Engineering Math II Midterm Exam 1

2023/04/13, 19:00–21:30

1. Let V be an inner product space, W be its finite-dimensional subspace, and W⊥ be the

orthogonal complement of W.

a) (10) Show that for any v ∈ V, there exist unique vectors w ∈ W and z ∈ W⊥
such that v = w + z. Furthermore, explain that w is the best approximation of
v on W.
(sol)
(existence) Since W is a finite dimension, let {v1 , v2 , ..., vk } be an orthonormal
Pk
basis for W. Let w = i=1 ⟨v, vi ⟩ vi , then w ∈ W.
For any jD(1 ≤ j ≤ k), E DP E
Pk k
⟨z, vj ⟩ = (v − i=1 ⟨v, vi ⟩ vi ), vi = ⟨v, vj ⟩ − i=1 ⟨v, vi ⟩ vi , vj
Pk
= ⟨v, vj ⟩ − i=1 ⟨v, vi ⟩ ⟨vi , vj ⟩ = ⟨v, vj ⟩ − ⟨v, vj ⟩ = 0
Therefore, z ∈ W⊥ (4 point)
(uniqueness) Suppose v = w + z = w′ + z ′ , where w, w′ ∈ W and z, z ′ ∈ W⊥ .
Then w − w′ = z ′ − z ∈ (W ∩ W⊥ ) = {0}.
Therefore, w = w′ , z = z ′ .(3 point)
(best approximation) For any x ∈ W,
||v − x||2 = ||w + z − x||2 = ||(w − x) + z||2 = ||w − x||2 + ||z||2 ≥ ||z||2 = ||v − w||2
This inequility is an equality, when ||w − x||2 + ||z||2 = ||z||2 .
It follows that ||w − x||2 = 0, and hence w = x.(3 point)
 ⊥
b) (5) Show that W⊥ = W. (Caution: It’s not guaranteed that V is finite-
dimensional.)
(sol)
If w ∈ W, then w is orthogonal to all elements of W⊥ . Therefore, w ∈ (W⊥ )⊥ .
∴ W ⊂ (W⊥ )⊥ (3 point)
/ W, there exist z ∈ W⊥ , which is ⟨w, z⟩ =
If w ∈ / (W⊥ )⊥ .
̸ 0. Therefore w ∈
∴ (W⊥ )⊥ ⊂ W(2 point)

2. Let T be a normal operator on a finite-dimensional inner product space V.

a) (5) Let λ be an eigenvalue of T, and v be an eigenvector corresponding to λ.

Show that v is also an eigenvector of T∗ corresponding to λ̄, where T∗ is the
adjoint operator of T.
(sol)
Let U = T − λI, then U∗ = T∗ − λ̄I.
UU∗ (x) = (T − λI)(T∗ − λ̄I)(x) = (TT∗ − λT∗ − λ̄T + λλ̄I)(x)
= (T∗ T − λT∗ − λ̄T + λλ̄I)(x) = (T∗ − λ̄I)(T − λI)(x) = U∗ U(x)
Therefore, U is a normal operator.(3 point)
0 = ||U(x)|| = ||U∗ (x)|| = ||(T∗ − λ̄I)(x)|| = ||T∗ (x) − λ̄x||.
∴ T∗ (x) = λ̄x and x is an eigenvector of T∗ .(2 point)
 b) (5) Show that two eigenvectors of T corresponding to the distinct eigenvalues are
orthogonal.
(sol)
Let two eigenvalues (λ1 , λ2 ) have corresponding eigenvectors (x1 , x2 ), respec-
tively.
λ1 ⟨x1 , x2 ⟩ = ⟨λ1 x1 , x2 ⟩ = ⟨T(x1 ), x2 ⟩ = ⟨x1 , T∗ (x2 )⟩ = x1 , λ̄2 x2 = λ2 ⟨x1 , x2 ⟩
Since λ1 ̸= λ2 , ⟨x1 , x2 ⟩ = 0.(5 point)

1
3. Consider z = (1 2 2)t ∈ R3 .
3
a) (5) Find the Householder matrix Hz , the matrix representing the symmetry send-
ing z to −z, and fixing all the vectors orthogonal to z.
(sol)  
7 −4 −4
 9 9 9 
Hz x = x − 2 ⟨x, z⟩ z =  −4 1 −8  x(5 point)
 9 9 9 
−4 −8 1
9 9 9

b) (5) Define T(v) = v × z. Find the matrix representation of T with respect to the
standard ordred basis S.
(sol)  
2 −2
î ĵ k̂ 0 3 3 

