 0\). With \(C_1, C_2 \in \mathbb{R}\), the first component of the general solution of Equation 6.1 is
- for two distinct real eigenvalues \(\lambda_1 \neq \lambda_2\) \[ y_1 = C_1 \mathrm{e}^{\lambda_1 x} + C_2 \mathrm{e}^{\lambda_2 x}, \]
- for one real eigenvalue \(\lambda_1 = \lambda_2 = \lambda\) with algebraic multiplicity 2 \[ y_1 = C_1 \mathrm{e}^{\lambda x} + C_2 x\,\mathrm{e}^{\lambda x}, \]
- for two complex conjugate eigenvalues \(\lambda_{1,2} = \alpha \pm \mathrm{j}\omega\) with \(\alpha, \omega \in \mathbb{R}\) and \(\omega \neq 0\) \[ y_1 = \mathrm{e}^{\alpha x} \bigl(C_1 \sin(\omega x) + C_2 \cos(\omega x)\bigr). \]
In all three cases the second component is \[ y_2 = \frac{1}{a_{12}} (y_1' - a_{11}y_1). \tag{6.3}\]
Proof. Solving the first equation \(y_1' = a_{11}y_1 + a_{12} y_2\) of the system for \(y_2\) gives Equation 6.3. Substituting Equation 6.3 into the second equation \(y_2' = a_{21} y_1 + a_{22} y_2\) (substitution method) and multiplying by \(a_{12}\) yields \[ y_1'' - (a_{11} + a_{22})\, y_1' + (a_{11} a_{22} - a_{12} a_{21})\, y_1 = 0. \] This is a homogeneous second-order linear ODE with constant coefficients whose characteristic equation is Equation 6.2. Its basis solutions in the three cases (Theorem 2.4, Equation 2.4, Equation 2.5) yield \(y_1\).
If \(a_{12} = 0\), the first equation contains only \(y_1\) and can be solved directly. As an alternative to the substitution method, the solution follows from the eigenvectors of the coefficient matrix \(A\) (Theorem 6.2). If a double eigenvalue has only one linearly independent eigenvector, alg. 6.1 supplies the missing solution.
Example 6.1 (System with Complex Eigenvalues (Real Solution Path)) Given is the system of first-order linear ODEs with constant coefficients \[\begin{align} y_1' &= y_1 + y_2, \\ y_2' &= - y_1 + y_2, \end{align}\] in matrix form \[ \begin{pmatrix} y_1' \\ y_2' \end{pmatrix} = \begin{pmatrix} 1 & 1 \\ -1 & 1 \end{pmatrix} \begin{pmatrix} y_1 \\ y_2 \end{pmatrix}. \] With the exponential ansatz \(y = v\,\mathrm{e}^{\lambda x}\) and the derivative \(y' = \lambda v\,\mathrm{e}^{\lambda x}\), substitution into the homogeneous system gives \[ \lambda v\,\mathrm{e}^{\lambda x} = \begin{pmatrix} 1 & 1 \\ -1 & 1 \end{pmatrix} v\,\mathrm{e}^{\lambda x} \] and, after division by \(\mathrm{e}^{\lambda x}\), \[ \lambda v = \begin{pmatrix} 1 & 1 \\ -1 & 1 \end{pmatrix} v \] the eigenvalue problem \[ \begin{pmatrix} 1-\lambda & 1 \\ -1 & 1-\lambda \end{pmatrix} v = 0. \tag{6.4}\] The characteristic equation of the coefficient matrix \(A\) is \[ \lambda^2 - 2 \lambda + 2 = 0, \] so that we obtain the complex eigenvalues \(\lambda_{1,2} = 1 \pm \mathrm{j}\).
