Math 54 Practice Final Solutions

May 7, 2023

1. Find the particular and general solutions for the differential equation
y ′′ + 2y ′ + y = 4e−t + 2t2 sin t.

SOLUTION: The auxiliary equation is

r2 + 2r + 1 = (r + 1)2 = 0,
so a general solution to the corresponding homogeneous equation is
c1 e−t + c2 te−t .

To find a solution to the nonhomogeneous equation, we consider each term separately. First note that
4e−t = 4t0 e−t . Since r = −1 is a double root of the auxiliary equation, our particular solution must
lie in
W = span{e−t , te−t , t2 e−t }
(the double root means we go up to t0+2 = t2 ). We have T (e−t ) = 0, T (te−t ) = 0, and T (t2 e−t ) = 2e−t .
Thus T (c1 e−t + c2 te−t + c3 t2 e−t ) = 4e−t can be solved by solving the linear system with augmented
matrix  
0 0 2 4
0 0 0 0 .
0 0 0 0
Solving this gives the first term in our nonhomogeneous solution as 2t2 e−t .
For the other terms, we note that t2 sin t = t2 e0t sin t. As 0 ± i is not a root of the auxiliary equation,
our corresponding solution will lie in
W = span{sin t, cos t, t sin t, t cos t, t2 sin t, t2 cos t}.
We can compute
T (sin t) = 2 cos t
T (cos t) = −2 sin t
T (t sin t) = 2 sin t + 2 cos t + 2t cos t
T (t cos t) = −2 sin t + 2 cos t − 2t sin t
T (t2 sin t) = 2 sin t + 4t sin t + 4t cos t + 2t2 cos t
T (t2 cos t) = 2 cos t − 4t sin t + 4t cos t + 2t2 sin t.

Thus we can solve T (yp ) = t2 sin t by solving the system with augmented matrix
 
0 −2 2 −2 2 0 0
2 0 2 2 0 2 0
 
0 0 0 −2 4 −4 0
 .
0 0 2 0 4 4 0
 
0 0 0 0 0 −2 2
0 0 0 0 2 0 0

1
 This gives the particular solution

−3 sin t + 2t sin t + 2t cos t − t2 cos t.

Combining these solutions gives the total particular solution

yp = 2t2 e−t − 3 sin t + 2t sin t + 2t cos t − t2 cos t,

so that any solution to our ODE can be written as

y = c1 e−t + c2 te−t + 2t2 e−t − 3 sin t + 2t sin t + 2t cos t − t2 cos t.

 
25 −15
2. ChatGPT tells us that the SVD of A = by finding the eigenvalues and eigenvectors of
−15 25
AT A and using those to compute U and V matrices.
(a) ChatGPT tells us that the eigenvalues are 700 and 100. Is this correct? If not, what is / are the
correct eigenvalue(s)?
√   √ 
2/2 − 2/2
(b) ChatGPT tells us that the eigenvectors are √ and √ . Is this correct? If not, what
2/2 2/2
is / are the correct eigenvector(s)?
(c) ChatGPT tells us that
√ √  √ √ 
2/2 −√ 2/2 2/2 −√ 2/2
U= √ and V = √ .
2/2 2/2 2/2 2/2

Is this correct? If not, what is / are the correct U and V?

SOLUTION: ChatGPT is not correct on any of these points. To see this, we compute the SVD of A.
Note that A = AT , and we have  
850 −750
AT A = .
−750 850
We can compute det(AT A −λI) = (λ − 1600)(λ − 100), so the eigenvalues are λ1 = 1600 and λ2 = 100.
−1
For λ1 = 1600, we see that is an eigenvector of AT A with eigenvalue λ1 . For λ2 = 100, we see
1
√
 
1
that is an eigenvector of AT A with eigenvalue λ2 . Computing the singular values (σi = λi ) and
1
normalizing the eigenvectors, we get
   √ √ 
40 0 −√ 2/2 √2/2
Σ= and V = .
0 10 2/2 2/2

Note that ChatGPT reverses the columns of V (which is wrong!). For the matrix U, we have
 √   √  √  √ 
1 − 2/2 − 2/2 1 2/2 2/2
⃗u1 = A √ = √ and ⃗u2 = A √ = √ ,
40 2/2 2/2 10 2/2 2/2
so that  √ √ 
− 2/2 √2/2 .
U= √
2/2 2/2

3. (a) Find a 5 × 6 matrix with rank 2 and nullity 4.

SOLUTION: We can take  
1 0 0 0 0 0
0 1 0 0 0 0
 
0 0 0 0 0 0
 .
0 0 0 0 0 0
0 0 0 0 0 0

2
 (b) Find a nonzero matrix A such that A2 ̸= 0 and A3 = 0.
SOLUTION: We can take  
0 1 0
0 0 2 ,
0 0 0
the matrix of the differentiation map D : P2 (R) → P2 (R).

(c) Find a non-invertible matrix with only one eigenvalue.

