CDS 140a Winter 2014 Homework 2
R. Murray  Issued: 14 Jan 2014 (Tue) 
ACM 101b/AM 125b/CDS 140a, Winter 2014  Due: 22 Jan 2014 (Wed) @ noon Turn in to box outside Steele House 
Note: In the upper left hand corner of the second page of your homework set, please put the number of hours that you spent on
this homework set (including reading).
 Perko, Section 1.8, problem 10
Suppose that the elementary blocks <amsmath>B</amsmath> in the Jordan form of the matrix <amsmath>A</amsmath>, have no ones or <amsmath>I_2</amsmath> blocks off the diagonal, so that they are of the form
<amsmath> B = \begin{bmatrix}
\lambda & 0 & 0 & \dots & 0 \\ 0 & \lambda & 0 & \dots & 0 \\ \dots & & & & & \\ 0 & \dots & & \lambda & 0 \\ 0 & \dots & & 0 & \lambda
\end{bmatrix} \qquad\text{or}\qquad B = \begin{bmatrix}
D & 0 & 0 & \dots & 0 \\ 0 & D & 0 & \dots & 0 \\ \dots & & & & & \\ 0 & \dots & & D & 0 \\ 0 & \dots & & 0 & D
\end{bmatrix}, \quad D = \begin{bmatrix} a & b \\ b & a \end{bmatrix}.
</amsmath>(a) Show that if all of the eigenvalues of <amsmath>A</amsmath> have nonpositive real parts, then for each <amsmath>x_0 \in {\mathbb R}^n</amsmath> there is a positive constant <amsmath>M</amsmath> such that <amsmath>x(t) \leq M</amsmath> for all <amsmath>t \geq 0</amsmath> where <amsmath>x(t)</amsmath> is the solution of the initial value problem.
(b) Show via a simple counterexample that this is not true if the Jordan blocks have nonzero off diagonal entries (with the same constraint on the eigenvalues).
 Perko, Section 1.9, problem 3 (modified): Consider the linear system
<amsmath> \dot x = \begin{bmatrix} a & 0 & 0 & 0\\ a & 0 & b & 0 \\ a & 0 & b & 0 \\ a & a & 0 & b \end{bmatrix} x,
</amsmath>where <amsmath>a,\,b > 0</amsmath>.
(a) Compute the solutions to the differential equation. You should provide the matrices used to transform the system to Jordan form along with the appropriate matrix exponential of the relevant Jordan form matrix (you don't need to multiply everything out to get the solution in the original basis).
 Note: you should show the various (regular and generalized) eigenvectors associated with each eigenvalue. OK to check your answer with MATLAB, but be sure to show that you know how to solve it by hand.
(b) Find the stable, unstable and center subspaces for this system (in the original coordinates).
 Perko, Section 1.9, problem 5, parts (c), (d2): Let <amsmath>A</amsmath> be an <amsmath>n \times n</amsmath> nonsingular matrix and let <amsmath>x(t)</amsmath> be the solution of the initial value problem (1) with <amsmath>x(0) = x_0</amsmath>. Show that
(c) If <amsmath>x_0 \in E^c</amsmath>, <amsmath>x_0 \neq 0</amsmath> and <amsmath>A</amsmath> is semisimple, then there are postive constants <amsmath>m</amsmath> and <amsmath>M</amsmath> such that for all <amsmath>t \in R</amsmath>, <amsmath>m \leq x(t) \leq M</amsmath>; Note: in the book, Perko defines <amsmath>\sim</amsmath> to mean "set subtraction". So <amsmath>E \sim \{0\}</amsmath> in the book is the set <amsmath>E</amsmath> minus the point 0.

Consider the system
<amsmath> \frac{dx}{dt} = \begin{bmatrix} 0 & 1 \\ 1 & 0 \end{bmatrix} x + \begin{bmatrix} 0 \\ 1 \end{bmatrix}u, \qquad y = \begin{bmatrix} 1 & 0 \end{bmatrix} x.
</amsmath>(a) Show that the unforced system (<amsmath>u = 0</amsmath>) is stable but not asymptotically stable.
(b) Given <amsmath>x(0) = x_0</amsmath> and <amsmath>u(t) = \cos(\omega*t)</amsmath>, solve for the output <amsmath>y(t)</amsmath>. Show that when <amsmath>\omega=1</amsmath> the output is unbounded.
Notes:
 The problems are transcribed above in case you don't have access to Perko. However, in the case of discrepancy, you should use Perko as the definitive source of the problem statement.
 There are a number of problems that can be solved using MATLAB. If you just give the answer with no explanation (or say "via MATLAB"), the TAs will take off points. Instead, you should show how the solutions can be worked out by hand, along the lines of what is done in the textbook. It is fine to check everything with MATLAB.