Chapter 4 Problem Set, Part 2: Complex and Repeated Eigenvalues, Spirals and Improper Nodes

Complete, step-by-step solutions to the eight exercises of the second Chapter 4 sheet: phase portraits and general solutions from complex eigen-data, canonical forms with Q = [u | w], independence of real and imaginary parts, building matrices from eigen-data, repeated eigenvalues with the Jordan chain, the nilpotent case and the full qualitative catalogue of linear portraits.

Source Document: sources/cuatrimestre-1/04-advanced-maths/unit-04-systems-of-odes/problemas/probls_ch4-ii_2627.pdf
Exercise 1

Eight Spiral and Centre Portraits from Complex Eigen-data

Original statement · probls_ch4-ii_2627.pdf
Complex Eigenvalues • Spirals and Centres • Eight Portraits
Statement

Draw the phase portrait for the equation $d\mathbf x/dt = A\mathbf x$, when the eigenvalues ($\lambda_\pm$) and eigenvectors ($\boldsymbol\eta_\pm$) of $A$ are as given. The sign of $\dot x$ is indicated when $x = 0$ and $y > 0$.

  • (i) $\lambda_\pm = 1 \pm 3i \text{ with } \boldsymbol\eta_\pm = (1,\ 2 \mp i),\ \dot x < 0$;
  • (ii) $\lambda_\pm = \pm 3i \text{ with } \boldsymbol\eta_\pm = (1 \pm 2i,\ 1 \mp 3i),\ \dot x < 0$;
  • (iii) $\lambda_\pm = -2 \pm i \text{ with } \boldsymbol\eta_\pm = (1 \mp i,\ 3 \pm i),\ \dot x > 0$;
  • (iv) $\lambda_\pm = -1 \pm i \text{ with } \boldsymbol\eta_\pm = (1,\ \pm i),\ \dot x > 0$;
  • (v) $\lambda_\pm = 2 \pm 2i \text{ with } \boldsymbol\eta_\pm = (\pm 3i,\ 5 \mp 4i),\ \dot x < 0$;
  • (vi) $\lambda_\pm = 5 \pm 3i \text{ with } \boldsymbol\eta_\pm = (2 \pm 5i,\ \mp i),\ \dot x < 0$;
  • (vii) $\lambda_\pm = \pm 7i \text{ with } \boldsymbol\eta_\pm = (1 \pm i,\ -1 \pm 2i),\ \dot x > 0$;
  • (viii) $\lambda_\pm = -13 \pm 17i \text{ with } \boldsymbol\eta_\pm = (\pm 6i - 8,\ 4 \mp 5i),\ \dot x > 0$.
Setup and classification

In each item $\lambda_\pm = \rho \pm i\omega$ with $\omega > 0$ and $\boldsymbol\eta_+ = \mathbf u + i\mathbf w$ is the eigenvector of $\lambda_+ = \rho + i\omega$ (the second eigenvector $\boldsymbol\eta_- = \bar{\boldsymbol\eta}_+$ carries no new information). The type is fixed by $\rho$: $\rho < 0$ stable spiral, $\rho > 0$ unstable spiral, $\rho = 0$ centre (Topic 4 Sec. 7.6). The sense of rotation follows from the vector field at $(0, 1)^T$: $\dot{\mathbf x} = A(0, 1)^T = (a_{12}, a_{22})^T$, and $\dot x = a_{12} > 0$ means clockwise, $\dot x < 0$ counter-clockwise (Topic 4 Sec. 7.8). The sign of $\dot x$ given in the statement is therefore the sense of rotation:

Item$\rho$$\omega$TypeGiven sign of $\dot x$ at $(0, 1)$Sense
(i)$1$$3$unstable spiral$<\ 0$counter-clockwise
(ii)$0$$3$centre$<\ 0$counter-clockwise
(iii)$-2$$1$stable spiral$>\ 0$clockwise
(iv)$-1$$1$stable spiral$>\ 0$clockwise
(v)$2$$2$unstable spiral$<\ 0$counter-clockwise
(vi)$5$$3$unstable spiral$<\ 0$counter-clockwise
(vii)$0$$7$centre$>\ 0$clockwise
(viii)$-13$$17$stable spiral$>\ 0$clockwise
Formulation

For each item the eigenvector gives $\mathbf u = \operatorname{Re}\boldsymbol\eta_+$ and $\mathbf w = \operatorname{Im}\boldsymbol\eta_+$, hence $Q = [\mathbf u \mid \mathbf w]$ and the canonical form $J = \begin{pmatrix} \rho & \omega \\ -\omega & \rho \end{pmatrix}$ (rotation clockwise in the $\mathbf z$-plane). The matrix $A = QJQ^{-1}$ is rebuilt, the given sign of $\dot x$ is checked against $a_{12}$, and the portrait is drawn in the $\mathbf z$-plane and in the $(x, y)$-plane ($\mathbf x = Q\mathbf z$).

Method of the class notes [Notes, p. 10]

Case 2 of the notes: with $\mathbf v_\pm = \mathbf v_1 \pm i\mathbf v_2$ (here $\mathbf u = \mathbf v_1$, $\mathbf w = \mathbf v_2$) and $Q = (\mathbf v_1 \mid \mathbf v_2)$, $\mathbf z = Q^{-1}\mathbf x$ satisfies $\dot{\mathbf z} = \begin{pmatrix} \rho & \omega \\ -\omega & \rho \end{pmatrix}\mathbf z$. In polar form $r = r(0)e^{\rho t}$ and $\theta = \theta(0) - \omega t$: clockwise in the $\mathbf z$-plane, a stable spiral for $\rho < 0$, an unstable spiral for $\rho > 0$ and a centre for $\rho = 0$. The sense of rotation in the $(x, y)$-plane is read, as in the example of the notes, from the sign of $\dot x$ on the positive $y$-axis.

Derivation of the eight portraits
(i) $\lambda_\pm = 1 \pm 3i \text{ with } \boldsymbol\eta_\pm = (1,\ 2 \mp i),\ \dot x < 0$

Split the eigenvector of $\lambda_+ = 1 + 3 i$ into real and imaginary parts, $\boldsymbol\eta_+ = \begin{pmatrix} 1 \\ 2 - i \end{pmatrix} = \mathbf u + i\mathbf w$, so $\rho = 1$, $\omega = 3$ and

$$ \mathbf u = \begin{pmatrix} 1 \\ 2 \end{pmatrix}, \qquad \mathbf w = \begin{pmatrix} 0 \\ -1 \end{pmatrix} $$

The matrix is not given; it is rebuilt from $A\mathbf u = \rho\mathbf u - \omega\mathbf w$, $A\mathbf w = \omega\mathbf u + \rho\mathbf w$ (Topic 4 Eq. 7.4), which say $AQ = QJ$ with $Q = [\mathbf u \mid \mathbf w]$, so $A = QJQ^{-1}$:

$$ Q = \begin{pmatrix} 1 & 0 \\ 2 & -1 \end{pmatrix}, \quad J = \begin{pmatrix} 1 & 3 \\ -3 & 1 \end{pmatrix}, \quad \det Q = -1, \quad Q^{-1} = \begin{pmatrix} 1 & 0 \\ 2 & -1 \end{pmatrix} $$

$$ \begin{aligned} (QJ)_{11} &= 1\cdot1 + 0\cdot(-3) = 1 \\ (QJ)_{12} &= 1\cdot3 + 0\cdot1 = 3 \\ (QJ)_{21} &= 2\cdot1 + (-1)\cdot(-3) = 5 \\ (QJ)_{22} &= 2\cdot3 + (-1)\cdot1 = 5 \end{aligned} $$

$$ QJ = \begin{pmatrix} 1 & 3 \\ 5 & 5 \end{pmatrix} $$

$$ \begin{aligned} (A)_{11} &= 1\cdot1 + 3\cdot2 = 7 \\ (A)_{12} &= 1\cdot0 + 3\cdot(-1) = -3 \\ (A)_{21} &= 5\cdot1 + 5\cdot2 = 15 \\ (A)_{22} &= 5\cdot0 + 5\cdot(-1) = -5 \end{aligned} $$

$$ A = QJQ^{-1} = \begin{pmatrix} 7 & -3 \\ 15 & -5 \end{pmatrix} $$

Check of the given sign: $\dot x$ at $(0, 1)$ equals $a_{12} = -3$, which is negative, in agreement with the statement ($\dot x < 0$).

How $A$ acts on $\mathbf u$ and $\mathbf w$

Separating the real and imaginary parts of $A\mathbf v = \lambda\mathbf v$ gives (Topic 4, Eq. 7.4) $A\mathbf u = \rho\mathbf u - \omega\mathbf w$ and $A\mathbf w = \omega\mathbf u + \rho\mathbf w$. Verification with the numbers:

$$ A\mathbf u = \begin{pmatrix} 1 \\ 5 \end{pmatrix}, \qquad \rho\mathbf u - \omega\mathbf w = 1\begin{pmatrix} 1 \\ 2 \end{pmatrix} - 3\begin{pmatrix} 0 \\ -1 \end{pmatrix} = \begin{pmatrix} 1 \\ 5 \end{pmatrix} $$

$$ A\mathbf w = \begin{pmatrix} 3 \\ 5 \end{pmatrix}, \qquad \omega\mathbf u + \rho\mathbf w = 3\begin{pmatrix} 1 \\ 2 \end{pmatrix} + 1\begin{pmatrix} 0 \\ -1 \end{pmatrix} = \begin{pmatrix} 3 \\ 5 \end{pmatrix} $$

Change of coordinates

$\mathbf u$ and $\mathbf w$ are linearly independent (they are not parallel when $\omega \neq 0$; Topic 4, Lemma of Sec. 7.3). With the columns in the order $\mathbf u, \mathbf w$,

$$ Q = [\mathbf u \mid \mathbf w] = \begin{pmatrix} 1 & 0 \\ 2 & -1 \end{pmatrix} $$

$\det Q = (1)(-1) - (0)(2) = -1$, which is not zero, so $Q$ is invertible. For a $2\times 2$ matrix $\begin{pmatrix} a & b \\ c & d \end{pmatrix}^{-1} = \frac{1}{ad - bc}\begin{pmatrix} d & -b \\ -c & a \end{pmatrix}$ (Robinson, Appendix B):

$$ Q^{-1} = \frac{1}{-1}\begin{pmatrix} -1 & 0 \\ -2 & 1 \end{pmatrix} = \begin{pmatrix} 1 & 0 \\ 2 & -1 \end{pmatrix} $$

Compute $AQ = [A\mathbf u \mid A\mathbf w]$ and multiply by $Q^{-1}$ on the left, entry by entry:

$$ AQ = \begin{pmatrix} 1 & 3 \\ 5 & 5 \end{pmatrix} $$

$$ \begin{aligned} (Q^{-1}AQ)_{11} &= 1\cdot1 + 0\cdot5 = 1 \\ (Q^{-1}AQ)_{12} &= 1\cdot3 + 0\cdot5 = 3 \\ (Q^{-1}AQ)_{21} &= 2\cdot1 + (-1)\cdot5 = -3 \\ (Q^{-1}AQ)_{22} &= 2\cdot3 + (-1)\cdot5 = 1 \end{aligned} $$

$$ Q^{-1}AQ = \begin{pmatrix} 1 & 3 \\ -3 & 1 \end{pmatrix} = J = \begin{pmatrix} \rho & \omega \\ -\omega & \rho \end{pmatrix} $$

Solution in the canonical plane

$$ \dot z_1 = z_1 + 3z_2, \qquad \dot z_2 = -3z_1 + z_2 $$

The two equations are coupled, but they say "scale and rotate", which polar coordinates express directly: $z_1 = r\cos\theta$, $z_2 = r\sin\theta$ (Topic 4, Sec. 7.5).

The radius. From $r^2 = z_1^2 + z_2^2$ the chain rule gives $2r\dot r = 2z_1\dot z_1 + 2z_2\dot z_2$. Substituting $\dot z_1 = \rho z_1 + \omega z_2$ and $\dot z_2 = -\omega z_1 + \rho z_2$:

$$ r\dot r = z_1(\rho z_1 + \omega z_2) + z_2(-\omega z_1 + \rho z_2) = \rho z_1^2 + \omega z_1z_2 - \omega z_1z_2 + \rho z_2^2 = \rho r^2 $$

The terms in $\omega$ cancel. Dividing by $r \neq 0$ (the origin is the equilibrium): $\dot r = \rho r$.

The angle. Differentiate $\tan\theta = z_2/z_1$ with the chain rule on the left and the quotient rule on the right: $\sec^2\theta\,\dot\theta = (z_1\dot z_2 - z_2\dot z_1)/z_1^2$. Since $\sec^2\theta = r^2/z_1^2$, $\dot\theta = (z_1\dot z_2 - z_2\dot z_1)/r^2$, and

$$ z_1\dot z_2 - z_2\dot z_1 = z_1(-\omega z_1 + \rho z_2) - z_2(\rho z_1 + \omega z_2) = -\omega z_1^2 + \rho z_1z_2 - \rho z_1z_2 - \omega z_2^2 = -\omega r^2 \;\Longrightarrow\; \dot\theta = -\omega $$

Integration. $\frac{d}{dt}\left(e^{-\rho t}r\right) = e^{-\rho t}(\dot r - \rho r) = 0$, and Barrow's rule between $0$ and $t$, $\left[e^{-\rho s}r(s)\right]_{0}^{t} = 0$, gives $r(t) = r_0e^{\rho t}$. For the angle, $\theta(t) - \theta_0 = \int_0^t(-\omega)\,ds = -\omega t$. Back to Cartesian coordinates, with $c_1 = r_0\cos\theta_0 = z_1(0)$, $c_2 = r_0\sin\theta_0 = z_2(0)$ and the subtraction formulas $\cos(\theta_0 - \omega t) = \cos\theta_0\cos\omega t + \sin\theta_0\sin\omega t$, $\sin(\theta_0 - \omega t) = \sin\theta_0\cos\omega t - \cos\theta_0\sin\omega t$:

$$ z_1(t) = e^{1t}\left(c_1\cos 3t + c_2\sin 3t\right), \qquad z_2(t) = e^{1t}\left(c_2\cos 3t - c_1\sin 3t\right) $$

Check by substitution (product rule and chain rule): $\dot z_1 = \rho z_1 + e^{\rho t}\omega\left(-c_1\sin\omega t + c_2\cos\omega t\right) = \rho z_1 + \omega z_2$ and $\dot z_2 = \rho z_2 + e^{\rho t}\omega\left(-c_2\sin\omega t - c_1\cos\omega t\right) = \rho z_2 - \omega z_1$, as required.

Back to the original coordinates

$\mathbf x = Q\mathbf z = z_1\mathbf u + z_2\mathbf w$, so

$$ \mathbf x(t) = e^{1t}\left[\left(c_1\cos 3t + c_2\sin 3t\right)\begin{pmatrix} 1 \\ 2 \end{pmatrix} + \left(c_2\cos 3t - c_1\sin 3t\right)\begin{pmatrix} 0 \\ -1 \end{pmatrix}\right] $$

$$ x(t) = c_{1} e^{t} \cos{3 t} + c_{2} e^{t} \sin{3 t}, \qquad y(t) = c_{1} e^{t} \sin{3 t} + 2 c_{1} e^{t} \cos{3 t} + 2 c_{2} e^{t} \sin{3 t} - c_{2} e^{t} \cos{3 t} $$

Phase portrait

$\rho = 1$ and $\omega = 3$: the origin is a unstable spiral. In the $\mathbf z$-plane $r = r_0e^{1t}$ and $\theta = \theta_0 - 3t$: the motion is always clockwise ($\dot\theta = -\omega < 0$), and one full turn takes $T = 2\pi/\omega$ while the radius is multiplied by $e^{2\pi\rho/\omega} = e^{\frac{2 \pi}{3}}$ (growing).

Sense of rotation in the $\mathbf x$-plane, tested directly on the vector field (Topic 4 Sec. 7.8): at $(0, 1)^T$, $\dot{\mathbf x} = A(0, 1)^T = \begin{pmatrix} -3 \\ -5 \end{pmatrix}$, so $\dot x = -3$ is $< 0$: to the left at the top of the plane: the rotation is counter-clockwise. Cross-check: $\det Q = -1$ is $< 0$, so $Q$ reverses orientation and the clockwise motion of the $\mathbf z$-plane becomes counter-clockwise in the $\mathbf x$-plane, in agreement.

z1 z2 z-plane
Canonical $\mathbf z$-plane (unstable spiral): clockwise, $\dot\theta = -\omega$.
Open the figure full size
x y x-plane
Original $(x, y)$-plane (unstable spiral), counter-clockwise.
Open the figure full size
(ii) $\lambda_\pm = \pm 3i \text{ with } \boldsymbol\eta_\pm = (1 \pm 2i,\ 1 \mp 3i),\ \dot x < 0$

Split the eigenvector of $\lambda_+ = 3 i$ into real and imaginary parts, $\boldsymbol\eta_+ = \begin{pmatrix} 1 + 2 i \\ 1 - 3 i \end{pmatrix} = \mathbf u + i\mathbf w$, so $\rho = 0$, $\omega = 3$ and

$$ \mathbf u = \begin{pmatrix} 1 \\ 1 \end{pmatrix}, \qquad \mathbf w = \begin{pmatrix} 2 \\ -3 \end{pmatrix} $$

The matrix is not given; it is rebuilt from $A\mathbf u = \rho\mathbf u - \omega\mathbf w$, $A\mathbf w = \omega\mathbf u + \rho\mathbf w$ (Topic 4 Eq. 7.4), which say $AQ = QJ$ with $Q = [\mathbf u \mid \mathbf w]$, so $A = QJQ^{-1}$:

$$ Q = \begin{pmatrix} 1 & 2 \\ 1 & -3 \end{pmatrix}, \quad J = \begin{pmatrix} 0 & 3 \\ -3 & 0 \end{pmatrix}, \quad \det Q = -5, \quad Q^{-1} = \begin{pmatrix} \frac{3}{5} & \frac{2}{5} \\ \frac{1}{5} & - \frac{1}{5} \end{pmatrix} $$

$$ \begin{aligned} (QJ)_{11} &= 1\cdot0 + 2\cdot(-3) = -6 \\ (QJ)_{12} &= 1\cdot3 + 2\cdot0 = 3 \\ (QJ)_{21} &= 1\cdot0 + (-3)\cdot(-3) = 9 \\ (QJ)_{22} &= 1\cdot3 + (-3)\cdot0 = 3 \end{aligned} $$

$$ QJ = \begin{pmatrix} -6 & 3 \\ 9 & 3 \end{pmatrix} $$

$$ \begin{aligned} (A)_{11} &= (-6)\cdot\frac{3}{5} + 3\cdot\frac{1}{5} = -3 \\ (A)_{12} &= (-6)\cdot\frac{2}{5} + 3\cdot(- \frac{1}{5}) = -3 \\ (A)_{21} &= 9\cdot\frac{3}{5} + 3\cdot\frac{1}{5} = 6 \\ (A)_{22} &= 9\cdot\frac{2}{5} + 3\cdot(- \frac{1}{5}) = 3 \end{aligned} $$

$$ A = QJQ^{-1} = \begin{pmatrix} -3 & -3 \\ 6 & 3 \end{pmatrix} $$

Check of the given sign: $\dot x$ at $(0, 1)$ equals $a_{12} = -3$, which is negative, in agreement with the statement ($\dot x < 0$).

How $A$ acts on $\mathbf u$ and $\mathbf w$

Separating the real and imaginary parts of $A\mathbf v = \lambda\mathbf v$ gives (Topic 4, Eq. 7.4) $A\mathbf u = \rho\mathbf u - \omega\mathbf w$ and $A\mathbf w = \omega\mathbf u + \rho\mathbf w$. Verification with the numbers:

$$ A\mathbf u = \begin{pmatrix} -6 \\ 9 \end{pmatrix}, \qquad \rho\mathbf u - \omega\mathbf w = 0\begin{pmatrix} 1 \\ 1 \end{pmatrix} - 3\begin{pmatrix} 2 \\ -3 \end{pmatrix} = \begin{pmatrix} -6 \\ 9 \end{pmatrix} $$

$$ A\mathbf w = \begin{pmatrix} 3 \\ 3 \end{pmatrix}, \qquad \omega\mathbf u + \rho\mathbf w = 3\begin{pmatrix} 1 \\ 1 \end{pmatrix} + 0\begin{pmatrix} 2 \\ -3 \end{pmatrix} = \begin{pmatrix} 3 \\ 3 \end{pmatrix} $$

Change of coordinates

$\mathbf u$ and $\mathbf w$ are linearly independent (they are not parallel when $\omega \neq 0$; Topic 4, Lemma of Sec. 7.3). With the columns in the order $\mathbf u, \mathbf w$,

$$ Q = [\mathbf u \mid \mathbf w] = \begin{pmatrix} 1 & 2 \\ 1 & -3 \end{pmatrix} $$

$\det Q = (1)(-3) - (2)(1) = -5$, which is not zero, so $Q$ is invertible. For a $2\times 2$ matrix $\begin{pmatrix} a & b \\ c & d \end{pmatrix}^{-1} = \frac{1}{ad - bc}\begin{pmatrix} d & -b \\ -c & a \end{pmatrix}$ (Robinson, Appendix B):

$$ Q^{-1} = \frac{1}{-5}\begin{pmatrix} -3 & -2 \\ -1 & 1 \end{pmatrix} = \begin{pmatrix} \frac{3}{5} & \frac{2}{5} \\ \frac{1}{5} & - \frac{1}{5} \end{pmatrix} $$

Compute $AQ = [A\mathbf u \mid A\mathbf w]$ and multiply by $Q^{-1}$ on the left, entry by entry:

$$ AQ = \begin{pmatrix} -6 & 3 \\ 9 & 3 \end{pmatrix} $$

$$ \begin{aligned} (Q^{-1}AQ)_{11} &= \frac{3}{5}\cdot(-6) + \frac{2}{5}\cdot9 = 0 \\ (Q^{-1}AQ)_{12} &= \frac{3}{5}\cdot3 + \frac{2}{5}\cdot3 = 3 \\ (Q^{-1}AQ)_{21} &= \frac{1}{5}\cdot(-6) + (- \frac{1}{5})\cdot9 = -3 \\ (Q^{-1}AQ)_{22} &= \frac{1}{5}\cdot3 + (- \frac{1}{5})\cdot3 = 0 \end{aligned} $$

$$ Q^{-1}AQ = \begin{pmatrix} 0 & 3 \\ -3 & 0 \end{pmatrix} = J = \begin{pmatrix} \rho & \omega \\ -\omega & \rho \end{pmatrix} $$

Solution in the canonical plane

$$ \dot z_1 = 3z_2, \qquad \dot z_2 = -3z_1 $$

The two equations are coupled, but they say "scale and rotate", which polar coordinates express directly: $z_1 = r\cos\theta$, $z_2 = r\sin\theta$ (Topic 4, Sec. 7.5).

The radius. From $r^2 = z_1^2 + z_2^2$ the chain rule gives $2r\dot r = 2z_1\dot z_1 + 2z_2\dot z_2$. Substituting $\dot z_1 = \rho z_1 + \omega z_2$ and $\dot z_2 = -\omega z_1 + \rho z_2$:

$$ r\dot r = z_1(\rho z_1 + \omega z_2) + z_2(-\omega z_1 + \rho z_2) = \rho z_1^2 + \omega z_1z_2 - \omega z_1z_2 + \rho z_2^2 = \rho r^2 $$

The terms in $\omega$ cancel. Dividing by $r \neq 0$ (the origin is the equilibrium): $\dot r = \rho r$.

The angle. Differentiate $\tan\theta = z_2/z_1$ with the chain rule on the left and the quotient rule on the right: $\sec^2\theta\,\dot\theta = (z_1\dot z_2 - z_2\dot z_1)/z_1^2$. Since $\sec^2\theta = r^2/z_1^2$, $\dot\theta = (z_1\dot z_2 - z_2\dot z_1)/r^2$, and

$$ z_1\dot z_2 - z_2\dot z_1 = z_1(-\omega z_1 + \rho z_2) - z_2(\rho z_1 + \omega z_2) = -\omega z_1^2 + \rho z_1z_2 - \rho z_1z_2 - \omega z_2^2 = -\omega r^2 \;\Longrightarrow\; \dot\theta = -\omega $$

Integration. $\frac{d}{dt}\left(e^{-\rho t}r\right) = e^{-\rho t}(\dot r - \rho r) = 0$, and Barrow's rule between $0$ and $t$, $\left[e^{-\rho s}r(s)\right]_{0}^{t} = 0$, gives $r(t) = r_0e^{\rho t}$. For the angle, $\theta(t) - \theta_0 = \int_0^t(-\omega)\,ds = -\omega t$. Back to Cartesian coordinates, with $c_1 = r_0\cos\theta_0 = z_1(0)$, $c_2 = r_0\sin\theta_0 = z_2(0)$ and the subtraction formulas $\cos(\theta_0 - \omega t) = \cos\theta_0\cos\omega t + \sin\theta_0\sin\omega t$, $\sin(\theta_0 - \omega t) = \sin\theta_0\cos\omega t - \cos\theta_0\sin\omega t$:

$$ z_1(t) = e^{0t}\left(c_1\cos 3t + c_2\sin 3t\right), \qquad z_2(t) = e^{0t}\left(c_2\cos 3t - c_1\sin 3t\right) $$

Check by substitution (product rule and chain rule): $\dot z_1 = \rho z_1 + e^{\rho t}\omega\left(-c_1\sin\omega t + c_2\cos\omega t\right) = \rho z_1 + \omega z_2$ and $\dot z_2 = \rho z_2 + e^{\rho t}\omega\left(-c_2\sin\omega t - c_1\cos\omega t\right) = \rho z_2 - \omega z_1$, as required.

Back to the original coordinates

$\mathbf x = Q\mathbf z = z_1\mathbf u + z_2\mathbf w$, so

$$ \mathbf x(t) = e^{0t}\left[\left(c_1\cos 3t + c_2\sin 3t\right)\begin{pmatrix} 1 \\ 1 \end{pmatrix} + \left(c_2\cos 3t - c_1\sin 3t\right)\begin{pmatrix} 2 \\ -3 \end{pmatrix}\right] $$

$$ x(t) = - 2 c_{1} \sin{3 t} + c_{1} \cos{3 t} + c_{2} \sin{3 t} + 2 c_{2} \cos{3 t}, \qquad y(t) = 3 c_{1} \sin{3 t} + c_{1} \cos{3 t} + c_{2} \sin{3 t} - 3 c_{2} \cos{3 t} $$

Phase portrait

$\rho = 0$ and $\omega = 3$: the origin is a centre. In the $\mathbf z$-plane $r = r_0e^{0t}$ and $\theta = \theta_0 - 3t$: the motion is always clockwise ($\dot\theta = -\omega < 0$), and the orbits are circles of period $T = 2\pi/\omega$ (closed orbits).

Sense of rotation in the $\mathbf x$-plane, tested directly on the vector field (Topic 4 Sec. 7.8): at $(0, 1)^T$, $\dot{\mathbf x} = A(0, 1)^T = \begin{pmatrix} -3 \\ 3 \end{pmatrix}$, so $\dot x = -3$ is $< 0$: to the left at the top of the plane: the rotation is counter-clockwise. Cross-check: $\det Q = -5$ is $< 0$, so $Q$ reverses orientation and the clockwise motion of the $\mathbf z$-plane becomes counter-clockwise in the $\mathbf x$-plane, in agreement.

z1 z2 z-plane
Canonical $\mathbf z$-plane (centre): clockwise, $\dot\theta = -\omega$.
Open the figure full size
x y x-plane
Original $(x, y)$-plane (centre), counter-clockwise.
Open the figure full size
(iii) $\lambda_\pm = -2 \pm i \text{ with } \boldsymbol\eta_\pm = (1 \mp i,\ 3 \pm i),\ \dot x > 0$

Split the eigenvector of $\lambda_+ = -2 + i$ into real and imaginary parts, $\boldsymbol\eta_+ = \begin{pmatrix} 1 - i \\ 3 + i \end{pmatrix} = \mathbf u + i\mathbf w$, so $\rho = -2$, $\omega = 1$ and

$$ \mathbf u = \begin{pmatrix} 1 \\ 3 \end{pmatrix}, \qquad \mathbf w = \begin{pmatrix} -1 \\ 1 \end{pmatrix} $$

The matrix is not given; it is rebuilt from $A\mathbf u = \rho\mathbf u - \omega\mathbf w$, $A\mathbf w = \omega\mathbf u + \rho\mathbf w$ (Topic 4 Eq. 7.4), which say $AQ = QJ$ with $Q = [\mathbf u \mid \mathbf w]$, so $A = QJQ^{-1}$:

$$ Q = \begin{pmatrix} 1 & -1 \\ 3 & 1 \end{pmatrix}, \quad J = \begin{pmatrix} -2 & 1 \\ -1 & -2 \end{pmatrix}, \quad \det Q = 4, \quad Q^{-1} = \begin{pmatrix} \frac{1}{4} & \frac{1}{4} \\ - \frac{3}{4} & \frac{1}{4} \end{pmatrix} $$

$$ \begin{aligned} (QJ)_{11} &= 1\cdot(-2) + (-1)\cdot(-1) = -1 \\ (QJ)_{12} &= 1\cdot1 + (-1)\cdot(-2) = 3 \\ (QJ)_{21} &= 3\cdot(-2) + 1\cdot(-1) = -7 \\ (QJ)_{22} &= 3\cdot1 + 1\cdot(-2) = 1 \end{aligned} $$

$$ QJ = \begin{pmatrix} -1 & 3 \\ -7 & 1 \end{pmatrix} $$

$$ \begin{aligned} (A)_{11} &= (-1)\cdot\frac{1}{4} + 3\cdot(- \frac{3}{4}) = - \frac{5}{2} \\ (A)_{12} &= (-1)\cdot\frac{1}{4} + 3\cdot\frac{1}{4} = \frac{1}{2} \\ (A)_{21} &= (-7)\cdot\frac{1}{4} + 1\cdot(- \frac{3}{4}) = - \frac{5}{2} \\ (A)_{22} &= (-7)\cdot\frac{1}{4} + 1\cdot\frac{1}{4} = - \frac{3}{2} \end{aligned} $$

$$ A = QJQ^{-1} = \begin{pmatrix} - \frac{5}{2} & \frac{1}{2} \\ - \frac{5}{2} & - \frac{3}{2} \end{pmatrix} $$

Check of the given sign: $\dot x$ at $(0, 1)$ equals $a_{12} = \frac{1}{2}$, which is positive, in agreement with the statement ($\dot x > 0$).

How $A$ acts on $\mathbf u$ and $\mathbf w$

Separating the real and imaginary parts of $A\mathbf v = \lambda\mathbf v$ gives (Topic 4, Eq. 7.4) $A\mathbf u = \rho\mathbf u - \omega\mathbf w$ and $A\mathbf w = \omega\mathbf u + \rho\mathbf w$. Verification with the numbers:

$$ A\mathbf u = \begin{pmatrix} -1 \\ -7 \end{pmatrix}, \qquad \rho\mathbf u - \omega\mathbf w = (-2)\begin{pmatrix} 1 \\ 3 \end{pmatrix} - 1\begin{pmatrix} -1 \\ 1 \end{pmatrix} = \begin{pmatrix} -1 \\ -7 \end{pmatrix} $$

$$ A\mathbf w = \begin{pmatrix} 3 \\ 1 \end{pmatrix}, \qquad \omega\mathbf u + \rho\mathbf w = 1\begin{pmatrix} 1 \\ 3 \end{pmatrix} + (-2)\begin{pmatrix} -1 \\ 1 \end{pmatrix} = \begin{pmatrix} 3 \\ 1 \end{pmatrix} $$

Change of coordinates

$\mathbf u$ and $\mathbf w$ are linearly independent (they are not parallel when $\omega \neq 0$; Topic 4, Lemma of Sec. 7.3). With the columns in the order $\mathbf u, \mathbf w$,

$$ Q = [\mathbf u \mid \mathbf w] = \begin{pmatrix} 1 & -1 \\ 3 & 1 \end{pmatrix} $$

$\det Q = (1)(1) - (-1)(3) = 4$, which is not zero, so $Q$ is invertible. For a $2\times 2$ matrix $\begin{pmatrix} a & b \\ c & d \end{pmatrix}^{-1} = \frac{1}{ad - bc}\begin{pmatrix} d & -b \\ -c & a \end{pmatrix}$ (Robinson, Appendix B):

$$ Q^{-1} = \frac{1}{4}\begin{pmatrix} 1 & 1 \\ -3 & 1 \end{pmatrix} = \begin{pmatrix} \frac{1}{4} & \frac{1}{4} \\ - \frac{3}{4} & \frac{1}{4} \end{pmatrix} $$

Compute $AQ = [A\mathbf u \mid A\mathbf w]$ and multiply by $Q^{-1}$ on the left, entry by entry:

$$ AQ = \begin{pmatrix} -1 & 3 \\ -7 & 1 \end{pmatrix} $$

$$ \begin{aligned} (Q^{-1}AQ)_{11} &= \frac{1}{4}\cdot(-1) + \frac{1}{4}\cdot(-7) = -2 \\ (Q^{-1}AQ)_{12} &= \frac{1}{4}\cdot3 + \frac{1}{4}\cdot1 = 1 \\ (Q^{-1}AQ)_{21} &= (- \frac{3}{4})\cdot(-1) + \frac{1}{4}\cdot(-7) = -1 \\ (Q^{-1}AQ)_{22} &= (- \frac{3}{4})\cdot3 + \frac{1}{4}\cdot1 = -2 \end{aligned} $$

$$ Q^{-1}AQ = \begin{pmatrix} -2 & 1 \\ -1 & -2 \end{pmatrix} = J = \begin{pmatrix} \rho & \omega \\ -\omega & \rho \end{pmatrix} $$

Solution in the canonical plane

$$ \dot z_1 = -2z_1 + z_2, \qquad \dot z_2 = -z_1 - 2z_2 $$

The two equations are coupled, but they say "scale and rotate", which polar coordinates express directly: $z_1 = r\cos\theta$, $z_2 = r\sin\theta$ (Topic 4, Sec. 7.5).

