ECON 205: Intermediate Mathematics for Economics

SMU School of Economics; Fall 2019

Answer Key to Homework Assignments
Takashi Kunimoto

November 26, 2019

You are required to submit your work at my dropbox (or what you call pigeon-
hole) AB48 on the 5th floor of SOE/SOSS building by 5:00pm on the date specified
below.

1 Homework 1 (Due Date: Aug 29, 2019)

Question 1.1 (12 points) Consider the following implications and decide whether
the implication is true or not. Let x and y be real numbers.

1. x = 5 and y = −3 ⇒ x + y = 2.

True.
2. x2 = 16 ⇒ x = 4.

False. x = −4 is consistent with x2 = 16.

3. (x − 3)2 (y + 2) > 0 ⇒ y > −2.

True. We must have (x − 3) 6= 0 and (y + 2) > 0 simultaneously. So, y > −2.

4. x3 = 8 ⇒ x = 2.

True. x3 −8 = 0 ⇒ (x−2)(x2 +2x+4) = 0. We know that x2 +2x+4 = 0 does

not have solutions in the real numbers. If we allow ourselves to use complex
numbers, there are solutions other than x = 2. But we are constrained to use
only real numbers here.

Question 1.2 (10 points) Let U = {1, 2, . . . , 11} be the universal set and define
A = {1, 4, 6} and B = {2, 11}. Answer the following:

1. A ∩ B?

A ∩ B = ∅.

1
 2. A ∪ B?

A ∪ B = {1, 2, 4, 6, 11}.

3. B c = U \B?

B c = {1, 3, 4, 5, 6, 7, 8, 9, 10}.

4. Ac = U \A?

Ac = {2, 3, 5, 7, 8, 9, 10, 11}.

Question 1.3 (13 points) Let f (x) = ax3 + bx2 + cx + d. Then, show that

f (x + h) − f (x)
= 3ax2 + 2bx + c + 3axh + ah2 + bh.
h
0
Also, find f (x).
We compute the following:

f (x + h) = a(x + h)3 + b(x + h)2 + c(x + h) + d

= a(x3 + 3x2 h + 3xh2 + h3 ) + b(x2 + 2xh + h2 ) + c(x + h) + d
f (x) = ax3 + bx2 + cx + d.

So,
f (x + h) − f (x) = a(3x2 h + 3xh2 + h3 ) + b(2xh + h2 ) + ch
Now, it is easy to see that we obtain what is desired as (f (x + h) − f (x))/h. Finally,

0 f (x + h) − f (x)
f (x) = lim = lim (3ax2 +2bx+c+3axh+ah2 +bh) = 3ax2 +2bx+c.
h→0 h h→0

Question 1.4 (10 points) The total cost of producing x units of a commodity is
C(x) = x3 − 90x2 + 7500x, where x ≥ 0. Answer the following questions.
0
1. Compute the marginal cost function, C (x).
0
C (x) = 3x2 − 180x + 7500.

2. For which value of x is the marginal cost the least?

0
We rearrange C (x) as follows:
0
C (x) = 3(x − 30)2 + 4800.

So, the marginal cost is minimized at x = 30.

Question 1.5 (15 points) Differentiate the following functions.

2
 √
1. f (x) = 35 x2 − 2x7 + 1
8 − x.

0 6 1
f (x) = x − 14x6 − x−1/2 .
5 2

2. f (x) = (2x2 − 1)(x4 − 1).

0
f (x) = 4x(x4 − 1) + (2x2 − 1)4x3 = 12x5 − 4x3 − 4x.

1
3. f (x) = x5 + (x5 + 1).


x
0
f (x) = (5x4 − x−2 )(x5 + 1) + (x5 + x−1 )5x4 = 10x9 + 5x4 + 4x3 − x−2 .

4. f (x) = ex x−2 .
 
0 x −2 x −3 x −2 −3 x 1 2
f (x) = e x + e × (−2)x = e (x − 2x )=e 2
− 3
x x

5. f (x) = (x + ex )2 .
0
f (x) = 2(x + ex )(1 + ex ).

Question 1.6 (20 points) Find the equations for the tangents to the graphs of the
following functions at the specified points.

1. y = 3 − x − x2 at x = 1.
dy
= −1 − 2x
dx
dy 

= −3.
dx x=1


So, we have
y − 1 = −3(x − 1) ⇔ y = −3x + 4,
which is the equation for the tangent to the graph.

2. y = (x2 − 1)/(x2 + 1) at x = 1.

dy 2x(x2 + 1) − (x2 − 1)2x 4x

= 2 2
= 2
dx (x + 1) (x + 1)2
dy 
= 1.
dx x=1


So, we have
y − 0 = (x − 1) ⇔ y = x − 1,
which is the equation for the tangent to the graph.

3
 1
+ 1 (x2 − 1) at x = 2.


3. y = x2

dy
= −2x−3 (x2 − 1) + (x−2 + 1)2x = 2(x−3 + x)
dx
dy 
= 17/4.
dx x=2


So, we have
5 17
y−
= (x − 2),
12 4
which is the equation for the tangent to the graph.
x4 +1
4. y = (x2 +1)(x+3)
at x = 0.

4x3 (x2 + 1)(x + 3) − (x4 + 1) 2x(x + 3) + (x2 + 1)


dy
=
dx (x2 + 1)2 (x + 3)2
x6 + 6x5 + 3x4 + 12x3 − 3x2 − 6x − 1
=
(x2 + 1)2 (x + 3)2
dy 
= −1/9.
dx x=0


So, we have
1 1
= − (x − 0),
y−
3 9
which is the equation for the tangent to the graph.

Question 1.7 (20 points) For each of the following functions, determine the in-
tervals where it is increasing.

1. y = 3x2 − 12x + 13.

We compute dy/dx:
dy
= 6x − 12 = 6(x − 2).
dx
So, the function is increasing if and only if x ≥ 2.

2. y = (x4 − 6x2 )/4.

We compute dy/dx:

dy 4x3 − 12x √ √
= = x3 − 3x = x(x2 − 3) = x(x + 3)(x − 3).
dx 4
√ √
Thus, the function is increasing if and only if either x ≥ 3 or − 3 ≤ x ≤ 0.

4
 3. y = 2x/(x2 + 2).

We compute dy/dx:
√ √
dy 2(x2 + 2) − 2x · 2x −2(x + 2)(x − 2)
= = .
dx (x2 + 2)2 (x2 + 2)2
√ √
Thus, the function is increasing if and only if − 2 ≤ x ≤ 2.

4. y = (x2 − x3 )/2(x + 1).

First, observe that this function is not well-defined when x = −1. We compute
dy/dx:

dy (2x − 3x2 )(x + 1) − (x2 − x3 ) −x(x2 + x − 1)

= =
dx 2(x + 1)2 (x + 1)2
 √  √ 
−x x + 1+2 5 x − (−1+2 5)
=
(x + 1)2
√
√ is increasing if and only if either x ≤ −(1 +
Thus, the function 5)/2 or
0 ≤ x ≤ (−1 + 5)/2.

2 Homework 2 (Due Date: Sep 5, 2019)

Question 2.1 (6 points) Compute the following:

1. dY /dt when Y = −3(V + 1)5 and V = t3 /3.

4
t3

dY dY dV
= · = −15(V + 1)4 · t2 = −15t2 +1 .
dt dV dt 3

2. dK/dt when K = ALα and L = bt + c (A, a, b, and c are positive constants).

dK dK dL
= · = αALα−1 · b = αbA(bt + c)α−1 .
dt dL dt

Question 2.2 (12 points) If u(z) denotes an individual’s utility of having income
(or consumption) z, then
00 0
R = −zu (z)/u (z)
is the coefficient of relative risk aversion. Compute R for the following utility
functions (where A1 , A2 , and ρ are positive constants with ρ 6= 1, and we assume
that z > 0):

5
 √
1. u(z) = z.
0 1 −1/2
u (z) = z
2
00 1
u (z) = − z −3/2
4
00 0
R = −zu (z)/u (z) = 1/2.

2. u(z) = A1 − A2 z −2 .
0
u (z) = 2A2 z −3
00
u (z) = −6A2 z −4
00 0
R = −zu (z)/u (z) = 3.

z 1−ρ
3. u(z) = A1 + A2 (1−ρ) .
0
u (z) = A2 z −ρ
00
u (z) = −ρA2 z −(1+ρ)
00 0
R = −zu (z)/u (z) = ρ.

Question 2.3 (12 points) Answer the following questions

00
1. Compute y for y = 3x3 + 2x − 1.
0
y = 9x2 + 2
00
y = 18x
000
2. Compute Y for Y = 1 − 2x2 + 6x3 .
0
Y = −4x + 18x2
00
Y = −4 + 36x
000
Y = 36.

3. Compute d3 z/dt3 for z = 120t − (1/3)t3 .

dz/dt = 120 − t2
d2 z/dt2 = −2t
d3 z/dt3 = −2.

4. Compute f (4) (1) for f (z) = 100z −4 .

0
f (z) = −400z −5
00
f (z) = 2000z −6
000
f (z) = −12000z −7
f (4) (z) = 84000z −8 .

6
Question 2.4 (12 points) Determine the following using the rules for limits:

1. limx→0 (3 + 2x2 ).

lim (3 + 2x2 ) = lim 3 + 2 lim (x · x)

x→0 x→0 x→0
= 3.

3+2x
2. limx→−1 x−1 .

3 + 2x 3 + 2 limx→−1 x
lim =
x→−1 x − 1 limx→−1 (x − 1)
1
= = −1/2.
−2

3. limx→2 (2x2 + 5)3 .

 3
lim (2x2 + 5)3 = 2 lim x2 + 5
x→2 x→2
= (23 + 5)3 = 133 = 2197.

4. limt→8 (5t2 + t2 − t3 /8).

1
lim(5t2 + t2 − t3 /8) = 5 lim t2 + lim t2 − lim t3
t→8 t→8 t→8 8 t→8
= 5 · 82 + 82 − 82 = 320.

(y+1)5 −y 5
5. limy→0 y+1 .

(y + 1)5 − y 5 (limy→0 (y + 1))5 − limy→0 y 5

lim =
y→0 y+1 limy→0 (y + 1)
1
= = 1.
1

1/z+2
6. limz→−2 z .

1/z + 2 limz→−2 1/z + 2

lim =
z→−2 z limz→−2 z
−1/2 + 2
= = −3/4.
−2
0
Question 2.5 (15 points) Use logarithmic differentiation to find f (x)/f (x) in the
following.

7
  1/3
x+1
1. f (x) = x−1 .
 
1 x+1 1
ln f (x) = ln [ln(x + 1) − ln(x − 1)]
=
3 x−1 3
0  
f (x) d ln f (x) 1 1 1 −2
= = − = .
f (x) dx 3 x+1 x−1 3(x + 1)(x − 1)

2. f (x) = x2x .

ln f (x) = 2x ln x
0
f (x) d ln f (x)
= = 2 ln x + 2 = 2(ln x + 1).
f (x) dx
√
3. f (x) = x − 2(x2 + 1)(x4 + 6).
1
ln f (x) = ln(x − 2) + ln(x2 + 1) + ln(x4 + 6)
2
0
f (x) d ln f (x) 1 2x 4x3
= = + 2 + 4 .
f (x) dx 2(x − 2) x + 1 x + 6

Question 2.6 (15 points) Find the intervals where the following functions are in-
creasing:

1. y = ln(4 − x2 ).