With \(\alpha = 1\) and \(\omega = 1\), Theorem 6.1 gives the first solution function \[ y_1 = \mathrm{e}^x (C_1 \sin(x) + C_2 \cos(x)). \] Its derivative is \[ y_1' = \mathrm{e}^x (C_1 \sin x + C_2 \cos x) + \mathrm{e}^x (C_1 \cos x - C_2 \sin x). \] With Equation 6.3 and \(a_{11} = 1\) as well as \(a_{12} = 1\) we obtain the second solution function \[\begin{align} y_2 &= y_1' - y_1 \\ &= \mathrm{e}^x (C_1 \sin x + C_2 \cos x) + \mathrm{e}^x (C_1 \cos x - C_2 \sin x) - \mathrm{e}^x (C_1 \sin(x) + C_2 \cos(x))\\ &= \mathrm{e}^x (C_1 \cos(x) - C_2 \sin(x)). \end{align}\] The solution functions are \[\begin{align} y_1 = \mathrm{e}^x (C_1 \sin(x) + C_2 \cos(x)), \\ y_2 = \mathrm{e}^x (C_1 \cos(x) - C_2 \sin(x)), \end{align}\] in vector form \[ y = \mathrm{e}^x \left[ C_1 \begin{pmatrix} \sin(x) \\ \cos(x) \end{pmatrix} + C_2 \begin{pmatrix} \cos(x) \\ -\sin(x) \end{pmatrix} \right] \qquad (C_1, C_2 \in \mathbb{R}). \tag{6.5}\]
Example 6.2 (System with Complex Eigenvalues (Complex Solution Path)) We now solve the system from Example 6.1 with \[ A = \begin{pmatrix} 1 & 1 \\ -1 & 1 \end{pmatrix} \] in the complex domain by means of the eigenvectors \(v_1\) and \(v_2\) of the coefficient matrix \(A\). We obtain them by substituting the eigenvalues into the eigenvalue problem (Equation 6.4). For the first eigenvalue \(\lambda_1 = 1 + \mathrm{j}\) the coefficient matrix reads \[ \begin{pmatrix} 1 - (1 + \mathrm{j}) & 1 \\ -1 & 1 - (1 + \mathrm{j}) \end{pmatrix} = \begin{pmatrix} -\mathrm{j} & 1 \\ -1 & -\mathrm{j} \end{pmatrix}. \] Multiplying the first equation by \(\mathrm{j}\) and adding the two equations yields \[ \begin{pmatrix} 1 & \mathrm{j} \\ 0 & 0 \end{pmatrix}. \] For the eigenvector \(v_1 = (\xi_1, \xi_2)^T\) we choose \(\xi_1 = 1\) (arbitrarily) and obtain from \[ \xi_1 + \mathrm{j}\xi_2 = 0 \] the second component \(\xi_2 = \mathrm{j}\). The eigenvector for the eigenvalue \(\lambda_1 = 1+\mathrm{j}\) is thus \[ v_1 = \begin{pmatrix} 1 \\ \mathrm{j} \end{pmatrix}. \] The eigenvector \(v_2\) for the eigenvalue \(\lambda_2 = 1-\mathrm{j}\) is computed correspondingly. This leads to \[ v_1 = \begin{pmatrix} 1 \\ \mathrm{j} \end{pmatrix}, \quad v_2 = \begin{pmatrix} \mathrm{j} \\ 1 \end{pmatrix}. \]
For the real solution we use the first eigenvalue \(\lambda_1 = 1 + \mathrm{j}\) and the associated eigenvector \(v_1\). Using Euler’s formula \[ \mathrm{e}^{(\alpha+\mathrm{j}\omega)x} = \mathrm{e}^{\alpha x} (\cos(\omega x) + \mathrm{j} \sin(\omega x)) \tag{6.6}\] we first set up the complex solution \(y\): \[ y = \mathrm{e}^{(1+\mathrm{j})x} \begin{pmatrix} 1 \\ \mathrm{j} \end{pmatrix} = \mathrm{e}^x (\cos(x) + \mathrm{j} \sin(x)) \begin{pmatrix} 1 \\ \mathrm{j} \end{pmatrix} = \mathrm{e}^x \begin{pmatrix} \cos(x) + \mathrm{j} \sin(x) \\ \mathrm{j} \cos(x) - \sin(x) \end{pmatrix}. \] Sorting the real and imaginary parts yields \[ y = \mathrm{e}^x \left[ \begin{pmatrix} \cos(x) \\ -\sin(x) \end{pmatrix} + \mathrm{j} \begin{pmatrix} \sin(x) \\ \cos(x) \end{pmatrix} \right]. \] The real basis solutions