The radius. From $r^2 = z_1^2 + z_2^2$ the chain rule gives $2r\dot r = 2z_1\dot z_1 + 2z_2\dot z_2$. Substituting $\dot z_1 = \rho z_1 + \omega z_2$ and $\dot z_2 = -\omega z_1 + \rho z_2$:

$$ r\dot r = z_1(\rho z_1 + \omega z_2) + z_2(-\omega z_1 + \rho z_2) = \rho z_1^2 + \omega z_1z_2 - \omega z_1z_2 + \rho z_2^2 = \rho r^2 $$

The terms in $\omega$ cancel. Dividing by $r \neq 0$ (the origin is the equilibrium): $\dot r = \rho r$.

The angle. Differentiate $\tan\theta = z_2/z_1$ with the chain rule on the left and the quotient rule on the right: $\sec^2\theta\,\dot\theta = (z_1\dot z_2 - z_2\dot z_1)/z_1^2$. Since $\sec^2\theta = r^2/z_1^2$, $\dot\theta = (z_1\dot z_2 - z_2\dot z_1)/r^2$, and

$$ z_1\dot z_2 - z_2\dot z_1 = z_1(-\omega z_1 + \rho z_2) - z_2(\rho z_1 + \omega z_2) = -\omega z_1^2 + \rho z_1z_2 - \rho z_1z_2 - \omega z_2^2 = -\omega r^2 \;\Longrightarrow\; \dot\theta = -\omega $$

Integration. $\frac{d}{dt}\left(e^{-\rho t}r\right) = e^{-\rho t}(\dot r - \rho r) = 0$, and Barrow's rule between $0$ and $t$, $\left[e^{-\rho s}r(s)\right]_{0}^{t} = 0$, gives $r(t) = r_0e^{\rho t}$. For the angle, $\theta(t) - \theta_0 = \int_0^t(-\omega)\,ds = -\omega t$. Back to Cartesian coordinates, with $c_1 = r_0\cos\theta_0 = z_1(0)$, $c_2 = r_0\sin\theta_0 = z_2(0)$ and the subtraction formulas $\cos(\theta_0 - \omega t) = \cos\theta_0\cos\omega t + \sin\theta_0\sin\omega t$, $\sin(\theta_0 - \omega t) = \sin\theta_0\cos\omega t - \cos\theta_0\sin\omega t$:

$$ z_1(t) = e^{-2t}\left(c_1\cos 1t + c_2\sin 1t\right), \qquad z_2(t) = e^{-2t}\left(c_2\cos 1t - c_1\sin 1t\right) $$

Check by substitution (product rule and chain rule): $\dot z_1 = \rho z_1 + e^{\rho t}\omega\left(-c_1\sin\omega t + c_2\cos\omega t\right) = \rho z_1 + \omega z_2$ and $\dot z_2 = \rho z_2 + e^{\rho t}\omega\left(-c_2\sin\omega t - c_1\cos\omega t\right) = \rho z_2 - \omega z_1$, as required.

Back to the original coordinates

$\mathbf x = Q\mathbf z = z_1\mathbf u + z_2\mathbf w$, so

$$ \mathbf x(t) = e^{-2t}\left[\left(c_1\cos 1t + c_2\sin 1t\right)\begin{pmatrix} 1 \\ 3 \end{pmatrix} + \left(c_2\cos 1t - c_1\sin 1t\right)\begin{pmatrix} -1 \\ 1 \end{pmatrix}\right] $$

$$ x(t) = c_{1} e^{-2 t} \sin{t} + c_{1} e^{-2 t} \cos{t} + c_{2} e^{-2 t} \sin{t} - c_{2} e^{-2 t} \cos{t}, \qquad y(t) = - c_{1} e^{-2 t} \sin{t} + 3 c_{1} e^{-2 t} \cos{t} + 3 c_{2} e^{-2 t} \sin{t} + c_{2} e^{-2 t} \cos{t} $$

Phase portrait

$\rho = -2$ and $\omega = 1$: the origin is a stable spiral. In the $\mathbf z$-plane $r = r_0e^{-2t}$ and $\theta = \theta_0 - 1t$: the motion is always clockwise ($\dot\theta = -\omega < 0$), and one full turn takes $T = 2\pi/\omega$ while the radius is multiplied by $e^{2\pi\rho/\omega} = e^{-4 \pi}$ (shrinking).

Sense of rotation in the $\mathbf x$-plane, tested directly on the vector field (Topic 4 Sec. 7.8): at $(0, 1)^T$, $\dot{\mathbf x} = A(0, 1)^T = \begin{pmatrix} \frac{1}{2} \\ - \frac{3}{2} \end{pmatrix}$, so $\dot x = \frac{1}{2}$ is $> 0$: to the right at the top of the plane: the rotation is clockwise. Cross-check: $\det Q = 4$ is $> 0$, so $Q$ preserves orientation and the clockwise motion of the $\mathbf z$-plane becomes clockwise in the $\mathbf x$-plane, in agreement.

z1 z2 z-plane
Canonical $\mathbf z$-plane (stable spiral): clockwise, $\dot\theta = -\omega$.
Open the figure full size
x y x-plane
Original $(x, y)$-plane (stable spiral), clockwise.
Open the figure full size
(iv) $\lambda_\pm = -1 \pm i \text{ with } \boldsymbol\eta_\pm = (1,\ \pm i),\ \dot x > 0$

Split the eigenvector of $\lambda_+ = -1 + i$ into real and imaginary parts, $\boldsymbol\eta_+ = \begin{pmatrix} 1 \\ i \end{pmatrix} = \mathbf u + i\mathbf w$, so $\rho = -1$, $\omega = 1$ and

$$ \mathbf u = \begin{pmatrix} 1 \\ 0 \end{pmatrix}, \qquad \mathbf w = \begin{pmatrix} 0 \\ 1 \end{pmatrix} $$

The matrix is not given; it is rebuilt from $A\mathbf u = \rho\mathbf u - \omega\mathbf w$, $A\mathbf w = \omega\mathbf u + \rho\mathbf w$ (Topic 4 Eq. 7.4), which say $AQ = QJ$ with $Q = [\mathbf u \mid \mathbf w]$, so $A = QJQ^{-1}$:

$$ Q = \begin{pmatrix} 1 & 0 \\ 0 & 1 \end{pmatrix}, \quad J = \begin{pmatrix} -1 & 1 \\ -1 & -1 \end{pmatrix}, \quad \det Q = 1, \quad Q^{-1} = \begin{pmatrix} 1 & 0 \\ 0 & 1 \end{pmatrix} $$

$$ \begin{aligned} (QJ)_{11} &= 1\cdot(-1) + 0\cdot(-1) = -1 \\ (QJ)_{12} &= 1\cdot1 + 0\cdot(-1) = 1 \\ (QJ)_{21} &= 0\cdot(-1) + 1\cdot(-1) = -1 \\ (QJ)_{22} &= 0\cdot1 + 1\cdot(-1) = -1 \end{aligned} $$

$$ QJ = \begin{pmatrix} -1 & 1 \\ -1 & -1 \end{pmatrix} $$

$$ \begin{aligned} (A)_{11} &= (-1)\cdot1 + 1\cdot0 = -1 \\ (A)_{12} &= (-1)\cdot0 + 1\cdot1 = 1 \\ (A)_{21} &= (-1)\cdot1 + (-1)\cdot0 = -1 \\ (A)_{22} &= (-1)\cdot0 + (-1)\cdot1 = -1 \end{aligned} $$

$$ A = QJQ^{-1} = \begin{pmatrix} -1 & 1 \\ -1 & -1 \end{pmatrix} $$

Check of the given sign: $\dot x$ at $(0, 1)$ equals $a_{12} = 1$, which is positive, in agreement with the statement ($\dot x > 0$).

How $A$ acts on $\mathbf u$ and $\mathbf w$

Separating the real and imaginary parts of $A\mathbf v = \lambda\mathbf v$ gives (Topic 4, Eq. 7.4) $A\mathbf u = \rho\mathbf u - \omega\mathbf w$ and $A\mathbf w = \omega\mathbf u + \rho\mathbf w$. Verification with the numbers:

$$ A\mathbf u = \begin{pmatrix} -1 \\ -1 \end{pmatrix}, \qquad \rho\mathbf u - \omega\mathbf w = (-1)\begin{pmatrix} 1 \\ 0 \end{pmatrix} - 1\begin{pmatrix} 0 \\ 1 \end{pmatrix} = \begin{pmatrix} -1 \\ -1 \end{pmatrix} $$

$$ A\mathbf w = \begin{pmatrix} 1 \\ -1 \end{pmatrix}, \qquad \omega\mathbf u + \rho\mathbf w = 1\begin{pmatrix} 1 \\ 0 \end{pmatrix} + (-1)\begin{pmatrix} 0 \\ 1 \end{pmatrix} = \begin{pmatrix} 1 \\ -1 \end{pmatrix} $$

Change of coordinates

$\mathbf u$ and $\mathbf w$ are linearly independent (they are not parallel when $\omega \neq 0$; Topic 4, Lemma of Sec. 7.3). With the columns in the order $\mathbf u, \mathbf w$,

$$ Q = [\mathbf u \mid \mathbf w] = \begin{pmatrix} 1 & 0 \\ 0 & 1 \end{pmatrix} $$

$\det Q = (1)(1) - (0)(0) = 1$, which is not zero, so $Q$ is invertible. For a $2\times 2$ matrix $\begin{pmatrix} a & b \\ c & d \end{pmatrix}^{-1} = \frac{1}{ad - bc}\begin{pmatrix} d & -b \\ -c & a \end{pmatrix}$ (Robinson, Appendix B):

$$ Q^{-1} = \frac{1}{1}\begin{pmatrix} 1 & 0 \\ 0 & 1 \end{pmatrix} = \begin{pmatrix} 1 & 0 \\ 0 & 1 \end{pmatrix} $$

Compute $AQ = [A\mathbf u \mid A\mathbf w]$ and multiply by $Q^{-1}$ on the left, entry by entry:

$$ AQ = \begin{pmatrix} -1 & 1 \\ -1 & -1 \end{pmatrix} $$

$$ \begin{aligned} (Q^{-1}AQ)_{11} &= 1\cdot(-1) + 0\cdot(-1) = -1 \\ (Q^{-1}AQ)_{12} &= 1\cdot1 + 0\cdot(-1) = 1 \\ (Q^{-1}AQ)_{21} &= 0\cdot(-1) + 1\cdot(-1) = -1 \\ (Q^{-1}AQ)_{22} &= 0\cdot1 + 1\cdot(-1) = -1 \end{aligned} $$

$$ Q^{-1}AQ = \begin{pmatrix} -1 & 1 \\ -1 & -1 \end{pmatrix} = J = \begin{pmatrix} \rho & \omega \\ -\omega & \rho \end{pmatrix} $$

Solution in the canonical plane

$$ \dot z_1 = -z_1 + z_2, \qquad \dot z_2 = -z_1 - z_2 $$

The two equations are coupled, but they say "scale and rotate", which polar coordinates express directly: $z_1 = r\cos\theta$, $z_2 = r\sin\theta$ (Topic 4, Sec. 7.5).

The radius. From $r^2 = z_1^2 + z_2^2$ the chain rule gives $2r\dot r = 2z_1\dot z_1 + 2z_2\dot z_2$. Substituting $\dot z_1 = \rho z_1 + \omega z_2$ and $\dot z_2 = -\omega z_1 + \rho z_2$:

$$ r\dot r = z_1(\rho z_1 + \omega z_2) + z_2(-\omega z_1 + \rho z_2) = \rho z_1^2 + \omega z_1z_2 - \omega z_1z_2 + \rho z_2^2 = \rho r^2 $$

The terms in $\omega$ cancel. Dividing by $r \neq 0$ (the origin is the equilibrium): $\dot r = \rho r$.

The angle. Differentiate $\tan\theta = z_2/z_1$ with the chain rule on the left and the quotient rule on the right: $\sec^2\theta\,\dot\theta = (z_1\dot z_2 - z_2\dot z_1)/z_1^2$. Since $\sec^2\theta = r^2/z_1^2$, $\dot\theta = (z_1\dot z_2 - z_2\dot z_1)/r^2$, and

$$ z_1\dot z_2 - z_2\dot z_1 = z_1(-\omega z_1 + \rho z_2) - z_2(\rho z_1 + \omega z_2) = -\omega z_1^2 + \rho z_1z_2 - \rho z_1z_2 - \omega z_2^2 = -\omega r^2 \;\Longrightarrow\; \dot\theta = -\omega $$

Integration. $\frac{d}{dt}\left(e^{-\rho t}r\right) = e^{-\rho t}(\dot r - \rho r) = 0$, and Barrow's rule between $0$ and $t$, $\left[e^{-\rho s}r(s)\right]_{0}^{t} = 0$, gives $r(t) = r_0e^{\rho t}$. For the angle, $\theta(t) - \theta_0 = \int_0^t(-\omega)\,ds = -\omega t$. Back to Cartesian coordinates, with $c_1 = r_0\cos\theta_0 = z_1(0)$, $c_2 = r_0\sin\theta_0 = z_2(0)$ and the subtraction formulas $\cos(\theta_0 - \omega t) = \cos\theta_0\cos\omega t + \sin\theta_0\sin\omega t$, $\sin(\theta_0 - \omega t) = \sin\theta_0\cos\omega t - \cos\theta_0\sin\omega t$:

$$ z_1(t) = e^{-1t}\left(c_1\cos 1t + c_2\sin 1t\right), \qquad z_2(t) = e^{-1t}\left(c_2\cos 1t - c_1\sin 1t\right) $$

Check by substitution (product rule and chain rule): $\dot z_1 = \rho z_1 + e^{\rho t}\omega\left(-c_1\sin\omega t + c_2\cos\omega t\right) = \rho z_1 + \omega z_2$ and $\dot z_2 = \rho z_2 + e^{\rho t}\omega\left(-c_2\sin\omega t - c_1\cos\omega t\right) = \rho z_2 - \omega z_1$, as required.

Back to the original coordinates

$\mathbf x = Q\mathbf z = z_1\mathbf u + z_2\mathbf w$, so

$$ \mathbf x(t) = e^{-1t}\left[\left(c_1\cos 1t + c_2\sin 1t\right)\begin{pmatrix} 1 \\ 0 \end{pmatrix} + \left(c_2\cos 1t - c_1\sin 1t\right)\begin{pmatrix} 0 \\ 1 \end{pmatrix}\right] $$

$$ x(t) = c_{1} e^{-t} \cos{t} + c_{2} e^{-t} \sin{t}, \qquad y(t) = - c_{1} e^{-t} \sin{t} + c_{2} e^{-t} \cos{t} $$

Phase portrait

$\rho = -1$ and $\omega = 1$: the origin is a stable spiral. In the $\mathbf z$-plane $r = r_0e^{-1t}$ and $\theta = \theta_0 - 1t$: the motion is always clockwise ($\dot\theta = -\omega < 0$), and one full turn takes $T = 2\pi/\omega$ while the radius is multiplied by $e^{2\pi\rho/\omega} = e^{-2 \pi}$ (shrinking).

Sense of rotation in the $\mathbf x$-plane, tested directly on the vector field (Topic 4 Sec. 7.8): at $(0, 1)^T$, $\dot{\mathbf x} = A(0, 1)^T = \begin{pmatrix} 1 \\ -1 \end{pmatrix}$, so $\dot x = 1$ is $> 0$: to the right at the top of the plane: the rotation is clockwise. Cross-check: $\det Q = 1$ is $> 0$, so $Q$ preserves orientation and the clockwise motion of the $\mathbf z$-plane becomes clockwise in the $\mathbf x$-plane, in agreement.

z1 z2 z-plane
Canonical $\mathbf z$-plane (stable spiral): clockwise, $\dot\theta = -\omega$.
Open the figure full size
x y x-plane
Original $(x, y)$-plane (stable spiral), clockwise.
Open the figure full size
(v) $\lambda_\pm = 2 \pm 2i \text{ with } \boldsymbol\eta_\pm = (\pm 3i,\ 5 \mp 4i),\ \dot x < 0$

Split the eigenvector of $\lambda_+ = 2 + 2 i$ into real and imaginary parts, $\boldsymbol\eta_+ = \begin{pmatrix} 3 i \\ 5 - 4 i \end{pmatrix} = \mathbf u + i\mathbf w$, so $\rho = 2$, $\omega = 2$ and

$$ \mathbf u = \begin{pmatrix} 0 \\ 5 \end{pmatrix}, \qquad \mathbf w = \begin{pmatrix} 3 \\ -4 \end{pmatrix} $$

The matrix is not given; it is rebuilt from $A\mathbf u = \rho\mathbf u - \omega\mathbf w$, $A\mathbf w = \omega\mathbf u + \rho\mathbf w$ (Topic 4 Eq. 7.4), which say $AQ = QJ$ with $Q = [\mathbf u \mid \mathbf w]$, so $A = QJQ^{-1}$:

$$ Q = \begin{pmatrix} 0 & 3 \\ 5 & -4 \end{pmatrix}, \quad J = \begin{pmatrix} 2 & 2 \\ -2 & 2 \end{pmatrix}, \quad \det Q = -15, \quad Q^{-1} = \begin{pmatrix} \frac{4}{15} & \frac{1}{5} \\ \frac{1}{3} & 0 \end{pmatrix} $$

$$ \begin{aligned} (QJ)_{11} &= 0\cdot2 + 3\cdot(-2) = -6 \\ (QJ)_{12} &= 0\cdot2 + 3\cdot2 = 6 \\ (QJ)_{21} &= 5\cdot2 + (-4)\cdot(-2) = 18 \\ (QJ)_{22} &= 5\cdot2 + (-4)\cdot2 = 2 \end{aligned} $$

$$ QJ = \begin{pmatrix} -6 & 6 \\ 18 & 2 \end{pmatrix} $$

$$ \begin{aligned} (A)_{11} &= (-6)\cdot\frac{4}{15} + 6\cdot\frac{1}{3} = \frac{2}{5} \\ (A)_{12} &= (-6)\cdot\frac{1}{5} + 6\cdot0 = - \frac{6}{5} \\ (A)_{21} &= 18\cdot\frac{4}{15} + 2\cdot\frac{1}{3} = \frac{82}{15} \\ (A)_{22} &= 18\cdot\frac{1}{5} + 2\cdot0 = \frac{18}{5} \end{aligned} $$

$$ A = QJQ^{-1} = \begin{pmatrix} \frac{2}{5} & - \frac{6}{5} \\ \frac{82}{15} & \frac{18}{5} \end{pmatrix} $$

Check of the given sign: $\dot x$ at $(0, 1)$ equals $a_{12} = - \frac{6}{5}$, which is negative, in agreement with the statement ($\dot x < 0$).

How $A$ acts on $\mathbf u$ and $\mathbf w$

Separating the real and imaginary parts of $A\mathbf v = \lambda\mathbf v$ gives (Topic 4, Eq. 7.4) $A\mathbf u = \rho\mathbf u - \omega\mathbf w$ and $A\mathbf w = \omega\mathbf u + \rho\mathbf w$. Verification with the numbers:

$$ A\mathbf u = \begin{pmatrix} -6 \\ 18 \end{pmatrix}, \qquad \rho\mathbf u - \omega\mathbf w = 2\begin{pmatrix} 0 \\ 5 \end{pmatrix} - 2\begin{pmatrix} 3 \\ -4 \end{pmatrix} = \begin{pmatrix} -6 \\ 18 \end{pmatrix} $$

$$ A\mathbf w = \begin{pmatrix} 6 \\ 2 \end{pmatrix}, \qquad \omega\mathbf u + \rho\mathbf w = 2\begin{pmatrix} 0 \\ 5 \end{pmatrix} + 2\begin{pmatrix} 3 \\ -4 \end{pmatrix} = \begin{pmatrix} 6 \\ 2 \end{pmatrix} $$

Change of coordinates

$\mathbf u$ and $\mathbf w$ are linearly independent (they are not parallel when $\omega \neq 0$; Topic 4, Lemma of Sec. 7.3). With the columns in the order $\mathbf u, \mathbf w$,

$$ Q = [\mathbf u \mid \mathbf w] = \begin{pmatrix} 0 & 3 \\ 5 & -4 \end{pmatrix} $$

$\det Q = (0)(-4) - (3)(5) = -15$, which is not zero, so $Q$ is invertible. For a $2\times 2$ matrix $\begin{pmatrix} a & b \\ c & d \end{pmatrix}^{-1} = \frac{1}{ad - bc}\begin{pmatrix} d & -b \\ -c & a \end{pmatrix}$ (Robinson, Appendix B):

$$ Q^{-1} = \frac{1}{-15}\begin{pmatrix} -4 & -3 \\ -5 & 0 \end{pmatrix} = \begin{pmatrix} \frac{4}{15} & \frac{1}{5} \\ \frac{1}{3} & 0 \end{pmatrix} $$

Compute $AQ = [A\mathbf u \mid A\mathbf w]$ and multiply by $Q^{-1}$ on the left, entry by entry:

$$ AQ = \begin{pmatrix} -6 & 6 \\ 18 & 2 \end{pmatrix} $$

$$ \begin{aligned} (Q^{-1}AQ)_{11} &= \frac{4}{15}\cdot(-6) + \frac{1}{5}\cdot18 = 2 \\ (Q^{-1}AQ)_{12} &= \frac{4}{15}\cdot6 + \frac{1}{5}\cdot2 = 2 \\ (Q^{-1}AQ)_{21} &= \frac{1}{3}\cdot(-6) + 0\cdot18 = -2 \\ (Q^{-1}AQ)_{22} &= \frac{1}{3}\cdot6 + 0\cdot2 = 2 \end{aligned} $$

$$ Q^{-1}AQ = \begin{pmatrix} 2 & 2 \\ -2 & 2 \end{pmatrix} = J = \begin{pmatrix} \rho & \omega \\ -\omega & \rho \end{pmatrix} $$

Solution in the canonical plane

$$ \dot z_1 = 2z_1 + 2z_2, \qquad \dot z_2 = -2z_1 + 2z_2 $$

The two equations are coupled, but they say "scale and rotate", which polar coordinates express directly: $z_1 = r\cos\theta$, $z_2 = r\sin\theta$ (Topic 4, Sec. 7.5).

The radius. From $r^2 = z_1^2 + z_2^2$ the chain rule gives $2r\dot r = 2z_1\dot z_1 + 2z_2\dot z_2$. Substituting $\dot z_1 = \rho z_1 + \omega z_2$ and $\dot z_2 = -\omega z_1 + \rho z_2$:

$$ r\dot r = z_1(\rho z_1 + \omega z_2) + z_2(-\omega z_1 + \rho z_2) = \rho z_1^2 + \omega z_1z_2 - \omega z_1z_2 + \rho z_2^2 = \rho r^2 $$

The terms in $\omega$ cancel. Dividing by $r \neq 0$ (the origin is the equilibrium): $\dot r = \rho r$.

The angle. Differentiate $\tan\theta = z_2/z_1$ with the chain rule on the left and the quotient rule on the right: $\sec^2\theta\,\dot\theta = (z_1\dot z_2 - z_2\dot z_1)/z_1^2$. Since $\sec^2\theta = r^2/z_1^2$, $\dot\theta = (z_1\dot z_2 - z_2\dot z_1)/r^2$, and

$$ z_1\dot z_2 - z_2\dot z_1 = z_1(-\omega z_1 + \rho z_2) - z_2(\rho z_1 + \omega z_2) = -\omega z_1^2 + \rho z_1z_2 - \rho z_1z_2 - \omega z_2^2 = -\omega r^2 \;\Longrightarrow\; \dot\theta = -\omega $$

Integration. $\frac{d}{dt}\left(e^{-\rho t}r\right) = e^{-\rho t}(\dot r - \rho r) = 0$, and Barrow's rule between $0$ and $t$, $\left[e^{-\rho s}r(s)\right]_{0}^{t} = 0$, gives $r(t) = r_0e^{\rho t}$. For the angle, $\theta(t) - \theta_0 = \int_0^t(-\omega)\,ds = -\omega t$. Back to Cartesian coordinates, with $c_1 = r_0\cos\theta_0 = z_1(0)$, $c_2 = r_0\sin\theta_0 = z_2(0)$ and the subtraction formulas $\cos(\theta_0 - \omega t) = \cos\theta_0\cos\omega t + \sin\theta_0\sin\omega t$, $\sin(\theta_0 - \omega t) = \sin\theta_0\cos\omega t - \cos\theta_0\sin\omega t$:

$$ z_1(t) = e^{2t}\left(c_1\cos 2t + c_2\sin 2t\right), \qquad z_2(t) = e^{2t}\left(c_2\cos 2t - c_1\sin 2t\right) $$

Check by substitution (product rule and chain rule): $\dot z_1 = \rho z_1 + e^{\rho t}\omega\left(-c_1\sin\omega t + c_2\cos\omega t\right) = \rho z_1 + \omega z_2$ and $\dot z_2 = \rho z_2 + e^{\rho t}\omega\left(-c_2\sin\omega t - c_1\cos\omega t\right) = \rho z_2 - \omega z_1$, as required.

Back to the original coordinates

$\mathbf x = Q\mathbf z = z_1\mathbf u + z_2\mathbf w$, so

$$ \mathbf x(t) = e^{2t}\left[\left(c_1\cos 2t + c_2\sin 2t\right)\begin{pmatrix} 0 \\ 5 \end{pmatrix} + \left(c_2\cos 2t - c_1\sin 2t\right)\begin{pmatrix} 3 \\ -4 \end{pmatrix}\right] $$

$$ x(t) = - 3 c_{1} e^{2 t} \sin{2 t} + 3 c_{2} e^{2 t} \cos{2 t}, \qquad y(t) = 4 c_{1} e^{2 t} \sin{2 t} + 5 c_{1} e^{2 t} \cos{2 t} + 5 c_{2} e^{2 t} \sin{2 t} - 4 c_{2} e^{2 t} \cos{2 t} $$

Phase portrait

$\rho = 2$ and $\omega = 2$: the origin is a unstable spiral. In the $\mathbf z$-plane $r = r_0e^{2t}$ and $\theta = \theta_0 - 2t$: the motion is always clockwise ($\dot\theta = -\omega < 0$), and one full turn takes $T = 2\pi/\omega$ while the radius is multiplied by $e^{2\pi\rho/\omega} = e^{2 \pi}$ (growing).

Sense of rotation in the $\mathbf x$-plane, tested directly on the vector field (Topic 4 Sec. 7.8): at $(0, 1)^T$, $\dot{\mathbf x} = A(0, 1)^T = \begin{pmatrix} - \frac{6}{5} \\ \frac{18}{5} \end{pmatrix}$, so $\dot x = - \frac{6}{5}$ is $< 0$: to the left at the top of the plane: the rotation is counter-clockwise. Cross-check: $\det Q = -15$ is $< 0$, so $Q$ reverses orientation and the clockwise motion of the $\mathbf z$-plane becomes counter-clockwise in the $\mathbf x$-plane, in agreement.

z1 z2 z-plane
Canonical $\mathbf z$-plane (unstable spiral): clockwise, $\dot\theta = -\omega$.
Open the figure full size
x y x-plane
Original $(x, y)$-plane (unstable spiral), counter-clockwise.
Open the figure full size
(vi) $\lambda_\pm = 5 \pm 3i \text{ with } \boldsymbol\eta_\pm = (2 \pm 5i,\ \mp i),\ \dot x < 0$

Split the eigenvector of $\lambda_+ = 5 + 3 i$ into real and imaginary parts, $\boldsymbol\eta_+ = \begin{pmatrix} 2 + 5 i \\ - i \end{pmatrix} = \mathbf u + i\mathbf w$, so $\rho = 5$, $\omega = 3$ and

$$ \mathbf u = \begin{pmatrix} 2 \\ 0 \end{pmatrix}, \qquad \mathbf w = \begin{pmatrix} 5 \\ -1 \end{pmatrix} $$

The matrix is not given; it is rebuilt from $A\mathbf u = \rho\mathbf u - \omega\mathbf w$, $A\mathbf w = \omega\mathbf u + \rho\mathbf w$ (Topic 4 Eq. 7.4), which say $AQ = QJ$ with $Q = [\mathbf u \mid \mathbf w]$, so $A = QJQ^{-1}$:

$$ Q = \begin{pmatrix} 2 & 5 \\ 0 & -1 \end{pmatrix}, \quad J = \begin{pmatrix} 5 & 3 \\ -3 & 5 \end{pmatrix}, \quad \det Q = -2, \quad Q^{-1} = \begin{pmatrix} \frac{1}{2} & \frac{5}{2} \\ 0 & -1 \end{pmatrix} $$

$$ \begin{aligned} (QJ)_{11} &= 2\cdot5 + 5\cdot(-3) = -5 \\ (QJ)_{12} &= 2\cdot3 + 5\cdot5 = 31 \\ (QJ)_{21} &= 0\cdot5 + (-1)\cdot(-3) = 3 \\ (QJ)_{22} &= 0\cdot3 + (-1)\cdot5 = -5 \end{aligned} $$

$$ QJ = \begin{pmatrix} -5 & 31 \\ 3 & -5 \end{pmatrix} $$

$$ \begin{aligned} (A)_{11} &= (-5)\cdot\frac{1}{2} + 31\cdot0 = - \frac{5}{2} \\ (A)_{12} &= (-5)\cdot\frac{5}{2} + 31\cdot(-1) = - \frac{87}{2} \\ (A)_{21} &= 3\cdot\frac{1}{2} + (-5)\cdot0 = \frac{3}{2} \\ (A)_{22} &= 3\cdot\frac{5}{2} + (-5)\cdot(-1) = \frac{25}{2} \end{aligned} $$

$$ A = QJQ^{-1} = \begin{pmatrix} - \frac{5}{2} & - \frac{87}{2} \\ \frac{3}{2} & \frac{25}{2} \end{pmatrix} $$

Check of the given sign: $\dot x$ at $(0, 1)$ equals $a_{12} = - \frac{87}{2}$, which is negative, in agreement with the statement ($\dot x < 0$).

How $A$ acts on $\mathbf u$ and $\mathbf w$

Separating the real and imaginary parts of $A\mathbf v = \lambda\mathbf v$ gives (Topic 4, Eq. 7.4) $A\mathbf u = \rho\mathbf u - \omega\mathbf w$ and $A\mathbf w = \omega\mathbf u + \rho\mathbf w$. Verification with the numbers:

$$ A\mathbf u = \begin{pmatrix} -5 \\ 3 \end{pmatrix}, \qquad \rho\mathbf u - \omega\mathbf w = 5\begin{pmatrix} 2 \\ 0 \end{pmatrix} - 3\begin{pmatrix} 5 \\ -1 \end{pmatrix} = \begin{pmatrix} -5 \\ 3 \end{pmatrix} $$

$$ A\mathbf w = \begin{pmatrix} 31 \\ -5 \end{pmatrix}, \qquad \omega\mathbf u + \rho\mathbf w = 3\begin{pmatrix} 2 \\ 0 \end{pmatrix} + 5\begin{pmatrix} 5 \\ -1 \end{pmatrix} = \begin{pmatrix} 31 \\ -5 \end{pmatrix} $$

Change of coordinates

$\mathbf u$ and $\mathbf w$ are linearly independent (they are not parallel when $\omega \neq 0$; Topic 4, Lemma of Sec. 7.3). With the columns in the order $\mathbf u, \mathbf w$,

$$ Q = [\mathbf u \mid \mathbf w] = \begin{pmatrix} 2 & 5 \\ 0 & -1 \end{pmatrix} $$

$\det Q = (2)(-1) - (5)(0) = -2$, which is not zero, so $Q$ is invertible. For a $2\times 2$ matrix $\begin{pmatrix} a & b \\ c & d \end{pmatrix}^{-1} = \frac{1}{ad - bc}\begin{pmatrix} d & -b \\ -c & a \end{pmatrix}$ (Robinson, Appendix B):

$$ Q^{-1} = \frac{1}{-2}\begin{pmatrix} -1 & -5 \\ 0 & 2 \end{pmatrix} = \begin{pmatrix} \frac{1}{2} & \frac{5}{2} \\ 0 & -1 \end{pmatrix} $$

Compute $AQ = [A\mathbf u \mid A\mathbf w]$ and multiply by $Q^{-1}$ on the left, entry by entry:

$$ AQ = \begin{pmatrix} -5 & 31 \\ 3 & -5 \end{pmatrix} $$

$$ \begin{aligned} (Q^{-1}AQ)_{11} &= \frac{1}{2}\cdot(-5) + \frac{5}{2}\cdot3 = 5 \\ (Q^{-1}AQ)_{12} &= \frac{1}{2}\cdot31 + \frac{5}{2}\cdot(-5) = 3 \\ (Q^{-1}AQ)_{21} &= 0\cdot(-5) + (-1)\cdot3 = -3 \\ (Q^{-1}AQ)_{22} &= 0\cdot31 + (-1)\cdot(-5) = 5 \end{aligned} $$

$$ Q^{-1}AQ = \begin{pmatrix} 5 & 3 \\ -3 & 5 \end{pmatrix} = J = \begin{pmatrix} \rho & \omega \\ -\omega & \rho \end{pmatrix} $$

Solution in the canonical plane

$$ \dot z_1 = 5z_1 + 3z_2, \qquad \dot z_2 = -3z_1 + 5z_2 $$

The two equations are coupled, but they say "scale and rotate", which polar coordinates express directly: $z_1 = r\cos\theta$, $z_2 = r\sin\theta$ (Topic 4, Sec. 7.5).

