Applied Mathematics Masters Examination

Fall 2018, August 21, 1–4 pm.

Each of the fifteen numbered questions is worth 20 points. All questions will be graded,
but your score for the examination will be the sum of your scores on your eight best
questions.

Real Analysis
Problem 1. Let f be a differentiable function function on interval [0, 3] and f 0 is contin-
uous on the same interval. Suppose we know that f (0) = 3, f (2) = −2, f (3) = 1.
• Show that there exists a point a ∈ [0, 3] such that f ( a) = 0;
• Show that there exists a point b ∈ [0, 3] such that f 0 (b) = 3;
• Show that there exists a point c ∈ [0, 3] such that f 0 (c) = 1;

Solution: By the IVT, since f (0) = 3 and f (2) = −2 the function must attain any
value in between, in particular, there exists a point a ∈ [0, 2] such that f ( a) = 0. By
f (3)− f (2)
the mean value theorem there exists b ∈ [2, 3] such that f 0 (b) = 3−2 = 3. Also, by
f (2)− f (0)
the MVT there exists a point d ∈ [0, 2] such that f 0 (d) = 2−0 = − 52 . Applying the
IVT to the derivative f 0 , we conclude that there should exist a point at which f 0 take
intermediate value 1 between 3 and − 52 . 

Problem 2. Prove using the definition of a limit that

cos(n)
lim √ = 0,
n→∞ n+1
√
lim h + 12 = 4.
h →4

Solution: 1. Let us fix an e > 0. We need to find N such that for all n > N we
have
cos(n)
√ < e.
n+1
Using that | cos(n)| ≤ 1 we estimate
cos(n) 1
√ ≤√ .
n+1 n+1
We thus will aim at a stronger inequality
1
√ < e.
n+1
1
Solving for n we obtain
1
− 1.
n>
e2
So, if we choose N = [ e12 ] then for all n > N we guarantee the desired inequality.

2. Let us fix an e > 0. We need to find δ > 0 such that for all |h − 4| < δ we have
√
| h + 12 − 4| < ε.
√
Let us multiply and divide by the conjugate h + 12 + 4:
√ | h − 4|
| h + 12 − 4| = √
h + 12 + 4
If δ1 = 1, then h > 3, and hence we continue
| h − 4| | h − 4|
√ <√ < | h − 4|.
h + 12 + 4 15 + 4
We need to make the last expression < ε. So, if δ2 = ε, then the condition |h − 4| < δ2
guarantees this. Picking δ = min{ε, 1} we guarantee with the final and intermediate
inequalities. This implies that if |h − 4| < δ, then
√
| h + 12 − 4| < ε.


Problem 3.
R1
• Let f : [0, 1] → R be a continuous function such that 0 f ( x )2 dx = 0. Prove that
f ( x ) = 0 for all x ∈ [0, 1].
• Give an example (proof is not required) of a bounded Riemann integrable func-
R1
tion g : [0, 1] → R with 0 g( x )2 dx = 0 and g(m/n) > 0 for all integers
1 ≤ m ≤ n.

Solution: Let f : [0, 1] → R be continuous. Assume, towards contradiction, that for some
a ∈ [0, 1] one has v = f (c) 6= 0. Let e < |v|/2. By continuity, there is some δ > 0 so that
for x ∈ (c − δ, c + δ) ∩ [0, 1] we have | f ( x ) − v| < e; in particular | f ( x )| > |v| − e > |v|/2.
Choose an interval [ a, b] ⊂ (c − δ, c + δ) ∩ [0, 1] with a < b. We have
Z 1 Z b
2
f ( x ) dx ≥ f ( x )2 dx ≥ (b − a) · v2 /4 > 0.
0 a
This contradiction proves that f ( x ) = 0 for all x ∈ [0, 1].
Let g( x ) = 0 for irrational x ∈ [0, 1], and g(m/n) = 1/n is known to be Riemann inte-
R1
grable with 0 g( x ) dx = 0. Since 0 ≤ g( x ) ≤ 1 on [0, 1], it follows that 0 ≤ g( x )2 ≤ g( x )
R1
and so 0 g( x )2 dx = 0. 

2
 Complex Analysis
Problem 4. Suppose f (z) is analytic on a domain D. Show that if f (z) is also analytic
on D, then f is constant.