are \[ \operatorname{Re}(y) = \mathrm{e}^x \begin{pmatrix} \cos(x) \\ -\sin(x) \end{pmatrix}, \quad \operatorname{Im}(y) = \mathrm{e}^x \begin{pmatrix} \sin(x) \\ \cos(x) \end{pmatrix}. \] Their linear combination is the general solution of the homogeneous system \[ y = \mathrm{e}^x \left[ C_1 \begin{pmatrix} \cos(x) \\ -\sin(x) \end{pmatrix} + C_2 \begin{pmatrix} \sin(x) \\ \cos(x) \end{pmatrix} \right] \qquad (C_1, C_2 \in \mathbb{R}). \tag{6.7}\]
Equation 6.5 and Equation 6.7 describe the same solution set. They transform into each other by interchanging the constants \(C_1\) and \(C_2\).
Example 6.3 (Computation with the Second Eigenvalue) Equation 6.5 and Equation 6.7 show that the solutions of the homogeneous system of first-order linear ODEs with constant coefficients can be determined by the eigenvalues and eigenvectors of the coefficient matrix \(A\). For the computation of the solution Equation 6.7 we used the first eigenvalue \(\lambda_1 = 1 + \mathrm{j}\) and the associated eigenvector \(v_1 = \begin{pmatrix} 1 \\ \mathrm{j} \end{pmatrix}\). We show that the same result is also obtained with the eigenvalue \(\lambda_2 = 1 - \mathrm{j}\) and the associated eigenvector \(v_2 = \begin{pmatrix} \mathrm{j} \\ 1 \end{pmatrix}\).
For the real solution we use the second eigenvalue \(\lambda_2 = 1 - \mathrm{j}\) and the associated eigenvector \(v_2 = \begin{pmatrix} \mathrm{j} \\ 1 \end{pmatrix}\). Using Euler’s formula (Equation 6.6) we first set up the complex solution \(y\) (note that \(\cos(-x) = \cos(x)\) and \(\sin(-x) = -\sin(x)\)): \[ y = \mathrm{e}^{(1-\mathrm{j})x} \begin{pmatrix} \mathrm{j} \\ 1 \end{pmatrix} = \mathrm{e}^x (\cos(x) - \mathrm{j} \sin(x)) \begin{pmatrix} \mathrm{j} \\ 1 \end{pmatrix} = \mathrm{e}^x \begin{pmatrix} \sin(x) + \mathrm{j} \cos(x) \\ \cos(x) - \mathrm{j} \sin(x) \end{pmatrix}. \] Sorting the real and imaginary parts yields \[ y = \mathrm{e}^x \left[ \begin{pmatrix} \sin(x) \\ \cos(x) \end{pmatrix} + \mathrm{j} \begin{pmatrix} \cos(x) \\ - \sin(x) \end{pmatrix} \right]. \] The real basis solutions are \[ \operatorname{Re}(y) = \mathrm{e}^x \begin{pmatrix} \sin(x) \\ \cos(x) \end{pmatrix}, \quad \operatorname{Im}(y) = \mathrm{e}^x \begin{pmatrix} \cos(x) \\ - \sin(x) \end{pmatrix}. \] These are the same basis solutions as in the computation with \(\lambda_1\), only interchanged: \(\operatorname{Re}(y)\) here is the former basis solution \(\operatorname{Im}(y)\) and vice versa. Assigning the constant \(C_1\) to the imaginary part and \(C_2\) to the real part, the general solution of the homogeneous system is \[ y = \mathrm{e}^x \left[ C_1 \begin{pmatrix} \cos(x) \\ -\sin(x) \end{pmatrix} + C_2 \begin{pmatrix} \sin(x) \\ \cos(x) \end{pmatrix} \right] \qquad (C_1, C_2 \in \mathbb{R}), \] that is, exactly Equation 6.7. The second eigenvalue therefore yields the same result. This is no coincidence: \(\lambda_2 = \overline{\lambda_1}\), and the associated complex solution is, up to a constant factor, the complex conjugate of the first one (Proposition 6.2).