We first check the domain of the function. The inside of logarithmic function
must be positive. So, we must satisfy 4−x2 > 0. This implies that −2 < x < 2.
0
Next, we compute y :
0 −2x
y = .
4 − x2
0
Then, y ≥ 0 if and only if either −2 < x ≤ 0 or x > 2. After we take into
account the domain of the function, we conclude that y is increasing if and
only if either −2 < x ≤ 0.

2. y = x3 ln x.
0
We compute y :
0
y = 3x2 ln x + x2 = x2 (3 ln x + 1) = x2 (ln x3 + 1).
0
Then, y ≥ 0 if and only if x ≥ e−1/3 . That is, y is increasing if and only if
x ≥ e−1/3 . Note that ln e−1 = − ln e = −1. Recall that f (x) = ln x is not
well-defined over the negative orthant of the real line.

8
 (1−ln x)2
3. y = 2x .
0
We compute y :

0 2(1 − ln x) · (−1/x) · x − (1 − ln x)2 −(ln x − 1)(ln x − 3)

y = = .
2x2 2x2
0
Then, y ≥ 0 if and only if e ≤ x ≤ e3 . Recall that f (x) = ln x is only defined
over the positive orthant of the real line. Hence, y is increasing if and only if
e ≤ x ≤ e3 .

Question 2.7 (20 points) Let f : R → R be a function defined by

f (x) = ln(2 + ex−3 ).

Answer the following questions.

1. Show that f is strictly increasing.

0 ex−3
f (x) = > 0.
2 + ex−3
So, f is strictly increasing.

2. Find the range of f .

The range of f is (ln 2, ∞).

3. Find an expression for the inverse function g of f .

y = ln(2 + ex−3 )
⇔ ey = 2 + ex−3
⇔ ex−3 = ey − 2
⇔ x − 3 = ln(ey − 2)
⇔ x = ln(ey − 2) + 3.

So, g(y) = ln(ey − 2) + 3.

4. Find the range of f over which g is well-defined.

Since the logarithmic function is only well-defined over the positive region of
the real line, we must have ey − 2 > 0. This implies that y > ln 2. So, the
range of f over which g is well-defined is (ln 2, ∞).
0 0
5. Verify that f (3) = 1/g (ln 3).
0
We compute g (y):
0 ey
g (y) = .
ey − 2

9
 So,
0 ey − 2
1/g (y) =
ey
0 0
and we have 1/g (ln 3) = 1/3. We also compute f (3) = 1/3.

Question 2.8 (8 points) Let f : R → R be defined by


3x − 2 if x ≤ 2
f (x) =
−x + 6 if x > 2.

Answer the following questions

1. Draw a graph of f .

We skip this.

2. Is f continuous at x = 2? If your answer is yes, verify it explicitly. If your

answer is no, provide a counterexample.

If we make x converge to 2 from the left, we have 3x − 2 → 4. If we make

x converge to 2 from the right, we have −x + 6 → 4. So, we confirm that
x → 2 ⇒ f (x) → 4. Thus, f (x) is continuous at x = 2.

3 Homework 3 (Due Date: Sep 12, 2019)

Question 3.1 (10 points) Prove that
√ 1
1+x≈1+ x
2
for x close to 0.
Let f (y) = y 1/2 . It is easy to see that f (y) is differentiable at all y. For x small
enough, we have
0 1
f (1 + x) ≈ f (1) + f (1)x = 1 + x.
2
Question 3.2 (10 points) Use the Intermediate Value Theorem to show that the
equation 10x4 − 15x3 + x2 − 1 = 0 has a solution between −1 and 1.
0
Let f (x) = 10x4 − 15x3 + x2 − 1. We obtain f (x) = 40x3 − 45x2 + 2x so that
f (x) is differentiable at all x. This implies that f (x) is continuous as well. We note
that f (−1) = 25 and f (1) = −5. Then, by Intermediate Value Theorem, there exists
c ∈ (−1, 1) such that f (c) = 0.

Question 3.3 (16 points) Find ∂z/∂x and ∂z/∂y for the following:

10
 1. z = x2 + 3y 2 .

∂z/∂x = 2x,
∂z/∂y = 6y.

2. z = xy.

∂z/∂x = y,
∂z/∂y = x.

3. z = 5x4 y 2 − 2xy 5 .

∂z/∂x = 20x3 y 2 − 2y 5 ,
∂z/∂y = 10x4 y − 10xy 4 .

4. z = ex+y .

∂z/∂x = ex+y ,
∂z/∂y = ex+y .

5. z = exy .

∂z/∂x = yexy ,
∂z/∂y = xexy .

6. z = ex /y.

∂z/∂x = ex /y,
∂z/∂y = −ex /y 2 .

7. z = ln(x + y).
1
∂z/∂x =
x+y
1
∂z/∂y = .
x+y

8. z = ln(xy).
1
∂z/∂x =
x
1
∂z/∂y = .
y

Question 3.4 (16 points) Find all first- and second-order partial derivatives for
the following:

11
1. z = x2 + e2y
∂z
= 2x
∂x
∂z
= 2e2y
∂y
∂2z
= 2
∂x2
∂2z
= 4e2y
∂y 2
2
∂ z ∂2z
= = 0.
∂x∂y ∂y∂x

2. z = y ln x
∂z
= y/x
∂x
∂z
= ln x
∂y
∂2z
= −y/x2
∂x2
∂2z
= 0
∂y 2
∂2z ∂2z
= = 1/x.
∂x∂y ∂y∂x

3. z = xy 2 − exy
∂z
= y 2 − yexy
∂x
∂z
= 2xy − xexy
∂y
∂2z
= −y 2 exy
∂x2
∂2z
= 2x − x2 exy
∂y 2
2
∂ z ∂2z
= = 2y − exy (xy + 1).
∂x∂y ∂y∂x

12
 4. z = xy
∂z
= yxy−1
∂x
∂z
= xy ln x
∂y
∂2z
= y(y − 1)xy−2
∂x2
∂2z
= xy (ln x)2
∂y 2
2
∂ z ∂2z
= = xy−1 + yxy−1 ln x = xy−1 (1 + y ln x).
∂x∂y ∂y∂x

Question 3.5 (10 points) Let z = ln(x2 + y 2 )/2. Show that ∂ 2 z/∂x2 + ∂ 2 z/∂y 2 =
0.

∂z 2x x
= = 2
∂x 2(x2 +y ) 2 x + y2
∂z 2y y
= = 2
∂y 2(x2 + y 2 ) x + y2
∂2z (x2 + y 2 ) − x · 2x y 2 − x2
= =
∂x2 (x2 + y 2 )2 (x2 + y 2 )2
∂2z (x2 + y 2 ) − y · 2y x2 − y 2
= =
∂y 2 (x2 + y 2 )2 (x2 + y 2 )2

Clearly, we obtain ∂ 2 z/∂x2 + ∂ 2 z/∂y 2 = 0.

Question 3.6 (9 points) Show that x2 + y 2 = 6 is a level curve of f (x, y) =

p
x2 + y 2 − x2 − y 2 + 2.
If x2 + y 2 = 6, we must have
p √ √
x2 + y 2 − x2 − y 2 + 2 = 6 − 6 + 2 = −4 + 6.

This implies that

√ if we choose all pairs of (x, y) such that x2 + y 2 = 6, we have
f (x, y) = −4 + 6, which is constant. Therefore, x + y 2 = 6 is a level curve of
2

f (x, y).

Question 3.7 (10 points) Verify that the points (−1, 5) and (1, 1) lie on the same
level curve for the function:

g(x, y) = (2x + y)3 − 2x + 5/y.

13
 g(−1, 5) = (−2 + 5)3 − 2(−1) + 5/5 = 33 + 2 + 1 = 27 + 3 = 30
g(1, 1) = (2 + 1)3 − 2 + 5 = 33 − 2 + 5 = 30.

So, (−1, 5) and (1, 1) lie on the same level curve.

Question 3.8 (10 points) Let F (x, y) = ex−2y − ln(2x2 y). Answer the following
questions.

1. For which points (x, y) is F (x, y) well-defined?

ex−2y is always well-defined for any values of x and y. However, the domain of
ln(2x2 y) must be positive. In order to have 2x2 y > 0, we must have x 6= 0 and
y > 0.

2. Show that (x, y) = (1, 1/2) lies on the level curve F (x, y) = 1.

F (1, 1/2) = e1−1 − ln(2 · 1 · (1/2)) = e0 − ln 1 = 1 − 0 = 1.

Question 3.9 (9 points) Let f (x, y, z) = xy + yz + zx. Answer the following

questions.

1. Calclulate f (−1, 2, 3).

f (−1, 2, 3) = (−1) · 2 + 2 · 3 + 3 · (−1) = −2 + 6 − 3 = 1.

2. Calculate f (a + 1, b + 1, c + 1) − f (a, b, c).

f (a + 1, b + 1, c + 1) = (a + 1)(b + 1) + (b + 1)(c + 1) + (c + 1)(a + 1)

= ab + bc + ca + 2(a + b + c) + 3
f (a, b, c) = ab + bc + ca.

So, f (a + 1, b + 1, c + 1) − f (a, b, c) = 2(a + b + c) + 3.

3. Show that f (tx, ty, tz) = t2 f (x, y, z) for all t > 0.

f (tx, ty, tz) = (tx)(ty) + (ty)(tz) + (tz)(tx) = t2 xy + t2 yz + t2 zx

= t2 (xy + yz + zx) = t2 f (x, y, z).

4 Homework 4 (Due Date: Sep 19, 2019)

Question 4.1 (12 points) Consider the functions: z = f (x, y), x = g(t, s) and
y = h(t, s). Answer the following questions.

14
1. Suppose that we are given

x = g(t, s) = (at + bs)2

√
y = h(t, s) = at + bs,

where a and b are constants. Find the differentials, dx and dy, in terms of dt
and ds.

∂g(t, s) ∂g(t, s)
dx = dt + ds
∂t ∂s
= 2a(at + bs)dt + 2b(at + bs)ds.
∂h(t, s) ∂h(t, s)
dy = dt + ds
∂t ∂s
a b
= (at + bs)−1/2 dt + (at + bs)−1/2 ds.
2 2

2. Suppose that we are given

z = f (x, y) = xα y 1−α
x = g(t, s) = (t + s)2
√
y = h(t, s) = t + s.

Using the chain rule, find ∂z/∂s when t = s = 1/2.