SOLUTION: The zero matrix is one such matrix.

(d) Find the matrix representation

    (in
 terms of the standard basis) of the linear transformation
1 3 2 −4
R2 → R2 sending to and to .
1 8 3 −1
SOLUTION: Note that
           
1 1 2 0 2 1
=3 − and = −2 .
0 1 3 1 3 1

Thus                 
1 1 2 1 2 3 −4 13
T =T 3 − = 3T −T =3 − = .
0 1 3 1 3 8 −1 25
Similarly                
1 2 1 2 1 −4 3 −10
T =T −2 =T − 2T = −2 = .
0 3 1 3 1 −1 8 −17
We obtain  
13 −10
T= .
25 −17

4. True / False. Justify your answer.

(a) A matrix A and its transpose AT have the same eigenvalues.
SOLUTION: True. The characteristic polynomial of AT is

det(AT − λI) = det(AT − λIT ) = det((A − λI)T ) = det(A − λI),

which is the characteristic polynomial of A. Since A and AT have the same characteristic poly-
nomial, they must also have the same eigenvalues.

(b) If the matrix A2 is invertible, then the matrix A3 is invertible.

SOLUTION: True. Write B for the inverse of A2 . Then

A3 (A3 B3 ) = A2 A2 A2 BBB = A2 A2 IBB = A2 IB = I,

so A3 B3 is the inverse of A3 , and in particular A3 is invertible.

(c) If U and W are subspaces of a vector space V, and we let

U + W = {⃗u + w
⃗ | ⃗u ∈ U, w
⃗ ∈ W}

(this is a subspace of V), then

dim(U + W) = dim U + dim W.

SOLUTION: False. For example, if U = W = V = R1 , then U + W = R1 , so dim(U + W) = 1

and dim U + dim W = 2.

3
 (d) The set {1, sin x, cos x, sin2 x, cos2 x} is linearly independent.
SOLUTION: False. We have the linear relation cos2 x + sin2 x = 1.

(e) If a matrix A has zero as one of its eigenvalues, then the dimension of the null space of A is at
least 1.
SOLUTION: True. If λ = 0 and A⃗v = λ⃗v = 0 (with ⃗v ̸= 0), then ⃗v is in the null space of A, so
dim Nul A ⩾ 1.
 
  3 −12 4 0
3 −12 4 −1 0 −2 0
5. Suppose A = −1 0 −2 and B =  .
−1 5 −1 0
−1 5 −1
0 0 0 0
(a) How do the characteristic equations of A and B differ?
SOLUTION: The characteristic polynomial of B is −λ times the characteristic polynomial of A.

(b) How do the rank and nullity of A and B differ?

SOLUTION: We have rank A = rank B and dim Nul A + 1 = dim Nul B.

(c) Find all the eigenvalues of A5 . (Hint: the eigenvalues of A are integers.)
SOLUTION: We can compute
dim(A − λI) = −λ3 + 2λ2 + 8λ − 2 = −(λ − 2)(λ − 1)(λ − 1)
by guessing and checking, so λ = ±1, 2. If we diagonalize A, say A = PDP−1 , then A5 = PD5 P−1 ,
and the eigenvalues of A5 are 15 = 1, (−1)5 = −1, and 25 = 32.

(d) What are the eigenvalues of A−1 ?

SOLUTION: If λ is an eigenvalue of A, then A⃗v = λ⃗v gives ⃗v = A−1 A⃗v = λA−1⃗v, hence A−1⃗v =
1
λ⃗v. The eigenvalues of A−1 are therefore the multiplicative inverses of those of A, namely 1, −1,
and 1/2.
6. Let V be the vector space of 2 × 2 matrices. Consider the map
   
2 3 2 3
T (A) = A−A .
5 7 5 7
(a) Show that A is a linear transformation.
SOLUTION: We have
   
2 3 2 3
T (αA + βB) = (αA + βB) − (αA + βB)
5 7 5 7
       
2 3 2 3 2 3 2 3
=α A−A +β B−B
5 7 5 7 5 7 5 7
= αT (A) + βT (B).
(b) Show that a linear transformation S is injective if and only if the null space of S is the zero vector
space.
SOLUTION: If ⃗u is injective, let ⃗u ∈ Nul S. Then T (⃗u) = T (⃗0) = ⃗0 forces ⃗u = ⃗0.
Conversely, suppose Nul S = ⃗0. If T (⃗u) = T (⃗v), then T (⃗u −⃗v) = ⃗0, hence ⃗u −⃗v = ⃗0, and finally we
see ⃗u = ⃗v.
(c) Is T injective?
SOLUTION: No. For example, we have
           
2 3 1 0 1 0 2 3 2 3 2 3 0 0
T (I) = − = − = .
5 7 0 1 0 1 5 7 5 7 5 7 0 0
Thus I ∈ Nul T , so Nul T ̸= {⃗0} and T cannot be injective.