The radius. From $r^2 = z_1^2 + z_2^2$ the chain rule gives $2r\dot r = 2z_1\dot z_1 + 2z_2\dot z_2$. Substituting $\dot z_1 = \rho z_1 + \omega z_2$ and $\dot z_2 = -\omega z_1 + \rho z_2$:

$$ r\dot r = z_1(\rho z_1 + \omega z_2) + z_2(-\omega z_1 + \rho z_2) = \rho z_1^2 + \omega z_1z_2 - \omega z_1z_2 + \rho z_2^2 = \rho r^2 $$

The terms in $\omega$ cancel. Dividing by $r \neq 0$ (the origin is the equilibrium): $\dot r = \rho r$.

The angle. Differentiate $\tan\theta = z_2/z_1$ with the chain rule on the left and the quotient rule on the right: $\sec^2\theta\,\dot\theta = (z_1\dot z_2 - z_2\dot z_1)/z_1^2$. Since $\sec^2\theta = r^2/z_1^2$, $\dot\theta = (z_1\dot z_2 - z_2\dot z_1)/r^2$, and

$$ z_1\dot z_2 - z_2\dot z_1 = z_1(-\omega z_1 + \rho z_2) - z_2(\rho z_1 + \omega z_2) = -\omega z_1^2 + \rho z_1z_2 - \rho z_1z_2 - \omega z_2^2 = -\omega r^2 \;\Longrightarrow\; \dot\theta = -\omega $$

Integration. $\frac{d}{dt}\left(e^{-\rho t}r\right) = e^{-\rho t}(\dot r - \rho r) = 0$, and Barrow's rule between $0$ and $t$, $\left[e^{-\rho s}r(s)\right]_{0}^{t} = 0$, gives $r(t) = r_0e^{\rho t}$. For the angle, $\theta(t) - \theta_0 = \int_0^t(-\omega)\,ds = -\omega t$. Back to Cartesian coordinates, with $c_1 = r_0\cos\theta_0 = z_1(0)$, $c_2 = r_0\sin\theta_0 = z_2(0)$ and the subtraction formulas $\cos(\theta_0 - \omega t) = \cos\theta_0\cos\omega t + \sin\theta_0\sin\omega t$, $\sin(\theta_0 - \omega t) = \sin\theta_0\cos\omega t - \cos\theta_0\sin\omega t$:

$$ z_1(t) = e^{5t}\left(c_1\cos 3t + c_2\sin 3t\right), \qquad z_2(t) = e^{5t}\left(c_2\cos 3t - c_1\sin 3t\right) $$

Check by substitution (product rule and chain rule): $\dot z_1 = \rho z_1 + e^{\rho t}\omega\left(-c_1\sin\omega t + c_2\cos\omega t\right) = \rho z_1 + \omega z_2$ and $\dot z_2 = \rho z_2 + e^{\rho t}\omega\left(-c_2\sin\omega t - c_1\cos\omega t\right) = \rho z_2 - \omega z_1$, as required.

Back to the original coordinates

$\mathbf x = Q\mathbf z = z_1\mathbf u + z_2\mathbf w$, so

$$ \mathbf x(t) = e^{5t}\left[\left(c_1\cos 3t + c_2\sin 3t\right)\begin{pmatrix} 2 \\ 0 \end{pmatrix} + \left(c_2\cos 3t - c_1\sin 3t\right)\begin{pmatrix} 5 \\ -1 \end{pmatrix}\right] $$

$$ x(t) = - 5 c_{1} e^{5 t} \sin{3 t} + 2 c_{1} e^{5 t} \cos{3 t} + 2 c_{2} e^{5 t} \sin{3 t} + 5 c_{2} e^{5 t} \cos{3 t}, \qquad y(t) = c_{1} e^{5 t} \sin{3 t} - c_{2} e^{5 t} \cos{3 t} $$

Phase portrait

$\rho = 5$ and $\omega = 3$: the origin is a unstable spiral. In the $\mathbf z$-plane $r = r_0e^{5t}$ and $\theta = \theta_0 - 3t$: the motion is always clockwise ($\dot\theta = -\omega < 0$), and one full turn takes $T = 2\pi/\omega$ while the radius is multiplied by $e^{2\pi\rho/\omega} = e^{\frac{10 \pi}{3}}$ (growing).

Sense of rotation in the $\mathbf x$-plane, tested directly on the vector field (Topic 4 Sec. 7.8): at $(0, 1)^T$, $\dot{\mathbf x} = A(0, 1)^T = \begin{pmatrix} - \frac{87}{2} \\ \frac{25}{2} \end{pmatrix}$, so $\dot x = - \frac{87}{2}$ is $< 0$: to the left at the top of the plane: the rotation is counter-clockwise. Cross-check: $\det Q = -2$ is $< 0$, so $Q$ reverses orientation and the clockwise motion of the $\mathbf z$-plane becomes counter-clockwise in the $\mathbf x$-plane, in agreement.

z1 z2 z-plane
Canonical $\mathbf z$-plane (unstable spiral): clockwise, $\dot\theta = -\omega$.
Open the figure full size
x y x-plane
Original $(x, y)$-plane (unstable spiral), counter-clockwise.
Open the figure full size
(vii) $\lambda_\pm = \pm 7i \text{ with } \boldsymbol\eta_\pm = (1 \pm i,\ -1 \pm 2i),\ \dot x > 0$

Split the eigenvector of $\lambda_+ = 7 i$ into real and imaginary parts, $\boldsymbol\eta_+ = \begin{pmatrix} 1 + i \\ -1 + 2 i \end{pmatrix} = \mathbf u + i\mathbf w$, so $\rho = 0$, $\omega = 7$ and

$$ \mathbf u = \begin{pmatrix} 1 \\ -1 \end{pmatrix}, \qquad \mathbf w = \begin{pmatrix} 1 \\ 2 \end{pmatrix} $$

The matrix is not given; it is rebuilt from $A\mathbf u = \rho\mathbf u - \omega\mathbf w$, $A\mathbf w = \omega\mathbf u + \rho\mathbf w$ (Topic 4 Eq. 7.4), which say $AQ = QJ$ with $Q = [\mathbf u \mid \mathbf w]$, so $A = QJQ^{-1}$:

$$ Q = \begin{pmatrix} 1 & 1 \\ -1 & 2 \end{pmatrix}, \quad J = \begin{pmatrix} 0 & 7 \\ -7 & 0 \end{pmatrix}, \quad \det Q = 3, \quad Q^{-1} = \begin{pmatrix} \frac{2}{3} & - \frac{1}{3} \\ \frac{1}{3} & \frac{1}{3} \end{pmatrix} $$

$$ \begin{aligned} (QJ)_{11} &= 1\cdot0 + 1\cdot(-7) = -7 \\ (QJ)_{12} &= 1\cdot7 + 1\cdot0 = 7 \\ (QJ)_{21} &= (-1)\cdot0 + 2\cdot(-7) = -14 \\ (QJ)_{22} &= (-1)\cdot7 + 2\cdot0 = -7 \end{aligned} $$

$$ QJ = \begin{pmatrix} -7 & 7 \\ -14 & -7 \end{pmatrix} $$

$$ \begin{aligned} (A)_{11} &= (-7)\cdot\frac{2}{3} + 7\cdot\frac{1}{3} = - \frac{7}{3} \\ (A)_{12} &= (-7)\cdot(- \frac{1}{3}) + 7\cdot\frac{1}{3} = \frac{14}{3} \\ (A)_{21} &= (-14)\cdot\frac{2}{3} + (-7)\cdot\frac{1}{3} = - \frac{35}{3} \\ (A)_{22} &= (-14)\cdot(- \frac{1}{3}) + (-7)\cdot\frac{1}{3} = \frac{7}{3} \end{aligned} $$

$$ A = QJQ^{-1} = \begin{pmatrix} - \frac{7}{3} & \frac{14}{3} \\ - \frac{35}{3} & \frac{7}{3} \end{pmatrix} $$

Check of the given sign: $\dot x$ at $(0, 1)$ equals $a_{12} = \frac{14}{3}$, which is positive, in agreement with the statement ($\dot x > 0$).

How $A$ acts on $\mathbf u$ and $\mathbf w$

Separating the real and imaginary parts of $A\mathbf v = \lambda\mathbf v$ gives (Topic 4, Eq. 7.4) $A\mathbf u = \rho\mathbf u - \omega\mathbf w$ and $A\mathbf w = \omega\mathbf u + \rho\mathbf w$. Verification with the numbers:

$$ A\mathbf u = \begin{pmatrix} -7 \\ -14 \end{pmatrix}, \qquad \rho\mathbf u - \omega\mathbf w = 0\begin{pmatrix} 1 \\ -1 \end{pmatrix} - 7\begin{pmatrix} 1 \\ 2 \end{pmatrix} = \begin{pmatrix} -7 \\ -14 \end{pmatrix} $$

$$ A\mathbf w = \begin{pmatrix} 7 \\ -7 \end{pmatrix}, \qquad \omega\mathbf u + \rho\mathbf w = 7\begin{pmatrix} 1 \\ -1 \end{pmatrix} + 0\begin{pmatrix} 1 \\ 2 \end{pmatrix} = \begin{pmatrix} 7 \\ -7 \end{pmatrix} $$

Change of coordinates

$\mathbf u$ and $\mathbf w$ are linearly independent (they are not parallel when $\omega \neq 0$; Topic 4, Lemma of Sec. 7.3). With the columns in the order $\mathbf u, \mathbf w$,

$$ Q = [\mathbf u \mid \mathbf w] = \begin{pmatrix} 1 & 1 \\ -1 & 2 \end{pmatrix} $$

$\det Q = (1)(2) - (1)(-1) = 3$, which is not zero, so $Q$ is invertible. For a $2\times 2$ matrix $\begin{pmatrix} a & b \\ c & d \end{pmatrix}^{-1} = \frac{1}{ad - bc}\begin{pmatrix} d & -b \\ -c & a \end{pmatrix}$ (Robinson, Appendix B):

$$ Q^{-1} = \frac{1}{3}\begin{pmatrix} 2 & -1 \\ 1 & 1 \end{pmatrix} = \begin{pmatrix} \frac{2}{3} & - \frac{1}{3} \\ \frac{1}{3} & \frac{1}{3} \end{pmatrix} $$

Compute $AQ = [A\mathbf u \mid A\mathbf w]$ and multiply by $Q^{-1}$ on the left, entry by entry:

$$ AQ = \begin{pmatrix} -7 & 7 \\ -14 & -7 \end{pmatrix} $$

$$ \begin{aligned} (Q^{-1}AQ)_{11} &= \frac{2}{3}\cdot(-7) + (- \frac{1}{3})\cdot(-14) = 0 \\ (Q^{-1}AQ)_{12} &= \frac{2}{3}\cdot7 + (- \frac{1}{3})\cdot(-7) = 7 \\ (Q^{-1}AQ)_{21} &= \frac{1}{3}\cdot(-7) + \frac{1}{3}\cdot(-14) = -7 \\ (Q^{-1}AQ)_{22} &= \frac{1}{3}\cdot7 + \frac{1}{3}\cdot(-7) = 0 \end{aligned} $$

$$ Q^{-1}AQ = \begin{pmatrix} 0 & 7 \\ -7 & 0 \end{pmatrix} = J = \begin{pmatrix} \rho & \omega \\ -\omega & \rho \end{pmatrix} $$

Solution in the canonical plane

$$ \dot z_1 = 7z_2, \qquad \dot z_2 = -7z_1 $$

The two equations are coupled, but they say "scale and rotate", which polar coordinates express directly: $z_1 = r\cos\theta$, $z_2 = r\sin\theta$ (Topic 4, Sec. 7.5).

The radius. From $r^2 = z_1^2 + z_2^2$ the chain rule gives $2r\dot r = 2z_1\dot z_1 + 2z_2\dot z_2$. Substituting $\dot z_1 = \rho z_1 + \omega z_2$ and $\dot z_2 = -\omega z_1 + \rho z_2$:

$$ r\dot r = z_1(\rho z_1 + \omega z_2) + z_2(-\omega z_1 + \rho z_2) = \rho z_1^2 + \omega z_1z_2 - \omega z_1z_2 + \rho z_2^2 = \rho r^2 $$

The terms in $\omega$ cancel. Dividing by $r \neq 0$ (the origin is the equilibrium): $\dot r = \rho r$.

The angle. Differentiate $\tan\theta = z_2/z_1$ with the chain rule on the left and the quotient rule on the right: $\sec^2\theta\,\dot\theta = (z_1\dot z_2 - z_2\dot z_1)/z_1^2$. Since $\sec^2\theta = r^2/z_1^2$, $\dot\theta = (z_1\dot z_2 - z_2\dot z_1)/r^2$, and

$$ z_1\dot z_2 - z_2\dot z_1 = z_1(-\omega z_1 + \rho z_2) - z_2(\rho z_1 + \omega z_2) = -\omega z_1^2 + \rho z_1z_2 - \rho z_1z_2 - \omega z_2^2 = -\omega r^2 \;\Longrightarrow\; \dot\theta = -\omega $$

Integration. $\frac{d}{dt}\left(e^{-\rho t}r\right) = e^{-\rho t}(\dot r - \rho r) = 0$, and Barrow's rule between $0$ and $t$, $\left[e^{-\rho s}r(s)\right]_{0}^{t} = 0$, gives $r(t) = r_0e^{\rho t}$. For the angle, $\theta(t) - \theta_0 = \int_0^t(-\omega)\,ds = -\omega t$. Back to Cartesian coordinates, with $c_1 = r_0\cos\theta_0 = z_1(0)$, $c_2 = r_0\sin\theta_0 = z_2(0)$ and the subtraction formulas $\cos(\theta_0 - \omega t) = \cos\theta_0\cos\omega t + \sin\theta_0\sin\omega t$, $\sin(\theta_0 - \omega t) = \sin\theta_0\cos\omega t - \cos\theta_0\sin\omega t$:

$$ z_1(t) = e^{0t}\left(c_1\cos 7t + c_2\sin 7t\right), \qquad z_2(t) = e^{0t}\left(c_2\cos 7t - c_1\sin 7t\right) $$

Check by substitution (product rule and chain rule): $\dot z_1 = \rho z_1 + e^{\rho t}\omega\left(-c_1\sin\omega t + c_2\cos\omega t\right) = \rho z_1 + \omega z_2$ and $\dot z_2 = \rho z_2 + e^{\rho t}\omega\left(-c_2\sin\omega t - c_1\cos\omega t\right) = \rho z_2 - \omega z_1$, as required.

Back to the original coordinates

$\mathbf x = Q\mathbf z = z_1\mathbf u + z_2\mathbf w$, so

$$ \mathbf x(t) = e^{0t}\left[\left(c_1\cos 7t + c_2\sin 7t\right)\begin{pmatrix} 1 \\ -1 \end{pmatrix} + \left(c_2\cos 7t - c_1\sin 7t\right)\begin{pmatrix} 1 \\ 2 \end{pmatrix}\right] $$

$$ x(t) = - c_{1} \sin{7 t} + c_{1} \cos{7 t} + c_{2} \sin{7 t} + c_{2} \cos{7 t}, \qquad y(t) = - 2 c_{1} \sin{7 t} - c_{1} \cos{7 t} - c_{2} \sin{7 t} + 2 c_{2} \cos{7 t} $$

Phase portrait

$\rho = 0$ and $\omega = 7$: the origin is a centre. In the $\mathbf z$-plane $r = r_0e^{0t}$ and $\theta = \theta_0 - 7t$: the motion is always clockwise ($\dot\theta = -\omega < 0$), and the orbits are circles of period $T = 2\pi/\omega$ (closed orbits).

Sense of rotation in the $\mathbf x$-plane, tested directly on the vector field (Topic 4 Sec. 7.8): at $(0, 1)^T$, $\dot{\mathbf x} = A(0, 1)^T = \begin{pmatrix} \frac{14}{3} \\ \frac{7}{3} \end{pmatrix}$, so $\dot x = \frac{14}{3}$ is $> 0$: to the right at the top of the plane: the rotation is clockwise. Cross-check: $\det Q = 3$ is $> 0$, so $Q$ preserves orientation and the clockwise motion of the $\mathbf z$-plane becomes clockwise in the $\mathbf x$-plane, in agreement.

z1 z2 z-plane
Canonical $\mathbf z$-plane (centre): clockwise, $\dot\theta = -\omega$.
Open the figure full size
x y x-plane
Original $(x, y)$-plane (centre), clockwise.
Open the figure full size
(viii) $\lambda_\pm = -13 \pm 17i \text{ with } \boldsymbol\eta_\pm = (\pm 6i - 8,\ 4 \mp 5i),\ \dot x > 0$

Split the eigenvector of $\lambda_+ = -13 + 17 i$ into real and imaginary parts, $\boldsymbol\eta_+ = \begin{pmatrix} -8 + 6 i \\ 4 - 5 i \end{pmatrix} = \mathbf u + i\mathbf w$, so $\rho = -13$, $\omega = 17$ and

$$ \mathbf u = \begin{pmatrix} -8 \\ 4 \end{pmatrix}, \qquad \mathbf w = \begin{pmatrix} 6 \\ -5 \end{pmatrix} $$

The matrix is not given; it is rebuilt from $A\mathbf u = \rho\mathbf u - \omega\mathbf w$, $A\mathbf w = \omega\mathbf u + \rho\mathbf w$ (Topic 4 Eq. 7.4), which say $AQ = QJ$ with $Q = [\mathbf u \mid \mathbf w]$, so $A = QJQ^{-1}$:

$$ Q = \begin{pmatrix} -8 & 6 \\ 4 & -5 \end{pmatrix}, \quad J = \begin{pmatrix} -13 & 17 \\ -17 & -13 \end{pmatrix}, \quad \det Q = 16, \quad Q^{-1} = \begin{pmatrix} - \frac{5}{16} & - \frac{3}{8} \\ - \frac{1}{4} & - \frac{1}{2} \end{pmatrix} $$

$$ \begin{aligned} (QJ)_{11} &= (-8)\cdot(-13) + 6\cdot(-17) = 2 \\ (QJ)_{12} &= (-8)\cdot17 + 6\cdot(-13) = -214 \\ (QJ)_{21} &= 4\cdot(-13) + (-5)\cdot(-17) = 33 \\ (QJ)_{22} &= 4\cdot17 + (-5)\cdot(-13) = 133 \end{aligned} $$

$$ QJ = \begin{pmatrix} 2 & -214 \\ 33 & 133 \end{pmatrix} $$

$$ \begin{aligned} (A)_{11} &= 2\cdot(- \frac{5}{16}) + (-214)\cdot(- \frac{1}{4}) = \frac{423}{8} \\ (A)_{12} &= 2\cdot(- \frac{3}{8}) + (-214)\cdot(- \frac{1}{2}) = \frac{425}{4} \\ (A)_{21} &= 33\cdot(- \frac{5}{16}) + 133\cdot(- \frac{1}{4}) = - \frac{697}{16} \\ (A)_{22} &= 33\cdot(- \frac{3}{8}) + 133\cdot(- \frac{1}{2}) = - \frac{631}{8} \end{aligned} $$

$$ A = QJQ^{-1} = \begin{pmatrix} \frac{423}{8} & \frac{425}{4} \\ - \frac{697}{16} & - \frac{631}{8} \end{pmatrix} $$

Check of the given sign: $\dot x$ at $(0, 1)$ equals $a_{12} = \frac{425}{4}$, which is positive, in agreement with the statement ($\dot x > 0$).

How $A$ acts on $\mathbf u$ and $\mathbf w$

Separating the real and imaginary parts of $A\mathbf v = \lambda\mathbf v$ gives (Topic 4, Eq. 7.4) $A\mathbf u = \rho\mathbf u - \omega\mathbf w$ and $A\mathbf w = \omega\mathbf u + \rho\mathbf w$. Verification with the numbers:

$$ A\mathbf u = \begin{pmatrix} 2 \\ 33 \end{pmatrix}, \qquad \rho\mathbf u - \omega\mathbf w = (-13)\begin{pmatrix} -8 \\ 4 \end{pmatrix} - 17\begin{pmatrix} 6 \\ -5 \end{pmatrix} = \begin{pmatrix} 2 \\ 33 \end{pmatrix} $$

$$ A\mathbf w = \begin{pmatrix} -214 \\ 133 \end{pmatrix}, \qquad \omega\mathbf u + \rho\mathbf w = 17\begin{pmatrix} -8 \\ 4 \end{pmatrix} + (-13)\begin{pmatrix} 6 \\ -5 \end{pmatrix} = \begin{pmatrix} -214 \\ 133 \end{pmatrix} $$

Change of coordinates

$\mathbf u$ and $\mathbf w$ are linearly independent (they are not parallel when $\omega \neq 0$; Topic 4, Lemma of Sec. 7.3). With the columns in the order $\mathbf u, \mathbf w$,

$$ Q = [\mathbf u \mid \mathbf w] = \begin{pmatrix} -8 & 6 \\ 4 & -5 \end{pmatrix} $$

$\det Q = (-8)(-5) - (6)(4) = 16$, which is not zero, so $Q$ is invertible. For a $2\times 2$ matrix $\begin{pmatrix} a & b \\ c & d \end{pmatrix}^{-1} = \frac{1}{ad - bc}\begin{pmatrix} d & -b \\ -c & a \end{pmatrix}$ (Robinson, Appendix B):

$$ Q^{-1} = \frac{1}{16}\begin{pmatrix} -5 & -6 \\ -4 & -8 \end{pmatrix} = \begin{pmatrix} - \frac{5}{16} & - \frac{3}{8} \\ - \frac{1}{4} & - \frac{1}{2} \end{pmatrix} $$

Compute $AQ = [A\mathbf u \mid A\mathbf w]$ and multiply by $Q^{-1}$ on the left, entry by entry:

$$ AQ = \begin{pmatrix} 2 & -214 \\ 33 & 133 \end{pmatrix} $$

$$ \begin{aligned} (Q^{-1}AQ)_{11} &= (- \frac{5}{16})\cdot2 + (- \frac{3}{8})\cdot33 = -13 \\ (Q^{-1}AQ)_{12} &= (- \frac{5}{16})\cdot(-214) + (- \frac{3}{8})\cdot133 = 17 \\ (Q^{-1}AQ)_{21} &= (- \frac{1}{4})\cdot2 + (- \frac{1}{2})\cdot33 = -17 \\ (Q^{-1}AQ)_{22} &= (- \frac{1}{4})\cdot(-214) + (- \frac{1}{2})\cdot133 = -13 \end{aligned} $$

$$ Q^{-1}AQ = \begin{pmatrix} -13 & 17 \\ -17 & -13 \end{pmatrix} = J = \begin{pmatrix} \rho & \omega \\ -\omega & \rho \end{pmatrix} $$

Solution in the canonical plane

$$ \dot z_1 = -13z_1 + 17z_2, \qquad \dot z_2 = -17z_1 - 13z_2 $$

The two equations are coupled, but they say "scale and rotate", which polar coordinates express directly: $z_1 = r\cos\theta$, $z_2 = r\sin\theta$ (Topic 4, Sec. 7.5).

The radius. From $r^2 = z_1^2 + z_2^2$ the chain rule gives $2r\dot r = 2z_1\dot z_1 + 2z_2\dot z_2$. Substituting $\dot z_1 = \rho z_1 + \omega z_2$ and $\dot z_2 = -\omega z_1 + \rho z_2$:

$$ r\dot r = z_1(\rho z_1 + \omega z_2) + z_2(-\omega z_1 + \rho z_2) = \rho z_1^2 + \omega z_1z_2 - \omega z_1z_2 + \rho z_2^2 = \rho r^2 $$

The terms in $\omega$ cancel. Dividing by $r \neq 0$ (the origin is the equilibrium): $\dot r = \rho r$.

The angle. Differentiate $\tan\theta = z_2/z_1$ with the chain rule on the left and the quotient rule on the right: $\sec^2\theta\,\dot\theta = (z_1\dot z_2 - z_2\dot z_1)/z_1^2$. Since $\sec^2\theta = r^2/z_1^2$, $\dot\theta = (z_1\dot z_2 - z_2\dot z_1)/r^2$, and

$$ z_1\dot z_2 - z_2\dot z_1 = z_1(-\omega z_1 + \rho z_2) - z_2(\rho z_1 + \omega z_2) = -\omega z_1^2 + \rho z_1z_2 - \rho z_1z_2 - \omega z_2^2 = -\omega r^2 \;\Longrightarrow\; \dot\theta = -\omega $$

Integration. $\frac{d}{dt}\left(e^{-\rho t}r\right) = e^{-\rho t}(\dot r - \rho r) = 0$, and Barrow's rule between $0$ and $t$, $\left[e^{-\rho s}r(s)\right]_{0}^{t} = 0$, gives $r(t) = r_0e^{\rho t}$. For the angle, $\theta(t) - \theta_0 = \int_0^t(-\omega)\,ds = -\omega t$. Back to Cartesian coordinates, with $c_1 = r_0\cos\theta_0 = z_1(0)$, $c_2 = r_0\sin\theta_0 = z_2(0)$ and the subtraction formulas $\cos(\theta_0 - \omega t) = \cos\theta_0\cos\omega t + \sin\theta_0\sin\omega t$, $\sin(\theta_0 - \omega t) = \sin\theta_0\cos\omega t - \cos\theta_0\sin\omega t$:

$$ z_1(t) = e^{-13t}\left(c_1\cos 17t + c_2\sin 17t\right), \qquad z_2(t) = e^{-13t}\left(c_2\cos 17t - c_1\sin 17t\right) $$

Check by substitution (product rule and chain rule): $\dot z_1 = \rho z_1 + e^{\rho t}\omega\left(-c_1\sin\omega t + c_2\cos\omega t\right) = \rho z_1 + \omega z_2$ and $\dot z_2 = \rho z_2 + e^{\rho t}\omega\left(-c_2\sin\omega t - c_1\cos\omega t\right) = \rho z_2 - \omega z_1$, as required.

Back to the original coordinates

$\mathbf x = Q\mathbf z = z_1\mathbf u + z_2\mathbf w$, so

$$ \mathbf x(t) = e^{-13t}\left[\left(c_1\cos 17t + c_2\sin 17t\right)\begin{pmatrix} -8 \\ 4 \end{pmatrix} + \left(c_2\cos 17t - c_1\sin 17t\right)\begin{pmatrix} 6 \\ -5 \end{pmatrix}\right] $$

$$ x(t) = - 6 c_{1} e^{-13 t} \sin{17 t} - 8 c_{1} e^{-13 t} \cos{17 t} - 8 c_{2} e^{-13 t} \sin{17 t} + 6 c_{2} e^{-13 t} \cos{17 t}, \qquad y(t) = 5 c_{1} e^{-13 t} \sin{17 t} + 4 c_{1} e^{-13 t} \cos{17 t} + 4 c_{2} e^{-13 t} \sin{17 t} - 5 c_{2} e^{-13 t} \cos{17 t} $$

Phase portrait

$\rho = -13$ and $\omega = 17$: the origin is a stable spiral. In the $\mathbf z$-plane $r = r_0e^{-13t}$ and $\theta = \theta_0 - 17t$: the motion is always clockwise ($\dot\theta = -\omega < 0$), and one full turn takes $T = 2\pi/\omega$ while the radius is multiplied by $e^{2\pi\rho/\omega} = e^{-\frac{26 \pi}{17}}$ (shrinking).

Sense of rotation in the $\mathbf x$-plane, tested directly on the vector field (Topic 4 Sec. 7.8): at $(0, 1)^T$, $\dot{\mathbf x} = A(0, 1)^T = \begin{pmatrix} \frac{425}{4} \\ - \frac{631}{8} \end{pmatrix}$, so $\dot x = \frac{425}{4}$ is $> 0$: to the right at the top of the plane: the rotation is clockwise. Cross-check: $\det Q = 16$ is $> 0$, so $Q$ preserves orientation and the clockwise motion of the $\mathbf z$-plane becomes clockwise in the $\mathbf x$-plane, in agreement.

z1 z2 z-plane
Canonical $\mathbf z$-plane (stable spiral): clockwise, $\dot\theta = -\omega$.
Open the figure full size
x y x-plane
Original $(x, y)$-plane (stable spiral), clockwise.
Open the figure full size
Interpretation and checks
Item$A$TypeSense of rotation
(i)$\begin{pmatrix} 7 & -3 \\ 15 & -5 \end{pmatrix}$unstable spiralcounter-clockwise
(ii)$\begin{pmatrix} -3 & -3 \\ 6 & 3 \end{pmatrix}$centrecounter-clockwise
(iii)$\begin{pmatrix} - \frac{5}{2} & \frac{1}{2} \\ - \frac{5}{2} & - \frac{3}{2} \end{pmatrix}$stable spiralclockwise
(iv)$\begin{pmatrix} -1 & 1 \\ -1 & -1 \end{pmatrix}$stable spiralclockwise
(v)$\begin{pmatrix} \frac{2}{5} & - \frac{6}{5} \\ \frac{82}{15} & \frac{18}{5} \end{pmatrix}$unstable spiralcounter-clockwise
(vi)$\begin{pmatrix} - \frac{5}{2} & - \frac{87}{2} \\ \frac{3}{2} & \frac{25}{2} \end{pmatrix}$unstable spiralcounter-clockwise
(vii)$\begin{pmatrix} - \frac{7}{3} & \frac{14}{3} \\ - \frac{35}{3} & \frac{7}{3} \end{pmatrix}$centreclockwise
(viii)$\begin{pmatrix} \frac{423}{8} & \frac{425}{4} \\ - \frac{697}{16} & - \frac{631}{8} \end{pmatrix}$stable spiralclockwise
  • In every item $\det Q$ and the sign of $a_{12}$ agree (the code asserts $\det Q > 0 \Leftrightarrow a_{12} > 0$), which is the Topic 4 rule: the $\mathbf z$-plane turns clockwise, and $Q$ reverses the sense exactly when $\det Q < 0$.
  • Centres (ii) and (vii) ($\rho = 0$): $\operatorname{tr}A = 0$ and $\det A = \omega^2 > 0$; the orbits are closed ellipses with period $2\pi/\omega$ ($2\pi/3$ and $2\pi/7$).
  • Spirals: (i), (v), (vi) unstable; (iii), (iv), (viii) stable. The growth/decay factor per turn is $e^{2\pi\rho/\omega}$, stated in each item; (viii) decays extremely fast ($e^{-26\pi/17} \approx 0.008$ per turn), so its spiral is nearly invisible after one turn.
  • The eigen-data are consistent: $A\boldsymbol\eta_+ = \lambda_+\boldsymbol\eta_+$ holds in each item by construction of $A$ (Eq. 7.4 verified entry by entry).
Exercise 2

General Solutions for the Eight Complex Eigen-data

Original statement · probls_ch4-ii_2627.pdf
Real General Solution • Euler's Formula
Statement

Write down the general solution of the equation $d\mathbf x/dt = A\mathbf x$, when the eigenvalues ($\lambda_\pm$) and eigenvectors ($\boldsymbol\eta_\pm$) of $A$ are those in the previous exercise.

Setup

The eight eigen-data are those of Exercise 1 (listed there). For each, $\boldsymbol\eta_+ = \mathbf u + i\mathbf w$ is the eigenvector of $\lambda_+ = \rho + i\omega$.