∂z ∂f (x, y) ∂x ∂f (x, y) ∂y
= +
∂s ∂x ∂s ∂y ∂s
∂f (x, y) ∂g(t, s) ∂f (x, y) ∂h(t, s)
= +
∂x ∂s ∂y ∂s
1
= αxα−1 y 1−α · 2(t + s) + (1 − α)xα y −α · (t + s)−1/2
2
xα−1 y −α n o
= 4αy(t + s)3/2 + (1 − α)x
2(t + s)1/2
(t + s)2α−2 (t + s)−α/2 n 5/2 2
o
= 4α(t + s) + (1 − α)(t + s)
2(t + s)1/2
(t + s)3α/2−3 n o
= 4α(t + s)2 + (1 − α)(t + s)3/2
2
1n o
= 4α(t + s)3α/2−1 + (1 − α)(t + s)3(α−1)/2
2
So, we obtain
∂z  1
= {4α + 1 − α}
∂s (t,s)=(1/2,1/2) 2

3α + 1
= .
2

15
Question 4.2 (7 points) Consider the equation:

exyz − 3xyz = 1.

Find ∂z/∂x when x = 1 and z = 1.

Define F (x, y, z) = exyz − 3xyz − 1 so that we have F (x, y, z) = 0. We first
compute the following:
∂F
= exyz · (yz) − 3yz = yz(exyz − 3);
∂x
∂F
= exyz · (xy) − 3xy = xy(exyz − 3).
∂z
Using the implicit function theorem, we obtain

∂z  yz(exyz − 3) 
= −
∂x (x,z)=(1,1) xy(exyz − 3) (x,z)=(1,1)
 
ey − 3
= − y = −1.
e −3
0
Question 4.3 (15 points) Using the implicit function theorem, find y for the fol-
00
lowing level curves. Also, find y using chain rule.

1. x2 y = 1

Define F (x, y) = x2 y − 1 so that we have F (x, y) = 0. We first compute the

following:
∂F
= 2xy
∂x
∂F
= x2 .
∂y

Assume that x 6= 0 so that ∂F/∂y 6= 0 is guaranteed. Using the implicit

function theorem, we have

0 ∂F/∂x 2y
y =− =− .
∂F/∂y x

We also define G(x, y) = − 2y

x . Using chain rule, we obtain
 
00 ∂G ∂G 0 2y 2 −2y 6y
y = + y = 2− · = 2.
∂x ∂y x x x x

2. x − y + 3xy = 2

16
 Define F (x, y) = x − y + 3xy − 2 so that we have F (x, y) = 0. We first compute
the following:
∂F
= 1 + 3y
∂x
∂F
= −1 + 3x
∂y

Assume that x 6= 1/3 so that ∂F/∂y 6= 0 is guaranteed. Using the implicit

function theorem, we have
0 1 + 3y 1 + 3y
y =− = .
−1 + 3x 1 − 3x
Define G(x, y) = −(1 + 3y)/(1 − 3x). Then, using chain rule, we obtain
00 ∂G ∂G 0 3(1 + 3y) 3(1 + 3y) 6(1 + 3y)
y = + y = 2
+ 2
= .
∂x ∂y (1 − 3x) (1 − 3x) (1 − 3x)2

3. y 5 − x6 = 0.

Define F (x, y) = y 5 − x6 so that we have F (x, y) = 0. We first compute the

following:
∂F
= −6x5
∂x
∂F
= 5y 4
∂y

Assume that y 6= 0 so that ∂F/∂y 6= 0 is guaranteed. Using the implicit

function theorem, we have
0 ∂F/∂x 6x5
y =− = 4.
∂F/∂y 5y
Define G(x, y) = 6x5 /5y 4 . Using chain rule, we obtain
6x4 24x5 6x5 6x4 24x6
 
00 ∂G ∂G 0
y = + y = 4 − · = 1− .
∂x ∂y y 5y 5 5y 4 y4 25y 5

Question 4.4 (14 points) Consider the following three matrices:

     
1 2 −3 3 −1 2 4 1 2
A= 5 0 2  ; B =  4 2 5  ; and C =  0 3 2  .
1 −1 1 2 0 3 1 −2 3
Find the matrices (i) A + B; (ii) A − B; (iii) AB; (iv) BA; (v) A(BC); and (vi)
(AB)C. You are required to provide the entire process of the derivation for each
small question.

17
    
1 + 3 2 − 1 −3 + 2 4 1 −1
A+B =  5+4 0+2 2+5 = 9 2 7 .
1 + 2 −1 + 0 1 + 3 3 −1 4
   
1 − 3 2 + 1 −3 − 2 −2 3 −5
A−B =  5−4 0−2 2 − 5  =  1 −2 −3 
1 − 2 −1 − 0 1 − 3 −1 −1 −2
 
1 · 3 + 2 · 4 + (−3) · 2 1 · (−1) + 2 · 2 + (−3) · 0 1 · 2 + 2 · 5 + (−3) · 3
AB =  5·3+0·4+2·2 5 · (−1) + 0 · 2 + 2 · 0 5·2+0·5+2·3 
1 · 3 + (−1) · 4 + 1 · 2 1 · (−1) + (−1) · 2 + 1 · 0 1 · 2 + (−1) · 5 + 1 · 3
 
5 3 3
=  19 −5 16  .
1 −3 0
 
3 · 1 + (−1) · 5 + 2 · 1 3 · 2 + (−1) · 0 + 2 · (−1) 3 · (−3) + (−1) · 2 + 2 · 1
BA =  4·1+2·5+5·1 4 · 2 + 2 · 0 + 5 · (−1) 4 · (−3) + 2 · 2 + 5 · 1 
2·1+0·5+3·1 2 · 2 + 0 · 0 + 3 · (−1) 2 · (−3) + 0 · 2 + 3 · 1
 
0 4 −9
=  19 3 −3  .
5 1 −3

 
3 · 4 + (−1) · 0 + 2 · 1 3 · 1 + (−1) · 3 + 2 · (−2) 3 · 2 + (−1) · 2 + 2 · 3
BC =  4·4+2·0+5·1 4 · 1 + 2 · 3 + 5 · (−2) 4·2+2·2+5·3 
2·4+0·0+3·1 2 · 1 + 0 · 3 + 3 · (−2) 2·2+0·2+3·3
 
14 −4 10
=  21 0 27  .
11 −4 13
  
1 2 −3 14 −4 10
A(BC) =  5 0 2   21 0 27 
1 −1 1 11 −4 13
 
1 · 14 + 2 · 21 + (−3) · 11 1 · (−4) + 2 · 0 + (−3) · (−4) 1 · 10 + 2 · 27 + (−3) · 13
=  5 · 14 + 0 · 21 + 2 · 11 5 · (−4) + 0 · 0 + 2 · (−4) 5 · 10 + 0 · 27 + 2 · 13 
1 · 14 + (−1) · 21 + 1 · 11 1 · (−4) + (−1) · 0 + 1 · (−4) 1 · 10 + (−1) · 27 + 1 · 13
 
23 8 25
=  92 −28 76  .
4 −8 −4

18
   
5 3 3 4 1 2
(AB)C =  19 −5 16   0 3 2 
1 −3 0 1 −2 3
 
5·4+3·0+3·1 5 · 1 + 3 · 3 + 3 · (−2) 5·2+3·2+3·3
=  19 · 4 + (−5) · 0 + 16 · 1 19 · 1 + (−5) · 3 + 16 · (−2) 19 · 2 + (−5) · 2 + 16 · 3 
1 · 4 + (−3) · 0 + 0 · 1 1 · 1 + (−3) · 3 + 0 · (−2) 1 · 2 + (−3) · 2 + 0 · 3
 
23 8 25
=  92 −28 76  .
4 −8 −4

Question 4.5 (8 points) Consider the following three matrices:

     
2 2 2 0 1 0
A= ; B= ; and I = .
1 5 3 2 0 1
Answer the following questions.

1. Find a matrix C satisfying (A − 2I)C = I.

     
2 2 2 0 0 2
A − 2I = − = .
1 5 0 2 1 3

    
0 2 c11 c12 2c21 2c22
(A − 2I)C = = .
1 3 c21 c22 c11 + 3c21 c12 + 3c22
If we want to have (A−2I)C = I, we must have c11 = −3/2; c12 = 1; c21 = 1/2;
and c22 = 0. Therefore,  
−3/2 1
C= .
1/2 0

2. Is there a matrix D satisfying (B − 2I)D = I? If your answer is yes, provide

such D. If your answer is no, argue why.
     
2 0 2 0 0 0
B − 2I = − = .
3 2 0 2 3 0
    
0 0 d11 d12 0 0
(B − 2I)D = = .
3 0 d21 d22 3d11 3d12
The above computation clearly shows that there is no such D.

Question 4.6 (6 points) Consider the following matrices:

   
a11 a12 b11 b12
A= and B = .
a21 a22 b21 b22
0 0 0
Prove (AB) = B A explicitly.

19
 Let  
a11 b11 + a12 b21 a11 b12 + a12 b22
AB =
a21 b11 + a22 b21 a21 b12 + a22 b22
and  
0 a11 b11 + a12 b21 a21 b11 + a22 b21
(AB) =
a11 b12 + a12 b22 a21 b12 + a22 b22
    
0 0 b11 b21 a11 a21 b11 a11 + b21 a12 b11 a21 + b21 a22 0
BA = = = (AB) .
b12 b22 a12 a22 b12 a11 + b22 a12 b12 a21 + b22 a22

Question 4.7 (8 points) For what values of a is

a2 − 1
 
a −3
 a+1 2 a2 + 4 
−3 4a −1

symmetric?
In order for the above matrix to be symmetric, we must have the following two
equations:
(1) a2 − 1 = a + 1 and (2) a2 + 4 = 4a.
For (1), we obtain (a − 2)(a + 1) = 0. For (2), we obtain (a − 2)2 = 0. To satisfy
these equations simultaneously, we must have a = 2.

Question 4.8 (12 points) Consider the following matrices:

   
1 −1 0 1 2 3
A= 1 3 2  and B =  2 3 4 .
1 2 1 0 1 −1

Calculate (i) AB; (ii) |A|; (iii) |B|; (iv) |A| · |B|; and (v) |AB|.

 
1 · 1 + (−1) · 2 + 0 · 0 1 · 2 + (−1) · 3 + 0 · 1 1 · 3 + (−1) · 4 + 0 · (−1)
AB =  1 · 1 + 3 · 2 + 2 · 0 1·2+3·3+2·1 1 · 3 + 3 · 4 + 2 · (−1) 
1·1+2·2+1·0 1·2+2·3+1·1 1 · 3 + 2 · 4 + 1 · (−1)
 
−1 −1 −1
=  7 13 13  .
5 9 10
   
 3 2 
1+1  1+2  1
 2 
|A| = 1 · (−1)  2 1  + (−1) · (−1)

 1 = −1 − 1 = −2.
1 
   
1+1  3 4  2+1  2 3 
 
|B| = 1 · (−1)  1 −1  + 2 · (−1) = −7 + 10 = 3.
 1 −1 

20
|A| · |B| = (−2) · 3 = −6.
     