6.1.3 Homogeneous Systems of First-Order Linear Differential Equations with Constant Coefficients of Arbitrary Order
The solution path described for \(n=2\) generalizes to systems of arbitrary order.
Definition 6.2 (Homogeneous System of Order \(n\)) An \(n \times n\) system of differential equations of the form \[ y' = A y \] with \(y = (y_1, \dots, y_n)^T\) and a constant matrix \(A \in \mathbb{R}^{n \times n}\) is called a homogeneous system of first-order linear ODEs with constant coefficients.
Theorem 6.2 (Solutions of the Homogeneous System of First-Order Linear ODEs with Constant Coefficients) Let \(y' = Ay\) be a homogeneous system according to Definition 6.2. The eigenvalues \(\lambda_i\) of \(A\) are the roots of the characteristic polynomial \[ p(\lambda) = \det(A - \lambda E) = 0. \tag{6.8}\] For each eigenvalue \(\lambda_i\) with eigenvector \(v_i\), i.e., \((A - \lambda_i E) v_i = 0\), \[ Y_i = \mathrm{e}^{\lambda_i x} v_i \qquad (i = 1, \ldots, n) \tag{6.9}\] is a solution of the system. If \(A\) has \(n\) linearly independent eigenvectors, the general solution of the homogeneous system is \[ y = C_1 \mathrm{e}^{\lambda_1 x} v_1 + \ldots + C_n \mathrm{e}^{\lambda_n x} v_n \qquad (C_1, \ldots, C_n \in \mathbb{R}). \]
Proof. That \(Y_i\) is a solution follows from Proposition 6.1, whose proof holds verbatim for \(n \times n\) matrices. Since the system is linear, every linear combination of solutions is again a solution. If \(v_1, \dots, v_n\) are linearly independent, then so are \(Y_1(x), \dots, Y_n(x)\) at every point \(x\). They therefore form a fundamental system of the \(n\)-dimensional solution space.
A summary of the computation of eigenvalues and eigenvectors can be found in Chapter 5.
If the eigenvalue \(\lambda\) is a root of multiplicity \(k\) of the characteristic polynomial (Equation 6.8), then \(k\) linearly independent solutions belong to \(\lambda\). Two cases require additional steps: complex eigenvalues (Proposition 6.2) and multiple eigenvalues with too few eigenvectors (alg. 6.1).
Proposition 6.2 (Real Solutions for Complex Eigenvalues) Let \(A\) be real and \(\lambda\) a complex eigenvalue of \(A\) with multiplicity \(k\). Then \(\overline{\lambda}\) is also an eigenvalue of \(A\). If \(Y_1, \dots, Y_k\) are the linearly independent solutions belonging to \(\lambda\), then the \(2k\) real functions \[ \operatorname{Re}(Y_\ell), \quad \operatorname{Im}(Y_\ell) \qquad (\ell = 1, \ldots, k) \] are linearly independent solutions of the system. They belong to the two eigenvalues \(\lambda\) and \(\overline{\lambda}\). The real and imaginary parts are computed with Euler’s formula (Equation 6.6).