Formulation: the real general solution

Complex solutions $e^{\lambda_+t}\boldsymbol\eta_+$ and $e^{\lambda_-t}\bar{\boldsymbol\eta}_+$ combine into real ones: choose conjugate coefficients, $\mathbf x = 2\operatorname{Re}\left[Ce^{(\rho + i\omega)t}(\mathbf u + i\mathbf w)\right]$ with $C = \alpha + i\beta$. Expanding $e^{i\omega t} = \cos\omega t + i\sin\omega t$ (Euler) and keeping the real part, with $c_1 = 2\alpha$, $c_2 = -2\beta$ (Topic 4 Eq. 7.3, Robinson Eq. 29.2):

$$ \mathbf x(t) = e^{\rho t}\Big[\left(c_1\cos\omega t + c_2\sin\omega t\right)\mathbf u + \left(c_2\cos\omega t - c_1\sin\omega t\right)\mathbf w\Big] $$

Check of the derivation of the real part: $(a + ib)(\mathbf u + i\mathbf w)$ has real part $a\mathbf u - b\mathbf w$ because $i\cdot i = -1$; with $a = \alpha\cos\omega t - \beta\sin\omega t$ and $b = \beta\cos\omega t + \alpha\sin\omega t$ this is $\frac12[(c_1\cos\omega t + c_2\sin\omega t)\mathbf u + (c_2\cos\omega t - c_1\sin\omega t)\mathbf w]$ after substituting $\alpha = c_1/2$, $\beta = -c_2/2$.

Method of the class notes [Notes, p. 10]

The notes write the real general solution as $\mathbf x = C_1\operatorname{Re}\{\mathbf v e^{\lambda t}\} + C_2\operatorname{Im}\{\mathbf v e^{\lambda t}\}$ with $\mathbf v = \mathbf v_1 + i\mathbf v_2$ and $\lambda = \rho + i\omega$. Since $\mathbf ve^{\lambda t} = e^{\rho t}(\cos\omega t + i\sin\omega t)(\mathbf v_1 + i\mathbf v_2)$,

$$ \operatorname{Re}\{\mathbf ve^{\lambda t}\} = e^{\rho t}(\cos\omega t\,\mathbf v_1 - \sin\omega t\,\mathbf v_2), \qquad \operatorname{Im}\{\mathbf ve^{\lambda t}\} = e^{\rho t}(\sin\omega t\,\mathbf v_1 + \cos\omega t\,\mathbf v_2), $$

so it is the formula above with $c_1 = C_1$ and $c_2 = C_2$.

Derivation: the eight solutions
(i) $\lambda_\pm = 1 \pm 3i \text{ with } \boldsymbol\eta_\pm = (1,\ 2 \mp i),\ \dot x < 0$

$\rho = 1$, $\omega = 3$, $\mathbf u = \begin{pmatrix} 1 \\ 2 \end{pmatrix}$, $\mathbf w = \begin{pmatrix} 0 \\ -1 \end{pmatrix}$. Substituting in the general formula,

$$ \mathbf x(t) = e^{1t}\left[\left(c_1\cos 3t + c_2\sin 3t\right)\begin{pmatrix} 1 \\ 2 \end{pmatrix} + \left(c_2\cos 3t - c_1\sin 3t\right)\begin{pmatrix} 0 \\ -1 \end{pmatrix}\right] $$

Component by component:

$$ x(t) = c_{1} e^{t} \cos{3 t} + c_{2} e^{t} \sin{3 t}, \qquad y(t) = c_{1} e^{t} \sin{3 t} + 2 c_{1} e^{t} \cos{3 t} + 2 c_{2} e^{t} \sin{3 t} - c_{2} e^{t} \cos{3 t} $$

Check: with $A = \begin{pmatrix} 7 & -3 \\ 15 & -5 \end{pmatrix}$ (Exercise 1), $\dot{\mathbf x} - A\mathbf x = \mathbf 0$ identically (verified symbolically).

(ii) $\lambda_\pm = \pm 3i \text{ with } \boldsymbol\eta_\pm = (1 \pm 2i,\ 1 \mp 3i),\ \dot x < 0$

$\rho = 0$, $\omega = 3$, $\mathbf u = \begin{pmatrix} 1 \\ 1 \end{pmatrix}$, $\mathbf w = \begin{pmatrix} 2 \\ -3 \end{pmatrix}$. Substituting in the general formula,

$$ \mathbf x(t) = e^{0t}\left[\left(c_1\cos 3t + c_2\sin 3t\right)\begin{pmatrix} 1 \\ 1 \end{pmatrix} + \left(c_2\cos 3t - c_1\sin 3t\right)\begin{pmatrix} 2 \\ -3 \end{pmatrix}\right] $$

Component by component:

$$ x(t) = - 2 c_{1} \sin{3 t} + c_{1} \cos{3 t} + c_{2} \sin{3 t} + 2 c_{2} \cos{3 t}, \qquad y(t) = 3 c_{1} \sin{3 t} + c_{1} \cos{3 t} + c_{2} \sin{3 t} - 3 c_{2} \cos{3 t} $$

Check: with $A = \begin{pmatrix} -3 & -3 \\ 6 & 3 \end{pmatrix}$ (Exercise 1), $\dot{\mathbf x} - A\mathbf x = \mathbf 0$ identically (verified symbolically).

(iii) $\lambda_\pm = -2 \pm i \text{ with } \boldsymbol\eta_\pm = (1 \mp i,\ 3 \pm i),\ \dot x > 0$

$\rho = -2$, $\omega = 1$, $\mathbf u = \begin{pmatrix} 1 \\ 3 \end{pmatrix}$, $\mathbf w = \begin{pmatrix} -1 \\ 1 \end{pmatrix}$. Substituting in the general formula,

$$ \mathbf x(t) = e^{-2t}\left[\left(c_1\cos 1t + c_2\sin 1t\right)\begin{pmatrix} 1 \\ 3 \end{pmatrix} + \left(c_2\cos 1t - c_1\sin 1t\right)\begin{pmatrix} -1 \\ 1 \end{pmatrix}\right] $$

Component by component:

$$ x(t) = c_{1} e^{-2 t} \sin{t} + c_{1} e^{-2 t} \cos{t} + c_{2} e^{-2 t} \sin{t} - c_{2} e^{-2 t} \cos{t}, \qquad y(t) = - c_{1} e^{-2 t} \sin{t} + 3 c_{1} e^{-2 t} \cos{t} + 3 c_{2} e^{-2 t} \sin{t} + c_{2} e^{-2 t} \cos{t} $$

Check: with $A = \begin{pmatrix} - \frac{5}{2} & \frac{1}{2} \\ - \frac{5}{2} & - \frac{3}{2} \end{pmatrix}$ (Exercise 1), $\dot{\mathbf x} - A\mathbf x = \mathbf 0$ identically (verified symbolically).

(iv) $\lambda_\pm = -1 \pm i \text{ with } \boldsymbol\eta_\pm = (1,\ \pm i),\ \dot x > 0$

$\rho = -1$, $\omega = 1$, $\mathbf u = \begin{pmatrix} 1 \\ 0 \end{pmatrix}$, $\mathbf w = \begin{pmatrix} 0 \\ 1 \end{pmatrix}$. Substituting in the general formula,

$$ \mathbf x(t) = e^{-1t}\left[\left(c_1\cos 1t + c_2\sin 1t\right)\begin{pmatrix} 1 \\ 0 \end{pmatrix} + \left(c_2\cos 1t - c_1\sin 1t\right)\begin{pmatrix} 0 \\ 1 \end{pmatrix}\right] $$

Component by component:

$$ x(t) = c_{1} e^{-t} \cos{t} + c_{2} e^{-t} \sin{t}, \qquad y(t) = - c_{1} e^{-t} \sin{t} + c_{2} e^{-t} \cos{t} $$

Check: with $A = \begin{pmatrix} -1 & 1 \\ -1 & -1 \end{pmatrix}$ (Exercise 1), $\dot{\mathbf x} - A\mathbf x = \mathbf 0$ identically (verified symbolically).

(v) $\lambda_\pm = 2 \pm 2i \text{ with } \boldsymbol\eta_\pm = (\pm 3i,\ 5 \mp 4i),\ \dot x < 0$

$\rho = 2$, $\omega = 2$, $\mathbf u = \begin{pmatrix} 0 \\ 5 \end{pmatrix}$, $\mathbf w = \begin{pmatrix} 3 \\ -4 \end{pmatrix}$. Substituting in the general formula,

$$ \mathbf x(t) = e^{2t}\left[\left(c_1\cos 2t + c_2\sin 2t\right)\begin{pmatrix} 0 \\ 5 \end{pmatrix} + \left(c_2\cos 2t - c_1\sin 2t\right)\begin{pmatrix} 3 \\ -4 \end{pmatrix}\right] $$

Component by component:

$$ x(t) = - 3 c_{1} e^{2 t} \sin{2 t} + 3 c_{2} e^{2 t} \cos{2 t}, \qquad y(t) = 4 c_{1} e^{2 t} \sin{2 t} + 5 c_{1} e^{2 t} \cos{2 t} + 5 c_{2} e^{2 t} \sin{2 t} - 4 c_{2} e^{2 t} \cos{2 t} $$

Check: with $A = \begin{pmatrix} \frac{2}{5} & - \frac{6}{5} \\ \frac{82}{15} & \frac{18}{5} \end{pmatrix}$ (Exercise 1), $\dot{\mathbf x} - A\mathbf x = \mathbf 0$ identically (verified symbolically).

(vi) $\lambda_\pm = 5 \pm 3i \text{ with } \boldsymbol\eta_\pm = (2 \pm 5i,\ \mp i),\ \dot x < 0$

$\rho = 5$, $\omega = 3$, $\mathbf u = \begin{pmatrix} 2 \\ 0 \end{pmatrix}$, $\mathbf w = \begin{pmatrix} 5 \\ -1 \end{pmatrix}$. Substituting in the general formula,

$$ \mathbf x(t) = e^{5t}\left[\left(c_1\cos 3t + c_2\sin 3t\right)\begin{pmatrix} 2 \\ 0 \end{pmatrix} + \left(c_2\cos 3t - c_1\sin 3t\right)\begin{pmatrix} 5 \\ -1 \end{pmatrix}\right] $$

Component by component:

$$ x(t) = - 5 c_{1} e^{5 t} \sin{3 t} + 2 c_{1} e^{5 t} \cos{3 t} + 2 c_{2} e^{5 t} \sin{3 t} + 5 c_{2} e^{5 t} \cos{3 t}, \qquad y(t) = c_{1} e^{5 t} \sin{3 t} - c_{2} e^{5 t} \cos{3 t} $$

Check: with $A = \begin{pmatrix} - \frac{5}{2} & - \frac{87}{2} \\ \frac{3}{2} & \frac{25}{2} \end{pmatrix}$ (Exercise 1), $\dot{\mathbf x} - A\mathbf x = \mathbf 0$ identically (verified symbolically).

(vii) $\lambda_\pm = \pm 7i \text{ with } \boldsymbol\eta_\pm = (1 \pm i,\ -1 \pm 2i),\ \dot x > 0$

$\rho = 0$, $\omega = 7$, $\mathbf u = \begin{pmatrix} 1 \\ -1 \end{pmatrix}$, $\mathbf w = \begin{pmatrix} 1 \\ 2 \end{pmatrix}$. Substituting in the general formula,

$$ \mathbf x(t) = e^{0t}\left[\left(c_1\cos 7t + c_2\sin 7t\right)\begin{pmatrix} 1 \\ -1 \end{pmatrix} + \left(c_2\cos 7t - c_1\sin 7t\right)\begin{pmatrix} 1 \\ 2 \end{pmatrix}\right] $$

Component by component:

$$ x(t) = - c_{1} \sin{7 t} + c_{1} \cos{7 t} + c_{2} \sin{7 t} + c_{2} \cos{7 t}, \qquad y(t) = - 2 c_{1} \sin{7 t} - c_{1} \cos{7 t} - c_{2} \sin{7 t} + 2 c_{2} \cos{7 t} $$

Check: with $A = \begin{pmatrix} - \frac{7}{3} & \frac{14}{3} \\ - \frac{35}{3} & \frac{7}{3} \end{pmatrix}$ (Exercise 1), $\dot{\mathbf x} - A\mathbf x = \mathbf 0$ identically (verified symbolically).

(viii) $\lambda_\pm = -13 \pm 17i \text{ with } \boldsymbol\eta_\pm = (\pm 6i - 8,\ 4 \mp 5i),\ \dot x > 0$

$\rho = -13$, $\omega = 17$, $\mathbf u = \begin{pmatrix} -8 \\ 4 \end{pmatrix}$, $\mathbf w = \begin{pmatrix} 6 \\ -5 \end{pmatrix}$. Substituting in the general formula,

$$ \mathbf x(t) = e^{-13t}\left[\left(c_1\cos 17t + c_2\sin 17t\right)\begin{pmatrix} -8 \\ 4 \end{pmatrix} + \left(c_2\cos 17t - c_1\sin 17t\right)\begin{pmatrix} 6 \\ -5 \end{pmatrix}\right] $$

Component by component:

$$ x(t) = - 6 c_{1} e^{-13 t} \sin{17 t} - 8 c_{1} e^{-13 t} \cos{17 t} - 8 c_{2} e^{-13 t} \sin{17 t} + 6 c_{2} e^{-13 t} \cos{17 t}, \qquad y(t) = 5 c_{1} e^{-13 t} \sin{17 t} + 4 c_{1} e^{-13 t} \cos{17 t} + 4 c_{2} e^{-13 t} \sin{17 t} - 5 c_{2} e^{-13 t} \cos{17 t} $$

Check: with $A = \begin{pmatrix} \frac{423}{8} & \frac{425}{4} \\ - \frac{697}{16} & - \frac{631}{8} \end{pmatrix}$ (Exercise 1), $\dot{\mathbf x} - A\mathbf x = \mathbf 0$ identically (verified symbolically).

Interpretation and checks
  • The two constants are real and are the initial coordinates in the $(\mathbf u, \mathbf w)$ basis: $(c_1, c_2)^T = Q^{-1}\mathbf x_0$.
  • Centres $(\rho = 0)$: (ii) and (vii) give purely periodic $\mathbf x(t)$ with period $2\pi/3$ and $2\pi/7$; spirals have an envelope $e^{\rho t}$.
  • Linear independence: $e^{\rho t}(\cos\omega t\,\mathbf u - \sin\omega t\,\mathbf w)$ and $e^{\rho t}(\sin\omega t\,\mathbf u + \cos\omega t\,\mathbf w)$ have Wronskian $e^{2\rho t}\det[\mathbf u\ \mathbf w] \neq 0$ since $\mathbf u, \mathbf w$ are independent (Exercise 4).
  • Dimensional check: the exponent $\rho t$ and the phase $\omega t$ are dimensionless when $\rho$ and $\omega$ are rates in s$^{-1}$ and $t$ in s.
Exercise 3

Canonical Form of Four Systems with Complex Eigenvalues

Original statement · probls_ch4-ii_2627.pdf
Canonical Form • Q = [u | w] • Four Matrices
Statement

For the following equations find the eigenvalues and eigenvectors of the matrix on the right-hand side, and hence find the coordinate transformation that will put the equations into their standard simple (canonical) form. Show that this transformation has the desired effect.

  • (i) $\dfrac{d\mathbf x}{dt} = \begin{pmatrix} 0 & -1 \\ 1 & -1 \end{pmatrix}\mathbf x$;
  • (ii) $\dfrac{d\mathbf x}{dt} = \begin{pmatrix} -2 & 3 \\ -6 & 4 \end{pmatrix}\mathbf x$;
  • (iii) $\dfrac{d\mathbf x}{dt} = \begin{pmatrix} -11 & -2 \\ 13 & -9 \end{pmatrix}\mathbf x$;
  • (iv) $\dfrac{d\mathbf x}{dt} = \begin{pmatrix} 7 & -5 \\ 10 & -3 \end{pmatrix}\mathbf x$.
Setup

The standard (canonical) form for a complex pair $\rho \pm i\omega$ is $\dot{\mathbf z} = J\mathbf z$ with $J = \begin{pmatrix} \rho & \omega \\ -\omega & \rho \end{pmatrix}$, reached by $\mathbf z = Q^{-1}\mathbf x$ with $Q = [\mathbf u \mid \mathbf w]$, $\mathbf u + i\mathbf w$ the eigenvector of $\rho + i\omega$ (Topic 4 Sec. 7.4). For every matrix $\Delta = (\operatorname{tr}A)^2 - 4\det A < 0$ (checked in each item), so the eigenvalues are complex conjugate.

Formulation

Steps for each matrix: characteristic equation, eigenvalue $\rho + i\omega$, eigenvector via the first row, $\mathbf u, \mathbf w$, relations (7.4), $Q$ and $Q^{-1}$, and the explicit product $Q^{-1}AQ = J$.

Method of the class notes [Notes, p. 10]

The construction of the notes: complex diagonalisation $A = PDP^{-1}$ with $P = (\mathbf v_+ \mid \mathbf v_-)$, then $Q = (\mathbf v_1 \mid \mathbf v_2) = PM$ with $M = \begin{pmatrix} 1/2 & -i/2 \\ 1/2 & i/2 \end{pmatrix}$, so that $Q^{-1}AQ = M^{-1}P^{-1}APM = M^{-1}DM = \begin{pmatrix} \rho & \omega \\ -\omega & \rho \end{pmatrix}$. The explicit product $Q^{-1}AQ$ computed for each matrix below checks this canonical form.

Derivation
(i) $A = \begin{pmatrix} 0 & -1 \\ 1 & -1 \end{pmatrix}$

Characteristic equation $\det(A - \lambda I) = 0$ (Topic 4, Eq. 27.8):

$$ \det(A - \lambda I) = \begin{vmatrix} 0 - \lambda & -1 \\ 1 & -1 - \lambda \end{vmatrix} = (0 - \lambda)(-1 - \lambda) - (-1)(1) $$

$$ = \lambda^2 - (0 + (-1))\lambda + (0\cdot(-1) - (-1)\cdot1) = \lambda^2 - (-1)\lambda + (1) = \lambda^{2} + \lambda + 1 $$

$\operatorname{tr}A = -1$, $\det A = 1$, discriminant $\Delta = (\operatorname{tr}A)^2 - 4\det A = 1 - 4 = -3$.

$\Delta < 0$: a complex-conjugate pair. $\lambda = \dfrac{-1 \pm \sqrt{-3}}{2} = \dfrac{-1 \pm \sqrt{3} i}{2}$, so

$$ \lambda_\pm = - \frac{1}{2} \pm \frac{\sqrt{3}}{2}\,i \qquad (\rho = - \frac{1}{2},\ \omega = \frac{\sqrt{3}}{2} > 0) $$

Eigenvector of the eigenvalue with positive imaginary part

For $\lambda = \rho + i\omega = - \frac{1}{2} + \frac{\sqrt{3} i}{2}$ the first row of $(A - \lambda I)\mathbf v = \mathbf 0$ reads $(\frac{1}{2} - \frac{\sqrt{3} i}{2})\,v_1 + (-1)\,v_2 = 0$ (the coefficient of $v_2$ is non-zero because $\Delta < 0$ forces $bc < 0$). Choose $v_1 = -1$; then $v_2 = -(\frac{1}{2} - \frac{\sqrt{3} i}{2}) = - \frac{1}{2} + \frac{\sqrt{3} i}{2}$:

$$ \mathbf v = \begin{pmatrix} -1 \\ - \frac{1}{2} + \frac{\sqrt{3} i}{2} \end{pmatrix} = \underbrace{\begin{pmatrix} -1 \\ - \frac{1}{2} \end{pmatrix}}_{\mathbf u} + i\,\underbrace{\begin{pmatrix} 0 \\ \frac{\sqrt{3}}{2} \end{pmatrix}}_{\mathbf w} $$

Check with the second row: $1\cdot(-1) + (- \frac{1}{2} - \frac{\sqrt{3} i}{2})(- \frac{1}{2} + \frac{\sqrt{3} i}{2}) = -1 + 1 = 0$ (using $i^2 = -1$).

How $A$ acts on $\mathbf u$ and $\mathbf w$

Separating the real and imaginary parts of $A\mathbf v = \lambda\mathbf v$ gives (Topic 4, Eq. 7.4) $A\mathbf u = \rho\mathbf u - \omega\mathbf w$ and $A\mathbf w = \omega\mathbf u + \rho\mathbf w$. Verification with the numbers:

$$ A\mathbf u = \begin{pmatrix} \frac{1}{2} \\ - \frac{1}{2} \end{pmatrix}, \qquad \rho\mathbf u - \omega\mathbf w = (- \frac{1}{2})\begin{pmatrix} -1 \\ - \frac{1}{2} \end{pmatrix} - \frac{\sqrt{3}}{2}\begin{pmatrix} 0 \\ \frac{\sqrt{3}}{2} \end{pmatrix} = \begin{pmatrix} \frac{1}{2} \\ - \frac{1}{2} \end{pmatrix} $$

$$ A\mathbf w = \begin{pmatrix} - \frac{\sqrt{3}}{2} \\ - \frac{\sqrt{3}}{2} \end{pmatrix}, \qquad \omega\mathbf u + \rho\mathbf w = \frac{\sqrt{3}}{2}\begin{pmatrix} -1 \\ - \frac{1}{2} \end{pmatrix} + (- \frac{1}{2})\begin{pmatrix} 0 \\ \frac{\sqrt{3}}{2} \end{pmatrix} = \begin{pmatrix} - \frac{\sqrt{3}}{2} \\ - \frac{\sqrt{3}}{2} \end{pmatrix} $$

Change of coordinates

$\mathbf u$ and $\mathbf w$ are linearly independent (they are not parallel when $\omega \neq 0$; Topic 4, Lemma of Sec. 7.3). With the columns in the order $\mathbf u, \mathbf w$,

$$ Q = [\mathbf u \mid \mathbf w] = \begin{pmatrix} -1 & 0 \\ - \frac{1}{2} & \frac{\sqrt{3}}{2} \end{pmatrix} $$

$\det Q = (-1)(\frac{\sqrt{3}}{2}) - (0)(- \frac{1}{2}) = - \frac{\sqrt{3}}{2}$, which is not zero, so $Q$ is invertible. For a $2\times 2$ matrix $\begin{pmatrix} a & b \\ c & d \end{pmatrix}^{-1} = \frac{1}{ad - bc}\begin{pmatrix} d & -b \\ -c & a \end{pmatrix}$ (Robinson, Appendix B):

$$ Q^{-1} = \frac{1}{- \frac{\sqrt{3}}{2}}\begin{pmatrix} \frac{\sqrt{3}}{2} & 0 \\ \frac{1}{2} & -1 \end{pmatrix} = \begin{pmatrix} -1 & 0 \\ - \frac{\sqrt{3}}{3} & \frac{2 \sqrt{3}}{3} \end{pmatrix} $$

Compute $AQ = [A\mathbf u \mid A\mathbf w]$ and multiply by $Q^{-1}$ on the left, entry by entry:

$$ AQ = \begin{pmatrix} \frac{1}{2} & - \frac{\sqrt{3}}{2} \\ - \frac{1}{2} & - \frac{\sqrt{3}}{2} \end{pmatrix} $$

$$ \begin{aligned} (Q^{-1}AQ)_{11} &= (-1)\cdot\frac{1}{2} + 0\cdot(- \frac{1}{2}) = - \frac{1}{2} \\ (Q^{-1}AQ)_{12} &= (-1)\cdot(- \frac{\sqrt{3}}{2}) + 0\cdot(- \frac{\sqrt{3}}{2}) = \frac{\sqrt{3}}{2} \\ (Q^{-1}AQ)_{21} &= (- \frac{\sqrt{3}}{3})\cdot\frac{1}{2} + \frac{2 \sqrt{3}}{3}\cdot(- \frac{1}{2}) = - \frac{\sqrt{3}}{2} \\ (Q^{-1}AQ)_{22} &= (- \frac{\sqrt{3}}{3})\cdot(- \frac{\sqrt{3}}{2}) + \frac{2 \sqrt{3}}{3}\cdot(- \frac{\sqrt{3}}{2}) = - \frac{1}{2} \end{aligned} $$

$$ Q^{-1}AQ = \begin{pmatrix} - \frac{1}{2} & \frac{\sqrt{3}}{2} \\ - \frac{\sqrt{3}}{2} & - \frac{1}{2} \end{pmatrix} = J = \begin{pmatrix} \rho & \omega \\ -\omega & \rho \end{pmatrix} $$

Solution in the canonical plane

$$ \dot z_1 = -\frac{1}{2}z_1 + \frac{\sqrt{3}}{2}z_2, \qquad \dot z_2 = -\frac{\sqrt{3}}{2}z_1 - \frac{1}{2}z_2 $$

The two equations are coupled, but they say "scale and rotate", which polar coordinates express directly: $z_1 = r\cos\theta$, $z_2 = r\sin\theta$ (Topic 4, Sec. 7.5).

The radius. From $r^2 = z_1^2 + z_2^2$ the chain rule gives $2r\dot r = 2z_1\dot z_1 + 2z_2\dot z_2$. Substituting $\dot z_1 = \rho z_1 + \omega z_2$ and $\dot z_2 = -\omega z_1 + \rho z_2$:

$$ r\dot r = z_1(\rho z_1 + \omega z_2) + z_2(-\omega z_1 + \rho z_2) = \rho z_1^2 + \omega z_1z_2 - \omega z_1z_2 + \rho z_2^2 = \rho r^2 $$

The terms in $\omega$ cancel. Dividing by $r \neq 0$ (the origin is the equilibrium): $\dot r = \rho r$.

The angle. Differentiate $\tan\theta = z_2/z_1$ with the chain rule on the left and the quotient rule on the right: $\sec^2\theta\,\dot\theta = (z_1\dot z_2 - z_2\dot z_1)/z_1^2$. Since $\sec^2\theta = r^2/z_1^2$, $\dot\theta = (z_1\dot z_2 - z_2\dot z_1)/r^2$, and

$$ z_1\dot z_2 - z_2\dot z_1 = z_1(-\omega z_1 + \rho z_2) - z_2(\rho z_1 + \omega z_2) = -\omega z_1^2 + \rho z_1z_2 - \rho z_1z_2 - \omega z_2^2 = -\omega r^2 \;\Longrightarrow\; \dot\theta = -\omega $$

Integration. $\frac{d}{dt}\left(e^{-\rho t}r\right) = e^{-\rho t}(\dot r - \rho r) = 0$, and Barrow's rule between $0$ and $t$, $\left[e^{-\rho s}r(s)\right]_{0}^{t} = 0$, gives $r(t) = r_0e^{\rho t}$. For the angle, $\theta(t) - \theta_0 = \int_0^t(-\omega)\,ds = -\omega t$. Back to Cartesian coordinates, with $c_1 = r_0\cos\theta_0 = z_1(0)$, $c_2 = r_0\sin\theta_0 = z_2(0)$ and the subtraction formulas $\cos(\theta_0 - \omega t) = \cos\theta_0\cos\omega t + \sin\theta_0\sin\omega t$, $\sin(\theta_0 - \omega t) = \sin\theta_0\cos\omega t - \cos\theta_0\sin\omega t$:

$$ z_1(t) = e^{- \frac{1}{2}t}\left(c_1\cos \frac{\sqrt{3}}{2}t + c_2\sin \frac{\sqrt{3}}{2}t\right), \qquad z_2(t) = e^{- \frac{1}{2}t}\left(c_2\cos \frac{\sqrt{3}}{2}t - c_1\sin \frac{\sqrt{3}}{2}t\right) $$

Check by substitution (product rule and chain rule): $\dot z_1 = \rho z_1 + e^{\rho t}\omega\left(-c_1\sin\omega t + c_2\cos\omega t\right) = \rho z_1 + \omega z_2$ and $\dot z_2 = \rho z_2 + e^{\rho t}\omega\left(-c_2\sin\omega t - c_1\cos\omega t\right) = \rho z_2 - \omega z_1$, as required.

Back to the original coordinates

$\mathbf x = Q\mathbf z = z_1\mathbf u + z_2\mathbf w$, so

$$ \mathbf x(t) = e^{- \frac{1}{2}t}\left[\left(c_1\cos \frac{\sqrt{3}}{2}t + c_2\sin \frac{\sqrt{3}}{2}t\right)\begin{pmatrix} -1 \\ - \frac{1}{2} \end{pmatrix} + \left(c_2\cos \frac{\sqrt{3}}{2}t - c_1\sin \frac{\sqrt{3}}{2}t\right)\begin{pmatrix} 0 \\ \frac{\sqrt{3}}{2} \end{pmatrix}\right] $$

$$ x(t) = - c_{1} e^{-\frac{t}{2}} \cos{\frac{\sqrt{3} t}{2}} - c_{2} e^{-\frac{t}{2}} \sin{\frac{\sqrt{3} t}{2}}, \qquad y(t) = - \frac{\sqrt{3} c_{1} e^{-\frac{t}{2}} \sin{\frac{\sqrt{3} t}{2}}}{2} - \frac{c_{1} e^{-\frac{t}{2}} \cos{\frac{\sqrt{3} t}{2}}}{2} - \frac{c_{2} e^{-\frac{t}{2}} \sin{\frac{\sqrt{3} t}{2}}}{2} + \frac{\sqrt{3} c_{2} e^{-\frac{t}{2}} \cos{\frac{\sqrt{3} t}{2}}}{2} $$

The canonical system is $\dot z_1 = - \frac{1}{2}z_1 + (\frac{\sqrt{3}}{2})z_2$, $\dot z_2 = -(\frac{\sqrt{3}}{2})z_1 + - \frac{1}{2}z_2$: a rotation clockwise with angular speed $\omega = \frac{\sqrt{3}}{2}$ combined with a radial factor $e^{- \frac{1}{2}t}$.

(ii) $A = \begin{pmatrix} -2 & 3 \\ -6 & 4 \end{pmatrix}$

Characteristic equation $\det(A - \lambda I) = 0$ (Topic 4, Eq. 27.8):

$$ \det(A - \lambda I) = \begin{vmatrix} -2 - \lambda & 3 \\ -6 & 4 - \lambda \end{vmatrix} = (-2 - \lambda)(4 - \lambda) - (3)(-6) $$

$$ = \lambda^2 - (-2 + 4)\lambda + (-2\cdot4 - 3\cdot(-6)) = \lambda^2 - (2)\lambda + (10) = \lambda^{2} - 2 \lambda + 10 $$

$\operatorname{tr}A = 2$, $\det A = 10$, discriminant $\Delta = (\operatorname{tr}A)^2 - 4\det A = 4 - 40 = -36$.

$\Delta < 0$: a complex-conjugate pair. $\lambda = \dfrac{2 \pm \sqrt{-36}}{2} = \dfrac{2 \pm 6 i}{2}$, so

$$ \lambda_\pm = 1 \pm 3\,i \qquad (\rho = 1,\ \omega = 3 > 0) $$

Eigenvector of the eigenvalue with positive imaginary part

For $\lambda = \rho + i\omega = 1 + 3 i$ the first row of $(A - \lambda I)\mathbf v = \mathbf 0$ reads $(-3 - 3 i)\,v_1 + 3\,v_2 = 0$ (the coefficient of $v_2$ is non-zero because $\Delta < 0$ forces $bc < 0$). Choose $v_1 = 3$; then $v_2 = -(-3 - 3 i) = 3 + 3 i$:

$$ \mathbf v = \begin{pmatrix} 3 \\ 3 + 3 i \end{pmatrix} = \underbrace{\begin{pmatrix} 3 \\ 3 \end{pmatrix}}_{\mathbf u} + i\,\underbrace{\begin{pmatrix} 0 \\ 3 \end{pmatrix}}_{\mathbf w} $$

Check with the second row: $-6\cdot3 + (3 - 3 i)(3 + 3 i) = -18 + 18 = 0$ (using $i^2 = -1$).