 −1 −1
|AB| = 7 · (−1)2+1   + 13 · (−1)2+2  −1 −1  + 13 · (−1)2+3  −1 −1 
    
9 10   5 10   5 9 
= 7 − 65 + 52 = −6.

Question 4.9 (8 points) Solve the equation:

 
 1−x 2 2 
 
 2
 1−x 2  = 0.

 2 2 1−x 

We thus have the following equation:

(1 − x)3 + 16 − 12(1 − x) = 0.

Define y = 1 − x. Then, the above equation can be rewritten as:

y 3 − 12y + 16 = 0.

This is decomposed into

(y − 2)2 (y + 4) = 0.
Taking into account y = 1 − x, we obtain x = −1, 5 as the solutions.

Question 4.10 (10 points) Use Cramer’s rule to solve the following system of
equations:

2x1 − 3x2 = 2
4x1 − 6x2 + x3 = 7
x1 + 10x2 = 1.

Let      
2 −3 0 x1 2
A =  4 −6 1  ; x =  x2  ; and b =  7  .
1 10 0 x3 1
Then, the given system of linear equations is expressed as Ax = b. We first claim
that |A| =
6 0.

|A| = 0 × |A13 | + 1 × |A23 | + 0 × |A33 |

 
2+3  2 −3 
 
= (−1)  1 10  = −(2 × 10 − (−3) × 1) = −23 6= 0.

21
Then, by Cramer’s rule, we obtain
 
 2 −3 0  .  
2+3  2 −3 
   
x1 =  7 −6 1  |A| = (−1)
 
 1 10  = 1.
 1 10 0 
 
 2 2 0 .  
  1  2 2 
x2 =  4 7 1  |A| =
  = 0.
 1 1 0  23  1 1 
 
 2 −3 2  .
  −1
x3 =  4 −6 7  |A| = (2 × |A11 | + 4 × |A21 | + |A31 |)
 1 10 1  23
      
−1 1+1  −6 7 
  2+1  −3 2 
  3+1  −3 2
 
= 2(−1)  10 1  + 4 × (−1)  10 1  + (−1)  −6 7

23 
−1
= {2(−6 − 70) − 4(−3 − 20) + (−21 + 12)}
23
−152 + 92 − 9
= = (−69)/(−23) = 3.
(−23)

5 Homework 5 (Due Date: Sep 26, 2019)

Note that we cannot return your work on this assignment before the in-
class quiz. So, I strongly encourage you to photocopy your work before
submitting it to the pigeonhole (dropbox).

Question 5.1 (15 points) Consider the following matrices:

   
1 3 −7 a b c
C =  2 5 1  and D =  −13 14 −15  ,
1 2 7 −1 1 −1

where a, b, and c are constant real numbers. Answer the following questions.

1. Find the determinant |D|.

     
1+1  14 −15  1+2  −13 −15  1+3  −13 14 
     
|D| = a(−1)  1 −1  + b(−1)  −1 −1  + c(−1)  −1 1 
= a(−14 + 15) − b(13 − 15) + c(−13 + 14) = a + 2b + c

22
2. Compute the matrix product CD.
  
1 3 −7 a b c
CD =  2 5 1   −13 14 −15 
1 2 7 −1 1 −1
 
a − 39 + 7 b + 42 − 7 c − 45 + 7
=  2a − 65 − 1 2b + 70 + 1 2c − 75 − 1 
a − 26 − 7 b + 28 + 7 c − 30 − 7
 
a − 32 b + 35 c − 38
=  2a − 66 2b + 71 2c − 76  .
a − 33 b + 35 c − 37

3. Find the values of a, b, and c for which D = C −1 .

If D = C −1 , we must have
   
a − 32 b + 35 c − 38 1 0 0
CD =  2a − 66 2b + 71 2c − 76  =  0 1 0  .
a − 33 b + 35 c − 37 0 0 1

This concludes that a = 33, b = −35, and c = 38.

4. Let  
1 0 2
A =  0 2 −2  ,
0 0 −1
and  
h1
H =  h2  ,
h3
where h1 , h2 , and h3 are constant real numbers. Show that there exists exactly
one 3 × 1 matrix Y such that AY = CH. (Hint: You do not need to find Y
for this question.)

If A−1 exists, we can derive Y uniquely in such a way that Y = A−1 CH. Since
we know that A−1 exists if and only if |A| 6= 0, it only remains to show that
|A| =
6 0. That is,
 
1+1  2 −2 
 
|A| = 1 · (−1)  0 −1  = −2 6= 0.

Thus, we are done.

5. Let B = C −1 AC and  
x1
X =  x2  .
x3

23
 Show that X = C −1 Y is the solution to the equation BX = H.

If B −1 exists, we have

X = B −1 H = (C −1 AC)−1 H (∵ B = C −1 AC)
= C −1 A−1 CH
= C −1 Y (∵ Y = A−1 CH)

Therefore, it only remains to show that B −1 exists. Since we know that B −1

exists if and only if |B| =
6 0. Thus, we compute the following:

|B| = |C −1 | · |A| · |C| (∵ B = C −1 AC)

= |D| · |A| · |C| (∵ D = C −1 )
= |D| · |A| · |C| (∵ |D| = 1 6= 0)
6 0 ⇔ D−1 exists, (ii) |C| =
6= 0 (∵ (i) |D| = 6 0 ⇔ C −1 = D exists, and (iii) |A| = −2)

So, we are done.

Question 5.2 (9 points) Calculate the differentials of the following:

1. z = x3 + y 3 .
dz = 3x2 dx + 3y 2 dy.
2
2. z = xey .
2 2
dz = ey dx + 2xyey dy.

3. z = ln(x2 − y 2 ).    
2x 2y
dz = dx − dy.
x − y2
2 x − y2
2

Question 5.3 (9 points) Let u = u(x, y). Answer the following questions.

1. Find dz expressed in terms of dx and dy when z = x2 u.

0 0
dz = (2xu + x2 ux )dx + x2 uy dy.

2. Find dz expressed in terms of dx and dy when z = u2 .

0 0
dz = 2uux dx + 2uuy dy.

3. Find dz expressed in terms of dx and dy when z = ln(xy + yu).

0
0
! !
y + yux x + u + yuy
dz = dx + dy
xy + yu xy + yu

24
Question 5.4 (9 points) Let a = (1, 2, 2), b = (0, 0, −3), and c = (−2, 4, −3).
Find the following.

1. a + b + c;
      
1 0 −2 −1
a + b + c =  2  +  0  +  4  =  6 .
2 −3 −3 −4

2. a − 2b + 2c;
       
1 0 −2 −3
a − 2b + 2c =  2  − 2  0  + 2  4  =  10  .
2 −3 −3 2

3. 3a + 2b − 3c.

      
1 0 −2 9
3a + 2b − 3c = 3  2  + 2  0  − 3  4  =  −6  .
2 −3 −3 9

Question 5.5 (9 points) Express the vector (4, −11) as a linear combination of
(2, −1) and (1, 4).
Set      
2 1 4
t +s = ,
−1 4 −11
where t, s ∈ R. This is equivalent to the following system of equations:

2t + s = 4,
−t + 4s = −11.

Solving this system, we obtain (t, s) = (3, −2). So,

     
2 1 4
3 −2 = ,
−1 4 −11

Question 5.6 (8 points) For what values of x are (x, −x−8, x, x) and (x, 1, −2, 1)
orthogonal?
We must have (x, −x − 8, x, x) · (x, 1, −2, 1) = 0. This is equivalent to

x2 − x − 8 − 2x + x = 0 ⇔ (x − 4)(x + 2) = 0.

Therefore, x = −2, 4.

Question 5.7 (10 points) Answer the following questions.

25
 1. Find the equation for the line that passes through (3, −2, 2) and (10, 2, 1).

Let a = (3, −2, 2) and b = (10, 2, 1). We compute b − a = (7, 4, −1). Thus,
the corresponding equation is

x = (3, −2, 2) + t(7, 4, −1),

where t ∈ R.

2. Find the equation for the line that passes through (1, 3, 1) and has the same
direction as (0, −1, 1).

The corresponding equation is given as

x = (1, 3, 1) + t(0, −1, 1),

where t ∈ R.

Question 5.8 (16 points) The following system of equations defines both u =
u(x, y) and v = v(x, y) as continuously differentiable functions of x and y around
the point P where (x, y, u, v) = (1, 1, 1, 1):

2uv + v 2 = 2x + y
u − v = x2 − y 2 .

Answer the following questions.

1. Find the differentials of u and v expressed in terms of the differentials of x and

y.

Define g1 (u, v, x, y) = 2uv+v 2 −2x−y = 0 and g2 (u, v, x, y) = u−v−x2 +y 2 = 0

as the implicit functions. First, we elaborate on dg1 = 0:
∂g1 ∂g1 ∂g1 ∂g1
dg1 = du + dv + dx + dy = 0
∂u ∂v ∂x ∂y
⇔ 2vdu + 2(u + v)dv − 2dx − dy = 0
⇔ 2vdu + 2(u + v)dv = 2dx + dy

Second, we elaborate on dg2 = 0:

∂g2 ∂g2 ∂g2 ∂g2
dg2 = du + dv + dx + dy = 0
∂u ∂v ∂x ∂y
⇔ du − dv − 2xdx + 2ydy = 0
⇔ du − dv = 2xdx − 2ydy

We summarize dg1 = 0 and dg2 = 0 in matrix form:

    
2v 2(u + v) du 2dx + dy
=
1 −1 dv 2xdx − 2ydy

26
2. Find ∂u/∂x and ∂v/∂x at P

We start from
    
2v 2(u + v) du 2dx + dy
= .
1 −1 dv 2xdx − 2ydy
Setting dy = 0, we obtain
    
2v 2(u + v) du 2dx
=
1 −1 dv 2xdx
    
2v 2(u + v) ∂u/∂x 2
⇔ =
1 −1 ∂v/∂x 2x

At P = (1, 1, 1, 1), we have

    
2 4 ∂u/∂x 2
= .
1 −1 ∂v/∂x 2
Since  
 2 4 
  = −6 6= 0,
 1 −1 
this matrix has its inverse. Therefore,
   −1       
∂u/∂x 2 4 2 1 1 4 2 5/3
= = = .
∂v/∂x 1 −1 2 6 1 −2 2 −1/3

3. Find ∂u/∂y and ∂v/∂y at P .

We start from
    
2v 2(u + v) du 2dx + dy
= .
1 −1 dv 2xdx − 2ydy
Setting dx = 0, we obtain
    
2v 2(u + v) du dy
=
1 −1 dv −2ydy
    
2v 2(u + v) ∂u/∂y 1
⇔ = .
1 −1 ∂v/∂y −2y

At P = (1, 1, 1, 1), we have

    
2 4 ∂u/∂y 1
= .
1 −1 ∂v/∂y −2
As we did in the previous question, we obtain
   −1       
∂u/∂y 2 4 1 1 1 4 1 −7/6
= = = .
∂v/∂y 1 −1 −2 6 1 −2 −2 5/6

27
 4. If x decreases by 0.1 and y increases by 0.2 from their values at P , what are
the approximate changes in u and v?