Proof. Proof sketch: Since \(A\) is real, complex conjugation of \(A v = \lambda v\) gives \(A \overline{v} = \overline{\lambda}\, \overline{v}\). Hence, together with \(Y_\ell\), \(\overline{Y_\ell}\) is also a solution. As linear combinations of solutions, the real and imaginary parts \(\operatorname{Re}(Y_\ell) = \tfrac{1}{2}(Y_\ell + \overline{Y_\ell})\) and \(\operatorname{Im}(Y_\ell) = \tfrac{1}{2\mathrm{j}}(Y_\ell - \overline{Y_\ell})\) are again solutions, and they span the same space as the \(2k\) complex solutions \(Y_\ell\), \(\overline{Y_\ell}\).
If \(\lambda\) is a root of multiplicity \(k\) but there are fewer than \(k\) linearly independent eigenvectors, the missing solutions are found with the following method.
Algorithm 6.1 (Brute-Force (“Hau-Ruck”) Method) Let \(\lambda\) be a root of multiplicity \(k\) of the characteristic polynomial of the \(n \times n\) matrix \(A\).
- Form the matrix \(B = A - \lambda E\).
- Form \(B^2, B^3, \ldots\) until some power has rank \(n-k\). Let this be the case for \(B^{k_0}\).
- Determine a basis \(v_1, \ldots, v_{k}\) of the solution space of \(B^{k_0}v = 0\).
- A system of solutions belonging to the eigenvalue \(\lambda\) of multiplicity \(k\) is then, for \(\ell=1,\ldots,k\), \[ Y_\ell = \mathrm{e}^{\lambda x} \left( v_\ell + x B v_\ell + \frac{x^2}{2} B^2v_\ell + \ldots + \frac{x^{k_0-1}}{(k_0-1)!} B^{k_0-1} v_\ell \right). \tag{6.10}\]
The method is taken from Furlan (1995).
Example 6.4 (Real Eigenvalues) Let \[ y' = \begin{pmatrix} 3 & -2 \\ -1 & 2 \end{pmatrix} y. \] The characteristic polynomial of the coefficient matrix \(A\) is \[ p(\lambda) = \det(A - \lambda E) = \det \begin{pmatrix} 3-\lambda & -2 \\ -1 & 2-\lambda \end{pmatrix} = \lambda^2 - 5 \lambda + 4 = 0, \] and the eigenvalues are \(\lambda_1 = 1\) and \(\lambda_2 = 4\). We obtain the eigenvectors \(v_1\) and \(v_2\) by substituting the eigenvalues into the eigenvalue problem \[ \begin{pmatrix} 2 & -2 \\ -1 & 1 \end{pmatrix} v_1 = 0, \quad \begin{pmatrix} -1 & -2 \\ -1 & -2 \end{pmatrix} v_2 = 0, \] so that we obtain the eigenvectors \(v_1 = \begin{pmatrix} 1 \\ 1 \end{pmatrix}\) and \(v_2 = \begin{pmatrix} -2 \\ 1 \end{pmatrix}\). By Theorem 6.2 the general solution of the homogeneous system is \[ y = C_1 \mathrm{e}^x \begin{pmatrix} 1 \\ 1 \end{pmatrix} + C_2 \mathrm{e}^{4x} \begin{pmatrix} -2 \\ 1 \end{pmatrix} \qquad (C_1, C_2 \in \mathbb{R}). \]