How $A$ acts on $\mathbf u$ and $\mathbf w$

Separating the real and imaginary parts of $A\mathbf v = \lambda\mathbf v$ gives (Topic 4, Eq. 7.4) $A\mathbf u = \rho\mathbf u - \omega\mathbf w$ and $A\mathbf w = \omega\mathbf u + \rho\mathbf w$. Verification with the numbers:

$$ A\mathbf u = \begin{pmatrix} 3 \\ -6 \end{pmatrix}, \qquad \rho\mathbf u - \omega\mathbf w = 1\begin{pmatrix} 3 \\ 3 \end{pmatrix} - 3\begin{pmatrix} 0 \\ 3 \end{pmatrix} = \begin{pmatrix} 3 \\ -6 \end{pmatrix} $$

$$ A\mathbf w = \begin{pmatrix} 9 \\ 12 \end{pmatrix}, \qquad \omega\mathbf u + \rho\mathbf w = 3\begin{pmatrix} 3 \\ 3 \end{pmatrix} + 1\begin{pmatrix} 0 \\ 3 \end{pmatrix} = \begin{pmatrix} 9 \\ 12 \end{pmatrix} $$

Change of coordinates

$\mathbf u$ and $\mathbf w$ are linearly independent (they are not parallel when $\omega \neq 0$; Topic 4, Lemma of Sec. 7.3). With the columns in the order $\mathbf u, \mathbf w$,

$$ Q = [\mathbf u \mid \mathbf w] = \begin{pmatrix} 3 & 0 \\ 3 & 3 \end{pmatrix} $$

$\det Q = (3)(3) - (0)(3) = 9$, which is not zero, so $Q$ is invertible. For a $2\times 2$ matrix $\begin{pmatrix} a & b \\ c & d \end{pmatrix}^{-1} = \frac{1}{ad - bc}\begin{pmatrix} d & -b \\ -c & a \end{pmatrix}$ (Robinson, Appendix B):

$$ Q^{-1} = \frac{1}{9}\begin{pmatrix} 3 & 0 \\ -3 & 3 \end{pmatrix} = \begin{pmatrix} \frac{1}{3} & 0 \\ - \frac{1}{3} & \frac{1}{3} \end{pmatrix} $$

Compute $AQ = [A\mathbf u \mid A\mathbf w]$ and multiply by $Q^{-1}$ on the left, entry by entry:

$$ AQ = \begin{pmatrix} 3 & 9 \\ -6 & 12 \end{pmatrix} $$

$$ \begin{aligned} (Q^{-1}AQ)_{11} &= \frac{1}{3}\cdot3 + 0\cdot(-6) = 1 \\ (Q^{-1}AQ)_{12} &= \frac{1}{3}\cdot9 + 0\cdot12 = 3 \\ (Q^{-1}AQ)_{21} &= (- \frac{1}{3})\cdot3 + \frac{1}{3}\cdot(-6) = -3 \\ (Q^{-1}AQ)_{22} &= (- \frac{1}{3})\cdot9 + \frac{1}{3}\cdot12 = 1 \end{aligned} $$

$$ Q^{-1}AQ = \begin{pmatrix} 1 & 3 \\ -3 & 1 \end{pmatrix} = J = \begin{pmatrix} \rho & \omega \\ -\omega & \rho \end{pmatrix} $$

Solution in the canonical plane

$$ \dot z_1 = z_1 + 3z_2, \qquad \dot z_2 = -3z_1 + z_2 $$

The two equations are coupled, but they say "scale and rotate", which polar coordinates express directly: $z_1 = r\cos\theta$, $z_2 = r\sin\theta$ (Topic 4, Sec. 7.5).

The radius. From $r^2 = z_1^2 + z_2^2$ the chain rule gives $2r\dot r = 2z_1\dot z_1 + 2z_2\dot z_2$. Substituting $\dot z_1 = \rho z_1 + \omega z_2$ and $\dot z_2 = -\omega z_1 + \rho z_2$:

$$ r\dot r = z_1(\rho z_1 + \omega z_2) + z_2(-\omega z_1 + \rho z_2) = \rho z_1^2 + \omega z_1z_2 - \omega z_1z_2 + \rho z_2^2 = \rho r^2 $$

The terms in $\omega$ cancel. Dividing by $r \neq 0$ (the origin is the equilibrium): $\dot r = \rho r$.

The angle. Differentiate $\tan\theta = z_2/z_1$ with the chain rule on the left and the quotient rule on the right: $\sec^2\theta\,\dot\theta = (z_1\dot z_2 - z_2\dot z_1)/z_1^2$. Since $\sec^2\theta = r^2/z_1^2$, $\dot\theta = (z_1\dot z_2 - z_2\dot z_1)/r^2$, and

$$ z_1\dot z_2 - z_2\dot z_1 = z_1(-\omega z_1 + \rho z_2) - z_2(\rho z_1 + \omega z_2) = -\omega z_1^2 + \rho z_1z_2 - \rho z_1z_2 - \omega z_2^2 = -\omega r^2 \;\Longrightarrow\; \dot\theta = -\omega $$

Integration. $\frac{d}{dt}\left(e^{-\rho t}r\right) = e^{-\rho t}(\dot r - \rho r) = 0$, and Barrow's rule between $0$ and $t$, $\left[e^{-\rho s}r(s)\right]_{0}^{t} = 0$, gives $r(t) = r_0e^{\rho t}$. For the angle, $\theta(t) - \theta_0 = \int_0^t(-\omega)\,ds = -\omega t$. Back to Cartesian coordinates, with $c_1 = r_0\cos\theta_0 = z_1(0)$, $c_2 = r_0\sin\theta_0 = z_2(0)$ and the subtraction formulas $\cos(\theta_0 - \omega t) = \cos\theta_0\cos\omega t + \sin\theta_0\sin\omega t$, $\sin(\theta_0 - \omega t) = \sin\theta_0\cos\omega t - \cos\theta_0\sin\omega t$:

$$ z_1(t) = e^{1t}\left(c_1\cos 3t + c_2\sin 3t\right), \qquad z_2(t) = e^{1t}\left(c_2\cos 3t - c_1\sin 3t\right) $$

Check by substitution (product rule and chain rule): $\dot z_1 = \rho z_1 + e^{\rho t}\omega\left(-c_1\sin\omega t + c_2\cos\omega t\right) = \rho z_1 + \omega z_2$ and $\dot z_2 = \rho z_2 + e^{\rho t}\omega\left(-c_2\sin\omega t - c_1\cos\omega t\right) = \rho z_2 - \omega z_1$, as required.

Back to the original coordinates

$\mathbf x = Q\mathbf z = z_1\mathbf u + z_2\mathbf w$, so

$$ \mathbf x(t) = e^{1t}\left[\left(c_1\cos 3t + c_2\sin 3t\right)\begin{pmatrix} 3 \\ 3 \end{pmatrix} + \left(c_2\cos 3t - c_1\sin 3t\right)\begin{pmatrix} 0 \\ 3 \end{pmatrix}\right] $$

$$ x(t) = 3 c_{1} e^{t} \cos{3 t} + 3 c_{2} e^{t} \sin{3 t}, \qquad y(t) = - 3 c_{1} e^{t} \sin{3 t} + 3 c_{1} e^{t} \cos{3 t} + 3 c_{2} e^{t} \sin{3 t} + 3 c_{2} e^{t} \cos{3 t} $$

The canonical system is $\dot z_1 = 1z_1 + (3)z_2$, $\dot z_2 = -(3)z_1 + 1z_2$: a rotation clockwise with angular speed $\omega = 3$ combined with a radial factor $e^{1t}$.

(iii) $A = \begin{pmatrix} -11 & -2 \\ 13 & -9 \end{pmatrix}$

Characteristic equation $\det(A - \lambda I) = 0$ (Topic 4, Eq. 27.8):

$$ \det(A - \lambda I) = \begin{vmatrix} -11 - \lambda & -2 \\ 13 & -9 - \lambda \end{vmatrix} = (-11 - \lambda)(-9 - \lambda) - (-2)(13) $$

$$ = \lambda^2 - (-11 + (-9))\lambda + (-11\cdot(-9) - (-2)\cdot13) = \lambda^2 - (-20)\lambda + (125) = \lambda^{2} + 20 \lambda + 125 $$

$\operatorname{tr}A = -20$, $\det A = 125$, discriminant $\Delta = (\operatorname{tr}A)^2 - 4\det A = 400 - 500 = -100$.

$\Delta < 0$: a complex-conjugate pair. $\lambda = \dfrac{-20 \pm \sqrt{-100}}{2} = \dfrac{-20 \pm 10 i}{2}$, so

$$ \lambda_\pm = -10 \pm 5\,i \qquad (\rho = -10,\ \omega = 5 > 0) $$

Eigenvector of the eigenvalue with positive imaginary part

For $\lambda = \rho + i\omega = -10 + 5 i$ the first row of $(A - \lambda I)\mathbf v = \mathbf 0$ reads $(-1 - 5 i)\,v_1 + (-2)\,v_2 = 0$ (the coefficient of $v_2$ is non-zero because $\Delta < 0$ forces $bc < 0$). Choose $v_1 = -2$; then $v_2 = -(-1 - 5 i) = 1 + 5 i$:

$$ \mathbf v = \begin{pmatrix} -2 \\ 1 + 5 i \end{pmatrix} = \underbrace{\begin{pmatrix} -2 \\ 1 \end{pmatrix}}_{\mathbf u} + i\,\underbrace{\begin{pmatrix} 0 \\ 5 \end{pmatrix}}_{\mathbf w} $$

Check with the second row: $13\cdot(-2) + (1 - 5 i)(1 + 5 i) = -26 + 26 = 0$ (using $i^2 = -1$).

How $A$ acts on $\mathbf u$ and $\mathbf w$

Separating the real and imaginary parts of $A\mathbf v = \lambda\mathbf v$ gives (Topic 4, Eq. 7.4) $A\mathbf u = \rho\mathbf u - \omega\mathbf w$ and $A\mathbf w = \omega\mathbf u + \rho\mathbf w$. Verification with the numbers:

$$ A\mathbf u = \begin{pmatrix} 20 \\ -35 \end{pmatrix}, \qquad \rho\mathbf u - \omega\mathbf w = (-10)\begin{pmatrix} -2 \\ 1 \end{pmatrix} - 5\begin{pmatrix} 0 \\ 5 \end{pmatrix} = \begin{pmatrix} 20 \\ -35 \end{pmatrix} $$

$$ A\mathbf w = \begin{pmatrix} -10 \\ -45 \end{pmatrix}, \qquad \omega\mathbf u + \rho\mathbf w = 5\begin{pmatrix} -2 \\ 1 \end{pmatrix} + (-10)\begin{pmatrix} 0 \\ 5 \end{pmatrix} = \begin{pmatrix} -10 \\ -45 \end{pmatrix} $$

Change of coordinates

$\mathbf u$ and $\mathbf w$ are linearly independent (they are not parallel when $\omega \neq 0$; Topic 4, Lemma of Sec. 7.3). With the columns in the order $\mathbf u, \mathbf w$,

$$ Q = [\mathbf u \mid \mathbf w] = \begin{pmatrix} -2 & 0 \\ 1 & 5 \end{pmatrix} $$

$\det Q = (-2)(5) - (0)(1) = -10$, which is not zero, so $Q$ is invertible. For a $2\times 2$ matrix $\begin{pmatrix} a & b \\ c & d \end{pmatrix}^{-1} = \frac{1}{ad - bc}\begin{pmatrix} d & -b \\ -c & a \end{pmatrix}$ (Robinson, Appendix B):

$$ Q^{-1} = \frac{1}{-10}\begin{pmatrix} 5 & 0 \\ -1 & -2 \end{pmatrix} = \begin{pmatrix} - \frac{1}{2} & 0 \\ \frac{1}{10} & \frac{1}{5} \end{pmatrix} $$

Compute $AQ = [A\mathbf u \mid A\mathbf w]$ and multiply by $Q^{-1}$ on the left, entry by entry:

$$ AQ = \begin{pmatrix} 20 & -10 \\ -35 & -45 \end{pmatrix} $$

$$ \begin{aligned} (Q^{-1}AQ)_{11} &= (- \frac{1}{2})\cdot20 + 0\cdot(-35) = -10 \\ (Q^{-1}AQ)_{12} &= (- \frac{1}{2})\cdot(-10) + 0\cdot(-45) = 5 \\ (Q^{-1}AQ)_{21} &= \frac{1}{10}\cdot20 + \frac{1}{5}\cdot(-35) = -5 \\ (Q^{-1}AQ)_{22} &= \frac{1}{10}\cdot(-10) + \frac{1}{5}\cdot(-45) = -10 \end{aligned} $$

$$ Q^{-1}AQ = \begin{pmatrix} -10 & 5 \\ -5 & -10 \end{pmatrix} = J = \begin{pmatrix} \rho & \omega \\ -\omega & \rho \end{pmatrix} $$

Solution in the canonical plane

$$ \dot z_1 = -10z_1 + 5z_2, \qquad \dot z_2 = -5z_1 - 10z_2 $$

The two equations are coupled, but they say "scale and rotate", which polar coordinates express directly: $z_1 = r\cos\theta$, $z_2 = r\sin\theta$ (Topic 4, Sec. 7.5).

The radius. From $r^2 = z_1^2 + z_2^2$ the chain rule gives $2r\dot r = 2z_1\dot z_1 + 2z_2\dot z_2$. Substituting $\dot z_1 = \rho z_1 + \omega z_2$ and $\dot z_2 = -\omega z_1 + \rho z_2$:

$$ r\dot r = z_1(\rho z_1 + \omega z_2) + z_2(-\omega z_1 + \rho z_2) = \rho z_1^2 + \omega z_1z_2 - \omega z_1z_2 + \rho z_2^2 = \rho r^2 $$

The terms in $\omega$ cancel. Dividing by $r \neq 0$ (the origin is the equilibrium): $\dot r = \rho r$.

The angle. Differentiate $\tan\theta = z_2/z_1$ with the chain rule on the left and the quotient rule on the right: $\sec^2\theta\,\dot\theta = (z_1\dot z_2 - z_2\dot z_1)/z_1^2$. Since $\sec^2\theta = r^2/z_1^2$, $\dot\theta = (z_1\dot z_2 - z_2\dot z_1)/r^2$, and

$$ z_1\dot z_2 - z_2\dot z_1 = z_1(-\omega z_1 + \rho z_2) - z_2(\rho z_1 + \omega z_2) = -\omega z_1^2 + \rho z_1z_2 - \rho z_1z_2 - \omega z_2^2 = -\omega r^2 \;\Longrightarrow\; \dot\theta = -\omega $$

Integration. $\frac{d}{dt}\left(e^{-\rho t}r\right) = e^{-\rho t}(\dot r - \rho r) = 0$, and Barrow's rule between $0$ and $t$, $\left[e^{-\rho s}r(s)\right]_{0}^{t} = 0$, gives $r(t) = r_0e^{\rho t}$. For the angle, $\theta(t) - \theta_0 = \int_0^t(-\omega)\,ds = -\omega t$. Back to Cartesian coordinates, with $c_1 = r_0\cos\theta_0 = z_1(0)$, $c_2 = r_0\sin\theta_0 = z_2(0)$ and the subtraction formulas $\cos(\theta_0 - \omega t) = \cos\theta_0\cos\omega t + \sin\theta_0\sin\omega t$, $\sin(\theta_0 - \omega t) = \sin\theta_0\cos\omega t - \cos\theta_0\sin\omega t$:

$$ z_1(t) = e^{-10t}\left(c_1\cos 5t + c_2\sin 5t\right), \qquad z_2(t) = e^{-10t}\left(c_2\cos 5t - c_1\sin 5t\right) $$

Check by substitution (product rule and chain rule): $\dot z_1 = \rho z_1 + e^{\rho t}\omega\left(-c_1\sin\omega t + c_2\cos\omega t\right) = \rho z_1 + \omega z_2$ and $\dot z_2 = \rho z_2 + e^{\rho t}\omega\left(-c_2\sin\omega t - c_1\cos\omega t\right) = \rho z_2 - \omega z_1$, as required.

Back to the original coordinates

$\mathbf x = Q\mathbf z = z_1\mathbf u + z_2\mathbf w$, so

$$ \mathbf x(t) = e^{-10t}\left[\left(c_1\cos 5t + c_2\sin 5t\right)\begin{pmatrix} -2 \\ 1 \end{pmatrix} + \left(c_2\cos 5t - c_1\sin 5t\right)\begin{pmatrix} 0 \\ 5 \end{pmatrix}\right] $$

$$ x(t) = - 2 c_{1} e^{-10 t} \cos{5 t} - 2 c_{2} e^{-10 t} \sin{5 t}, \qquad y(t) = - 5 c_{1} e^{-10 t} \sin{5 t} + c_{1} e^{-10 t} \cos{5 t} + c_{2} e^{-10 t} \sin{5 t} + 5 c_{2} e^{-10 t} \cos{5 t} $$

The canonical system is $\dot z_1 = -10z_1 + (5)z_2$, $\dot z_2 = -(5)z_1 + -10z_2$: a rotation clockwise with angular speed $\omega = 5$ combined with a radial factor $e^{-10t}$.

(iv) $A = \begin{pmatrix} 7 & -5 \\ 10 & -3 \end{pmatrix}$

Characteristic equation $\det(A - \lambda I) = 0$ (Topic 4, Eq. 27.8):

$$ \det(A - \lambda I) = \begin{vmatrix} 7 - \lambda & -5 \\ 10 & -3 - \lambda \end{vmatrix} = (7 - \lambda)(-3 - \lambda) - (-5)(10) $$

$$ = \lambda^2 - (7 + (-3))\lambda + (7\cdot(-3) - (-5)\cdot10) = \lambda^2 - (4)\lambda + (29) = \lambda^{2} - 4 \lambda + 29 $$

$\operatorname{tr}A = 4$, $\det A = 29$, discriminant $\Delta = (\operatorname{tr}A)^2 - 4\det A = 16 - 116 = -100$.

$\Delta < 0$: a complex-conjugate pair. $\lambda = \dfrac{4 \pm \sqrt{-100}}{2} = \dfrac{4 \pm 10 i}{2}$, so

$$ \lambda_\pm = 2 \pm 5\,i \qquad (\rho = 2,\ \omega = 5 > 0) $$

Eigenvector of the eigenvalue with positive imaginary part

For $\lambda = \rho + i\omega = 2 + 5 i$ the first row of $(A - \lambda I)\mathbf v = \mathbf 0$ reads $(5 - 5 i)\,v_1 + (-5)\,v_2 = 0$ (the coefficient of $v_2$ is non-zero because $\Delta < 0$ forces $bc < 0$). Choose $v_1 = -5$; then $v_2 = -(5 - 5 i) = -5 + 5 i$:

$$ \mathbf v = \begin{pmatrix} -5 \\ -5 + 5 i \end{pmatrix} = \underbrace{\begin{pmatrix} -5 \\ -5 \end{pmatrix}}_{\mathbf u} + i\,\underbrace{\begin{pmatrix} 0 \\ 5 \end{pmatrix}}_{\mathbf w} $$

Check with the second row: $10\cdot(-5) + (-5 - 5 i)(-5 + 5 i) = -50 + 50 = 0$ (using $i^2 = -1$).

How $A$ acts on $\mathbf u$ and $\mathbf w$

Separating the real and imaginary parts of $A\mathbf v = \lambda\mathbf v$ gives (Topic 4, Eq. 7.4) $A\mathbf u = \rho\mathbf u - \omega\mathbf w$ and $A\mathbf w = \omega\mathbf u + \rho\mathbf w$. Verification with the numbers:

$$ A\mathbf u = \begin{pmatrix} -10 \\ -35 \end{pmatrix}, \qquad \rho\mathbf u - \omega\mathbf w = 2\begin{pmatrix} -5 \\ -5 \end{pmatrix} - 5\begin{pmatrix} 0 \\ 5 \end{pmatrix} = \begin{pmatrix} -10 \\ -35 \end{pmatrix} $$

$$ A\mathbf w = \begin{pmatrix} -25 \\ -15 \end{pmatrix}, \qquad \omega\mathbf u + \rho\mathbf w = 5\begin{pmatrix} -5 \\ -5 \end{pmatrix} + 2\begin{pmatrix} 0 \\ 5 \end{pmatrix} = \begin{pmatrix} -25 \\ -15 \end{pmatrix} $$

Change of coordinates

$\mathbf u$ and $\mathbf w$ are linearly independent (they are not parallel when $\omega \neq 0$; Topic 4, Lemma of Sec. 7.3). With the columns in the order $\mathbf u, \mathbf w$,

$$ Q = [\mathbf u \mid \mathbf w] = \begin{pmatrix} -5 & 0 \\ -5 & 5 \end{pmatrix} $$

$\det Q = (-5)(5) - (0)(-5) = -25$, which is not zero, so $Q$ is invertible. For a $2\times 2$ matrix $\begin{pmatrix} a & b \\ c & d \end{pmatrix}^{-1} = \frac{1}{ad - bc}\begin{pmatrix} d & -b \\ -c & a \end{pmatrix}$ (Robinson, Appendix B):

$$ Q^{-1} = \frac{1}{-25}\begin{pmatrix} 5 & 0 \\ 5 & -5 \end{pmatrix} = \begin{pmatrix} - \frac{1}{5} & 0 \\ - \frac{1}{5} & \frac{1}{5} \end{pmatrix} $$

Compute $AQ = [A\mathbf u \mid A\mathbf w]$ and multiply by $Q^{-1}$ on the left, entry by entry:

$$ AQ = \begin{pmatrix} -10 & -25 \\ -35 & -15 \end{pmatrix} $$

$$ \begin{aligned} (Q^{-1}AQ)_{11} &= (- \frac{1}{5})\cdot(-10) + 0\cdot(-35) = 2 \\ (Q^{-1}AQ)_{12} &= (- \frac{1}{5})\cdot(-25) + 0\cdot(-15) = 5 \\ (Q^{-1}AQ)_{21} &= (- \frac{1}{5})\cdot(-10) + \frac{1}{5}\cdot(-35) = -5 \\ (Q^{-1}AQ)_{22} &= (- \frac{1}{5})\cdot(-25) + \frac{1}{5}\cdot(-15) = 2 \end{aligned} $$

$$ Q^{-1}AQ = \begin{pmatrix} 2 & 5 \\ -5 & 2 \end{pmatrix} = J = \begin{pmatrix} \rho & \omega \\ -\omega & \rho \end{pmatrix} $$

Solution in the canonical plane

$$ \dot z_1 = 2z_1 + 5z_2, \qquad \dot z_2 = -5z_1 + 2z_2 $$

The two equations are coupled, but they say "scale and rotate", which polar coordinates express directly: $z_1 = r\cos\theta$, $z_2 = r\sin\theta$ (Topic 4, Sec. 7.5).

The radius. From $r^2 = z_1^2 + z_2^2$ the chain rule gives $2r\dot r = 2z_1\dot z_1 + 2z_2\dot z_2$. Substituting $\dot z_1 = \rho z_1 + \omega z_2$ and $\dot z_2 = -\omega z_1 + \rho z_2$:

$$ r\dot r = z_1(\rho z_1 + \omega z_2) + z_2(-\omega z_1 + \rho z_2) = \rho z_1^2 + \omega z_1z_2 - \omega z_1z_2 + \rho z_2^2 = \rho r^2 $$

The terms in $\omega$ cancel. Dividing by $r \neq 0$ (the origin is the equilibrium): $\dot r = \rho r$.

The angle. Differentiate $\tan\theta = z_2/z_1$ with the chain rule on the left and the quotient rule on the right: $\sec^2\theta\,\dot\theta = (z_1\dot z_2 - z_2\dot z_1)/z_1^2$. Since $\sec^2\theta = r^2/z_1^2$, $\dot\theta = (z_1\dot z_2 - z_2\dot z_1)/r^2$, and

$$ z_1\dot z_2 - z_2\dot z_1 = z_1(-\omega z_1 + \rho z_2) - z_2(\rho z_1 + \omega z_2) = -\omega z_1^2 + \rho z_1z_2 - \rho z_1z_2 - \omega z_2^2 = -\omega r^2 \;\Longrightarrow\; \dot\theta = -\omega $$

Integration. $\frac{d}{dt}\left(e^{-\rho t}r\right) = e^{-\rho t}(\dot r - \rho r) = 0$, and Barrow's rule between $0$ and $t$, $\left[e^{-\rho s}r(s)\right]_{0}^{t} = 0$, gives $r(t) = r_0e^{\rho t}$. For the angle, $\theta(t) - \theta_0 = \int_0^t(-\omega)\,ds = -\omega t$. Back to Cartesian coordinates, with $c_1 = r_0\cos\theta_0 = z_1(0)$, $c_2 = r_0\sin\theta_0 = z_2(0)$ and the subtraction formulas $\cos(\theta_0 - \omega t) = \cos\theta_0\cos\omega t + \sin\theta_0\sin\omega t$, $\sin(\theta_0 - \omega t) = \sin\theta_0\cos\omega t - \cos\theta_0\sin\omega t$:

$$ z_1(t) = e^{2t}\left(c_1\cos 5t + c_2\sin 5t\right), \qquad z_2(t) = e^{2t}\left(c_2\cos 5t - c_1\sin 5t\right) $$

Check by substitution (product rule and chain rule): $\dot z_1 = \rho z_1 + e^{\rho t}\omega\left(-c_1\sin\omega t + c_2\cos\omega t\right) = \rho z_1 + \omega z_2$ and $\dot z_2 = \rho z_2 + e^{\rho t}\omega\left(-c_2\sin\omega t - c_1\cos\omega t\right) = \rho z_2 - \omega z_1$, as required.

Back to the original coordinates

$\mathbf x = Q\mathbf z = z_1\mathbf u + z_2\mathbf w$, so

$$ \mathbf x(t) = e^{2t}\left[\left(c_1\cos 5t + c_2\sin 5t\right)\begin{pmatrix} -5 \\ -5 \end{pmatrix} + \left(c_2\cos 5t - c_1\sin 5t\right)\begin{pmatrix} 0 \\ 5 \end{pmatrix}\right] $$

$$ x(t) = - 5 c_{1} e^{2 t} \cos{5 t} - 5 c_{2} e^{2 t} \sin{5 t}, \qquad y(t) = - 5 c_{1} e^{2 t} \sin{5 t} - 5 c_{1} e^{2 t} \cos{5 t} - 5 c_{2} e^{2 t} \sin{5 t} + 5 c_{2} e^{2 t} \cos{5 t} $$

The canonical system is $\dot z_1 = 2z_1 + (5)z_2$, $\dot z_2 = -(5)z_1 + 2z_2$: a rotation clockwise with angular speed $\omega = 5$ combined with a radial factor $e^{2t}$.

Interpretation and checks
  • In each item the explicit product $Q^{-1}AQ$ reproduces $J$ exactly (entry-by-entry arithmetic above).
  • (i) has an irrational $\omega = \sqrt3/2$ (eigenvalues $-\tfrac12 \pm \tfrac{\sqrt3}{2}i$, with the surds carried exactly): stable spiral. (ii) $1 \pm 3i$, unstable spiral; (iii) $-10 \pm 5i$, stable spiral (fast decay, ratio $|\rho|/\omega = 2$); (iv) $2 \pm 5i$, unstable spiral.
  • The orientation of $Q$ ($\det Q$ sign) decides whether the clockwise rotation of the $\mathbf z$-plane stays clockwise in the $\mathbf x$-plane; it is $b\omega$ here (positive iff $b > 0$).
  • The eigenvalues of each matrix satisfy $\lambda_+ + \lambda_- = \operatorname{tr}A$ and $\lambda_+\lambda_- = \det A$ (checked in the text).
Exercise 4

Real and Imaginary Parts of a Complex Eigenvector Are Independent

Original statement · probls_ch4-ii_2627.pdf
Linear Independence • Complex Eigenvectors
Statement

In the previous chapter we used the result that eigenvectors which correspond to distinct eigenvalues are linearly independent. Use this result to show that the real and imaginary parts of complex eigenvectors are linearly independent.

Setup and hypotheses

Let $A$ be a real $2\times 2$ matrix with a non-real eigenvalue $\lambda = \rho + i\omega$ ($\omega \neq 0$) and eigenvector $\mathbf v = \mathbf u + i\mathbf w \in \mathbb C^2$, $\mathbf u, \mathbf w \in \mathbb R^2$. To show: $\mathbf u$ and $\mathbf w$ are linearly independent over $\mathbb R$, i.e. $\alpha\mathbf u + \beta\mathbf w = \mathbf 0$ with $\alpha, \beta \in \mathbb R$ implies $\alpha = \beta = 0$.

Formulation

Ingredients: (a) $\bar{\mathbf v} = \mathbf u - i\mathbf w$ is an eigenvector for $\bar\lambda$ (conjugate $A\mathbf v = \lambda\mathbf v$ and use $\bar A = A$ since $A$ is real); (b) $\lambda \neq \bar\lambda$ because $\omega \neq 0$; (c) eigenvectors of distinct eigenvalues are linearly independent (over $\mathbb C$) (Topic 4, Appendix B argument).

Method of the class notes [Notes, p. 10]

Route of the notes: $\mathbf v_1 = \tfrac12(\mathbf v_+ + \mathbf v_-)$ and $\mathbf v_2 = -\tfrac i2(\mathbf v_+ - \mathbf v_-)$, that is $Q = (\mathbf v_1 \mid \mathbf v_2) = PM$ with $P = (\mathbf v_+ \mid \mathbf v_-)$ and $M = \begin{pmatrix} 1/2 & -i/2 \\ 1/2 & i/2 \end{pmatrix}$. $P$ is invertible because $\mathbf v_\pm$ belong to the distinct eigenvalues $\rho \pm i\omega$ ($\omega \neq 0$), and $\det M = \frac{i}{4} + \frac{i}{4} = \frac{i}{2} \neq 0$. So $\det Q = \det P\,\det M \neq 0$ and the real vectors $\mathbf v_1$, $\mathbf v_2$ are independent: the argument below, written with matrices.

Derivation

Step 1: $\mathbf v$ and $\bar{\mathbf v}$ are independent

By (a), (b) and the quoted result, $\mathbf v$ and $\bar{\mathbf v}$ are linearly independent in $\mathbb C^2$: $\gamma\mathbf v + \delta\bar{\mathbf v} = \mathbf 0$ with $\gamma, \delta \in \mathbb C$ forces $\gamma = \delta = 0$.

Step 2: express $\mathbf u, \mathbf w$ through $\mathbf v, \bar{\mathbf v}$

$$ \mathbf u = \frac{\mathbf v + \bar{\mathbf v}}{2}, \qquad \mathbf w = \frac{\mathbf v - \bar{\mathbf v}}{2i} $$

Step 3: a real relation must be trivial

Suppose $\alpha\mathbf u + \beta\mathbf w = \mathbf 0$ with $\alpha, \beta \in \mathbb R$. Substitute Step 2 and collect the coefficients of $\mathbf v$ and $\bar{\mathbf v}$:

$$ \alpha\frac{\mathbf v + \bar{\mathbf v}}{2} + \beta\frac{\mathbf v - \bar{\mathbf v}}{2i} = \underbrace{\left(\frac{\alpha}{2} + \frac{\beta}{2i}\right)}_{\gamma}\mathbf v + \underbrace{\left(\frac{\alpha}{2} - \frac{\beta}{2i}\right)}_{\delta}\bar{\mathbf v} = \mathbf 0 $$

By Step 1, $\gamma = \delta = 0$. Using $\frac1i = -i$: $\gamma = \frac{\alpha - i\beta}{2}$ and $\delta = \frac{\alpha + i\beta}{2}$. Then $\gamma = 0$ means $\alpha - i\beta = 0$; with $\alpha, \beta$ real, real and imaginary parts give $\alpha = 0$ and $\beta = 0$. $\blacksquare$

Check against the direct argument

Alternative (Topic 4 Sec. 7.3): if $\mathbf w = k\mathbf u$ (or $\mathbf u = \mathbf 0$), $\mathbf v$ would be a complex multiple of one real vector $\mathbf r$, forcing $A\mathbf r = \lambda\mathbf r$ with $A\mathbf r$ real, so $\lambda$ would be real: contradiction. The two proofs agree; the one above only uses the independence of eigenvectors of distinct eigenvalues, as the exercise requests.

Interpretation and checks
  • Consequence: $Q = [\mathbf u \mid \mathbf w]$ is invertible, which is what makes the change of coordinates $\mathbf z = Q^{-1}\mathbf x$ possible for complex eigenvalues (Exercises 3 and 1).
  • Numerical check with the eigenvector of Topic 4 Example 7.1, $\mathbf v = (5, -1 + 3i)^T$: $\mathbf u = (5, -1)^T$, $\mathbf w = (0, 3)^T$, $\det[\mathbf u\ \mathbf w] = 15 \neq 0$.
  • The argument fails for a real eigenvalue ($\lambda = \bar\lambda$): then $\mathbf w = \mathbf 0$ can occur, consistent with Step 1 breaking down.
  • In dimension $n$ the same idea shows that the real and imaginary parts of eigenvectors of different conjugate pairs together with real eigenvectors remain independent.
Exercise 5

Building a Matrix from a Complex Pair of Eigenvalues and Eigenvectors

Original statement · probls_ch4-ii_2627.pdf, p. 2
Inverse Problem • A = Q J Q⁻¹ • Three Matrices
Statement

Show how to construct a matrix with a complex conjugate pair of eigenvalues $\lambda_\pm = \rho \pm i\omega$ and corresponding eigenvectors $\boldsymbol\eta_\pm = \mathbf v_1 \pm i\mathbf v_2$. Hence find the matrices with the following eigenvalues and eigenvectors:

  • (i) $\lambda_\pm = 3 \pm 3i,\ \boldsymbol\eta_\pm = (2 \pm i,\ 1 \mp i)$;
  • (ii) $\lambda_\pm = \pm 3i,\ \boldsymbol\eta_\pm = (\pm i,\ 3 \pm 2i)$;
  • (iii) $\lambda_\pm = -2 \pm i,\ \boldsymbol\eta_\pm = (1 \pm i,\ 1 \mp i)$.
Setup and hypotheses

We look for a real matrix $A$ ($2\times 2$) with $A\boldsymbol\eta_+ = \lambda_+\boldsymbol\eta_+$ where $\boldsymbol\eta_+ = \mathbf v_1 + i\mathbf v_2$ and $\lambda_+ = \rho + i\omega$, with $\mathbf v_1, \mathbf v_2$ real and independent and $\omega \neq 0$. Because $A$ is real, $A\bar{\boldsymbol\eta}_+ = \bar\lambda_+\bar{\boldsymbol\eta}_+$ follows automatically. (Note on the statement: the sheet writes $\boldsymbol\eta_\pm = \mathbf v_1 \pm i\mathbf v_2$ for the eigenvectors of $\lambda_\pm = \rho \pm i\omega$, i.e. $\mathbf v_1 = \mathbf u$, $\mathbf v_2 = \mathbf w$ of Topic 4.)