We use the following approximation:

∂u  ∂u  5 7 1
∆u|(1,1,1,1) ≈ ∆x + ∆y = × 0.1 − × 0.2 = −
∂x (1,1,1,1) ∂y (1,1,1,1) 3 6 15
 
∂v 
 ∂v 
 1 5 2
∆v|(1,1,1,1) ≈ ∆x + ∆y = − × 0.1 + × 0.2 =
∂x (1,1,1,1) ∂y (1,1,1,1) 3 6 15
 

Question 5.9 (15 points) The equation system:

ln(x + u) + uv − y 2 ev + y = 0
u2 − xv = v
defines u and v as continuously differentiable functions of x and y around the point
P where (x, y, u, v) = (2, 1, −1, 0). Answer the following questions.

1. Differentiate the system of equations with respect to x and evaluate them at

(x, y, u, v) = (2, 1, −1, 0). (Hint: Use the following formula:
∂ v 0
(x ) = vxv−1 + (xv ln x) · vx ).
∂x
We first differentiate the system with respect to x:
0
1 + ux 0 0 0
+ ux v + u · vx − y 2 ev · vx = 0
x+u
0 0 0
2u · ux − vxv−1 − (xv ln x)vx = vx .
Evaluating the above derived system at P , we obtain
0 0 0
1 + ux − vx − vx = 0
0 0 0
−2ux − (ln 2)vx = vx
0 0
2. Find the values of the partial derivatives ux and vx at P .
0 1 + ln 2
ux (2, 1) = −
5 + ln 2
0 2
vx (2, 1) = .
5 + ln 2
3. Differentiate the system of equations with respect to y and evaluate them at
(x, y, u, v) = (2, 1, −1, 0).
We first differentiate the system with respect to y:
0
uy 0 0 0
+ uy v + uvy − 2yev − y 2 ev · vy + 1 = 0
x+u
0 0 0
2u · uy − (xv ln x) · vy = vy .

28
 Next, evaluating the above derived system at P , we obtain
0 0
uy − 2vy − 1 = 0
0 0 0
−2uy − (ln 2)vy = vy .

0 0
4. Find the values of the partial derivatives uy and vy at P .

0 1 + ln 2
uy (2, 1) =
5 + ln 2
0 2
vy (2, 1) = − .
5 + ln 2

5. Find an approximate value of u(1.99, 1.02).

Using the linear approximation, we have

10 0 1
u(1.99, 1.02) ≈ u(2, 1) − ux (2, 1) · + uy (2, 1) ·
100 50
1 + ln 2 1 1 + ln 2 1
= −1 + · + ·
5 + ln 2 100 5 + ln 2 50
3(1 + ln 2)
= −1 + .
100(5 + ln 2)

Question 5.10 (an additional question, not for the grade of HW#5) A firm’s
cost per unit of output is given by the function:
wr
c(w, r) =
w+r
where r denote the rental price of capital and w denote the wage rate. Answer the
following questions.

1. Show that the function c(·) is a homogeneous function of w and r.

Fix t > 0. Then,

(tw)(tr) t2 (wr) wr
c(tw, tr) = = =t = tc(w, r).
tw + tr t(w + r) w+r
So, c(·) is homogenous of degree one.

2. Show that the function c(·) satisfies Euler’s theorem for homogeneous functions.

We first compute the following:

0 r(w + r) − wr · 1 r2
c1 (w, r) = =
(w + r)2 (w + r)2
0 w(w + r) − wr · 1 w2
c2 (w, r) = = .
(w + r)2 (w + r)2

29
 Second, we compute

0 0 wr2 + w2 r wr(w + r) wr
wc1 (w, r) + rc2 (w, r) = 2
= 2
= = c(w, r).
(w + r) (w + r) w+r
0 0
This implies that wc1 (w, r) + rc2 (w, r) = kc(w, r) where k = 1. Thus, Euler’s
theorem holds.

3. Find all the second-order partial derivatives of c(·), and show that they are all
homogeneous of the same degree.
00 ∂ 2
c11 (w, r) =(r (w + r)−2 ) = −2r2 (w + r)−3
∂w
00 ∂
c22 (w, r) = (w2 (w + r)−2 ) = −2w2 (w + r)−3
∂r
00 00 ∂ 2
c12 (w, r) = c21 (w, r) = (r (w + r)−2 ) = 2r(w + r)−2 + r2 · (−2)(w + r)−3
∂r
= 2r(w + r)−3 (w + r − r) = 2wr(w + r)−3

Fix t > 0. We compute the following:

00 t2 00 00
c11 (tw, tr) = −2(tr)2 (tw + tr)−3 =c (w, r) = t−1 c11 (w, r)
t3 11
00 t2 00 00
c22 (tw, tr) = −2(tw)2 (tw + tr)−3 = 3 c22 (w, r) = t−1 c22 (w, r)
t
00 00 t2 0 00
c12 (w, r) = c21 (w, r) = 2(tw)(tr)(tw + tr)−3 = 3 c12 (w, r) = t−1 c12 (w, r)
t
This shows that all the second derivatives of c(·) are homogenous of degree -1.

6 Homework 6 (Due Date: Oct 17, 2019)

Question 6.1 (9 points) Find the linear approximation about (0, 0) for the follow-
ing:
The linear approximation to f (x, y) about (x0 , y 0 ) is
0 0
f (x, y) ≈ f (x0 , y 0 )+df1 (x0 , y0 ) = f (x0 , y 0 )+f1 (x0 , y 0 ) (x − x0 ) +f2 (x0 , y 0 ) (y − y 0 ) .
| {z } | {z }
dx dy

1. f (x, y) = (x + 1)5 (y + 1)6

We first compute
0 0
f1 (x, y) = 5(x + 1)4 (y + 1)6 and f2 (x, y) = 6(x + 1)5 (y + 1)5 .

We therefore have
f (x, y) ≈ 1 + 5x + 6y.

30
 √
2. f (x, y) = 1+x+y

We first compute
0 1 0 1
f1 (x, y) = (1 + x + y)−1/2 and f2 (x, y) = (1 + x + y)−1/2 .
2 2
We therefore have
1 1
f (x, y) ≈ 1 + x + y.
2 2
3. f (x, y) = ex ln(1 + y).

We first compute
0 0 ex
f1 (x, y) = ex ln(1 + y) and f2 (x, y) = .
1+y
We therefore have
f (x, y) ≈ 0 + 0 · x + y = y.

Question 6.2 (10 points) Find the tangent planes to the following surfaces at the
indicated points:

1. z = x2 + y 2 at (1, 2, 5)

We compute
0 0
∇f (x, y) = (f1 , f2 ) = (2x, 2y).
Thus, the tangent plane at (1, 2, 5) is
0 0
{(x, y, z) ∈ R3 | f1 (1, 2)(x − 1) + f2 (1, 2)(y − 2) = z − 5}
= {(x, y, z) ∈ R3 | 2(x − 1) + 4(y − 2) = z − 5}
= {(x, y, z) ∈ R3 | 2x + 4y − 5 = z}.

2. z = (y − x2 )(y − 2x2 ) at (1, 3, 2).

We compute
0
f1 (x, y) = −2x(y − 2x2 ) − 4x(y − x2 ) = 2x(4x2 − 3y)
0
f2 (x, y) = y − 2x2 + y − x2 = 2y − 3x2 .

Thus, the tangent plane at (1, 3, 2) is

0 0
{(x, y, z) ∈ R3 | f1 (1, 3)(x − 1) + f2 (1, 3)(y − 2) = z − 2}
= {(x, y, z) ∈ R3 | − 10(x − 1) + 3(y − 3) = z − 2}
= {(x, y, z) ∈ R3 | − 10x + 3y + 3 = z}.

Question 6.3 (18 points) Find the following integrals:

31
 (t3 + 2t − 3)dt
R
1.
t4
= + t2 − 3t + C.
4
(x − 1)2 dx
R
2.
x3
Z
= (x2 − 2x + 1)dx = − x2 + x + C.
3
R
3. (x − 1)(x + 2)dx

x3 x2
Z
= (x2 + x − 2)dx = + − 2x + C.
3 2

(x + 2)3 dx
R
4.

x4
Z
= (x3 + 6x2 + 12x + 8)dx = + 2x3 + 6x2 + 8x + C.
4

(e3x − e2x + ex )dx

R
5.
e3x e2x
= − + ex + C.
3 2
x3 −3x+4
R
6. x dx.

x3
Z  
2 4
= x −3+ dx = − 3x + 4 ln x + C.
x 3

Note that C stands for a constant of integration.

Question 6.4 (8 points) Prove

Z
2x ln(x2 + a2 )dx = (x2 + a2 ) ln(x2 + a2 ) − x2 + C.

Proof : Let F (x) = (x2 + a2 ) ln(x2 + a2 ) − x2 + C. So, we compute the following:

0 1
F (x) = 2x ln(x2 + a2 ) + (x2 + a2 ) · · 2x − 2x = 2x ln(x2 + a2 ).
x2 + a2

Since 2x ln(x2 +a2 ) is a continuous function, by the fundamental theorem of calculus,

we complete the proof. 

Question 6.5 (8 points) Find the general form of a function f whose second deriva-
00 0
tive f (x) is x2 . If we require in addition that f (0) = 1 and f (0) = −1, what is
f (x)?

32
 We first observe
x3
Z Z
0 00
f (x) = f (x)dx = x2 dx = + C1 ,
3
0
where C1 is a constant of integration. Since f (0) = −1, we obtain C1 = −1. So,
0
f (x) = x3 /3 − 1. Furthermore,
Z  3
x4
Z 
0 x
f (x) = f (x)dx = − 1 dx = − x + C2 ,
3 12

where C2 is a constant of integration. Since f (0) = 1, we have C2 = 1. Hence,

x4
f (x) = − x + 1.
12
Question 6.6 (16 points) Use integration by parts to evaluate the following:
Let us recall the following formula:
Z b Z b
0 0
f (x)g (x)dx = [f (x)g(x)]ba − f (x)g(x)dx.
a a

Z 1
1. x ln(x + 2)dx
−1
Z 2
2. x2x dx
0
Z 1
3. x2 ex dx
0
3 √
Z
4. x 1 + xdx.
0
0
1. Set f (x) = x and g(x) = (x + 2) ln(x + 2) − x. Note that g (x) = ln(x + 2). By
integration by parts, we have
Z 1 Z 1
1
x ln(x + 2)dx = [x{(x + 2) ln(x + 2) − x}]−1 − ((x + 2) ln(x + 2) − x) dx
−1 −1
Z 1 Z 1 Z 1
= 3 ln 3 − x ln(x + 2)dx − 2 ln(x + 2)dx + xdx.
−1 −1 −1

This implies
1 1
x2
Z 
2 x ln(x + 2)dx = 3 ln 3 − 2 [(x + 2) ln(x + 2) − x]1−1 +
−1 2 −1
= 3 ln 3 − 2 (3 ln 3 − 2) = −3 ln 3 + 4.