Example 6.5 (Double Eigenvalues) We determine the general solution of the system \[ y' = \begin{pmatrix} 3 & 2 \\ -2 & -1 \end{pmatrix} y. \] The characteristic polynomial of the coefficient matrix \(A\) is \[ p(\lambda) = \det(A - \lambda E) = \det \begin{pmatrix} 3-\lambda & 2 \\ -2 & -1-\lambda \end{pmatrix} = \lambda^2 - 2 \lambda +1 = 0, \] which yields the double eigenvalue \(\lambda = 1\). We obtain the eigenvector \(v\) by substituting the eigenvalue into the eigenvalue problem \[ \begin{pmatrix} 2 & 2 \\ -2 & -2 \end{pmatrix} v = 0, \] so \(v = \begin{pmatrix} 1 \\ -1 \end{pmatrix}\). We computed it only as a check. We determine the solution with the brute-force method (alg. 6.1) and set \[ B = A - E = \begin{pmatrix} 2 & 2 \\ -2 & -2 \end{pmatrix}. \] The matrix \(B\) is raised to successive powers until a power \(B^{k_0}\) reaches the rank \(n-k\). Here \(n=2\) and \(k=2\). \(B^1\) has rank 1 and \(B^2\) has rank 0, so that \(k_0 = 2\). Thus \(B^2 v = 0\) has \(k=2\) linearly independent solutions, which we can choose freely. We set \[ v_1 = \begin{pmatrix} 1 \\ 0 \end{pmatrix}, \quad v_2 = \begin{pmatrix} 0 \\ 1 \end{pmatrix}. \] We substitute these vectors into Equation 6.10 and obtain for the first solution (\(v_1\)) \[ Y_1 = \mathrm{e}^x \left[ \begin{pmatrix} 1 \\ 0 \end{pmatrix} + x \begin{pmatrix} 2 & 2 \\ -2 & -2 \end{pmatrix} \begin{pmatrix} 1 \\ 0 \end{pmatrix} \right], \] that is, \[ Y_1 = \mathrm{e}^x \begin{pmatrix} 1 \\ 0 \end{pmatrix} + \mathrm{e}^x \begin{pmatrix} 2 \\ -2 \end{pmatrix} x = \mathrm{e}^x \begin{pmatrix} 1 + 2x \\ -2x \end{pmatrix}. \] For the second solution (\(v_2\)) we obtain \[ Y_2 = \mathrm{e}^x \left[ \begin{pmatrix} 0 \\ 1 \end{pmatrix} + x \begin{pmatrix} 2 & 2 \\ -2 & -2 \end{pmatrix} \begin{pmatrix} 0 \\ 1 \end{pmatrix} \right], \] that is, \[ Y_2 = \mathrm{e}^x \begin{pmatrix} 0 \\ 1 \end{pmatrix} + \mathrm{e}^x \begin{pmatrix} 2 \\ -2 \end{pmatrix} x = \mathrm{e}^x \begin{pmatrix} 2x \\ 1 - 2x \end{pmatrix}. \] The general solution of the homogeneous system is the linear combination of the two solutions: \[ y = C_1 \mathrm{e}^x \begin{pmatrix} 1 + 2x \\ -2x \end{pmatrix} + C_2 \mathrm{e}^x \begin{pmatrix} 2x \\ 1 - 2x \end{pmatrix} \qquad (C_1, C_2 \in \mathbb{R}). \] Every linear combination of the basis solutions \(Y_1\) and \(Y_2\) is again a solution of the homogeneous system. Their difference gives the basis solution \[ Y_1 - Y_2 = \mathrm{e}^x \begin{pmatrix} 1 \\ -1 \end{pmatrix}. \] This is the solution belonging to the eigenvector \(v\), which we computed at the beginning. The general solution of the homogeneous system can therefore also be written as \[ y = C_1 \mathrm{e}^x \begin{pmatrix} 1 \\ -1 \end{pmatrix} + C_2 \mathrm{e}^x \begin{pmatrix} 1 + 2x \\ -2x \end{pmatrix} = \mathrm{e}^x\left[ C_1 \begin{pmatrix} 1 \\ -1 \end{pmatrix} + C_2 \begin{pmatrix} 1 + 2x \\ -2x \end{pmatrix} \right] \qquad (C_1, C_2 \in \mathbb{R}) \]