Formulation: the construction

Expand $A(\mathbf v_1 + i\mathbf v_2) = (\rho + i\omega)(\mathbf v_1 + i\mathbf v_2)$ and separate real and imaginary parts (both $A\mathbf v_j$ are real vectors, $i^2 = -1$):

$$ A\mathbf v_1 + iA\mathbf v_2 = (\rho\mathbf v_1 - \omega\mathbf v_2) + i(\omega\mathbf v_1 + \rho\mathbf v_2) \;\Longrightarrow\; A\mathbf v_1 = \rho\mathbf v_1 - \omega\mathbf v_2,\quad A\mathbf v_2 = \omega\mathbf v_1 + \rho\mathbf v_2 $$

Collect the two vector equations in one matrix equation, with $Q = [\mathbf v_1 \mid \mathbf v_2]$ (columns): the first column of $QJ$ is $\rho\mathbf v_1 - \omega\mathbf v_2$ for $J = \begin{pmatrix} \rho & \omega \\ -\omega & \rho \end{pmatrix}$ and the second is $\omega\mathbf v_1 + \rho\mathbf v_2$. Hence $AQ = QJ$, and since $\mathbf v_1, \mathbf v_2$ are independent (Exercise 4) $Q$ is invertible:

$$ \boxed{A = QJQ^{-1}, \qquad Q = [\mathbf v_1 \mid \mathbf v_2], \qquad J = \begin{pmatrix} \rho & \omega \\ -\omega & \rho \end{pmatrix}} $$

This $A$ is real (a product of real matrices), has the eigenvalues of $J$, $\rho \pm i\omega$ (similar matrices have the same characteristic polynomial), and $A\boldsymbol\eta_+ = \lambda_+\boldsymbol\eta_+$ by construction.

Method of the class notes [Notes, p. 10]

The identities of the notes read backwards: $A = PDP^{-1}$ and $Q = PM$ give $A = QM^{-1}DMQ^{-1} = QJQ^{-1}$ with $J = M^{-1}DM = \begin{pmatrix} \rho & \omega \\ -\omega & \rho \end{pmatrix}$. It is the matrix of the construction above, which only needs the real and imaginary parts of $A\boldsymbol\eta_+ = \lambda_+\boldsymbol\eta_+$.

Derivation of the three matrices
(i) $\lambda_\pm = 3 \pm 3i,\ \boldsymbol\eta_\pm = (2 \pm i,\ 1 \mp i)$

$\boldsymbol\eta_+ = \begin{pmatrix} 2 + i \\ 1 - i \end{pmatrix} = \begin{pmatrix} 2 \\ 1 \end{pmatrix} + i\begin{pmatrix} 1 \\ -1 \end{pmatrix}$, so $\mathbf v_1 = \begin{pmatrix} 2 \\ 1 \end{pmatrix}$, $\mathbf v_2 = \begin{pmatrix} 1 \\ -1 \end{pmatrix}$, $\rho = 3$, $\omega = 3$.

$$ Q = \begin{pmatrix} 2 & 1 \\ 1 & -1 \end{pmatrix}, \quad J = \begin{pmatrix} 3 & 3 \\ -3 & 3 \end{pmatrix}, \quad \det Q = -3, \quad Q^{-1} = \begin{pmatrix} \frac{1}{3} & \frac{1}{3} \\ \frac{1}{3} & - \frac{2}{3} \end{pmatrix} $$

$$ \begin{aligned} (QJ)_{11} &= 2\cdot3 + 1\cdot(-3) = 3 \\ (QJ)_{12} &= 2\cdot3 + 1\cdot3 = 9 \\ (QJ)_{21} &= 1\cdot3 + (-1)\cdot(-3) = 6 \\ (QJ)_{22} &= 1\cdot3 + (-1)\cdot3 = 0 \end{aligned} $$

$$ \begin{aligned} (A)_{11} &= 3\cdot\frac{1}{3} + 9\cdot\frac{1}{3} = 4 \\ (A)_{12} &= 3\cdot\frac{1}{3} + 9\cdot(- \frac{2}{3}) = -5 \\ (A)_{21} &= 6\cdot\frac{1}{3} + 0\cdot\frac{1}{3} = 2 \\ (A)_{22} &= 6\cdot\frac{1}{3} + 0\cdot(- \frac{2}{3}) = 2 \end{aligned} $$

$$ A = QJQ^{-1} = \begin{pmatrix} 4 & -5 \\ 2 & 2 \end{pmatrix} $$

Verification

$$ A\boldsymbol\eta_+ = \begin{pmatrix} 4 & -5 \\ 2 & 2 \end{pmatrix}\begin{pmatrix} 2 + i \\ 1 - i \end{pmatrix} = \begin{pmatrix} 4\,(2 + i) + (-5)\,(1 - i) = 3 + 9 i \\ 2\,(2 + i) + 2\,(1 - i) = 6 \end{pmatrix} = (3 + 3 i)\,\begin{pmatrix} 2 + i \\ 1 - i \end{pmatrix} $$

Characteristic polynomial: $\lambda^2 - (6)\lambda + (18) = 0$ has $\operatorname{tr}A = 6 = 2\rho$ and $\det A = 18 = \rho^2 + \omega^2$, hence $\lambda = 3 \pm i\sqrt{9}$, i.e. $\rho \pm i\omega$ with $\rho = 3$, $\omega = 3$.

(ii) $\lambda_\pm = \pm 3i,\ \boldsymbol\eta_\pm = (\pm i,\ 3 \pm 2i)$

$\boldsymbol\eta_+ = \begin{pmatrix} i \\ 3 + 2 i \end{pmatrix} = \begin{pmatrix} 0 \\ 3 \end{pmatrix} + i\begin{pmatrix} 1 \\ 2 \end{pmatrix}$, so $\mathbf v_1 = \begin{pmatrix} 0 \\ 3 \end{pmatrix}$, $\mathbf v_2 = \begin{pmatrix} 1 \\ 2 \end{pmatrix}$, $\rho = 0$, $\omega = 3$.

$$ Q = \begin{pmatrix} 0 & 1 \\ 3 & 2 \end{pmatrix}, \quad J = \begin{pmatrix} 0 & 3 \\ -3 & 0 \end{pmatrix}, \quad \det Q = -3, \quad Q^{-1} = \begin{pmatrix} - \frac{2}{3} & \frac{1}{3} \\ 1 & 0 \end{pmatrix} $$

$$ \begin{aligned} (QJ)_{11} &= 0\cdot0 + 1\cdot(-3) = -3 \\ (QJ)_{12} &= 0\cdot3 + 1\cdot0 = 0 \\ (QJ)_{21} &= 3\cdot0 + 2\cdot(-3) = -6 \\ (QJ)_{22} &= 3\cdot3 + 2\cdot0 = 9 \end{aligned} $$

$$ \begin{aligned} (A)_{11} &= (-3)\cdot(- \frac{2}{3}) + 0\cdot1 = 2 \\ (A)_{12} &= (-3)\cdot\frac{1}{3} + 0\cdot0 = -1 \\ (A)_{21} &= (-6)\cdot(- \frac{2}{3}) + 9\cdot1 = 13 \\ (A)_{22} &= (-6)\cdot\frac{1}{3} + 9\cdot0 = -2 \end{aligned} $$

$$ A = QJQ^{-1} = \begin{pmatrix} 2 & -1 \\ 13 & -2 \end{pmatrix} $$

Verification

$$ A\boldsymbol\eta_+ = \begin{pmatrix} 2 & -1 \\ 13 & -2 \end{pmatrix}\begin{pmatrix} i \\ 3 + 2 i \end{pmatrix} = \begin{pmatrix} 2\,(i) + (-1)\,(3 + 2 i) = -3 \\ 13\,(i) + (-2)\,(3 + 2 i) = -6 + 9 i \end{pmatrix} = (3 i)\,\begin{pmatrix} i \\ 3 + 2 i \end{pmatrix} $$

Characteristic polynomial: $\lambda^2 - (0)\lambda + (9) = 0$ has $\operatorname{tr}A = 0 = 2\rho$ and $\det A = 9 = \rho^2 + \omega^2$, hence $\lambda = 0 \pm i\sqrt{9}$, i.e. $\rho \pm i\omega$ with $\rho = 0$, $\omega = 3$.

(iii) $\lambda_\pm = -2 \pm i,\ \boldsymbol\eta_\pm = (1 \pm i,\ 1 \mp i)$

$\boldsymbol\eta_+ = \begin{pmatrix} 1 + i \\ 1 - i \end{pmatrix} = \begin{pmatrix} 1 \\ 1 \end{pmatrix} + i\begin{pmatrix} 1 \\ -1 \end{pmatrix}$, so $\mathbf v_1 = \begin{pmatrix} 1 \\ 1 \end{pmatrix}$, $\mathbf v_2 = \begin{pmatrix} 1 \\ -1 \end{pmatrix}$, $\rho = -2$, $\omega = 1$.

$$ Q = \begin{pmatrix} 1 & 1 \\ 1 & -1 \end{pmatrix}, \quad J = \begin{pmatrix} -2 & 1 \\ -1 & -2 \end{pmatrix}, \quad \det Q = -2, \quad Q^{-1} = \begin{pmatrix} \frac{1}{2} & \frac{1}{2} \\ \frac{1}{2} & - \frac{1}{2} \end{pmatrix} $$

$$ \begin{aligned} (QJ)_{11} &= 1\cdot(-2) + 1\cdot(-1) = -3 \\ (QJ)_{12} &= 1\cdot1 + 1\cdot(-2) = -1 \\ (QJ)_{21} &= 1\cdot(-2) + (-1)\cdot(-1) = -1 \\ (QJ)_{22} &= 1\cdot1 + (-1)\cdot(-2) = 3 \end{aligned} $$

$$ \begin{aligned} (A)_{11} &= (-3)\cdot\frac{1}{2} + (-1)\cdot\frac{1}{2} = -2 \\ (A)_{12} &= (-3)\cdot\frac{1}{2} + (-1)\cdot(- \frac{1}{2}) = -1 \\ (A)_{21} &= (-1)\cdot\frac{1}{2} + 3\cdot\frac{1}{2} = 1 \\ (A)_{22} &= (-1)\cdot\frac{1}{2} + 3\cdot(- \frac{1}{2}) = -2 \end{aligned} $$

$$ A = QJQ^{-1} = \begin{pmatrix} -2 & -1 \\ 1 & -2 \end{pmatrix} $$

Verification

$$ A\boldsymbol\eta_+ = \begin{pmatrix} -2 & -1 \\ 1 & -2 \end{pmatrix}\begin{pmatrix} 1 + i \\ 1 - i \end{pmatrix} = \begin{pmatrix} (-2)\,(1 + i) + (-1)\,(1 - i) = -3 - i \\ 1\,(1 + i) + (-2)\,(1 - i) = -1 + 3 i \end{pmatrix} = (-2 + i)\,\begin{pmatrix} 1 + i \\ 1 - i \end{pmatrix} $$

Characteristic polynomial: $\lambda^2 - (-4)\lambda + (5) = 0$ has $\operatorname{tr}A = -4 = 2\rho$ and $\det A = 5 = \rho^2 + \omega^2$, hence $\lambda = -2 \pm i\sqrt{1}$, i.e. $\rho \pm i\omega$ with $\rho = -2$, $\omega = 1$.

Interpretation and checks
  • The matrix is not unique if only the eigenvalues are prescribed, but it is unique once $\boldsymbol\eta_+$ is given (up to the complex scalar multiple that every eigenvector admits: rescaling $\boldsymbol\eta_+$ by $c = c_r + ic_i$ changes $(\mathbf v_1, \mathbf v_2)$ and leaves $A$ unchanged, since $A\bar{\boldsymbol\eta}_+$, $A\boldsymbol\eta_+$ still determine $A$ on the basis $\{\boldsymbol\eta_+, \bar{\boldsymbol\eta}_+\}$ of $\mathbb C^2$).
  • For the three items $\operatorname{tr}A = 6, 0, -4$ and $\det A = 18, 9, 5$ respectively, which are $2\rho$ and $\rho^2 + \omega^2$ as the construction predicts.
  • (ii) has $\rho = 0$: a centre. $A = \begin{pmatrix} 2 & -1 \\ 13 & -2 \end{pmatrix}$ has trace zero and determinant $9 = \omega^2$.
  • This is the inverse of Exercise 3: from $A$ to $(Q, J)$ there, from $(Q, J)$ to $A$ here.
Exercise 6

Five Systems with a Repeated Eigenvalue: Stars and Improper Nodes

Original statement · probls_ch4-ii_2627.pdf, p. 2
Repeated Eigenvalue • Jordan Chain • Stars and Improper Nodes
Statement

Find the eigenvalues and eigenvectors of the matrices occurring in the following equations, and draw the phase portrait. Find also the coordinate transformation that will put the equation into canonical form and show that this works. Write down the general solution.

  • (i) $\dfrac{d\mathbf x}{dt} = \begin{pmatrix} 5 & -4 \\ 1 & 1 \end{pmatrix}\mathbf x$;
  • (ii) $\dfrac{d\mathbf x}{dt} = \begin{pmatrix} -6 & 2 \\ -2 & -2 \end{pmatrix}\mathbf x$;
  • (iii) $\dfrac{d\mathbf x}{dt} = \begin{pmatrix} -3 & -1 \\ 1 & -5 \end{pmatrix}\mathbf x$;
  • (iv) $\dfrac{d\mathbf x}{dt} = \begin{pmatrix} 13 & 0 \\ 0 & 13 \end{pmatrix}\mathbf x$;
  • (v) $\dfrac{d\mathbf x}{dt} = \begin{pmatrix} 7 & -4 \\ 1 & 3 \end{pmatrix}\mathbf x$.
Setup

For every matrix the discriminant $\Delta = (\operatorname{tr}A)^2 - 4\det A$ is zero (checked in each item), so there is one repeated eigenvalue $\lambda = \tfrac12\operatorname{tr}A$ with algebraic multiplicity $m_a = 2$. The geometric multiplicity $m_g = \dim\ker(A - \lambda I)$ is $2$ iff $A = \lambda I$ (a star) and $1$ otherwise (an improper node with a Jordan chain), Topic 4 Sec. 8.2 to 8.4. The canonical form is $J = \begin{pmatrix} \lambda & 1 \\ 0 & \lambda \end{pmatrix}$ with $Q = [\mathbf v_1 \mid \mathbf v_2]$, $(A - \lambda I)\mathbf v_2 = \mathbf v_1$; for a star there is nothing to transform.

Formulation

Algorithm (Topic 4 Sec. 8.10): $\lambda = \tfrac12\operatorname{tr}A$; $N = A - \lambda I$; if $N = O$ star, $\mathbf x = e^{\lambda t}\mathbf x_0$; otherwise $\mathbf v_1 \in \ker N$, solve $N\mathbf v_2 = \mathbf v_1$, $Q = [\mathbf v_1 \mid \mathbf v_2]$, $\mathbf z = Q^{-1}\mathbf x$, $z_1 = (c_1 + c_2t)e^{\lambda t}$, $z_2 = c_2e^{\lambda t}$, $\mathbf x = Q\mathbf z$.

Method of the class notes [Notes, p. 10]

Case 3 of the notes. 3.1, two independent eigenvectors: $A = P(\lambda I)P^{-1} = \lambda I$, so $\mathbf x = \mathbf x_0e^{\lambda t}$, a star (item (iv)). 3.2, one eigenvector: the notes choose $\mathbf v_2$ independent of the eigenvector and set $\mathbf v_1 = (A - \lambda I)\mathbf v_2$, which is an eigenvector; then $A\mathbf v_2 = \mathbf v_1 + \lambda\mathbf v_2$, $A = P\begin{pmatrix} \lambda & 1 \\ 0 & \lambda \end{pmatrix}P^{-1}$, $z_2 = Ce^{\lambda t}$, $z_1 = (Ct + D)e^{\lambda t}$ and $\mathbf x = (Ct + D)e^{\lambda t}\mathbf v_1 + Ce^{\lambda t}\mathbf v_2$, an improper node.

The algorithm above uses the same relation $N\mathbf v_2 = \mathbf v_1$ the other way round: it fixes the eigenvector $\mathbf v_1$ first and solves for $\mathbf v_2$ ($c_1 = D$, $c_2 = C$).

Derivation
(i) $A = \begin{pmatrix} 5 & -4 \\ 1 & 1 \end{pmatrix}$

Characteristic equation $\det(A - \lambda I) = 0$ (Topic 4, Eq. 27.8):

$$ \det(A - \lambda I) = \begin{vmatrix} 5 - \lambda & -4 \\ 1 & 1 - \lambda \end{vmatrix} = (5 - \lambda)(1 - \lambda) - (-4)(1) $$

$$ = \lambda^2 - (5 + 1)\lambda + (5\cdot1 - (-4)\cdot1) = \lambda^2 - (6)\lambda + (9) = \lambda^{2} - 6 \lambda + 9 $$

$\operatorname{tr}A = 6$, $\det A = 9$, discriminant $\Delta = (\operatorname{tr}A)^2 - 4\det A = 36 - 36 = 0$.

$\Delta = 0$: the characteristic polynomial is a perfect square, $(\lambda - 3)^2 = \lambda^{2} - 6 \lambda + 9$, so

$$ \lambda_1 = \lambda_2 = \lambda = \tfrac12\operatorname{tr}A = 3 \quad (m_a = 2) $$

Geometric multiplicity: $N = A - \lambda I = \begin{pmatrix} 2 & -4 \\ 1 & -2 \end{pmatrix}$.

$N \neq O$ and $\det N = (2)(-2) - (-4)(1) = 0$, so $N$ has rank 1: $m_g = 1$ and there is a single eigendirection (improper node). By Cayley-Hamilton $N^2 = O$ (Topic 4, Sec. 8.4):

$$ \begin{aligned} (N^2)_{11} &= 2\cdot2 + (-4)\cdot1 = 0 \\ (N^2)_{12} &= 2\cdot(-4) + (-4)\cdot(-2) = 0 \\ (N^2)_{21} &= 1\cdot2 + (-2)\cdot1 = 0 \\ (N^2)_{22} &= 1\cdot(-4) + (-2)\cdot(-2) = 0 \end{aligned} $$

Eigenvector

For $\lambda = 3$: $(A - (3)I)\mathbf{v_1} = \mathbf 0$, that is

$$ \begin{pmatrix} 2 & -4 \\ 1 & -2 \end{pmatrix}\begin{pmatrix} v_1 \\ v_2 \end{pmatrix} = \begin{pmatrix} 0 \\ 0 \end{pmatrix} $$

$$ \begin{aligned} 2v_1 - 4v_2 = 0\quad (\text{row 1}) \\ 1v_1 - 2v_2 = 0\quad (\text{row 2}) \end{aligned} $$

The coefficient matrix is singular ($\det(A - \lambda I) = 0$), so the rows are proportional (row 2 equals $\frac{1}{2}$ times row 1) and express a single condition. Row 1 reads $2v_1 + (-4)v_2 = 0$, so $v_1 = 2\,v_2$: every eigenvector is a multiple of $(2, 1)^T$. The smallest integer representative is

$$ \mathbf{v_1} = \begin{pmatrix} 2 \\ 1 \end{pmatrix} $$

Check: $A\mathbf{v_1} = \begin{pmatrix} 6 \\ 3 \end{pmatrix} = 3\mathbf{v_1}$.

Generalised eigenvector

Solve $(A - \lambda I)\mathbf v_2 = \mathbf v_1$, i.e. $N\mathbf v_2 = \mathbf v_1$ (Eq. 8.2). The system is solvable because $\operatorname{im}N = \ker N$ (Sec. 8.4):

$$ \begin{pmatrix} 2 & -4 \\ 1 & -2 \end{pmatrix}\begin{pmatrix} p \\ q \end{pmatrix} = \begin{pmatrix} 2 \\ 1 \end{pmatrix} $$

$$ \begin{aligned} 2p - 4q &= 2 \\ 1p - 2q &= 1 \end{aligned} $$

The two equations are proportional (the right-hand sides are in the same ratio as the rows, because $\mathbf v_1 \in \operatorname{im}N$), so one is enough. The solution is not unique: $\mathbf v_2 + s\mathbf v_1$ also works for every $s$, because $N\mathbf v_1 = \mathbf 0$.

Take $q = 0$. Row 1 gives $2p = 2$, so $p = 1$:

$$ \mathbf v_2 = \begin{pmatrix} 1 \\ 0 \end{pmatrix} $$

Check: $N\mathbf v_2 = \begin{pmatrix} 2 \\ 1 \end{pmatrix} = \mathbf v_1$, equivalently $A\mathbf v_2 = \lambda\mathbf v_2 + \mathbf v_1 = \begin{pmatrix} 5 \\ 1 \end{pmatrix}$ (Eq. 8.2).

Change of coordinates

$$ Q = [\mathbf v_1 \mid \mathbf v_2] = \begin{pmatrix} 2 & 1 \\ 1 & 0 \end{pmatrix} $$

$\det Q = (2)(0) - (1)(1) = -1$, which is not zero, so $Q$ is invertible. For a $2\times 2$ matrix $\begin{pmatrix} a & b \\ c & d \end{pmatrix}^{-1} = \frac{1}{ad - bc}\begin{pmatrix} d & -b \\ -c & a \end{pmatrix}$ (Robinson, Appendix B):

$$ Q^{-1} = \frac{1}{-1}\begin{pmatrix} 0 & -1 \\ -1 & 2 \end{pmatrix} = \begin{pmatrix} 0 & 1 \\ 1 & -2 \end{pmatrix} $$

$$ A\mathbf v_1 = \begin{pmatrix} 6 \\ 3 \end{pmatrix} = 3\mathbf v_1, \qquad A\mathbf v_2 = \begin{pmatrix} 5 \\ 1 \end{pmatrix} = \mathbf v_1 + 3\mathbf v_2 $$

$$ AQ = \begin{pmatrix} 6 & 5 \\ 3 & 1 \end{pmatrix} $$

$$ \begin{aligned} (Q^{-1}AQ)_{11} &= 0\cdot6 + 1\cdot3 = 3 \\ (Q^{-1}AQ)_{12} &= 0\cdot5 + 1\cdot1 = 1 \\ (Q^{-1}AQ)_{21} &= 1\cdot6 + (-2)\cdot3 = 0 \\ (Q^{-1}AQ)_{22} &= 1\cdot5 + (-2)\cdot1 = 3 \end{aligned} $$

$$ Q^{-1}AQ = \begin{pmatrix} 3 & 1 \\ 0 & 3 \end{pmatrix} = J = \begin{pmatrix} \lambda & 1 \\ 0 & \lambda \end{pmatrix} $$

Solution in the canonical plane

$$ \dot z_1 = 3z_1 + z_2, \qquad \dot z_2 = 3z_2 $$

Second equation: $\frac{d}{dt}(e^{-\lambda t}z_2) = e^{-\lambda t}(\dot z_2 - \lambda z_2) = 0$, so $e^{-\lambda t}z_2$ is constant: $z_2 = c_2 e^{3t}$. First equation with this $z_2$: $\dot z_1 - \lambda z_1 = c_2 e^{3t}$. The integrating factor is $e^{-\lambda t}$: by the product rule $\frac{d}{dt}(e^{-\lambda t}z_1) = e^{-\lambda t}(\dot z_1 - \lambda z_1) = c_2$. Barrow's rule between $0$ and $t$:

$$ \left[e^{-\lambda s}z_1(s)\right]_{s=0}^{s=t} = \int_0^t c_2\,ds = c_2 t \;\Longrightarrow\; e^{-\lambda t}z_1(t) - z_1(0) = c_2 t $$

$$ z_1(t) = (c_1 + c_2 t)e^{3t}, \qquad z_2(t) = c_2 e^{3t} \qquad (c_1 = z_1(0),\ c_2 = z_2(0)) $$

Back to the original coordinates

$\mathbf x = Q\mathbf z = z_1\mathbf v_1 + z_2\mathbf v_2$, hence (Topic 4, Eq. 8.5)

$$ \mathbf x(t) = e^{3t}\left[c_1\begin{pmatrix} 2 \\ 1 \end{pmatrix} + c_2\left(t\begin{pmatrix} 2 \\ 1 \end{pmatrix} + \begin{pmatrix} 1 \\ 0 \end{pmatrix}\right)\right] $$

$$ x(t) = 2 c_{1} e^{3 t} + 2 c_{2} t e^{3 t} + c_{2} e^{3 t}, \qquad y(t) = c_{1} e^{3 t} + c_{2} t e^{3 t} $$

Phase portrait

$\lambda = 3$ is repeated with a single eigendirection and $A \neq \lambda I$: the origin is an unstable improper node (Topic 4 Sec. 8.8 and 8.9). In the $\mathbf z$-plane the $z_1$-axis is invariant ($z_2 \equiv 0$, $z_1 = c_1e^{3t}$, arrows away from the origin); the other orbits satisfy $z_1 = z_2\left[K + \frac{1}{\lambda}\ln|z_2|\right]$, are S-shaped, tangent to the $z_1$-axis at the origin and parallel to it far away. Since $\dot z_1 = \lambda z_1 + z_2 = z_2$ on the $z_2$-axis, orbits cross it moving to the right in the upper half-plane and to the left in the lower one.

Orientation in the $\mathbf x$-plane, from $N\mathbf x = (A - \lambda I)\mathbf x = z_2\mathbf v_1$ (Topic 4 Sec. 8.9): at $(0, 1)^T$, $N(0, 1)^T = \begin{pmatrix} -4 \\ -2 \end{pmatrix} = -2\,\mathbf v_1$, so there the orbit drifts along $-\mathbf v_1$. The orbits are tangent to the eigenline at the origin and bend into an S around it.

z1 z2 z1 z-plane
Canonical $\mathbf z$-plane (unstable improper node).
Open the figure full size
x y v1 x-plane
Original $(x, y)$-plane (unstable improper node); dashed: the eigenline with $\mathbf v_1 = (2, 1)$.
Open the figure full size
(ii) $A = \begin{pmatrix} -6 & 2 \\ -2 & -2 \end{pmatrix}$

Characteristic equation $\det(A - \lambda I) = 0$ (Topic 4, Eq. 27.8):

$$ \det(A - \lambda I) = \begin{vmatrix} -6 - \lambda & 2 \\ -2 & -2 - \lambda \end{vmatrix} = (-6 - \lambda)(-2 - \lambda) - (2)(-2) $$

$$ = \lambda^2 - (-6 + (-2))\lambda + (-6\cdot(-2) - 2\cdot(-2)) = \lambda^2 - (-8)\lambda + (16) = \lambda^{2} + 8 \lambda + 16 $$

$\operatorname{tr}A = -8$, $\det A = 16$, discriminant $\Delta = (\operatorname{tr}A)^2 - 4\det A = 64 - 64 = 0$.

$\Delta = 0$: the characteristic polynomial is a perfect square, $(\lambda + 4)^2 = \lambda^{2} + 8 \lambda + 16$, so

$$ \lambda_1 = \lambda_2 = \lambda = \tfrac12\operatorname{tr}A = -4 \quad (m_a = 2) $$

Geometric multiplicity: $N = A - \lambda I = \begin{pmatrix} -2 & 2 \\ -2 & 2 \end{pmatrix}$.

$N \neq O$ and $\det N = (-2)(2) - (2)(-2) = 0$, so $N$ has rank 1: $m_g = 1$ and there is a single eigendirection (improper node). By Cayley-Hamilton $N^2 = O$ (Topic 4, Sec. 8.4):

$$ \begin{aligned} (N^2)_{11} &= (-2)\cdot(-2) + 2\cdot(-2) = 0 \\ (N^2)_{12} &= (-2)\cdot2 + 2\cdot2 = 0 \\ (N^2)_{21} &= (-2)\cdot(-2) + 2\cdot(-2) = 0 \\ (N^2)_{22} &= (-2)\cdot2 + 2\cdot2 = 0 \end{aligned} $$

Eigenvector

For $\lambda = -4$: $(A - (-4)I)\mathbf{v_1} = \mathbf 0$, that is

$$ \begin{pmatrix} -2 & 2 \\ -2 & 2 \end{pmatrix}\begin{pmatrix} v_1 \\ v_2 \end{pmatrix} = \begin{pmatrix} 0 \\ 0 \end{pmatrix} $$

$$ \begin{aligned} -2v_1 + 2v_2 = 0\quad (\text{row 1}) \\ -2v_1 + 2v_2 = 0\quad (\text{row 2}) \end{aligned} $$

The coefficient matrix is singular ($\det(A - \lambda I) = 0$), so the rows are proportional (row 2 is identical to row 1) and express a single condition. Row 1 reads $-2v_1 + 2v_2 = 0$, so $v_1 = v_2$: every eigenvector is a multiple of $(1, 1)^T$. The smallest integer representative is

$$ \mathbf{v_1} = \begin{pmatrix} 1 \\ 1 \end{pmatrix} $$

Check: $A\mathbf{v_1} = \begin{pmatrix} -4 \\ -4 \end{pmatrix} = -4\mathbf{v_1}$.

Generalised eigenvector

Solve $(A - \lambda I)\mathbf v_2 = \mathbf v_1$, i.e. $N\mathbf v_2 = \mathbf v_1$ (Eq. 8.2). The system is solvable because $\operatorname{im}N = \ker N$ (Sec. 8.4):

$$ \begin{pmatrix} -2 & 2 \\ -2 & 2 \end{pmatrix}\begin{pmatrix} p \\ q \end{pmatrix} = \begin{pmatrix} 1 \\ 1 \end{pmatrix} $$

$$ \begin{aligned} -2p + 2q &= 1 \\ -2p + 2q &= 1 \end{aligned} $$

The two equations are proportional (the right-hand sides are in the same ratio as the rows, because $\mathbf v_1 \in \operatorname{im}N$), so one is enough. The solution is not unique: $\mathbf v_2 + s\mathbf v_1$ also works for every $s$, because $N\mathbf v_1 = \mathbf 0$.

Take $p = 0$. Row 1 gives $2q = 1$, so $q = \frac{1}{2}$:

$$ \mathbf v_2 = \begin{pmatrix} 0 \\ \frac{1}{2} \end{pmatrix} $$

Check: $N\mathbf v_2 = \begin{pmatrix} 1 \\ 1 \end{pmatrix} = \mathbf v_1$, equivalently $A\mathbf v_2 = \lambda\mathbf v_2 + \mathbf v_1 = \begin{pmatrix} 1 \\ -1 \end{pmatrix}$ (Eq. 8.2).

Change of coordinates

$$ Q = [\mathbf v_1 \mid \mathbf v_2] = \begin{pmatrix} 1 & 0 \\ 1 & \frac{1}{2} \end{pmatrix} $$

$\det Q = (1)(\frac{1}{2}) - (0)(1) = \frac{1}{2}$, which is not zero, so $Q$ is invertible. For a $2\times 2$ matrix $\begin{pmatrix} a & b \\ c & d \end{pmatrix}^{-1} = \frac{1}{ad - bc}\begin{pmatrix} d & -b \\ -c & a \end{pmatrix}$ (Robinson, Appendix B):

$$ Q^{-1} = \frac{1}{\frac{1}{2}}\begin{pmatrix} \frac{1}{2} & 0 \\ -1 & 1 \end{pmatrix} = \begin{pmatrix} 1 & 0 \\ -2 & 2 \end{pmatrix} $$

$$ A\mathbf v_1 = \begin{pmatrix} -4 \\ -4 \end{pmatrix} = (-4)\mathbf v_1, \qquad A\mathbf v_2 = \begin{pmatrix} 1 \\ -1 \end{pmatrix} = \mathbf v_1 + (-4)\mathbf v_2 $$

$$ AQ = \begin{pmatrix} -4 & 1 \\ -4 & -1 \end{pmatrix} $$

$$ \begin{aligned} (Q^{-1}AQ)_{11} &= 1\cdot(-4) + 0\cdot(-4) = -4 \\ (Q^{-1}AQ)_{12} &= 1\cdot1 + 0\cdot(-1) = 1 \\ (Q^{-1}AQ)_{21} &= (-2)\cdot(-4) + 2\cdot(-4) = 0 \\ (Q^{-1}AQ)_{22} &= (-2)\cdot1 + 2\cdot(-1) = -4 \end{aligned} $$

$$ Q^{-1}AQ = \begin{pmatrix} -4 & 1 \\ 0 & -4 \end{pmatrix} = J = \begin{pmatrix} \lambda & 1 \\ 0 & \lambda \end{pmatrix} $$

Solution in the canonical plane

$$ \dot z_1 = -4z_1 + z_2, \qquad \dot z_2 = -4z_2 $$

Second equation: $\frac{d}{dt}(e^{-\lambda t}z_2) = e^{-\lambda t}(\dot z_2 - \lambda z_2) = 0$, so $e^{-\lambda t}z_2$ is constant: $z_2 = c_2 e^{-4t}$. First equation with this $z_2$: $\dot z_1 - \lambda z_1 = c_2 e^{-4t}$. The integrating factor is $e^{-\lambda t}$: by the product rule $\frac{d}{dt}(e^{-\lambda t}z_1) = e^{-\lambda t}(\dot z_1 - \lambda z_1) = c_2$. Barrow's rule between $0$ and $t$:

$$ \left[e^{-\lambda s}z_1(s)\right]_{s=0}^{s=t} = \int_0^t c_2\,ds = c_2 t \;\Longrightarrow\; e^{-\lambda t}z_1(t) - z_1(0) = c_2 t $$

$$ z_1(t) = (c_1 + c_2 t)e^{-4t}, \qquad z_2(t) = c_2 e^{-4t} \qquad (c_1 = z_1(0),\ c_2 = z_2(0)) $$

Back to the original coordinates

$\mathbf x = Q\mathbf z = z_1\mathbf v_1 + z_2\mathbf v_2$, hence (Topic 4, Eq. 8.5)

$$ \mathbf x(t) = e^{-4t}\left[c_1\begin{pmatrix} 1 \\ 1 \end{pmatrix} + c_2\left(t\begin{pmatrix} 1 \\ 1 \end{pmatrix} + \begin{pmatrix} 0 \\ \frac{1}{2} \end{pmatrix}\right)\right] $$

$$ x(t) = c_{1} e^{-4 t} + c_{2} t e^{-4 t}, \qquad y(t) = c_{1} e^{-4 t} + c_{2} t e^{-4 t} + \frac{c_{2} e^{-4 t}}{2} $$

Phase portrait

$\lambda = -4$ is repeated with a single eigendirection and $A \neq \lambda I$: the origin is an stable improper node (Topic 4 Sec. 8.8 and 8.9). In the $\mathbf z$-plane the $z_1$-axis is invariant ($z_2 \equiv 0$, $z_1 = c_1e^{-4t}$, arrows towards the origin); the other orbits satisfy $z_1 = z_2\left[K + \frac{1}{\lambda}\ln|z_2|\right]$, are S-shaped, tangent to the $z_1$-axis at the origin and parallel to it far away. Since $\dot z_1 = \lambda z_1 + z_2 = z_2$ on the $z_2$-axis, orbits cross it moving to the right in the upper half-plane and to the left in the lower one.

