ELEC505/ELEC405 Handout #5

Linear System Theory Thursday, December 10, 2020

Homework Set #4
Due: Wednesday, December 23, 2020.

1. Complex vs. Real

Let
• V1 = C n , i.e., the n-dimensional complex vector space, equipped with the inner
product

< x, y >C = y∗ x, (1)

• V2 = <2n , i.e., the 2n-dimensional real vector space, equipped with the inner
product

< u, v >R = vT u, (2)

• Γ : C n → <2n is a mapping between the two vector spaces, where

" #
<e{x}
Γ(x) = , (3)
Im{x}

• A ∈ C n×n define a linear mapping f : C n → C n , such that for any x ∈ C n

f (x) = Ax, (4)

• g : <2n → <2n is the mapping corresponding to f such that, for any x ∈ V1

g(Γ(x)) = Γ(f (x)). (5)

If we explicitly write the linear mapping g as

g(u) = Au, ∀u ∈ <2n , (6)

(a) Write A in terms of A.

(b) What type of matrix is A, if A is Hermitian?
(c) What type of matrix is A, if A is Unitary?
(d) What type of matrix is A, if A is Skew-Hermitian?
(e) For x, y ∈ C n , write < Γ(x), Γ(y) >R , in terms of < ., . >C .
2. Cayley Transformation
Prove that if A is skew hermitian then

U = (I − A)(I + A)−1 (7)

is unitary.

1
3. Yet another way to obtain a unitary matrix
Prove that if PV is an orthogonal projection matrix, then U = I − 2PV is unitary.

4. Rigid Transformations
Let R(θ) represent a 2 × 2 rotation matrix
" #
cos(θ) − sin(θ)
R(θ) = , (8)
sin(θ) cos(θ)

where given a position vector x ∈ <2×1 , R(θ)x corresponds to its θ degree counter-
clockwise rotated version.

(a) Find the eigenvalue decomposition for R(θ).

(b) Given θ1 , θ2 ∈ <, can we claim that R(θ1 )R(θ2 ) = R(θ2 )R(θ1 )?
(c) A rigid transformation is defined in terms of a rotation R(θ) and a translation
t ∈ <2 as

y = R(θ)x + t. (9)

Note through a vector space extension trick, we can write

" # " #" #
y R(θ) t x
= , (10)
1 0 1 1
| {z }
T(θ,t)

so that the rigid transformation is defined as a matrix multiplication for the

extended vectors.
Find the eigenvalue decomposition for T(θ, t).
(d) Can we claim that T(θ1 , t1 )T(θ2 , t2 ) = T(θ2 , t2 )T(θ1 , t1 ) in general? How about
for T(θ, t1 )T(θ, t2 ) = T(θ, t2 )T(θ, t1 )? How about T(0, t1 )T(0, t2 ) = T(0, t2 )T(0, t1 )?
Can you explain your answers based on what is expected from a rigid motion?

5. Missed Programming huh?... This is simple, don’t worry!

Generate a 10 × 10 random complex unitary matrix in Matlab/Python/Julia. Only
built-in functions you are allowed to use are eig and randn. Everytime your code is
executed, it is expected to generate a different random unitary matrix. No online code
submission is needed: just write down your code on your solution sheet.