Tz (v) = v1 −2 1  v(5
v2 v3 = 
 3 0 3 
point)
1 2 2 2 −1
3 3 3 3 3 0
π
c) (10) Let T be the rotation about the axis z of the angle (counterclockwise).
2
Find the matrix representation of T with respect to S.
(sol)
Supposeorthnormal basis β= {x,  y, z}. 
cos π2 − sin π2 0 0 −1 0
   
π
[T]β = 
 sin 2 cos π2 0 = 1 0 0(3 point)
 
0 0 1 0 0 1
  
  0 −1 0 xt
β t t t
  
[T]S = [I]Sβ [T]β [I]S = x y z    t
1 0 0 y  = −xy +yx +zz (3 point)
0 0 1 zt
Since (−xy + yx )z = 0,(−xy + yx )x= y, (−xyt + yxt )y = −x,
t t t t

−2 2
0 3 3 

−xyt + yxt = −[Tz ]S =   3
2
0 −1 
3 
−2 1
0
  3 3 
1 1 2 2
3   9 9 9
zzt =  2 1
3 3
2
3
2
3
=
9
2 4
9
4
9
2 2 4 4
3  9 9 9   
−2 2 1 2 2 1 −4 8
0 3 3 
 9 9 9  9 9 9
[T]S =  2 −1  + 2 4 4 =  8 4 1 (4 point)
 3 0 3  9 9 9  9 9 9
−2 1 2 4 4 −4 7 4
3 3 0 9 9 9 9 9 9
  
2 −1 −3
 
4. 3 3 −2. That is, find
a) (10) Give the QR-decomposition of the matrix A =  
6 0 2
an orthogonal matrix Q (with determinant 1) and an upper-triangular matrix R
satisfying A = QR. You don’t have to show its uniqueness.
(sol)      
2 −1 −3
     
a1 = 3 , a2 =  3  , a3 = −2
    

6 0 2
Use Gram Schmidt
  Orthogonalization
  Process,
2 2
 
1  
 
1  
u1 = a1 = 7 3 , e1 = 7 3
6 6
   
−3 −3
⟨u1 ,a2 ⟩ 3 
   
1 
u2 = a2 − ||u1 ||2 u1 = 7  6 , e2 = 7  6 


−2 −2
   
−6 −6
⟨u1 ,a3 ⟩ ⟨u2 ,a3 ⟩
 
4 
 
1 
u3 = a3 − ||u1 ||2 u1 − ||u2 ||2 u2 = 7 −2 , e3 = 7 −2


3 3
 
  2 −3 −6
 
Q = e1 |e2 |e3 = 71  3 6 −2 (5 point),

6 −2 3
   
a1 · e1 a2 · e1 a3 · e1 7 1 0
   
R=  0 a2 · e2 a3 · e2   = 0 3 −1(5 point)
 
0 0 a3 · e3 0 0 4
b) (10) Explain that Q in a) represents a rotation, and determine the axis, and angle
of rotation.
(sol)
Since QQT = QT Q = I, Q is orthogonal and det(Q) = 1 , Q is rotation matrix.
(4 point)
det(Q − λI) = (−343λ2 + 196λ − 343)(λ − 1) = 0
eigenvalue
 λ=1       
3 6 3 6
−5 −3 −6 −5 −3 −6 1 5 5 1 5 5 
      
−14 −28 
Q−I =   3 −1 −2 →  3 −1 −2 → 3 −1 −2 → 0
    
5 5 
6 −2 −4 0 0 0 0 0 0 0 0 0
   
1 35 65 1 0 0
   
→ 0 1 2 
 → 0 1 2
 
0 0 0 0 0 0
 
0
1 
 
eigenvector v1 = 5  2 
√

−1
Since Qv1 = v1 , axis of rotation is v1 .(3 point)
  
1
 
v2 = 0

 is perpendicular to v1 .
0
 
2
7 2
3
 7  , cos θ = Qv2 · v2 =
Qv2 =  7
6
7
Therefore, rotation angle θ = arccos 27 . (3 point)