Orientation in the $\mathbf x$-plane, from $N\mathbf x = (A - \lambda I)\mathbf x = z_2\mathbf v_1$ (Topic 4 Sec. 8.9): at $(0, 1)^T$, $N(0, 1)^T = \begin{pmatrix} 2 \\ 2 \end{pmatrix} = 2\,\mathbf v_1$, so there the orbit drifts along $+\mathbf v_1$. The orbits are tangent to the eigenline at the origin and bend into an S around it.

z1 z2 z1 z-plane
Canonical $\mathbf z$-plane (stable improper node).
Open the figure full size
x y v1 x-plane
Original $(x, y)$-plane (stable improper node); dashed: the eigenline with $\mathbf v_1 = (1, 1)$.
Open the figure full size
(iii) $A = \begin{pmatrix} -3 & -1 \\ 1 & -5 \end{pmatrix}$

Characteristic equation $\det(A - \lambda I) = 0$ (Topic 4, Eq. 27.8):

$$ \det(A - \lambda I) = \begin{vmatrix} -3 - \lambda & -1 \\ 1 & -5 - \lambda \end{vmatrix} = (-3 - \lambda)(-5 - \lambda) - (-1)(1) $$

$$ = \lambda^2 - (-3 + (-5))\lambda + (-3\cdot(-5) - (-1)\cdot1) = \lambda^2 - (-8)\lambda + (16) = \lambda^{2} + 8 \lambda + 16 $$

$\operatorname{tr}A = -8$, $\det A = 16$, discriminant $\Delta = (\operatorname{tr}A)^2 - 4\det A = 64 - 64 = 0$.

$\Delta = 0$: the characteristic polynomial is a perfect square, $(\lambda + 4)^2 = \lambda^{2} + 8 \lambda + 16$, so

$$ \lambda_1 = \lambda_2 = \lambda = \tfrac12\operatorname{tr}A = -4 \quad (m_a = 2) $$

Geometric multiplicity: $N = A - \lambda I = \begin{pmatrix} 1 & -1 \\ 1 & -1 \end{pmatrix}$.

$N \neq O$ and $\det N = (1)(-1) - (-1)(1) = 0$, so $N$ has rank 1: $m_g = 1$ and there is a single eigendirection (improper node). By Cayley-Hamilton $N^2 = O$ (Topic 4, Sec. 8.4):

$$ \begin{aligned} (N^2)_{11} &= 1\cdot1 + (-1)\cdot1 = 0 \\ (N^2)_{12} &= 1\cdot(-1) + (-1)\cdot(-1) = 0 \\ (N^2)_{21} &= 1\cdot1 + (-1)\cdot1 = 0 \\ (N^2)_{22} &= 1\cdot(-1) + (-1)\cdot(-1) = 0 \end{aligned} $$

Eigenvector

For $\lambda = -4$: $(A - (-4)I)\mathbf{v_1} = \mathbf 0$, that is

$$ \begin{pmatrix} 1 & -1 \\ 1 & -1 \end{pmatrix}\begin{pmatrix} v_1 \\ v_2 \end{pmatrix} = \begin{pmatrix} 0 \\ 0 \end{pmatrix} $$

$$ \begin{aligned} 1v_1 - 1v_2 = 0\quad (\text{row 1}) \\ 1v_1 - 1v_2 = 0\quad (\text{row 2}) \end{aligned} $$

The coefficient matrix is singular ($\det(A - \lambda I) = 0$), so the rows are proportional (row 2 is identical to row 1) and express a single condition. Row 1 reads $1v_1 + (-1)v_2 = 0$, so $v_1 = v_2$: every eigenvector is a multiple of $(1, 1)^T$. The smallest integer representative is

$$ \mathbf{v_1} = \begin{pmatrix} 1 \\ 1 \end{pmatrix} $$

Check: $A\mathbf{v_1} = \begin{pmatrix} -4 \\ -4 \end{pmatrix} = -4\mathbf{v_1}$.

Generalised eigenvector

Solve $(A - \lambda I)\mathbf v_2 = \mathbf v_1$, i.e. $N\mathbf v_2 = \mathbf v_1$ (Eq. 8.2). The system is solvable because $\operatorname{im}N = \ker N$ (Sec. 8.4):

$$ \begin{pmatrix} 1 & -1 \\ 1 & -1 \end{pmatrix}\begin{pmatrix} p \\ q \end{pmatrix} = \begin{pmatrix} 1 \\ 1 \end{pmatrix} $$

$$ \begin{aligned} 1p - 1q &= 1 \\ 1p - 1q &= 1 \end{aligned} $$

The two equations are proportional (the right-hand sides are in the same ratio as the rows, because $\mathbf v_1 \in \operatorname{im}N$), so one is enough. The solution is not unique: $\mathbf v_2 + s\mathbf v_1$ also works for every $s$, because $N\mathbf v_1 = \mathbf 0$.

Take $p = 0$. Row 1 gives $-1q = 1$, so $q = -1$:

$$ \mathbf v_2 = \begin{pmatrix} 0 \\ -1 \end{pmatrix} $$

Check: $N\mathbf v_2 = \begin{pmatrix} 1 \\ 1 \end{pmatrix} = \mathbf v_1$, equivalently $A\mathbf v_2 = \lambda\mathbf v_2 + \mathbf v_1 = \begin{pmatrix} 1 \\ 5 \end{pmatrix}$ (Eq. 8.2).

Change of coordinates

$$ Q = [\mathbf v_1 \mid \mathbf v_2] = \begin{pmatrix} 1 & 0 \\ 1 & -1 \end{pmatrix} $$

$\det Q = (1)(-1) - (0)(1) = -1$, which is not zero, so $Q$ is invertible. For a $2\times 2$ matrix $\begin{pmatrix} a & b \\ c & d \end{pmatrix}^{-1} = \frac{1}{ad - bc}\begin{pmatrix} d & -b \\ -c & a \end{pmatrix}$ (Robinson, Appendix B):

$$ Q^{-1} = \frac{1}{-1}\begin{pmatrix} -1 & 0 \\ -1 & 1 \end{pmatrix} = \begin{pmatrix} 1 & 0 \\ 1 & -1 \end{pmatrix} $$

$$ A\mathbf v_1 = \begin{pmatrix} -4 \\ -4 \end{pmatrix} = (-4)\mathbf v_1, \qquad A\mathbf v_2 = \begin{pmatrix} 1 \\ 5 \end{pmatrix} = \mathbf v_1 + (-4)\mathbf v_2 $$

$$ AQ = \begin{pmatrix} -4 & 1 \\ -4 & 5 \end{pmatrix} $$

$$ \begin{aligned} (Q^{-1}AQ)_{11} &= 1\cdot(-4) + 0\cdot(-4) = -4 \\ (Q^{-1}AQ)_{12} &= 1\cdot1 + 0\cdot5 = 1 \\ (Q^{-1}AQ)_{21} &= 1\cdot(-4) + (-1)\cdot(-4) = 0 \\ (Q^{-1}AQ)_{22} &= 1\cdot1 + (-1)\cdot5 = -4 \end{aligned} $$

$$ Q^{-1}AQ = \begin{pmatrix} -4 & 1 \\ 0 & -4 \end{pmatrix} = J = \begin{pmatrix} \lambda & 1 \\ 0 & \lambda \end{pmatrix} $$

Solution in the canonical plane

$$ \dot z_1 = -4z_1 + z_2, \qquad \dot z_2 = -4z_2 $$

Second equation: $\frac{d}{dt}(e^{-\lambda t}z_2) = e^{-\lambda t}(\dot z_2 - \lambda z_2) = 0$, so $e^{-\lambda t}z_2$ is constant: $z_2 = c_2 e^{-4t}$. First equation with this $z_2$: $\dot z_1 - \lambda z_1 = c_2 e^{-4t}$. The integrating factor is $e^{-\lambda t}$: by the product rule $\frac{d}{dt}(e^{-\lambda t}z_1) = e^{-\lambda t}(\dot z_1 - \lambda z_1) = c_2$. Barrow's rule between $0$ and $t$:

$$ \left[e^{-\lambda s}z_1(s)\right]_{s=0}^{s=t} = \int_0^t c_2\,ds = c_2 t \;\Longrightarrow\; e^{-\lambda t}z_1(t) - z_1(0) = c_2 t $$

$$ z_1(t) = (c_1 + c_2 t)e^{-4t}, \qquad z_2(t) = c_2 e^{-4t} \qquad (c_1 = z_1(0),\ c_2 = z_2(0)) $$

Back to the original coordinates

$\mathbf x = Q\mathbf z = z_1\mathbf v_1 + z_2\mathbf v_2$, hence (Topic 4, Eq. 8.5)

$$ \mathbf x(t) = e^{-4t}\left[c_1\begin{pmatrix} 1 \\ 1 \end{pmatrix} + c_2\left(t\begin{pmatrix} 1 \\ 1 \end{pmatrix} + \begin{pmatrix} 0 \\ -1 \end{pmatrix}\right)\right] $$

$$ x(t) = c_{1} e^{-4 t} + c_{2} t e^{-4 t}, \qquad y(t) = c_{1} e^{-4 t} + c_{2} t e^{-4 t} - c_{2} e^{-4 t} $$

Phase portrait

$\lambda = -4$ is repeated with a single eigendirection and $A \neq \lambda I$: the origin is an stable improper node (Topic 4 Sec. 8.8 and 8.9). In the $\mathbf z$-plane the $z_1$-axis is invariant ($z_2 \equiv 0$, $z_1 = c_1e^{-4t}$, arrows towards the origin); the other orbits satisfy $z_1 = z_2\left[K + \frac{1}{\lambda}\ln|z_2|\right]$, are S-shaped, tangent to the $z_1$-axis at the origin and parallel to it far away. Since $\dot z_1 = \lambda z_1 + z_2 = z_2$ on the $z_2$-axis, orbits cross it moving to the right in the upper half-plane and to the left in the lower one.

Orientation in the $\mathbf x$-plane, from $N\mathbf x = (A - \lambda I)\mathbf x = z_2\mathbf v_1$ (Topic 4 Sec. 8.9): at $(0, 1)^T$, $N(0, 1)^T = \begin{pmatrix} -1 \\ -1 \end{pmatrix} = -1\,\mathbf v_1$, so there the orbit drifts along $-\mathbf v_1$. The orbits are tangent to the eigenline at the origin and bend into an S around it.

z1 z2 z1 z-plane
Canonical $\mathbf z$-plane (stable improper node).
Open the figure full size
x y v1 x-plane
Original $(x, y)$-plane (stable improper node); dashed: the eigenline with $\mathbf v_1 = (1, 1)$.
Open the figure full size
(iv) $A = \begin{pmatrix} 13 & 0 \\ 0 & 13 \end{pmatrix}$

Characteristic equation $\det(A - \lambda I) = 0$ (Topic 4, Eq. 27.8):

$$ \det(A - \lambda I) = \begin{vmatrix} 13 - \lambda & 0 \\ 0 & 13 - \lambda \end{vmatrix} = (13 - \lambda)(13 - \lambda) - (0)(0) $$

$$ = \lambda^2 - (13 + 13)\lambda + (13\cdot13 - 0\cdot0) = \lambda^2 - (26)\lambda + (169) = \lambda^{2} - 26 \lambda + 169 $$

$\operatorname{tr}A = 26$, $\det A = 169$, discriminant $\Delta = (\operatorname{tr}A)^2 - 4\det A = 676 - 676 = 0$.

$\Delta = 0$: the characteristic polynomial is a perfect square, $(\lambda - 13)^2 = \lambda^{2} - 26 \lambda + 169$, so

$$ \lambda_1 = \lambda_2 = \lambda = \tfrac12\operatorname{tr}A = 13 \quad (m_a = 2) $$

Geometric multiplicity: $N = A - \lambda I = \begin{pmatrix} 0 & 0 \\ 0 & 0 \end{pmatrix}$.

$N = O$: every vector is an eigenvector ($m_g = 2$) and $A = \lambda I$ (Topic 4, Sec. 8.3). The system is already uncoupled in every basis.

$$ \dot x = 13x, \qquad \dot y = 13y $$

Each equation is separable, $\frac{d}{dt}(e^{-\lambda t}x_i) = 0$, so $x_i = c_i e^{13t}$ and

$$ \mathbf x(t) = e^{13t}\mathbf x_0 = c_1 e^{13t}\begin{pmatrix} 1 \\ 0 \end{pmatrix} + c_2 e^{13t}\begin{pmatrix} 0 \\ 1 \end{pmatrix} $$

Phase portrait

$A = 13I$: every line through the origin is invariant and every solution moves along its own ray, $\mathbf x(t) = e^{13t}\mathbf x_0$, away from the origin (the origin is a unstable star). The picture is the same in every basis, so there is no separate canonical plane.

x y x-plane
Unstable star: straight rays out of the origin.
Open the figure full size
(v) $A = \begin{pmatrix} 7 & -4 \\ 1 & 3 \end{pmatrix}$

Characteristic equation $\det(A - \lambda I) = 0$ (Topic 4, Eq. 27.8):

$$ \det(A - \lambda I) = \begin{vmatrix} 7 - \lambda & -4 \\ 1 & 3 - \lambda \end{vmatrix} = (7 - \lambda)(3 - \lambda) - (-4)(1) $$

$$ = \lambda^2 - (7 + 3)\lambda + (7\cdot3 - (-4)\cdot1) = \lambda^2 - (10)\lambda + (25) = \lambda^{2} - 10 \lambda + 25 $$

$\operatorname{tr}A = 10$, $\det A = 25$, discriminant $\Delta = (\operatorname{tr}A)^2 - 4\det A = 100 - 100 = 0$.

$\Delta = 0$: the characteristic polynomial is a perfect square, $(\lambda - 5)^2 = \lambda^{2} - 10 \lambda + 25$, so

$$ \lambda_1 = \lambda_2 = \lambda = \tfrac12\operatorname{tr}A = 5 \quad (m_a = 2) $$

Geometric multiplicity: $N = A - \lambda I = \begin{pmatrix} 2 & -4 \\ 1 & -2 \end{pmatrix}$.

$N \neq O$ and $\det N = (2)(-2) - (-4)(1) = 0$, so $N$ has rank 1: $m_g = 1$ and there is a single eigendirection (improper node). By Cayley-Hamilton $N^2 = O$ (Topic 4, Sec. 8.4):

$$ \begin{aligned} (N^2)_{11} &= 2\cdot2 + (-4)\cdot1 = 0 \\ (N^2)_{12} &= 2\cdot(-4) + (-4)\cdot(-2) = 0 \\ (N^2)_{21} &= 1\cdot2 + (-2)\cdot1 = 0 \\ (N^2)_{22} &= 1\cdot(-4) + (-2)\cdot(-2) = 0 \end{aligned} $$

Eigenvector

For $\lambda = 5$: $(A - (5)I)\mathbf{v_1} = \mathbf 0$, that is

$$ \begin{pmatrix} 2 & -4 \\ 1 & -2 \end{pmatrix}\begin{pmatrix} v_1 \\ v_2 \end{pmatrix} = \begin{pmatrix} 0 \\ 0 \end{pmatrix} $$

$$ \begin{aligned} 2v_1 - 4v_2 = 0\quad (\text{row 1}) \\ 1v_1 - 2v_2 = 0\quad (\text{row 2}) \end{aligned} $$

The coefficient matrix is singular ($\det(A - \lambda I) = 0$), so the rows are proportional (row 2 equals $\frac{1}{2}$ times row 1) and express a single condition. Row 1 reads $2v_1 + (-4)v_2 = 0$, so $v_1 = 2\,v_2$: every eigenvector is a multiple of $(2, 1)^T$. The smallest integer representative is

$$ \mathbf{v_1} = \begin{pmatrix} 2 \\ 1 \end{pmatrix} $$

Check: $A\mathbf{v_1} = \begin{pmatrix} 10 \\ 5 \end{pmatrix} = 5\mathbf{v_1}$.

Generalised eigenvector

Solve $(A - \lambda I)\mathbf v_2 = \mathbf v_1$, i.e. $N\mathbf v_2 = \mathbf v_1$ (Eq. 8.2). The system is solvable because $\operatorname{im}N = \ker N$ (Sec. 8.4):

$$ \begin{pmatrix} 2 & -4 \\ 1 & -2 \end{pmatrix}\begin{pmatrix} p \\ q \end{pmatrix} = \begin{pmatrix} 2 \\ 1 \end{pmatrix} $$

$$ \begin{aligned} 2p - 4q &= 2 \\ 1p - 2q &= 1 \end{aligned} $$

The two equations are proportional (the right-hand sides are in the same ratio as the rows, because $\mathbf v_1 \in \operatorname{im}N$), so one is enough. The solution is not unique: $\mathbf v_2 + s\mathbf v_1$ also works for every $s$, because $N\mathbf v_1 = \mathbf 0$.

Take $q = 0$. Row 1 gives $2p = 2$, so $p = 1$:

$$ \mathbf v_2 = \begin{pmatrix} 1 \\ 0 \end{pmatrix} $$

Check: $N\mathbf v_2 = \begin{pmatrix} 2 \\ 1 \end{pmatrix} = \mathbf v_1$, equivalently $A\mathbf v_2 = \lambda\mathbf v_2 + \mathbf v_1 = \begin{pmatrix} 7 \\ 1 \end{pmatrix}$ (Eq. 8.2).

Change of coordinates

$$ Q = [\mathbf v_1 \mid \mathbf v_2] = \begin{pmatrix} 2 & 1 \\ 1 & 0 \end{pmatrix} $$

$\det Q = (2)(0) - (1)(1) = -1$, which is not zero, so $Q$ is invertible. For a $2\times 2$ matrix $\begin{pmatrix} a & b \\ c & d \end{pmatrix}^{-1} = \frac{1}{ad - bc}\begin{pmatrix} d & -b \\ -c & a \end{pmatrix}$ (Robinson, Appendix B):

$$ Q^{-1} = \frac{1}{-1}\begin{pmatrix} 0 & -1 \\ -1 & 2 \end{pmatrix} = \begin{pmatrix} 0 & 1 \\ 1 & -2 \end{pmatrix} $$

$$ A\mathbf v_1 = \begin{pmatrix} 10 \\ 5 \end{pmatrix} = 5\mathbf v_1, \qquad A\mathbf v_2 = \begin{pmatrix} 7 \\ 1 \end{pmatrix} = \mathbf v_1 + 5\mathbf v_2 $$

$$ AQ = \begin{pmatrix} 10 & 7 \\ 5 & 1 \end{pmatrix} $$

$$ \begin{aligned} (Q^{-1}AQ)_{11} &= 0\cdot10 + 1\cdot5 = 5 \\ (Q^{-1}AQ)_{12} &= 0\cdot7 + 1\cdot1 = 1 \\ (Q^{-1}AQ)_{21} &= 1\cdot10 + (-2)\cdot5 = 0 \\ (Q^{-1}AQ)_{22} &= 1\cdot7 + (-2)\cdot1 = 5 \end{aligned} $$

$$ Q^{-1}AQ = \begin{pmatrix} 5 & 1 \\ 0 & 5 \end{pmatrix} = J = \begin{pmatrix} \lambda & 1 \\ 0 & \lambda \end{pmatrix} $$

Solution in the canonical plane

$$ \dot z_1 = 5z_1 + z_2, \qquad \dot z_2 = 5z_2 $$

Second equation: $\frac{d}{dt}(e^{-\lambda t}z_2) = e^{-\lambda t}(\dot z_2 - \lambda z_2) = 0$, so $e^{-\lambda t}z_2$ is constant: $z_2 = c_2 e^{5t}$. First equation with this $z_2$: $\dot z_1 - \lambda z_1 = c_2 e^{5t}$. The integrating factor is $e^{-\lambda t}$: by the product rule $\frac{d}{dt}(e^{-\lambda t}z_1) = e^{-\lambda t}(\dot z_1 - \lambda z_1) = c_2$. Barrow's rule between $0$ and $t$:

$$ \left[e^{-\lambda s}z_1(s)\right]_{s=0}^{s=t} = \int_0^t c_2\,ds = c_2 t \;\Longrightarrow\; e^{-\lambda t}z_1(t) - z_1(0) = c_2 t $$

$$ z_1(t) = (c_1 + c_2 t)e^{5t}, \qquad z_2(t) = c_2 e^{5t} \qquad (c_1 = z_1(0),\ c_2 = z_2(0)) $$

Back to the original coordinates

$\mathbf x = Q\mathbf z = z_1\mathbf v_1 + z_2\mathbf v_2$, hence (Topic 4, Eq. 8.5)

$$ \mathbf x(t) = e^{5t}\left[c_1\begin{pmatrix} 2 \\ 1 \end{pmatrix} + c_2\left(t\begin{pmatrix} 2 \\ 1 \end{pmatrix} + \begin{pmatrix} 1 \\ 0 \end{pmatrix}\right)\right] $$

$$ x(t) = 2 c_{1} e^{5 t} + 2 c_{2} t e^{5 t} + c_{2} e^{5 t}, \qquad y(t) = c_{1} e^{5 t} + c_{2} t e^{5 t} $$

Phase portrait

$\lambda = 5$ is repeated with a single eigendirection and $A \neq \lambda I$: the origin is an unstable improper node (Topic 4 Sec. 8.8 and 8.9). In the $\mathbf z$-plane the $z_1$-axis is invariant ($z_2 \equiv 0$, $z_1 = c_1e^{5t}$, arrows away from the origin); the other orbits satisfy $z_1 = z_2\left[K + \frac{1}{\lambda}\ln|z_2|\right]$, are S-shaped, tangent to the $z_1$-axis at the origin and parallel to it far away. Since $\dot z_1 = \lambda z_1 + z_2 = z_2$ on the $z_2$-axis, orbits cross it moving to the right in the upper half-plane and to the left in the lower one.

Orientation in the $\mathbf x$-plane, from $N\mathbf x = (A - \lambda I)\mathbf x = z_2\mathbf v_1$ (Topic 4 Sec. 8.9): at $(0, 1)^T$, $N(0, 1)^T = \begin{pmatrix} -4 \\ -2 \end{pmatrix} = -2\,\mathbf v_1$, so there the orbit drifts along $-\mathbf v_1$. The orbits are tangent to the eigenline at the origin and bend into an S around it.

z1 z2 z1 z-plane
Canonical $\mathbf z$-plane (unstable improper node).
Open the figure full size
x y v1 x-plane
Original $(x, y)$-plane (unstable improper node); dashed: the eigenline with $\mathbf v_1 = (2, 1)$.
Open the figure full size
Interpretation and checks
Item$\lambda$Type
(i)$3$unstable improper node
(ii)$-4$stable improper node
(iii)$-4$stable improper node
(iv)$13$unstable star
(v)$5$unstable improper node
  • (i) is exactly the worked Example 8.1 of Topic 4 (Robinson Example 30.1): $\lambda = 3$, $\mathbf v_1 = (2, 1)^T$, $\mathbf v_2 = (1, 0)^T$ and $J = \begin{pmatrix} 3 & 1 \\ 0 & 3 \end{pmatrix}$ — the result reproduces it step by step.
  • In each improper node $Q^{-1}AQ = J$ was multiplied out entry by entry; $N^2 = O$ was checked (Cayley-Hamilton).
  • The general solution $\mathbf x = e^{\lambda t}[c_1\mathbf v_1 + c_2(t\mathbf v_1 + \mathbf v_2)]$ contains the secular term $te^{\lambda t}\mathbf v_1$, which makes all orbits tangent to $\mathbf v_1$ at both ends (S-shaped).
  • (iv) is a multiple of the identity: every vector is an eigenvector and the orbits are rays; the canonical form is $A$ itself, and any basis (for instance $Q = I$) works.
  • A repeated eigenvalue is a borderline case between nodes ($\Delta > 0$) and spirals ($\Delta < 0$).
Exercise 7

The Nilpotent Case: Zero as a Repeated Eigenvalue

Original statement · probls_ch4-ii_2627.pdf, p. 2
Nilpotent Matrix • Shear • Line of Equilibria
Statement

Suppose that a non-zero matrix $A$ has zero as a repeated eigenvalue, with eigenvector $\mathbf v$. Then we can change to coordinates referred to $\mathbf v_2$ and $\mathbf v_1 = A\mathbf v_2$, where $\mathbf v_2$ is any vector in a different direction to $\mathbf v$, so that $\tilde{\mathbf x} = \tilde x\mathbf v_1 + \tilde y\mathbf v_2$. The equation $d\mathbf x/dt = A\mathbf x$ becomes $\dfrac{d\tilde{\mathbf x}}{dt} = \begin{pmatrix} 0 & 1 \\ 0 & 0 \end{pmatrix}\tilde{\mathbf x}$, hence,

$$ \frac{d\tilde x}{dt} = \tilde y \quad\text{and}\quad \frac{d\tilde y}{dt} = 0. $$

  • (i) Solve the equations for $\tilde x(t)$ and $\tilde y(t)$, and hence write down the general solution for $\mathbf x(t)$.
  • (ii) Draw the phase diagram in the $(\tilde x, \tilde y)$ and $(x, y)$ planes.
  • (iii) Draw the phase diagram for the equation $\dfrac{d\mathbf x}{dt} = \begin{pmatrix} -1 & 1 \\ -1 & 1 \end{pmatrix}\mathbf x$.
Setup and hypotheses

$A \neq O$ is $2\times 2$ with the double eigenvalue $0$: $\operatorname{tr}A = 0$, $\det A = 0$, so by Cayley-Hamilton $A^2 = O$ ($A^2 - (\operatorname{tr}A)A + (\det A)I = O$). Since $A \neq O$, $\ker A$ is one-dimensional, spanned by $\mathbf v$: $m_g = 1$ (Topic 4 Sec. 8.4 with $\lambda = 0$, $N = A$).

Why the proposed basis works: take $\mathbf v_2 \notin \operatorname{span}\{\mathbf v\}$. Then $A\mathbf v_2 \neq \mathbf 0$ (otherwise $\mathbf v_2$ would be in $\ker A$), and $\mathbf v_1 = A\mathbf v_2$ satisfies $A\mathbf v_1 = A^2\mathbf v_2 = \mathbf 0$, so $\mathbf v_1 \in \ker A = \operatorname{span}\{\mathbf v\}$ and $\mathbf v_1 \neq \mathbf 0$ is an eigenvector. $\mathbf v_1, \mathbf v_2$ are independent (not parallel), and

$$ A\mathbf v_1 = \mathbf 0 = 0\cdot\mathbf v_1 + 0\cdot\mathbf v_2, \qquad A\mathbf v_2 = \mathbf v_1 = 1\cdot\mathbf v_1 + 0\cdot\mathbf v_2 \;\Longrightarrow\; Q^{-1}AQ = \begin{pmatrix} 0 & 1 \\ 0 & 0 \end{pmatrix} $$

Formulation

With $Q = [\mathbf v_1 \mid \mathbf v_2]$ and $\mathbf x = Q\tilde{\mathbf x} = \tilde x\mathbf v_1 + \tilde y\mathbf v_2$ the canonical system is $\dot{\tilde x} = \tilde y$, $\dot{\tilde y} = 0$ (the Jordan block of Topic 4 Eq. 8.3 with $\lambda = 0$).

Method of the class notes [Notes, p. 10]

Case 3.2 of the notes with $\lambda = 0$: $\mathbf v_1 = (A - 0I)\mathbf v_2 = A\mathbf v_2$, exactly the change of coordinates of the statement. The solution of the notes becomes $z_2 = C$, $z_1 = Ct + D$ and $\mathbf x = (Ct + D)\mathbf v_1 + C\mathbf v_2$: horizontal lines in the $\tilde{\mathbf x}$-plane, with the line of $\mathbf v_1$ ($C = 0$) made of equilibria.

Derivation
(i) Solution

The second equation $\dot{\tilde y} = 0$ integrates to $\tilde y(t) = c_2$ (a constant). Substituting in the first, $\dot{\tilde x} = c_2$, and Barrow's rule between $0$ and $t$ gives $\tilde x(t) - \tilde x(0) = \int_0^t c_2\,ds = c_2t$:

$$ \tilde x(t) = c_1 + c_2t, \qquad \tilde y(t) = c_2 \qquad (c_1 = \tilde x(0),\ c_2 = \tilde y(0)) $$

Back to the original coordinates, $\mathbf x = \tilde x\mathbf v_1 + \tilde y\mathbf v_2$:

$$ \boxed{\mathbf x(t) = (c_1 + c_2t)\,\mathbf v_1 + c_2\,\mathbf v_2} $$

Check: $\dot{\mathbf x} = c_2\mathbf v_1$ and $A\mathbf x = (c_1 + c_2t)A\mathbf v_1 + c_2A\mathbf v_2 = \mathbf 0 + c_2\mathbf v_1$. They agree. Unlike the case $\lambda \neq 0$ there is no exponential: the secular term grows only linearly.

(ii) Phase diagrams

In the $(\tilde x, \tilde y)$ plane the orbits are the horizontal lines $\tilde y = c_2$ (since $\tilde y$ is constant). On the line $\tilde y = c_2 > 0$ the abscissa $\tilde x = c_1 + c_2t$ increases at constant speed $c_2$: motion to the right; for $c_2 < 0$ to the left. The $\tilde x$-axis ($c_2 = 0$) consists of equilibria. There are no curved orbits: the speed is proportional to the height $\tilde y$ (a shear flow).

x̃ ỹ x̃ canonical plane
$(\tilde x, \tilde y)$-plane: horizontal lines $\tilde y = c_2$ traversed to the right above the axis and to the left below it; the $\tilde x$-axis is all equilibria.
Open the figure full size

In the $(x, y)$ plane, $\mathbf x = \tilde x\mathbf v_1 + \tilde y\mathbf v_2$ maps horizontal lines to lines parallel to $\mathbf v_1$, offset from the line $\operatorname{span}\{\mathbf v_1\}$ of equilibria by $c_2\mathbf v_2$; they are traversed along $+\mathbf v_1$ on the side of $\operatorname{span}\{\mathbf v_1\}$ where $\mathbf v_2$ points and along $-\mathbf v_1$ on the other side. The $(x, y)$-diagram of the following item shows a concrete example.