Solution: Write f (z) = u( x, y) + iv( x, y) in terms of its real and imaginary parts. Then
we have that f (z) = u( x, y) + iv( x, y) and f (z) = u( x, y) − iv( x, y) are both analytic
on D. So by the Cauchy-Riemann equations applied to f (z), we have ∂ x u = ∂y v and
∂y u = −∂ x v on D, while by the Cauchy-Riemann equations applied to f (z), we have
∂ x u = −∂y v and ∂y u = ∂ x v on D. The equations ∂ x u = ∂y v and ∂ x u = −∂y v give
∂y v = ∂ x u = −∂y v on D, so that ∂y v = 0 on D. Since ∂ x u = ∂y v, we also have ∂ x u = 0
on D.
The other two equations similarly give ∂y u = ∂ x v = 0 on D. Hence ∇u and ∇v are
both identically zero on D. Hence u and v are constant functions, implying f (z) is con-
stant on D. 

Problem 5. Find all z ∈ C such that the principal value of zi is equal to i.

Solution: By definition, zi = eiLog(z) . Writing z = reiθ for θ ∈ (−π, π ], we have

Log(z) = ln r + iθ. Thus zi = eiLog(z) = ei ln r−θ = e−θ ei ln r . Also, i = e 2 . For e−θ ei ln r to
πi

equal e 2 , we must have e−θ = 1 and ln r = π2 + 2πk for some k ∈ Z. Hence θ = 0 and
πi

r = e 2 +2πk for some k ∈ Z. We conclude that the z for which zi = i are exactly the z of
π

the form e 2 +2πk for some k ∈ Z. 

π

Problem 6. Using residues, determine the value of the following integral:

Z ∞
sin x
dx
−∞ x2 + 2x + 2

Solution: The integral is the imaginary part of

Z ∞
eix
dx
−∞ x2 + 2x + 2
The residue theorem applies to integrals of this form. The zeroes of the denominator oc-
cur at x = −1 ± i, and only the zero −1 + i occurs above the x axis. Hence by the residue
iz eiz
theorem, the integral is given by 2πi times the residue of z2 +e2z+2 = (z−(−1+i))( z−(−1−i ))
at z = −1 + i. Since the pole is of order 1, the residue is given by inserting z = −1 + i
eiz i (−1+i )
into (z−(− 1−i ))
, obtaining e 2i . So the integral is 2πi times this, or πei(−1+i) = πe e−i =
e (cos 1 − i sin 1). Hence the original integral is the imaginary part of this, or − e sin 1.
π π

3
 Applied Differential Equations
Problem 7. Find the general solution to the system
x 00 + 4x =3 sin t,
x 0 − y00 + y =2 cos t.

Solution: We solve the first equation by undetermined coefficients. The homogeneous

solution is
xh (t) = c0 cos(2t) + c1 sin(2t).
We look for a particular solution in the form x p = c2 sin t + c3 cos t. Inserting it to the
first equation yields
−c2 sin t − c3 cos t + 4c2 sin t + 4c3 cos t = 3 sin t
which implies c2 = 1 and c3 = 0. Thus a general solution for x is
x (t) = c0 cos(2t) + c1 sin(2t) + sin t.
Now we consider y equation. Inserting x (t) to the second equation leads to
(1) −y00 + y = cos t + 2c0 sin(2t) − 2c1 cos(2t).
The homogeneous solution is
y h ( t ) = c2 e t + c3 e − t .
We look for a particular solution in the form y p (t) = c4 cos(2t) + c5 sin(2t) + c6 cos t +
c7 sin t. Inserting y p to (1) leads to
5c4 cos(2t) + 5c5 sin(2t) + 2c6 cos t + 2c7 sin t = cos t + 2c0 sin(2t) − 2c1 cos(2t)
which implies
2 2 1
c4 = − c1 , c5 = c0 , c6 = , c7 = 0.
5 5 2
Thus the general solution for y is
1 2 2
y ( t ) = c2 e t + c3 e − t + cos t + c0 sin(2t) − c1 cos(2t)
2 5 5


Problem 8. Find a second solution y2 to the differential equation

t2 y00 + 2ty0 − 2y = 0
given that y1 (t) = t is a solution. Verify that y1 and y2 are linearly independent.

Solution: Let y2 (t) = tv(t) for some function v(t) which will be determined. Assume
y2 (t) is a solution to the linear differential equation. Thus,
t2 y200 + 2ty20 − 2y2 = 0.
4
Inserting the form y2 (t) = tv(t) to the equation above yields
4
(2) v00 + v0 = 0.
t
Rewriting equation (2) gives
v00 4
0
=− .
v t
Integrating both sides, we obtain
ln |v0 | = −4 ln |t| + c1
for a constant c1 . Thus we have
v 0 = c 2 t −4
for another constant c2 . Integrating one more time leads to
v ( t ) = c 3 t −3 + c 4 .
Thus,
y2 (t) = tv(t) = c3 t−2 + c4 t
is a solution as well. In particular, one can choose c3 = 1, c4 = 0. So y2 (t) = t−2 .
One can check that there are no constants A and B such that
Ay1 (t) + By2 (t) = 0, ∀t.
Thus, y1 and y2 are linearly independent. 