 
2 −1 1
 
5. a) (10) Give the spectral decomposition of the matrix A = 
−1 2 .
1
1 1 2
(sol)
2−λ −1 1
det(A − λI) = −1 2−λ 1 = −λ(λ − 3)2
1 1 2−λ
λ1 = 0        
−1 −1
2 −1 1 1 2
1
2 1 2
1
2 1 −1
2
1
2
       
3 3
A−λ1 I =  −1 2 1 → −1 2 1  → −1 2 1  → 0 2
     
2
1 1 2 1 1 2 0 3 3 0 3 3
   
1 −12
1
2 1 0 1
  
→ 0 1 1  → 0 1 1
  

0 0 0 0 0 0
 
1
 
1 
v1 = 3  1 
√
(2 point)
−1
λ2 = 3    
−1 −1 1 1 1 −1
   
A − λ2 I =  −1 −1 1  → 0 0 0 
  
1 1 −1 0 0 0
   
1 1
 
1 
 
1  
v2 = 2 −1(2 point), v3 = 6 1(2 point)
√  √

0 2
A = λ1 v1 · v1t + λ2 v2 · v2t + λ2 v3· v3t   
1 1 −1 1 −1 0 1 1 2
     
= λ1 13 
 1 1 −1  + λ2 1 −1 1 0 + λ2 1 1 1 2(4 point)
 2   6  
−1 −1 1 0 0 0 2 2 4
 
1 0 1
 
b) (10) Find the Moore-Penrose pseudoinverse of B =  1 −1 0.

0 1 1
(sol)  
2 −1 1
 
BT B =  −1 2 1 = A . As obtained in a) eigenvalue λ1 = 0 , λ2 = 3 ,

1 1 2
      
1 1 1
     
v1 = √1
 1  , v2 = √1 −1, v3 = √1  1 
3
  2   6  
−1 0 −2
     
1 1 1
1 1
  1 1
  1
 
u 2 = √λ B · v 2 = √6 
 2  , u3 = λ2 B · v3 = 2 0, u1 = 3 −1
 √ √   √  
2

−1 1 −1
   
√1 √1 √1 0 0 0
 −13 2
6 2
P  √ 
U =  √3 √6
 0 (2 point) ,
 = 0
 3 0  (2 point),
−1 −1 1
√ 
√ √ √ 0 0 3
 3 6 2
√1 √1 √1
 13 −1
2 6
V = √1 (2 point)
 √3 √
2 6
−1 √2
√
3
0
6     
√1 √1 √1 √1 −1 −1 1 1
0 0 0 √ √ 0
P†  13 2 6    13 3 3 3 3 
B† = V U† = √
 3
−1
√
2
√1 
6 
0 √1
3
0   √6
 √2
6
−1 
√
6
=
0
−1
3
1
3
−1 2 √1 √1 √1 1 1
√
3
0 √
6
0 0 3 2
0 2 3 0 3
(4 point)

c) (10) Consider the inconsistent linear system Bx = (2 2 2)t . Find the least
square error solution x0 of the minimal norm. That is, find x0 minimizing
||Bx − (2 2 2)t ||2 , and explain your solution has the minimal norm.
(sol)   
4
2 3
  
z= 2, x0 = B † z =  0 (5 point)
   
4
2 3
Since Bx0 = BB † z, x0 is the orthogonal projectionofz on R(T)(5 point)
2
 
Therefore, Bx0 is the vector in R(T) nearest z =  2

2

6. Let f be a function, and F be an vector field on R3 . (Assume they are C ∞ .)

a) (5) Explain Leibniz’s Rule holds: ∇ · (f F) = ∇f · F + f (∇ · F). One may derive

this by using the Cartesian system.
(sol)
∂ ∂ ∂
∇ · (f F) = ∇ · (f F1 î + f F2 ĵ + f F3 k̂) = ∂x (f F1 )î + ∂y (f F2 )ĵ + ∂z (f F3 )k̂
∂f ∂f ∂f
= ∂x F1 + ∂y F2 + ∂z F3 + f ( ∂F
∂x +
1 ∂F2 ∂F3
∂y + ∂z ) = ∇f · F + f (∇ · F)(point 5)

b) (5) Let D be a 3-dimensional region with the boundary ∂D as a closed orientable

regular surface, and u be a C ∞ function on D. Show that
  
u∇2 udv = (u∇u) · ds − |∇u|2 dv.
D ∂D D

(One may call this Integration by parts.)