33
Hence,
1
−3 ln 3
Z
x ln(x + 2)dx = + 2.
−1 2
0
2. Set f (x) = x and g(x) = 2x / ln 2. Note that g (x) = 2x . By integration by parts,
we have
Z 2  x 2 Z 2 x  x 2
x 2 2 8 2
x2 dx = x − dx = −
0 ln 2 0 0 ln 2 ln 2 (ln 2)2 0
 
8 4 1 8 3
= − − = − .
ln 2 (ln 2)2 (ln 2)2 ln 2 (ln 2)2

3. Set f (x) = x2 and g(x) = ex . By integration by parts, we have

Z 1 Z 1 Z 1
2 x
 2 x 1 x
x e dx = x e 0 − 2 xe dx = e − 2 xex dx (∗)
0 0 0

Set h(x) = x and k(x) = ex . By integration by parts, we have

Z 1 Z 1
xex dx = [xex ]10 − ex dx = e − [ex ]10 = e − (e − 1) = 1 (∗∗)
0 0

Plugging (∗∗) into (∗), we obtain

Z 1
x2 ex dx = e − 2.
0
0
4. Set f (x) = x and g(x) = 2(1 + x)3/2 /3. Note that g (x) = (1 + x)1/2 . By
integration by parts, we have
√
Z 3 i3 2 Z 3
2h 3/2
x 1 + xdx = x(1 + x) − (1 + x)3/2 dx
0 3 0 3 0
2  4 h i3
= 3 · (22 )3/2 − 0 − (1 + x)5/2
3 15 0
2 4  2 5/2 
= (3 · 8) − (2 ) − 1
3 15
48 4 1 116
= − (32 − 1) = (240 − 124) = .
3 15 15 15
Question 6.7 (15 points) Calculate the following integrals.

Z 1
1. (x4 − x9 )(x5 − 1)12 dx
0
Z
ln x
2. √ dx
x
Z 4
dx
3. p √ .
0 1+ x

34
1. First, we observe the following:
Z 1 Z 1 Z 1
(x4 −x9 )(x5 −1)12 dx = − x4 (x5 −1)(x5 −1)12 dx = − x4 (x5 −1)13 dx. (∗)
0 0 0

Set u = x5 − 1. Then, we have du/dx = 5x4 so that

1 du
x4 = .
5 dx
Besides, x = 0 ⇔ u = −1 and x = 1 ⇔ u = 0. By integration by substitution, we
have Z 1 0
1 0 13 1 u14
Z 
1
x4 (x5 − 1)13 dx = u du = = − . (∗∗)
0 5 −1 5 14 −1 70
Plugging (∗∗) into (∗), we obtain
Z 1
1
(x4 − x9 )(x5 − 1)12 dx = .
0 70
√ √
2. Set u = x. Then, x = u2 and du/dx = 1/2 x. So,

dx
√ = 2du.
x

By integration by substitution, we have

√ √
Z Z Z
ln x 2
√ dx = 2 ln(u )du = 4 ln(u)du = 4(u ln u − u) + C = 4 x(ln( x) − 1) + C.
x
√ √
3. Set u = 1 + x. Then, we have du/dx = 1/2 x = 1/2(u − 1). Then,

dx = 2(u − 1)du.

Besides, x = 0 ⇔ u = 1 and x = 4 ⇔ u = 3. By integration by substitution, we have

Z 4 Z 3 Z 3
dx −1/2
p √ = 2 u (u − 1)du = 2 (u1/2 − u−1/2 )du
0 1+ x 1 1

2 3/2 3
Z 3 Z 3   h i3
1/2 −1/2
= 2 u du − 2 u du = 2 u − 2 2u1/2
1 1 3 1 1
2 3/2
    4 √ √
= 2· 3 − 1 − 2 · 2 31/2 − 1 = (3 3 − 1) − 4( 3 − 1)
3 3
= 8/3.

Question 6.8 (8 points) The value v0 of a new car depreciates continuously at the
annual rate of 10% – that is, v(t) = v0 e−δt where δ = 0.1 is the rate of depreciation.
How many years does it take for the car to lose 90% of its original value?

35
 What we want is to find t to satisfy
v0
= v0 e−δt .
10
This is equivalent to
1 = 10e−δt .
Taking the natural logarithm on both hand sides, we obtain
t
ln 1 = ln(10e−δt ) = ln 10 − δt = ln 10 − .
10
Since ln 1 = 0, we obtain
t = 10 ln(10) ≈ 23.02.

Question 6.9 (8 points) An author is to be paid royalties for publishing a book.

Two alternatives offers are made:
1. The author can be paid $21,000 immediately.
2. There can be 5 equal annual payments of $4,600, the first being paid at once.
Which of these offers will be more valuable if the interest rate is 6% per annum?
We compute the present discounted value of Option 2:
4600 4600 4600 4600
P DV2 = 4600 + + + +
1 + 0.06 (1 + 0.06)2 (1 + 0.06)3 (1 + 0.06)4
≈ 4600 + 4339.62 + 4093.98 + 3862.25 + 3643.63 = 20539.48
Clearly, the present discounted value of Option 1 is P DV1 = 21000. So, Option
1 is better than Option 2.

7 Homework 7 (Due Date: Oct 24, 2019)

Question 7.1 (10 points) Consider
2x2
f (x) = .
x4 + 1
Answer the following questions.
0
1. Derive f (x).
0 4x(x4 + 1) − 2x2 · 4x3 −4x(x4 − 1)
f (x) = =
(x4 + 1)2 (x4 + 1)2

2. Find the maximum value of f on [0, ∞).

Note that (x4 + 1) > 0. So, since f is defined over [0, ∞], we conclude that f
is strictly increasing if and only if x ∈ (0, 1). This implies that f achieves the
maximum at x = 1. Therefore, the maximum of f is 1.

36
Question 7.2 (9 points) Find possible local extreme (maximum or minimum) points
for f (x) = e3x − 6ex , where x ∈ R = (−∞, ∞).
We check the first-order condition of f :
0
f (x) = 3e3x − 6ex = 3ex (e2x − 2) = 0.

Since ex is always positive for any x ∈ R, the only stationary point is obtained as
follows: √
e2x − 2 ⇔ 2x = ln 2 ⇔ x∗ = ln( 2).
0 √ 0 √
Since f √ (x) < 0 if x < ln( 2) and f (x) > 0 if x > ln( 2), we conclude that
x∗ = ln( 2) is a local minimum point of f .

Question 7.3 (16 points) Consider

x2 + 1
f (x) = ,
x
which is defined over [1/2, 2]. Answer the following questions.

1. Show that f achieves the maximum and minimum over [1/2, 2].

Since f is differentiable, it must be continuous. In addition, [1/2, 2] is a

nonempty compact set. Once again, by the extreme-value theorem, there must
exist the maximum and minimum points of f on [1/2, 2].

2. Find all the stationary points of f over (1/2, 2).

0 2x · x − (x2 + 1) (x + 1)(x − 1)
f (x) = 2
= = 0.
x x2
So, x = −1, 1 are the stationary points of f but the domain of f is [1/2, 2], we
only consider x = 1 as the only candidate for extreme points on the interior of
[1/2, 2]. We obtain f (1) = 2.

3. Compute f (1/2) and f (2).

We compute:
1/4 + 1 5
f (1/2) = =
1/2 2
4+1 5
f (2) = = .
2 2

4. Find the maximum point and minimum point of f over [1/2, 2].

Given the results we obtained so far, x = 1 is the minimum point of f and

x = 1/2 and x = 2 are the maximum points of f on [1/2, 2] so that the
maximum of f is 5/2.

37
Question 7.4 (15 points) Let f be the function defined over [1, e3 ] as

f (x) = (ln x)3 − 2(ln x)2 + ln x.

Answer the following questions.

1. Compute f (e1/3 ), f (e2 ), and f (e3 ).

4
f (e1/3 ) = (1/3)3 − 2(1/3)2 + 1/3 =
27
f (e2 ) = 23 − 2 · 22 + 2 = 2
f (e3 ) = 33 − 2 · 32 + 3 = 27 − 18 + 3 = 12.

2. Find the zeros of f (x).

We obtain the following factorization:

f (x) = ln x(ln x − 1)2 .

So, x = 1, e are the zeros of f .

3. Find the extreme (local maximum or minimum) points of f .

0
We first compute f (x):
0 1 1 1
f (x) = 3(ln x)2 · − 4(ln x) · +
x x x
1
3(ln x)2 − 4 ln x + 1

=
x
1
= (3 ln x − 1)(ln x − 1).
x
First of all, for ln x to be well-defined, we must have x > 0. When we solve
0
f (x) = 0, we obtain

3 ln x = 1 ⇔ x = e1/3 > 1.
ln x = 1 ⇔ x = e.

Since f is defined over [1, e3 ],

> 0 if x > e or 0 < x < e1/3


0
f (x)
< 0 If e1/3 < x < e.

Thus, x = e1/3 is a local maximum and x = e is a local minimum point of f .

Since f is clearly continuous and [1, e3 ] is a compact set, by the extreme-value

theorem, there exists the global maximum and minimum points of f over [1, e3 ].

38
 We thus compare the following values:

f (1) = 0
4
f (e1/3 ) =
27
f (e) = 0
f (e3 ) = 12.

Hence, x = 1, e are the global minimum points of f and x = e3 is the global

maximum point of f over [1, e3 ].

Question 7.5 (12 points) The price a firm obtains for a commodity varies with
demand q according to the inverse demand function, p(q) = 18 − 0.006q. The total
cost function is C(q) = 0.004q 2 + 4q + 4500. Answer the following questions.

1. Derive the firm’s profit function π(q).

π(q) = p(q)q−C(q) = (18−0.006q)q−(0.004q 2 +4q+4500) = −0.01q 2 +14q−4500.

2. Find the value of q that maximizes profit.

We first derive the first-order condition:

0
π (q) = −0.02q + 14 = 0 ⇔ q = 700.

We claim that q = 700 is the profit maximizing output. To show this, we check
the second-order condition: for any q,
00
π (q) = −0.02 < 0.

So, q = 700 is the value of q that maximizes the profit.

dq p ∗
3. Let η = dp q be the price elasticity of demand. Find the particular value of q
at which η = −1.

Since p(q) is strictly decreasing, it is a one-to-one mapping. Thus, this guaran-

tees the existence of the inverse function q(p). By the inverse function theorem,
we have
dq 1 1
= 0 =− .
dp p (q) 0.006
So, it only remains to solve the following equation:
1 18 − 0.006q
− · = −1
0.006 q
After solving this out, we obtain q ∗∗ = 1500.

39
Question 7.6 (16 points) Consider the following function:
2 √
 
f (x) = 1 + x + 6.
x
Answer the following questions.
1. Find the domain of f and the intervals where f (x) is positive.