Problem 9. Given the system

du
=u − v − u3 − uv2 ,
dt
dv
=u + v − v3 − u2 v.
dt
Show that: (a) the system has a unique equilibrium point; (b) the equilibrium is unstable.

Solution: (a) Set the right hand side of the equations to 0,

u − v − u3 − uv2 =0
u + v − v3 − u2 v =0.
Multiplying the first by v and the second by u, and adding them leads to
u2 + v2 = 0
which implies u = 0, v = 0 is the only solution. Thus, (0, 0) is the unique equilibrium
point.
(b) Take V (u, v) = 21 u2 + 12 v2 , with constant a, c satisfying a < 2c. It is easy to check
V (0, 0) = 0; V ( x, y) > 0, ∀(u, v) 6= (0, 0). Therefore, V is positive definite. We can
compute
dV
V ∗ (u, v) := = (u2 + v2 )[1 − (u2 + v2 )]
dt
5
which is positive in a small neighborhood of (0, 0). Thus, V ∗ is positive definite on
{(u, v)|u2 + v2 = 21 }. It implies the equilibrium point at the origin is unstable. 

Applied Partial Differential Equations

Problem 10. Derive the general form of the solution for the following initial-boundary
value problem (with c > 0):
∂2 u 2
2∂ u
−c = 0, 0 < x < 1,
∂t2 ∂x2
subject to
u(t, 0) = 0, u(t, 1) = 5,
and
∂u
u(0, x ) = f ( x ), (0, x ) = 0,
∂t
where f ( x ) is a given function.

Solution: In order to take into account the inhomogeneous boundary condition at x = 1,

we write v(t, x ) = u(t, x ) − 5x. Then, one easily sees that v(t, x ) solves the wave equation
with homogenous boundary conditions: v(t, 0) = v(t, 1) = 0 and (modified) initial data:
v(0, x ) = f ( x ) − 5x, ∂t v(0, x ) = 0.
The method of eigenfunction expansion (or, separation of variables) can thus be applied
to v(t, x ) = ψ(t)φ( x ) which yields:
φ00 ( x ) + λφ( x ) = 0, ψ00 (t) + λc2 ψ(t) = 0,
with separation constant λ. The homogenous b.c. require φ(0) = φ(1) = 0, and thus:
λn = (nπ )2 , φn ( x ) = sin (nπx ) n = 1, 2, 3, . . . .
With this in mind, the solution for ψ reads
p p
ψn (t) = an cos( λn ct) + bn sin( λn ct).
In summary, we find the following general form of the solution u(t, x ):
∞
∑


u(t, x ) = 5x + an cos (nπct) + bn sin (nπct) sin (nπx ) ,
n =1

where the coefficients are given by:

Z 1
an = ( f ( x ) − 5x ) sin (nπx ) dx, bn = 0, n = 1, 2, . . . .
0
Here, the condition bn = 0 follows from differentiating the solution formula w.r.t. t and
using ∂t u(0, x ) = 0. 

6
Problem 11. Consider the following heat type equation
∂u ∂2 u
− 2 = u, x ∈ R, t > 0,
∂t ∂x
with initial data
u(0, x ) = f ( x ),
Derive the general solution to this problem for integrable f .

Solution: We use the Fourier transformation defined by

Z
û(t, k) ≡ F (u(t, ·)) = u(t, x )e−ikx dx.
R
In Fourier space the equation reads
∂t û + k2 û = û, û(0, k ) = fˆ(k ),
The general solution in Fourier space is
2 )t
û(t, k ) = e(1−k fˆ(k ).
Computing the Fourier inverse of this Gaussian-type function yields
 2
  2 et − x2 /4t
F −1 e (1− k ) t ( x ) = e t F −1 e − k t ( x ) = √ e .
4πt
Thus, we find the general solution given by the Fourier-convolution formula:
et  −|·|2 /4t et
 Z
2
(A) u(t, x ) = √ e ∗ f (x) = √ e−( x−y) /4t f (y) dy.
4πt 4πt R


Problem 12. Consider the following eigenvalue problem

d2 1
2
ϕ − 2 ϕ = λϕ, x ∈ ( a, b),
dx x
subject to homogenous Dirichlet conditions: ϕ( a) = ϕ(b) = 0.
Prove that all eigenvalues satisfy λ ≤ 0.