(sol)
 Using Leibniz’s Rule, ∇ · (u∇u) = (∇u) · (∇u) + u(∇ · ∇u) = |∇u|2 + u∇2 u.
  
∇ · (u∇u)dv = |∇u|2 dv + u∇2 udv
D D D

According to the divergence theorem,

 
∇ · (u∇u)dv = (u∇u) · ds
D ∂D

Therefore,   
2
u∇ udv = (u∇u) · ds − |∇u|2 dv.
D ∂D D

(point 5)

7. a) (10) Let F = fx î + fy ĵ + fz k̂ = fR R̂ + fθ θ̂ + fϕ ϕ̂, represented in Cartesian and

spherical coordinates, respectively. Write the relations between (fx fy fz )t and
(fR fθ fϕ )t in a matrix form. Here, θ and ϕ denote the zenith angle and the
azimuth angle, respectively.
(sol)
fR = F · R̂ = fx î · R̂ + fy ĵ · R̂ + fz k̂ · R̂ = fx (sin θ cos ϕ) + fy (sin θ sin ϕ) + fz cos θ

fθ = F · θ̂ = fx î · θ̂ + fy ĵ · θ̂ + fz k̂ · θ̂ = fx (cos θ cos ϕ) + fy (cos θ sin ϕ) + fz (− sin θ)

fϕ = F · ϕ̂ = fx
î · ϕ̂ + fy ĵ · ϕ̂ + fz k̂ · ϕ̂ = fx (− sin
 ϕ) + fy (cos ϕ) + fz (0)
sin θ cos ϕ sin θ sin ϕ cos θ
t
  t t
(fR fθ fϕ ) = cos θ cos ϕ cos θ sin ϕ  · (fx fy fz ) = A · (fx fy fz )
− sin θ

− sin ϕ cos ϕ 0
 
sin θ cos ϕ cos θ cos ϕ − sin ϕ
t −1 t
  t
(fx fy fz ) = A (fR fθ fϕ ) = 
 sin θ sin ϕ cos θ sin ϕ  (fR fθ fϕ )
cos ϕ 
cos θ − sin θ 0

2
b) (10) Let F = 3 sin θR̂ + cos ϕθ̂. Evaluate the outward flux F · ds, where D
∂D
π
is the region enclosed by R = 2 and θ = for z ≥ 0.
6
(sol1)
According to the divergence theorem,
 
F · ds = (∇ · F)dv
∂D D

1 ∂ 2 1 ∂ 1 ∂fϕ 6 cos2 θ cos ϕ

∇·F= R2 ∂R (R fR ) + R sin θ ∂θ (sin θfθ ) + R sin θ ∂ϕ = R sin θ + R sin θ

  π  2π  2
6 6 cos2 θ cos ϕ 2
(∇ · F)dv = ( sin θ + )R sin θdRdϕdθ
D 0 0 0 R R sin θ
 π
6
 2π  2
= (6R sin2 θ + R cos θ cos2 ϕ)dRdϕdθ
0 0 0
 π
6
 2π
= (12 sin2 θ + 2 cos θ cos2 ϕ)dϕdθ
0 0
  π
6
= (24π sin2 θ + 2π cos θ cos2 ϕ)dθ
0
√
= 2π 2 − 3 3π + π

(sol2)
∂D = ∂Dtop + ∂Dside
  π
6
 2π  π
6
 2π
F·ds = 2
(3 sin θR̂+cos ϕθ̂)·(R sin θdϕdθR̂) = 2
12 sin2 θdϕdθ
∂Dtop 0 0 0 0

 π
6 √
= 24π sin2 θdθ = 2π 2 − 3 3π
0
  2π  2  2π  2
1
F·ds = (3 sin θR̂+cos2 ϕθ̂)·(R sin θdRdϕθ̂) = R cos2 ϕdRdϕ
∂Dside 0 0 0 0 2
 2π
= cos2 ϕdϕ = π
0
   √
F · ds = F · ds + F · ds = 2π 2 − 3 3π + π
∂D ∂Dtop ∂Dside

8. (출석문제, 5) 한자 ‘山’이 의미하는 것은?

(sol)
수업 중 이야기가 산으로 간다는 의미