First of all, f is not well-defined when x = 0. Second, the inside the square-
root must be nonnegative, we must have x ≥ −6. Therefore, the domain of f
is
D = [−6, 0) ∪ (0, ∞).
In order for f (x) to be positive, we must have the following conditions:
2
x > −6 and 1 + > 0.
x
These two inequalities imply that x < −2. So, f (x) is positive if and only if
x ∈ (−6, −2) and x > 0.
2. Find all stationary points of f .
0
We derive f (x):
2√
 
0 2 1
f (x) = − 2 x + 6 + 1 + · √
x x 2 x+6
   
1 4 2 1 4 24 2
= √ − 2 (x + 6) + 1 + = √ − − 2 +1+
2 x+6 x x 2 x+6 x x x
1 1
x2 − 2x − 24 = 2 √


= √ (x − 6)(x + 4).
2x2 x + 6 2x x + 6
Thus, x = −4, 6 are the stationary points of f .
3. Examine the limit of f (x) when x → 0 from both the left and the right. In
addition, examine the limit of f (x) when x → +∞.
√
When x → 0 regardless from the left or right, x + 6 → 6. So, everything
depends on (1+2/x). When x → 0 from the left, f (x) converges to −∞. When
x → 0 from the right, f (x) converges
√ to ∞. When x → +∞, (1 + 2/x) → 1.
So, now everything depends on x + 6. So, x → ∞, f (x) → +∞.
4. Are the stationary points you found are the global maximum or minimum
points? If so, verify it. If not, argue why.

We compute the following:

√
2
f (−4) =
2√
8 3
f (6) = .
3

40
 By the previous question, we know that limx→0− f (x) = −∞ and limx→0+ f (x) =
∞. Thus, the stationary points are neither global maximum nor minimum
points.
Question 7.7 (12 points) Let f be the function defined over (−∞, ∞) as
3
f (x) = x3 + x2 − 6x + 10.
2
Answer the following questions.
1. Find the stationary points of f and determine the intervals where f is nonde-
creasing.
0
We first compute f (x):
0
f (x) = 3x2 + 3x − 6 = 3(x2 + x − 2) = 3(x + 2)(x − 1).
0
From the first-order condition f (x) = 0, x = −2, 1 are the stationary points.
0
Since f is nondecreasing if and only if f ≥ 0, we conclude that f (x) is nonde-
creasing if and only if either x ∈ (−∞, −2] or [1, ∞).
2. Find the inflection point for f .
00
We compute f (x):
00
f (x) = 6x + 3 = 3(2x + 1).
00 00
Since f (x) ≥ 0 if x ≥ −1/2 and f (x) ≤ 0 if x ≤ −1/2, x = −1/2 is the
inflection point for f .
Question 7.8 (10 points) A firm’s production function is f (L) = 12L2 − L3 /12,
where L denotes the number of workers, with L ∈ [0, 200]. Answer the following
questions.
1. What size of the work force (L∗ ) maximizes output q = f (L)?
0
We first compute f (L):
0 L2 L
f (L) = 24L − = (96 − L)
4 4
0
From the first-order condition f (L) = 0, we claim L∗ = 96 ∈ [0, 200]. Next,
00
we compute f (L):
00 L 1
f (L) = 24 − = (48 − L).
2 2
00 ∗ ∗
Therefore, we have f (L ) < 0 so that L = 96 is the local maximum of f . We
compute the following:
f (0) = 0
f (96) = 36, 864
f (200) = −186, 666

41
 Since f is clearly continuous and [0, 200] is a nonempty compact set, f has the
maximum and minimum point of f over [0, 200]. So, L∗ = 96 is the maximum
point of f .

2. What size of the work force (L∗∗ ) maximizes output per worker, f (L)/L?
0
Let F (L) = f (L)/L = 12L − L2 /12. We compute F (L):

0 L 1
F (L) = 12 − = (72 − L).
6 6
00
We claim L∗∗ = 72 ∈ [0, 200]. We next compute f (L) = −1/6 for any L. So,
L∗∗ = 72 is the global maximum of F .

8 Homework 8 (Due Date: Oct 31, 2019)

Question 8.1 (10 points) Investigate the definiteness of the following quadratic
forms:

1. Q(x1 , x2 ) = 5x21 − 2x1 x2 + x22

Q(x1 , x2 ) = 5x21 − x1 x2 − x2 x1 + x22

  
5 −1 x1 0
= (x1 , x2 ) = x Ax.
−1 1 x2

Since a11 = 5 > 0 and |A| = 5 − 1 = 4 > 0, Q(x1 , x2 ) is positive definite.

2. Q(x1 , x2 ) = −4x21 + 12x1 x2 − 9x22 .

Q(x1 , x2 ) = −4x21 + 6x1 x2 + 6x2 x1 − 9x22

  
−4 6 x1
= (x1 , x2 )
6 −9 x2
  
a11 a12 x1
= (x1 , x2 ) .
a21 a22 x2

Since a11 = −4 < 0, a22 = −9 < 0, and a11 a22 − a12 a21 = 0, Q(x1 , x2 ) is
negative semidefinite.

Question 8.2 (14 points) The function f defined for all (x, y) ∈ R2 is give as
follows:
f (x, y) = x2 + y 2 − 6x + 8y + 35.
Answer the following questions.

1. Find a local minimum point of f .

42
 First, we obtain the FOC:
0
f1 (x, y) = 2x − 6 = 0,
0
f2 (x, y) = 2y + 8 = 0.
Thus, we obtain the unique stationary point (x∗ , y ∗ ) = (3, −4). Next, we
compute the second derivatives of f :
00
f11 (x, y) = 2,
00
f22 (x, y) = 2,
00 00
f12 (x, y) = f21 (x, y) = 0.
Hence, we obtain the Hessian matrix as follows:
 
2 0
H(x, y) = .
0 2
00
Since f11 = 2 > 0 and |H| = 4 > 0, H(x, y) is positive definite in all (x, y).
Thus, (x∗ , y ∗ ) = (3, −4) is the local minimum point of f .
2. Show that f (x, y) can be written in the form f (x, y) = (x − 3)2 + (y + 4)2 + 10.
Explain why this shows that the local minimum point you found in the previous
question is indeed the (global) minimum point of f .
(x − 3)2 + (y + 4)2 + 10 = x2 − 6x + 9 + y 2 + 8y + 16 + 10
= x2 + y 2 − 6x + 8y + 35 = f (x, y).
Since H(x, y) is positive definite not only at (x∗ , y ∗ ) = (3, −4) but also all
other points, (x∗ , y ∗ ) = (3, −4) is the global minimum point of f .
Question 8.3 (12 points) The demands for a monopolist’s two products are de-
termined by the equations:
px = 25 − x,
py = 24 − 2y,
where px and py are the prices per unit of the two goods, respectively, and x and y
are the corresponding quantities. The costs of producing and selling x units of the
first good and y units of the other are given as
C(x, y) = 3x2 + 3xy + y 2 .
Answer the following questions.
1. Find the monopolist’s profit π(x, y) from selling x units of the first good and
y units of the other good.
π(x, y) = px x + py y − C(x, y)
= (25 − x)x + (24 − 2y)y − 3x2 − 3xy − y 2
= −4x2 − 3y 2 − 3xy + 25x + 24y.

43
 2. Find the local maximum values of x and y for π(x, y).

We first check the FOC:

0
π1 (x, y) = −8x − 3y + 25 = 0,
0
π2 (x, y) = −6y − 3x + 24 = 0.

Solving the above system of equations, we obtain the unique stationary point
(x∗ , y ∗ ) = (2, 3). Second, we compute the second derivatives of π:
00
π11 (x, y) = −8,
00
π22 (x, y) = −6,
00 00
π12 (x, y) = π21 (x, y) = −3.

Then, we obtain the Hessian matrix:

 
−8 −3
H(x, y) = .
−3 −6
00
Since π11 = −8 < 0 and |H| = 48 − 9 = 39 > 0, H(x, y) is negative definite at
all (x, y). Thus, (x∗ , y ∗ ) = (2, 3) is the local maximum point of π.
3. Verify that the local maximum values of x and y are in fact the (global) max-
imum point of π(x, y).

Since H(x, y) is negative definite at not only (x∗ , y ∗ ) but also all other points,
(x∗ , y ∗ ) = (2, 3) is the global maximum point of π.

Question 8.4 (12 points) Consider the following function:

f (x, y) = x2 + 2xy 2 + 2y 2 .

Answer the following questions.

1. Find all the first- and second-order derivatives of f .

0
f1 = 2x + 2y 2 ,
0
f2 = 4xy + 4y,
00
f11 = 2,
00
f22 = 4x + 4,
00 00
f12 = f21 = 4y.

2. Find all the stationary points of f .

To find the stationary points, we must solve the following system of equations:
0
f1 = 2x + 2y 2 = 2(x + y 2 ) = 0,
0
f2 = 4xy + 4y = 4y(x + 1) = 0.

44
 From the second equation, we must have two cases: (1) y = 0 and (2) x = −1.

0
Case (1): y = 0 and f1 = 0 imply that x = 0. So, (0, 0) is a stationary point.

0
Case (2): x = −1 and f1 = 0 imply that y 2 = 1. So, (−1, −1) and (−1, 1) are
stationary points of f .

So, there are three stationary points: (0, 0), (−1, −1), and (−1, 1).

3. Classify these stationary points of f in the sense that you clarify whether it is
a local maximum, local minimum, or saddle point.

We compute the Hessian matrix:

 
2 4y
H(x, y) =
4y 4x + 4

At each stationary point, we have

 
2 0
H(0, 0) = ,
0 4
 
2 −4
H(−1, −1) = ,
−4 0
 
2 4
H(−1, 1) = .
4 0

00
Since f11 (0, 0) = 2 > 0 and |H(0, 0)| = 8 > 0, H(0, 0) is positive definite.
Thus, (0, 0) is a local minimum point of f .
Since |H(−1, −1)| = |H(−1, 1)| = −16 < 0, both (−1, −1) and (0, 0) are saddle
points.

Question 8.5 (12 points) Consider the following three functions: (1) f (x, y) =
−x4 − y 4 ; (2) f (x, y) = x4 + y 4 ; and (3) f (x, y) = x3 + y 3 . Answer the following
questions.

1. Prove that the origin (0, 0) is a stationary point for each one of these functions.
0 0
(1) We compute the following: f1 = −4x3 and f2 = −4y 3 . So, (0, 0) is the
unique stationary point of f (x, y) = x4 − y 4 .
0 0
(2) We compute the following: f1 = 4x3 = 0 and f2 = 4y 3 = 0. So, (0, 0) is
the unique stationary point of f (x, y) = x4 + y 4 .
0 0
(3) We compute the following: f1 = 3x2 and f2 = 3y 2 . So, (0, 0) is the unique
stationary point of f (x, y) = x3 + y 3 .

45
 2. For each one of these functions, derive the Hessian matrix H.

−12x2
 
(1) 0
H (x, y) = ,
0 −12y 2
12x2
 
(2) 0
H (x, y) = ,
0 12y 2
 
(3) 6x 0
H (x, y) = .
0 6y

3. Prove that |H| = 0 evaluated at (x, y) = (0, 0) for each one of these functions.

This is trivial because H (1) , H (2) , and H (3) all are zero matrices at (x, y) =
(0, 0).