Solution: We multiply the equation by ϕ and integrate w.r.t. x ∈ [ a, b]:

Z b Z b Z b
00 1 2
(∗) ϕ ϕ dx − 2
ϕ dx = λ ϕ2 dx.
a a x a
By integrating by parts, we find that the first term on the right hand side is equal to
Z b Z b Z b
b
ϕ00 ϕ dx = ϕ0 ϕ a − ( ϕ0 )2 dx = − ( ϕ0 )2 dx,
a a a
since, by assumption, ϕ( a) = ϕ(b) = 0. Hence (∗) is equivalent to
Z b  Z b
0 2 1 2
− ( ϕ ) + 2 ϕ dx = λ ϕ2 dx.
a x a
7
Since the right hand side is always non-positive, the left hand side has to be non-positive
too, which consequently implies λ ≤ 0. 

Numerical Analysis
Problem 13. Consider the equation
x4 − 6x3 + 12x2 − 8x = 0.
(1) For the solution x = 0 which will converge faster, the Bisection Method or New-
ton’s Method. Why? Be explicit and quantitative.
(2) For the solution x = 2 which will converge faster, the Bisection Method or New-
ton’s Method. Why? Be explicit and quantitative.

Solution: Define f ( x ) = x4 − 6x3 + 12x2 − 8x. Recall that Section 1.1.2 of Sauer tells
us that Bisection converges linearly with factor S = 1/2, while Theorem 1.11 says that
if f 0 (r ) 6= 0 then Newton’s Method converges quadratically. However, Theorem 1.12
shows that if r is a root of multiplicity m then Newton’s Method converges at a linear
rate with factor S = (m − 1)/m.
(1) It is easy to verify that f (0) = 0 (so that 0 is a root) and that f 0 (0) = −8 6= 0
so that Newton’s Method converges quadratically. This is faster than Bisection’s
linear rate of convergence S = 1/2.
(2) It can be shown that f (2) = 0 (so that 2 is a root), while f 0 (2) = f 00 (2) = 0 and
f 000 (2) = 12 6= 0. This indicates that 2 is a root of multiplicity m = 3 and thus
Newton’s Method converges linearly with rate S = (m − 1)/m = 2/3 which is
slower than Bisection’s linear rate of convergence S = 1/2.


Problem 14. Assume that the polynomial P5 ( x ) interpolates f ( x ) = e x at the 6 evenly

spaced points {0, 1/5, 2/5, 3/5, 4/5, 1}.
(1) Find an upper bound for the error | f (1/2) − P5 (1/2)|.
(2) How many decimal
√ places can you guarantee to be correct if P5 (1/2) is used to
approximate e?

Solution: A theorem (Sauer, Theorem 3.4) tells us that if P( x ) is the (degree (n − 1)

or less) interpolating polynomial fitting the n points ( x1 , y1 ), . . . , ( xn , yn ), then the inter-
polation error is
( x − x1 )( x − x2 ) · · · ( x − xn ) (n)
f ( x ) − P( x ) = f ( c ),
n!
where c lies between the smallest and largest of { x, x1 , . . . , xn }. Here we have n = 6 and
it is easy to see that f (6) ( x ) = e x . So, the maximum value that | f (6) (c)| has for these
points is e.
8
 (1) From the theorem
   
1 1
f −P ≤
2 2
      
1 1 1 1 2 1 3 1 4 1
2 −0 2 − 5 2 − 5 2 − 5 2 − 5 2 −1 5e
e = .
6! 8 × 106

(2) From above error is approximately 10−6 so about six.



Problem 15. Find c0 , c1 , c2 such that the rule

Z 2
f ( x ) dx ≈ c0 f (0) + c1 f (1) + c2 f (2)
0
has degree of precision greater than one.

Solution: To guarantee degree of precision greater than one using three constants, we
demand that the rule be exact for the monomials {1, x, x2 }. This gives
Z 2
2= 1 dx = c0 (1) + c1 (1) + c2 (1),
0
Z 2
2= x dx = c0 (0) + c1 (1) + c2 (2),
0
Z 2
8/3 = x2 dx = c0 (0) + c1 (1) + c2 (4).
0
The latter two equations become
    
1 2 c1 2
= ,
1 4 c2 8/3
which has solution c1 = 4/3 and c2 = 1/3. The first equation gives c0 = 2 − c1 − c2 =
1/3.