Example 6.6 (Complex Eigenvalues) We determine the general solution of the system \[ y' = \begin{pmatrix} 3 & 2 \\ -2 & 1 \end{pmatrix} y. \] The characteristic polynomial of the coefficient matrix \(A\) is \[ p(\lambda) = \det(A - \lambda E) = \det \begin{pmatrix} 3-\lambda & 2 \\ -2 & 1-\lambda \end{pmatrix} = \lambda^2 - 4 \lambda + 7 = 0, \] which yields the complex eigenvalues \(\lambda_{1,2} = 2 \pm \mathrm{j}\sqrt{3}\). For the first eigenvalue \(\lambda_1 = 2 + \mathrm{j}\sqrt{3}\), solving the eigenvalue problem \[ \begin{pmatrix} 1 - \mathrm{j}\sqrt{3} & 2 \\ -2 & - (1 + \mathrm{j}\sqrt{3}) \end{pmatrix} v_1 = 0 \] gives the eigenvector \(v_1 = \begin{pmatrix} -2 \\ 1- \mathrm{j} \sqrt{3} \end{pmatrix}\). We do not need the second eigenvalue for the real solution. It can serve as a check. We substitute the first eigenvalue and the associated eigenvector into Equation 6.9: \[ y = \mathrm{e}^{(2+\mathrm{j}\sqrt{3})x} \begin{pmatrix} -2 \\ 1- \mathrm{j} \sqrt{3} \end{pmatrix}. \] Euler’s formula (Equation 6.6) gives \[ y = \mathrm{e}^{2x} \left[ \cos(\sqrt{3}x) + \mathrm{j} \sin(\sqrt{3}x) \right] \begin{pmatrix} -2 \\ 1- \mathrm{j} \sqrt{3} \end{pmatrix}. \] Multiplying out with the eigenvector yields \[ y = \mathrm{e}^{2x} \begin{pmatrix} -2 \cos(\sqrt{3}x) - 2\mathrm{j} \sin(\sqrt{3}x) \\ \cos(\sqrt{3}x) - \mathrm{j} \sqrt{3} \cos(\sqrt{3}x) + \mathrm{j} \sin(\sqrt{3}x) + \sqrt{3} \sin(\sqrt{3}x) \end{pmatrix}, \] sorted by real and imaginary parts, \[ y = \mathrm{e}^{2x} \left[ \begin{pmatrix} -2 \cos(\sqrt{3}x) \\ \cos(\sqrt{3}x) + \sqrt{3} \sin (\sqrt{3}x) \end{pmatrix} + \mathrm{j} \begin{pmatrix} -2 \sin(\sqrt{3}x) \\ - \sqrt{3} \cos(\sqrt{3}x) + \sin(\sqrt{3}x) \end{pmatrix} \right]. \] A real fundamental system is thus \[ \operatorname{Re}(y) = \mathrm{e}^{2x} \begin{pmatrix} -2 \cos(\sqrt{3}x) \\ \cos(\sqrt{3}x) + \sqrt{3} \sin (\sqrt{3}x) \end{pmatrix}, \quad \operatorname{Im}(y) = \mathrm{e}^{2x} \begin{pmatrix} -2 \sin(\sqrt{3}x) \\ - \sqrt{3} \cos(\sqrt{3}x) + \sin(\sqrt{3}x) \end{pmatrix}. \] The general solution of the homogeneous system is their linear combination \[ y = \mathrm{e}^{2x} \left[ C_1 \begin{pmatrix} -2 \cos(\sqrt{3}x) \\ \cos(\sqrt{3}x) + \sqrt{3} \sin (\sqrt{3}x) \end{pmatrix} + C_2 \begin{pmatrix} -2 \sin(\sqrt{3}x) \\ - \sqrt{3} \cos(\sqrt{3}x) + \sin(\sqrt{3}x) \end{pmatrix} \right] \qquad (C_1, C_2 \in \mathbb{R}). \]