4
 (d) Is T bijective?
SOLUTION: No, because T is not injective.
7. (a) Find the Fourier series of f(x) = ex on the interval [−π, π].
SOLUTION: We can compute
Z
1 π x eπ − e−π
a0 = e dx =
π −π π
and
Z
1 π x
an = e cos(nx)dx
π −π
1 ex
= [cos(nx) − n sin(nx)]π
−π
π 1 + n2
1 (−1)n (eπ − e−π )
= 2
(eπ (−1)n − e−π (−1)n ) = .
π(1 + n ) π(1 + n2 )
Similarly
Z
1 π x
bn = e sin(nx)dx
π −π
1 ex
= [sin(nx) − n cos(nx)]π −π
π 1 + n2
1 π n −π n −n(−1)n (eπ − e−π )
= (e (−n)(−1) − e (−n)(−1) ) = .
π(1 + n2 ) π(1 + n2 )
Thus our Fourier series on [−π, π] is
∞ 
eπ − e−π X (−1)n (eπ − e−π ) −n(−1)n (eπ − e−π )

ex = + cos(nx) + sin(nx) .
2π n=1
π(1 + n2 ) π(1 + n2 )

(b) Sketch the 2π-periodic even extension of f(x) = ex . Find the Fourier cosine series of f(x) = ex on
the interval (0, π).
SOLUTION: A sketch is provided in Fig. 1.

Figure 1: Sketch of the even extension of f(x) = ex .

For the Fourier cosine series, we compute

Zπ
2 2(eπ − 1)
a0 = ex dx =
π 0 π
and
Zπ
2 2 2
ex cos(nx)dx = eπ cos(nπ) − e0 cos 0 = (eπ (−1)n − 1).


an =
π 0 π(n2+ 1) 2
π(n + 1)
Thus our Fourier cosine series is
X∞
!
eπ − 1 2 eπ (−1)n − 1
+ cos(nx) .
π π n=1
n2 + 1

5
 (c) Sketch the 2π-periodic odd extension of f(x) = ex . Find the Fourier sine series of f(x) = ex on
the interval (0, π).
SOLUTION: A sketch is provided in Fig. 2.

Figure 2: Sketch of the odd extension of f(x) = ex .

For the Fourier sine series, we compute

Z
2 π x
bn = e sin(nx)dx
π 0
 
2 −n
= [ex cos(nx)]π
0
π n2 + 1
−2n 2n
= 2
(eπ cos(nπ) − 1) = (−eπ (−1)n + 1) .
π(n + 1) π(n2 + 1)

Thus our Fourier sine series is

∞
2 X n
(1 − eπ (−1)n ) sin(nx).
π n=1 n2 + 1

8. (a) To find the best fit line y = β0 + β1 x given points (x1 , y1 ), (x2 , y2 ), . . . , (xn , yn ), we find the least
squares solution to the system of equations

β0 + β1 x1 = y1
β0 + β1 x2 = y2
...
β0 + β1 xn = yn .

Write this system of equations as a matrix equation. Find the best fit line given the points (1, 0),
(2, 1), (3, 1).
SOLUTION: The matrix equation is
   
1 x1 y1
1 x2     y2 
 0 β
..  β =  ..  .
  
 ..
. .  1  . 
1 xn yn

For the specific collection of points mentioned, this becomes

   
1 1   0
β
1 2 0 = 1 .
β1
1 3 1

6
 To find a least-squares solution, we consider the equation
   
  1 1     0
1 1 1  β 1 1 1  
1 2 0 = 1 ,
1 2 3 β1 1 2 3
1 3 1

or equivalently     
3 6 β0 2
= .
6 14 β1 5
We can solve to get β0 = −1/3 and β1 = 1/2. Our line of best fit is then y = − 13 + 12 x.

(b) For the best fit cubic curve y = β0 +β1 x+β2 x2 +β3 x3 given points (x1 , y1 ), (x2 , y2 ), . . . , (xn , yn ),
write the system of equations as well as the matrix equation for which we seek the least squares
solution.
SOLUTION: The system of equations is

β0 + β1 x1 + β2 x21 + β3 x31 = y1
.. ..
.=.
β0 + β1 xn + β2 x2n + β3 x3n = yn .

In matrix form this is

x21 x31 β0 
   
1 x1 y1
1 x2 x22 x32     y2 
 β1   
 = . .

 .. .. .. ..  
. . . .  β2  .. 
1 xn x2n x3 β3 yn
n

(c) Find the best fit cubic curve given the points (−2, −5), (−1, 1), (0, −3), (1, 2), (2, 7).
SOLUTION: The matrix equation here is
   
1 −2 4 −8   −5
1 −1 1 −1 β0 1
1 0 0 0  β1  = −3 .
    
1 1 1 1  β 2
    
2
β3
1 2 4 8 7

We can find the cubic of best fit using least-squares as before: the solution is y = β0 + β1 x +
β2 x2 + β3 x3 , where
3 1 1 5
β0 = − , β1 = − , β2 = , β3 = .
5 3 2 6