(iii) The matrix $\begin{pmatrix} -1 & 1 \\ -1 & 1 \end{pmatrix}$

$\operatorname{tr}A = -1 + 1 = 0$ and $\det A = (-1)(1) - (1)(-1) = 0$, so the characteristic equation is $\lambda^2 = 0$: the double eigenvalue $0$, and $A \neq O$. Check $A^2 = O$:

$$ \begin{aligned} (A^2)_{11} &= (-1)\cdot(-1) + 1\cdot(-1) = 0 \\ (A^2)_{12} &= (-1)\cdot1 + 1\cdot1 = 0 \\ (A^2)_{21} &= (-1)\cdot(-1) + 1\cdot(-1) = 0 \\ (A^2)_{22} &= (-1)\cdot1 + 1\cdot1 = 0 \end{aligned} $$

Eigenvector: $A\mathbf v = \mathbf 0$ gives $-v_1 + v_2 = 0$ (both rows are the same equation up to sign), so $\mathbf v = (1, 1)^T$. Following the exercise, take any $\mathbf v_2$ not parallel to $\mathbf v$, for instance $\mathbf v_2 = (1, 0)^T$, and set $\mathbf v_1 = A\mathbf v_2$:

$$ \mathbf v_1 = A\mathbf v_2 = \begin{pmatrix} -1 & 1 \\ -1 & 1 \end{pmatrix}\begin{pmatrix} 1 \\ 0 \end{pmatrix} = \begin{pmatrix} -1 \\ -1 \end{pmatrix} $$

$\mathbf v_1 = (-1, -1)^T$ is indeed an eigenvector ($A\mathbf v_1 = (1 - 1, 1 - 1)^T = \mathbf 0$), a multiple of $\mathbf v$.

$$ Q = [\mathbf v_1 \mid \mathbf v_2] = \begin{pmatrix} -1 & 1 \\ -1 & 0 \end{pmatrix} $$

$\det Q = (-1)(0) - (1)(-1) = 1$, which is not zero, so $Q$ is invertible. For a $2\times 2$ matrix $\begin{pmatrix} a & b \\ c & d \end{pmatrix}^{-1} = \frac{1}{ad - bc}\begin{pmatrix} d & -b \\ -c & a \end{pmatrix}$ (Robinson, Appendix B):

$$ Q^{-1} = \frac{1}{1}\begin{pmatrix} 0 & -1 \\ 1 & -1 \end{pmatrix} = \begin{pmatrix} 0 & -1 \\ 1 & -1 \end{pmatrix} $$

$$ AQ = [A\mathbf v_1 \mid A\mathbf v_2] = [\mathbf 0 \mid \mathbf v_1] = \begin{pmatrix} 0 & -1 \\ 0 & -1 \end{pmatrix} $$

$$ \begin{aligned} (Q^{-1}AQ)_{11} &= 0\cdot0 + (-1)\cdot0 = 0 \\ (Q^{-1}AQ)_{12} &= 0\cdot(-1) + (-1)\cdot(-1) = 1 \\ (Q^{-1}AQ)_{21} &= 1\cdot0 + (-1)\cdot0 = 0 \\ (Q^{-1}AQ)_{22} &= 1\cdot(-1) + (-1)\cdot(-1) = 0 \end{aligned} $$

$$ Q^{-1}AQ = \begin{pmatrix} 0 & 1 \\ 0 & 0 \end{pmatrix} $$

By part (i) the general solution is $\mathbf x(t) = (c_1 + c_2t)\mathbf v_1 + c_2\mathbf v_2$:

$$ x(t) = - c_{1} - c_{2} t + c_{2}, \qquad y(t) = - c_{1} - c_{2} t $$

Check: $\dot x = - c_{2}$ and $-x + y = - c_{2}$; $\dot y = - c_{2}$ and $-x + y$ again: both equations $\dot x = -x + y$, $\dot y = -x + y$ hold.

The coordinates are $\tilde{\mathbf x} = Q^{-1}\mathbf x$: $\tilde x = -y$, $\tilde y = x - y$ (first row $(0, -1)$, second row $(1, -1)$ of $Q^{-1}$). Every orbit is a straight line parallel to $\mathbf v = (1, 1)$; the line $y = x$ ($\tilde y = 0$) consists of equilibria.

Direction: $\dot{\mathbf x} = (y - x)(1, 1)^T$. Above the line $y = x$ ($y > x$) the motion is along $+(1, 1)$ (up and to the right); below it, along $-(1, 1)$ (down and to the left). The speed $|y - x|\sqrt2$ grows with the distance to the line of equilibria.

Phase portrait

$\lambda = 0$: every point of the eigenline through $\mathbf v_1 = (-1, -1)$ is an equilibrium. In the $\mathbf z$-plane $z_2 = c_2$ is constant and $z_1 = c_1 + c_2t$ moves with constant speed $c_2$: horizontal lines, to the right if $z_2 > 0$ and to the left if $z_2 < 0$. Under $\mathbf x = Q\mathbf z$ they become lines parallel to $\mathbf v_1$: on the side of the eigenline where $\mathbf v_2$ points (for $z_2 > 0$) they are traversed along $+\mathbf v_1$, on the opposite side along $-\mathbf v_1$.

Orientation in the $\mathbf x$-plane, from $N\mathbf x = (A - \lambda I)\mathbf x = z_2\mathbf v_1$ (Topic 4 Sec. 8.9): at $(0, 1)^T$, $N(0, 1)^T = \begin{pmatrix} 1 \\ 1 \end{pmatrix} = -1\,\mathbf v_1$, so there the orbit drifts along $-\mathbf v_1$. The orbits are the straight lines parallel to $\mathbf v_1$.

z1 z2 z1 z-plane
Canonical $\mathbf z$-plane (line of equilibria (nilpotent shear)).
Open the figure full size
x y v1 x-plane
Original $(x, y)$-plane (line of equilibria (nilpotent shear)); dashed: the eigenline with $\mathbf v_1 = (-1, -1)$.
Open the figure full size
Interpretation and checks
  • This is the limit $\lambda \to 0$ of the improper node $J = \begin{pmatrix} \lambda & 1 \\ 0 & \lambda \end{pmatrix}$: the factor $e^{\lambda t}$ disappears and only the shear remains (the origin of the trace-determinant plane): the flow is $\mathbf x(t) = \mathbf x_0 + t\,A\mathbf x_0$ (the series of $e^{At}$ stops after the linear term because $A^2 = O$): in (iii), $\mathbf x(t) = \mathbf x_0 + t(y_0 - x_0)(1, 1)^T$, which matches the direction and speed found.
  • Check at a point: $\mathbf x_0 = (1, 0)$: formula gives $\mathbf x(t) = (1, 0) + t(0 - 1)(1, 1) = (1 - t, -t)$, and the general solution with $c_2 = \tilde y(0) = x_0 - y_0 = 1$, $c_1 = \tilde x(0) = -y_0 = 0$ gives $x = -(0 + t)(1) + 1 = 1 - t$ and $y = -t$. They coincide.
  • $\operatorname{tr}A = \det A = 0$ is the origin of the trace-determinant plane (Topic 4 Sec. 9.1), where all the portrait types meet.
  • The solutions are polynomials in $t$, so every orbit not on the line of equilibria leaves to infinity linearly in time in both directions.
Exercise 8

Nine Qualitative Phase Portraits from Partial Data

Original statement · probls_ch4-ii_2627.pdf, p. 2
Qualitative Portraits • Classification • Nine Cases
Statement

Draw the phase portrait for the equation $d\mathbf x/dt = A\mathbf x$ when the eigenvalues and eigenvectors of $A$ are as given:

  • (i) $\lambda_1 = 3$ with $\mathbf v_1 = (1, 1)$ and $\lambda_2 = -2$ with $\mathbf v_2 = (1, -2)$;
  • (ii) complex conjugate eigenvalues $\lambda_\pm = -1 \pm 3i$, with $\dot x < 0$ when $x = 0$ and $y > 0$;
  • (iii) a single eigenvalue $\lambda = 13$ with eigenvector $(3, 2)$, and $\dot x > 0$ when $x = 0$ and $y > 0$;
  • (iv) $\lambda_1 = -2$ with $\mathbf v_1 = (2, 1)$ and $\lambda_2 = -3$ with $\mathbf v_2 = (1, -1)$;
  • (v) a single eigenvalue $\lambda = -3$ with eigenvector $(1, -1)$, and $\dot x > 0$ when $x = 0$ and $y > 0$;
  • (vi) $\lambda_\pm = \pm 2i$, where $\dot y < 0$ when $y = 0$ and $x > 0$;
  • (vii) $\lambda_1 = 1$ with $\mathbf v_1 = (3, 2)$ and $\lambda_2 = 5$ with $\mathbf v_2 = (1, -4)$;
  • (viii) $\lambda_\pm = 5 \pm i$, and $\dot y > 0$ when $y = 0$ and $x > 0$;
  • (ix) a single eigenvalue $\lambda = -7$, with the matrix $A$ being a multiple of the identity.
Setup: what the data determine

The data fix the type of the origin and the position of the eigenlines, which is all that the phase portrait of a linear system needs (Topic 4 Sec. 9, Robinson Ch. 31). A matrix consistent with the data is built in each item, to draw the portrait and to check the sign information numerically; any other matrix with the same data has the same portrait up to a linear distortion of the plane.

DataOrigin is aExtra information used
two real eigenvalues, opposite signssaddleeigenlines; stable = negative eigenvalue
two real eigenvalues, same signnodetangent to the eigenvector of the smaller $\vert \lambda\vert $; stable iff $\lambda < 0$
complex $\rho \pm i\omega$, $\rho \neq 0$spiralstable iff $\rho < 0$; sense from the sign of $\dot x$ at $(0, 1)$ or $\dot y$ at $(1, 0)$
imaginary $\pm i\omega$centreclosed orbits; sense from the sign condition
single eigenvalue, one eigenvectorimproper nodeS-shaped orbits along the eigenline; side/orientation from the sign condition
single eigenvalue, $A = \lambda I$starstraight rays
Formulation: how the sign conditions are read

At the point $(0, 1)$ the vector field is $A(0, 1)^T = (a_{12}, a_{22})^T$, so the sign of $\dot x$ there is the sign of $a_{12}$: positive means motion to the right at the top of the plane, which is clockwise; negative means counter-clockwise. At $(1, 0)$ the field is $(a_{11}, a_{21})^T$, so the sign of $\dot y$ there is the sign of $a_{21}$: negative means downwards on the right-hand side, i.e. clockwise; positive means counter-clockwise (Topic 4 Sec. 7.8).

Method of the class notes [Notes, pp. 9–10]

The classification of the notes decides each item. Two real eigenvalues (Case 1) give a node or a saddle, read from the sign table of $\lambda_2/\lambda_1$. A complex pair (Case 2) gives a spiral or a centre from the sign of $\rho$, with the sense of rotation read from the sign of $\dot x$ or $\dot y$ on an axis, as in the example of the notes. A repeated eigenvalue (Case 3) gives a star when $A = \lambda I$ and an improper node, with one eigenvector, otherwise.

Derivation of the nine portraits
(i) $\lambda_1 = 3$ with $\mathbf v_1 = (1, 1)$ and $\lambda_2 = -2$ with $\mathbf v_2 = (1, -2)$

Two real eigenvalues with independent eigenvectors ($\det[\mathbf v_1\ \mathbf v_2] = -3 \neq 0$). A consistent matrix is $A = Q\Lambda Q^{-1} = \begin{pmatrix} \frac{4}{3} & \frac{5}{3} \\ \frac{10}{3} & - \frac{1}{3} \end{pmatrix}$ (checked: $A\mathbf v_1 = \begin{pmatrix} 3 \\ 3 \end{pmatrix} = 3\mathbf v_1$, $A\mathbf v_2 = \begin{pmatrix} -2 \\ 4 \end{pmatrix} = -2\mathbf v_2$).

Phase portrait

$\lambda_1 = 3$ and $\lambda_2 = -2$ are of opposite signs: the origin is a saddle (Robinson Ch. 28, Topic 4 Sec. 6 and 9).

In the canonical $\mathbf z$-plane, eliminating $t$ from $z_1 = c_1e^{3t}$ and $z_2 = c_2e^{-2t}$ (for $c_1 \neq 0$): $|z_1|/|c_1| = e^{\lambda_1t}$, so $e^{t} = (|z_1|/|c_1|)^{1/\lambda_1}$ and $|z_2| = |c_2|e^{\lambda_2t} = |c_2|\,(|z_1|/|c_1|)^{\lambda_2/\lambda_1}$, i.e. $z_2 = K\,|z_1|^{\lambda_2/\lambda_1} = K\,|z_1|^{- \frac{2}{3}}$ with $K = c_2|c_1|^{-\lambda_2/\lambda_1}$ (equivalently, integrate the separable equation $\frac{dz_2}{dz_1} = \frac{\lambda_2z_2}{\lambda_1z_1}$ by the chain rule, as in Exercise 10). The axes $z_1 = 0$ and $z_2 = 0$ are invariant: a solution that starts on an axis stays on it.

The exponent $\lambda_2/\lambda_1 = - \frac{2}{3}$ is negative, so the orbits are hyperbola-like. Only solutions starting on the eigenline of the negative eigenvalue, the line through $(1, -2)$ (the stable manifold), reach the origin as $t \to +\infty$; the line through $(1, 1)$ (positive eigenvalue) is the unstable manifold. All other orbits come in close to the stable line, then leave along the unstable one.

The map $\mathbf x = Q\mathbf z$ sends the $z_1$-axis to the line through $\mathbf v_1 = (1, 1)$ and the $z_2$-axis to the line through $\mathbf v_2 = (1, -2)$. Straight lines stay straight and the origin stays fixed, so the $\mathbf x$-portrait is the canonical one sheared onto the eigenlines. On $\mathbf v_1$: $\dot{\mathbf x} = A\mathbf v_1 = 3\mathbf v_1$ points outward; on $\mathbf v_2$: $\dot{\mathbf x} = -2\mathbf v_2$ points inward.

z1 z2 z1 z2 z-plane
Canonical $\mathbf z$-plane (saddle): invariant axes and orbits $z_2 = K|z_1|^{- \frac{2}{3}}$.
Open the figure full size
x y v1 v2 x-plane
Original $(x, y)$-plane (saddle): dashed lines are the eigenlines $\mathbf v_1 = (1, 1)$ and $\mathbf v_2 = (1, -2)$.
Open the figure full size
(ii) complex conjugate eigenvalues $\lambda_\pm = -1 \pm 3i$, with $\dot x < 0$ when $x = 0$ and $y > 0$

$\dot x < 0$ at $(0, 1)$ gives $a_{12} < 0$: leftwards at the top of the plane, i.e. counter-clockwise. A consistent matrix with eigenvalues $-1 \pm 3i$ is $A = \begin{pmatrix} -1 & -3 \\ 3 & -1 \end{pmatrix}$ ($\operatorname{tr}A = -2$, $\det A = 10$, $\lambda = -1 \pm 3i$; $a_{12} = -3 < 0$).

Phase portrait

$\rho = -1$ and $\omega = 3$: the origin is a stable spiral. In the $\mathbf z$-plane $r = r_0e^{-1t}$ and $\theta = \theta_0 - 3t$: the motion is always clockwise ($\dot\theta = -\omega < 0$), and one full turn takes $T = 2\pi/\omega$ while the radius is multiplied by $e^{2\pi\rho/\omega} = e^{-\frac{2 \pi}{3}}$ (shrinking).

Sense of rotation in the $\mathbf x$-plane, tested directly on the vector field (Topic 4 Sec. 7.8): at $(0, 1)^T$, $\dot{\mathbf x} = A(0, 1)^T = \begin{pmatrix} -3 \\ -1 \end{pmatrix}$, so $\dot x = -3$ is $< 0$: to the left at the top of the plane: the rotation is counter-clockwise. Cross-check: $\det Q = -9$ is $< 0$, so $Q$ reverses orientation and the clockwise motion of the $\mathbf z$-plane becomes counter-clockwise in the $\mathbf x$-plane, in agreement.

z1 z2 z-plane
Canonical $\mathbf z$-plane (stable spiral): clockwise, $\dot\theta = -\omega$.
Open the figure full size
x y x-plane
Original $(x, y)$-plane (stable spiral), counter-clockwise.
Open the figure full size
(iii) a single eigenvalue $\lambda = 13$ with eigenvector $(3, 2)$, and $\dot x > 0$ when $x = 0$ and $y > 0$

Single eigenvalue $13$ with one eigenvector $(3, 2)^T$: $A \neq 13I$ (a star would have two independent eigenvectors), so $N = A - 13I$ is non-zero, $N^2 = O$ and $\ker N = \operatorname{span}\{(3, 2)^T\}$. Hence $N = c\begin{pmatrix} 3 \\ 2 \end{pmatrix}(2, -3) = c\begin{pmatrix} 6 & -9 \\ 4 & -6 \end{pmatrix}$ (rows orthogonal to $(3, 2)$ so that $N(3, 2)^T = \mathbf 0$), and $a_{12} = N_{12} = -9c$. The condition $\dot x > 0$ at $(0, 1)$ means $a_{12} > 0$, so $c < 0$; take $c = -1$: $N = \begin{pmatrix} -6 & 9 \\ -4 & 6 \end{pmatrix}$, $A = 13I + N = \begin{pmatrix} 7 & 9 \\ -4 & 19 \end{pmatrix}$.

Phase portrait

$\lambda = 13$ is repeated with a single eigendirection and $A \neq \lambda I$: the origin is an unstable improper node (Topic 4 Sec. 8.8 and 8.9). In the $\mathbf z$-plane the $z_1$-axis is invariant ($z_2 \equiv 0$, $z_1 = c_1e^{13t}$, arrows away from the origin); the other orbits satisfy $z_1 = z_2\left[K + \frac{1}{\lambda}\ln|z_2|\right]$, are S-shaped, tangent to the $z_1$-axis at the origin and parallel to it far away. Since $\dot z_1 = \lambda z_1 + z_2 = z_2$ on the $z_2$-axis, orbits cross it moving to the right in the upper half-plane and to the left in the lower one.

Orientation in the $\mathbf x$-plane, from $N\mathbf x = (A - \lambda I)\mathbf x = z_2\mathbf v_1$ (Topic 4 Sec. 8.9): at $(0, 1)^T$, $N(0, 1)^T = \begin{pmatrix} 9 \\ 6 \end{pmatrix} = 3\,\mathbf v_1$, so there the orbit drifts along $+\mathbf v_1$. The orbits are tangent to the eigenline at the origin and bend into an S around it.

z1 z2 z1 z-plane
Canonical $\mathbf z$-plane (unstable improper node).
Open the figure full size
x y v1 x-plane
Original $(x, y)$-plane (unstable improper node); dashed: the eigenline with $\mathbf v_1 = (3, 2)$.
Open the figure full size
(iv) $\lambda_1 = -2$ with $\mathbf v_1 = (2, 1)$ and $\lambda_2 = -3$ with $\mathbf v_2 = (1, -1)$

Two real eigenvalues with independent eigenvectors ($\det[\mathbf v_1\ \mathbf v_2] = -3 \neq 0$). A consistent matrix is $A = Q\Lambda Q^{-1} = \begin{pmatrix} - \frac{7}{3} & \frac{2}{3} \\ \frac{1}{3} & - \frac{8}{3} \end{pmatrix}$ (checked: $A\mathbf v_1 = \begin{pmatrix} -4 \\ -2 \end{pmatrix} = -2\mathbf v_1$, $A\mathbf v_2 = \begin{pmatrix} -3 \\ 3 \end{pmatrix} = -3\mathbf v_2$).

Phase portrait

$\lambda_1 = -2$ and $\lambda_2 = -3$ are both negative: the origin is a stable node (Robinson Ch. 28, Topic 4 Sec. 6 and 9).

In the canonical $\mathbf z$-plane, eliminating $t$ from $z_1 = c_1e^{-2t}$ and $z_2 = c_2e^{-3t}$ (for $c_1 \neq 0$): $|z_1|/|c_1| = e^{\lambda_1t}$, so $e^{t} = (|z_1|/|c_1|)^{1/\lambda_1}$ and $|z_2| = |c_2|e^{\lambda_2t} = |c_2|\,(|z_1|/|c_1|)^{\lambda_2/\lambda_1}$, i.e. $z_2 = K\,|z_1|^{\lambda_2/\lambda_1} = K\,|z_1|^{\frac{3}{2}}$ with $K = c_2|c_1|^{-\lambda_2/\lambda_1}$ (equivalently, integrate the separable equation $\frac{dz_2}{dz_1} = \frac{\lambda_2z_2}{\lambda_1z_1}$ by the chain rule, as in Exercise 10). The axes $z_1 = 0$ and $z_2 = 0$ are invariant: a solution that starts on an axis stays on it.

The exponent $\lambda_2/\lambda_1 = \frac{3}{2}$ is positive, so the orbits are parabola-like curves through the origin. At the origin they are tangent to the eigenvector of the eigenvalue of smaller modulus, $|\lambda| = 2$, i.e. to the line through $(2, 1)$ (they arrive there as $t \to +\infty$), and far from the origin they become parallel to the other eigenvector (the fast direction). The arrows point towards the origin on both eigenlines.

The map $\mathbf x = Q\mathbf z$ sends the $z_1$-axis to the line through $\mathbf v_1 = (2, 1)$ and the $z_2$-axis to the line through $\mathbf v_2 = (1, -1)$. Straight lines stay straight and the origin stays fixed, so the $\mathbf x$-portrait is the canonical one sheared onto the eigenlines. On $\mathbf v_1$: $\dot{\mathbf x} = A\mathbf v_1 = -2\mathbf v_1$ points inward; on $\mathbf v_2$: $\dot{\mathbf x} = -3\mathbf v_2$ points inward.

z1 z2 z1 z2 z-plane
Canonical $\mathbf z$-plane (stable node): invariant axes and orbits $z_2 = K|z_1|^{\frac{3}{2}}$.
Open the figure full size
x y v1 v2 x-plane
Original $(x, y)$-plane (stable node): dashed lines are the eigenlines $\mathbf v_1 = (2, 1)$ and $\mathbf v_2 = (1, -1)$.
Open the figure full size
(v) a single eigenvalue $\lambda = -3$ with eigenvector $(1, -1)$, and $\dot x > 0$ when $x = 0$ and $y > 0$

Single eigenvalue $-3$ with one eigenvector $(1, -1)^T$: $A \neq -3I$ and $N = A + 3I \neq O$ with $\ker N = \operatorname{span}\{(1, -1)^T\}$. Hence $N = c\begin{pmatrix} 1 & 1 \\ -1 & -1 \end{pmatrix}$ and $a_{12} = N_{12} = c$. The condition $\dot x > 0$ at $(0, 1)$ means $a_{12} > 0$, so $c > 0$; take $c = 1$: $N = \begin{pmatrix} 1 & 1 \\ -1 & -1 \end{pmatrix}$, $A = -3I + N = \begin{pmatrix} -2 & 1 \\ -1 & -4 \end{pmatrix}$.

Phase portrait

$\lambda = -3$ is repeated with a single eigendirection and $A \neq \lambda I$: the origin is an stable improper node (Topic 4 Sec. 8.8 and 8.9). In the $\mathbf z$-plane the $z_1$-axis is invariant ($z_2 \equiv 0$, $z_1 = c_1e^{-3t}$, arrows towards the origin); the other orbits satisfy $z_1 = z_2\left[K + \frac{1}{\lambda}\ln|z_2|\right]$, are S-shaped, tangent to the $z_1$-axis at the origin and parallel to it far away. Since $\dot z_1 = \lambda z_1 + z_2 = z_2$ on the $z_2$-axis, orbits cross it moving to the right in the upper half-plane and to the left in the lower one.

Orientation in the $\mathbf x$-plane, from $N\mathbf x = (A - \lambda I)\mathbf x = z_2\mathbf v_1$ (Topic 4 Sec. 8.9): at $(0, 1)^T$, $N(0, 1)^T = \begin{pmatrix} 1 \\ -1 \end{pmatrix} = 1\,\mathbf v_1$, so there the orbit drifts along $+\mathbf v_1$. The orbits are tangent to the eigenline at the origin and bend into an S around it.

z1 z2 z1 z-plane
Canonical $\mathbf z$-plane (stable improper node).
Open the figure full size
x y v1 x-plane
Original $(x, y)$-plane (stable improper node); dashed: the eigenline with $\mathbf v_1 = (1, -1)$.
Open the figure full size
(vi) $\lambda_\pm = \pm 2i$, where $\dot y < 0$ when $y = 0$ and $x > 0$

$\dot y = a_{21}x + a_{22}y$ at $(1, 0)$ equals $a_{21}$, and $\dot y < 0$ there, so $a_{21} < 0$: at the right-hand side of the plane the motion is downwards, i.e. clockwise. A consistent matrix with eigenvalues $\pm 2i$ and $a_{21} < 0$ is $A = \begin{pmatrix} 0 & 2 \\ -2 & 0 \end{pmatrix}$ ($\operatorname{tr}A = 0$, $\det A = 4$, eigenvalues $\pm 2i$; $a_{21} = -2 < 0$).

Phase portrait

$\rho = 0$ and $\omega = 2$: the origin is a centre. In the $\mathbf z$-plane $r = r_0e^{0t}$ and $\theta = \theta_0 - 2t$: the motion is always clockwise ($\dot\theta = -\omega < 0$), and the orbits are circles of period $T = 2\pi/\omega$ (closed orbits).

Sense of rotation in the $\mathbf x$-plane, tested directly on the vector field (Topic 4 Sec. 7.8): at $(0, 1)^T$, $\dot{\mathbf x} = A(0, 1)^T = \begin{pmatrix} 2 \\ 0 \end{pmatrix}$, so $\dot x = 2$ is $> 0$: to the right at the top of the plane: the rotation is clockwise. Cross-check: $\det Q = 4$ is $> 0$, so $Q$ preserves orientation and the clockwise motion of the $\mathbf z$-plane becomes clockwise in the $\mathbf x$-plane, in agreement.

z1 z2 z-plane
Canonical $\mathbf z$-plane (centre): clockwise, $\dot\theta = -\omega$.
Open the figure full size
x y x-plane
Original $(x, y)$-plane (centre), clockwise.
Open the figure full size
(vii) $\lambda_1 = 1$ with $\mathbf v_1 = (3, 2)$ and $\lambda_2 = 5$ with $\mathbf v_2 = (1, -4)$

Two real eigenvalues with independent eigenvectors ($\det[\mathbf v_1\ \mathbf v_2] = -14 \neq 0$). A consistent matrix is $A = Q\Lambda Q^{-1} = \begin{pmatrix} \frac{11}{7} & - \frac{6}{7} \\ - \frac{16}{7} & \frac{31}{7} \end{pmatrix}$ (checked: $A\mathbf v_1 = \begin{pmatrix} 3 \\ 2 \end{pmatrix} = 1\mathbf v_1$, $A\mathbf v_2 = \begin{pmatrix} 5 \\ -20 \end{pmatrix} = 5\mathbf v_2$).

Phase portrait

$\lambda_1 = 1$ and $\lambda_2 = 5$ are both positive: the origin is a unstable node (Robinson Ch. 28, Topic 4 Sec. 6 and 9).

In the canonical $\mathbf z$-plane, eliminating $t$ from $z_1 = c_1e^{1t}$ and $z_2 = c_2e^{5t}$ (for $c_1 \neq 0$): $|z_1|/|c_1| = e^{\lambda_1t}$, so $e^{t} = (|z_1|/|c_1|)^{1/\lambda_1}$ and $|z_2| = |c_2|e^{\lambda_2t} = |c_2|\,(|z_1|/|c_1|)^{\lambda_2/\lambda_1}$, i.e. $z_2 = K\,|z_1|^{\lambda_2/\lambda_1} = K\,|z_1|^{5}$ with $K = c_2|c_1|^{-\lambda_2/\lambda_1}$ (equivalently, integrate the separable equation $\frac{dz_2}{dz_1} = \frac{\lambda_2z_2}{\lambda_1z_1}$ by the chain rule, as in Exercise 10). The axes $z_1 = 0$ and $z_2 = 0$ are invariant: a solution that starts on an axis stays on it.

The exponent $\lambda_2/\lambda_1 = 5$ is positive, so the orbits are parabola-like curves through the origin. At the origin they are tangent to the eigenvector of the eigenvalue of smaller modulus, $|\lambda| = 1$, i.e. to the line through $(3, 2)$ (they leave from there (the limit $t \to -\infty$)), and far from the origin they become parallel to the other eigenvector (the fast direction). The arrows point away from the origin on both eigenlines.

The map $\mathbf x = Q\mathbf z$ sends the $z_1$-axis to the line through $\mathbf v_1 = (3, 2)$ and the $z_2$-axis to the line through $\mathbf v_2 = (1, -4)$. Straight lines stay straight and the origin stays fixed, so the $\mathbf x$-portrait is the canonical one sheared onto the eigenlines. On $\mathbf v_1$: $\dot{\mathbf x} = A\mathbf v_1 = 1\mathbf v_1$ points outward; on $\mathbf v_2$: $\dot{\mathbf x} = 5\mathbf v_2$ points outward.

z1 z2 z1 z2 z-plane
Canonical $\mathbf z$-plane (unstable node): invariant axes and orbits $z_2 = K|z_1|^{5}$.
Open the figure full size
x y v1 v2 x-plane
Original $(x, y)$-plane (unstable node): dashed lines are the eigenlines $\mathbf v_1 = (3, 2)$ and $\mathbf v_2 = (1, -4)$.
Open the figure full size
(viii) $\lambda_\pm = 5 \pm i$, and $\dot y > 0$ when $y = 0$ and $x > 0$

$\dot y > 0$ at $(1, 0)$ gives $a_{21} > 0$: upwards on the right-hand side, i.e. counter-clockwise. A consistent matrix with eigenvalues $5 \pm i$ is $A = \begin{pmatrix} 5 & -1 \\ 1 & 5 \end{pmatrix}$ ($\operatorname{tr}A = 10$, $\det A = 26$, $\lambda = 5 \pm i$; $a_{21} = 1 > 0$).

Phase portrait

$\rho = 5$ and $\omega = 1$: the origin is a unstable spiral. In the $\mathbf z$-plane $r = r_0e^{5t}$ and $\theta = \theta_0 - 1t$: the motion is always clockwise ($\dot\theta = -\omega < 0$), and one full turn takes $T = 2\pi/\omega$ while the radius is multiplied by $e^{2\pi\rho/\omega} = e^{10 \pi}$ (growing).

Sense of rotation in the $\mathbf x$-plane, tested directly on the vector field (Topic 4 Sec. 7.8): at $(0, 1)^T$, $\dot{\mathbf x} = A(0, 1)^T = \begin{pmatrix} -1 \\ 5 \end{pmatrix}$, so $\dot x = -1$ is $< 0$: to the left at the top of the plane: the rotation is counter-clockwise. Cross-check: $\det Q = -1$ is $< 0$, so $Q$ reverses orientation and the clockwise motion of the $\mathbf z$-plane becomes counter-clockwise in the $\mathbf x$-plane, in agreement.

z1 z2 z-plane
Canonical $\mathbf z$-plane (unstable spiral): clockwise, $\dot\theta = -\omega$.
Open the figure full size
x y x-plane
Original $(x, y)$-plane (unstable spiral), counter-clockwise.
Open the figure full size
(ix) a single eigenvalue $\lambda = -7$, with the matrix $A$ being a multiple of the identity

$A$ is a multiple of the identity with eigenvalue $-7$: $A = \begin{pmatrix} -7 & 0 \\ 0 & -7 \end{pmatrix}$.

Phase portrait

$A = -7I$: every line through the origin is invariant and every solution moves along its own ray, $\mathbf x(t) = e^{-7t}\mathbf x_0$, towards the origin (the origin is a stable star). The picture is the same in every basis, so there is no separate canonical plane.

x y x-plane
Stable star: straight rays into the origin.
Open the figure full size
Interpretation and checks
ItemOriginStability
(i)saddleunstable
(ii)spiral, counter-clockwisestable
(iii)improper nodeunstable
(iv)node, tangent to $\mathbf v_1 = (2, 1)$stable
(v)improper nodestable
(vi)centre, clockwisestable, not asymptotically
(vii)node, tangent to $\mathbf v_1 = (3, 2)$unstable
(viii)spiral, counter-clockwiseunstable
(ix)starasymptotically stable
  • The nine items cover the nine pictures of Robinson's classification (Ch. 31, Fig. 31.1, as reproduced in Topic 4 Fig. 4.11; the unstable star is the time reversal of the stable one): saddle (i); stable node (iv); unstable node (vii); stable spiral (ii); unstable spiral (viii); centre (vi); stable star (ix); stable improper node (v); unstable improper node (iii).
  • The tangency of nodes was verified on the numerical orbits: the orbits of (iv) are tangent at the origin to $\mathbf v_1 = (2, 1)$ ($|\lambda| = 2 < 3$) and those of (vii) to $\mathbf v_1 = (3, 2)$ ($1 < 5$).
  • In (iii) and (v) the sign condition only chooses the orientation of the S-shaped orbits (on which side of the eigenline the orbit moves along $+\mathbf v_1$); the eigenline and the type do not depend on it.
  • Trace-determinant check: (i) $\det A = -6 < 0$ saddle; (iv) $\operatorname{tr} = -5$, $\det = 6$, $\Delta = 1 > 0$ stable node; (vii) $\operatorname{tr} = 6$, $\det = 5$ unstable node; (iii) $\operatorname{tr}^2 = 4\det = 676$; (ix) $\operatorname{tr}^2 = 4\det = 196$.
  • Spiral (viii) has $\rho/\omega = 5$: one turn multiplies the radius by $e^{2\pi\rho/\omega} = e^{10\pi} \approx 4\times 10^{13}$, so it is a genuine spiral that, on the scale of the figure, looks like almost radial curves (the winding is invisible after the first fraction of a turn); in (ii), $\rho/\omega = -\tfrac13$ the winding is clearly visible.