4. By studying the functions directly, prove that the origin (0, 0) is respectively
a (global or local) maximum point for (1), a (global or local) minimum point
for (2), and a saddle point for (3).

In all three cases, we observe f (0, 0) = 0.

(1) For any (x, y) 6= (0, 0), we obtain f (x, y) = −x4 − y 4 < 0. So, (0, 0) is the
global maximum point.

(2) For any (x, y) 6= (0, 0), we obtain f (x, y) = x4 + y 4 > 0. So, (0, 0) the
global minimum point.

(3) Let ε > 0. Then, we have f (ε, ε) = 2ε3 > 0 and f (−ε, −ε) = −2ε3 < 0, no
matter how small ε can be. So, (0, 0) is the saddle point.

Question 8.6 (12 points) Suppose a monopolist faces two types of markets, called
Market 1 and Market 2, respectively. These two markets are characterized their
respective (inverse) demand function as follows:

p1 = 200 − 2q1 ,
p2 = 180 − 4q2 ,

where p1 and p2 are the prices of the good at Market 1 and Market 2, respectively,
and q1 and q2 are the corresponding quantities. The cost function is

C(q1 , q2 ) = 20(q1 + q2 ).

Answer the following questions.

1. How much should be sold in the two markets to maximize total profit? What
are the corresponding prices?

46
 We set up the profit function of this firm:
π(q1 , q2 ) = p1 q1 + p2 q2 − 20(q1 + q2 )
= (200 − 2q1 )q1 + (180 − 4q2 )q2 − 20q1 − 20q2
= −2q12 − 4q22 + 180q1 + 160q2
= −2(q1 − 45)2 − 4 (q2 − 20)2 + 4050 + 1600
= −2(q1 − 45)2 − 4 (q2 − 20)2 + 5650.
The profit maximizing input vector is (q1 , q2 ) = (45, 20). Therefore, the corre-
sponding prices are (p1 , p2 ) = (200−90, 180−80) = (110, 100). The maximized
total profit is 5, 650.
2. How much profit is lost if it becomes illegal to discriminate prices?

First, we must have p1 = p2 so that 200 − 2q1 = 180 − 4q2 . This leads to the
relation: q1 = 2q2 + 10. With this relation between q1 and q2 , we express the
profit function of the firm as a function of q2 only:
π(q2 ) = −2(2q2 + 10)2 − 4q22 + 180(2q2 + 10) + 160q2
= −2(4q22 + 40q2 + 100) − 4q22 + 360q2 + 1800 + 160q2
= −12q22 + 440q2 + 1600
55 2
 
12100
= −12 q2 − + 1600 +
3 3
55 2 16900
 
= −12 q2 − +
3 3
 2
55 1
= −12 q2 − + 5633 + .
3 3
Therefore, the lost profit is (16 + 2/3).

Question 8.7 (18 points) Let

f (x, y) = 4x − 2x2 − 2y 2 ,
and
D = (x, y) ∈ R2 | x2 + y 2 ≤ 25 .


We consider the following maximization problem:

max f (x, y) subject to (x, y) ∈ D.
(x,y)∈R2

Answer the following questions.

1. Show that D is a bounded set.

p
It is easy to see that for any point (x, y) ∈ D, we have x2 + y 2 ≤ 5. So, the
closed ball around (0, 0) with radius 5 contains D. That is, D is bounded.

47
2. Show that f is a continuous function.

Observe that f consists of the sum of polynomials, all of which, 4x, −2x2 , and
−2y 2 are continuous. So, f is also continuous.

3. Show that D is a closed set.

Let g(x, y) = x2 + y 2 . Since g(x, y) is a polynomial function, it is a continuous

function. Then, we can define

D = {(x, y) ∈ R2 | g(x, y) ≤ 25}

This shows that D is a closed set.

4. Find the stationary points of f .

We solve the following system of equations:

0
f1 = 4 − 4x = 0,
0
f2 = −4y = 0.

Therefore, (x∗ , y ∗ ) = (1, 0) is the unique stationary point of f .

5. Argue why the maximization problem has the maximum and minimum points
of f over D.

Since D is both closed and bounded, it is also compact. Moreover, we know

that f is continuous. By the extreme-value theorem, this optmization problem
has the maximum and minimum points of f over D.

6. Find the maximum and minimum points of f over D.

First, we compute f (x∗ , y ∗ ) = f (1, 0) = 2. Next, we consider the boundary

points of D where x2 + y 2 = 25. So, we have y 2 = 25 − x2 . Plugging this into
f (x, y), we obtain

f (x, 25 − x2 ) = 4x − 2x2 − 2(25 − x2 ) = 4x − 50,

where −5 ≤ x ≤ 5. Taking into account that −5 ≤ x ≤ 5, (5, 0) achieves

the maximum of f over the boundary of D so that its maximum value is −30.
Similarly, (−5, 0) achieves the minimum of f over the boundary of D so that
the minimum value is −70.
Comparing these extreme values on the interior and boundary of D, we can
conclude that (1, 0) is the maximum point and (−5, 0) is the minimum point
of f over D.

48
Question 8.8 (10 points) Using the Lagrangian method, find the only solution
candidate to the following opmization problem with an equality constraint:

max 10x1/2 y 1/3 subject to 2x + 4y = m,

(x,y)∈R2

where m > 0 is a constant.

Set up the Lagrangian:

L = 10x1/2 y 1/3 − λ(2x + 4y − m).

We derive the FOCs of the Lagrangian:

0
L1 = 5x−1/2 y 1/3 − 2λ = 0,
0 10 1/2 −2/3
L2 = x y − 4λ = 0,
3
0
L3 = −(2x + 4y − m) = 0.

From the first two equations, we obtain x = 3y. Plugging x = 3y into the equality
constraint, we obtain y = m/10. So, we also have x = 3m/10. Thus, the candidate
for the solution to the optimization problem is (x∗ , y ∗ ) = (3m/10, m/10).

9 Homework 9 (Due Date: Nov 14, 2019)

Question 9.1 (20 points) Solve the following optimization problem:

max 3xy subject to x2 + y 2 = 8.

(x,y)∈R2

We set up the Lagrangian:

L(x, y, λ) = 3xy − λ(x2 + y 2 − 8).

The FOCs of the Lagrangian and the constraint are:

0
L1 = 3y − 2λx = 0, (1)
0
L2 = 3x − 2λy = 0, (2)
x2 + y 2 = 8. (3)

From (1) and (2), we obtain x2 = y 2 . Plugging this relation into the con-
straint, we obtain x2 = 4, or x = ±2. So, there are four solution candidates:
(x, y) = (2, 2), (2, −2), (−2, 2), (−2, −2). When plugging these values into (1), we
also obtain the associated value of λ. In sum, the solution candidates (x∗ , y ∗ , λ∗ ) are
(i) (2, 2, 3/2); (ii) (2, −2, −3/2); (iii) (−2, 2, −3/2); and (iv) (−2, −2, 3/2).
Since the objective function is 3xy, we can ignore (ii) and (iii), the solution
candidates are reduced to (x∗ , y ∗ , λ∗ ) = (2, 2, 3/2), (−2, −2, 3/2). Define g(x, y) =
x2 + y 2 . We compute
∇g(x, y) = (2x, 2y)

49
Note that the constraint qualification is satisfied unless (x, y) = (0, 0). Moreover,
(0, 0) is not a feasible point because 02 + 02 = 0 6= 8. So, the constraint qualification
is indeed satisfied for this problem.
Define D = {(x, y) ∈ R2 |g(x, y) = 8} as the set of feasible points in this problem.
We claim that D is compact. It is easy to see that g(x, y) = x2 + y 2 is a continuous
function so that D is closed. Moreover, D is bounded because any (x, y) ∈ D satisfies
x2 + y 2 ≤ 8. So, D is compact.
Since the objective function f (x, y) = 3xy is continuous and D is compact, by
the extreme value theorem, the constrained maximization problem has a solution.
We thus conclude that the solutions are (x∗ , y ∗ ) = (2, 2), (−2, −2).

Question 9.2 (20 points) Consider the following optimization problem:

√ √
max 2x + 3y subject to x + y = 5.
(x,y)∈R2

Answer the following questions.

1. Using the Lagrangian method, solve the optimization problem.

We set up the Lagrangian:

√ √
L(x, y, λ) = 2x + 3y − λ[ x + y − 5].

The FOCs of the Lagrangian and the constraint are:

0 λ
L1 = 2 − √ = 0, (1)
2 x
0 λ
L2 = 3 − √ = 0, (2)
2 y
√ √
x + y = 5. (3)
√ √
We obtain x = 3 y/2 from (1) and (2); y = 4 from (3); and x = 9 from
√ √
x = 3 y/2 and y = 4. Plugging (x, y) = (9, 4) into (1), we obtain λ = 12.
Therefore, (x∗ , y ∗ , λ∗ ) = (9, 4, 12) is the unique solution candidate that follows
the Lagrangian method.
2. Show that the Lagrangian method suggests the wrong answer.

At (x, y, λ) = (9, 4, 12), the value of the objective function is 30. Let us pro-
pose (x̂, ŷ) = (0, 25). Clearly, (x̂, ŷ) satisfies the constraint. Moreover, the
corresponding value of the objective function is 75. This implies that one can
find a better solution than the one we found using the Lagrangian method.
3. Find the solution to the optimization problem.

First, for the constraint function to be well defined, the domain of the objective
√ √
function must be x ≥ 0 and y ≥ 0. By the constraint, we obtain y = 5 − x.

50
 Since the objective function f (x, y) is 2x + 3y, we must set x = 0 and y = 25
to maximize it. So, we conclude that the solution is (x, y) = (0, 25).

4. Explain why the Lagrangian method does not work.

√ √
Define g(x, y) = x + y. We compute
 
0 0 1 1
∇g(x, y) = (g1 (x, y), g2 (x, y)) = √ , √ .
2 x 2 y

Clearly, we do not satisfy the constraint qualification at a feasible point (x, y) =

√
(0, 25) because 1/2 x is not well-defined at x = 0. So, the reason why the
Lagrangian method does not work is that the constraint qualification fails.

Question 9.3 (20 points) A firm uses two inputs x and y to produce its single
1 3
output. It has a production function: 4x 4 y 4 . This firm is constrained to use input
x = 1000. Suppose that the price of output p is 1 and input prices are also 1. Answer
the following questions.

1. Set up the Lagrangian function for the profit maximization problem.

We first define the profit function of the firm:

π(x, y) = 4x1/4 y 3/4 − x − y.

Therefore, the firm’s optmization problem becomes

max 4x1/4 y 3/4 − x − y subject to x = 1000.

x,y

We set up the Lagrangian:

L(x, y, λ) = 4x1/4 y 3/4 − x − y − λ(x − 1000).

2. Find (x∗ , y ∗ , λ∗ ) that satisfies the first-order conditions of the Lagrangian func-
tion.

The FOCs of the Lagrangian and the constraint are:

0
L1 = x−3/4 y 3/4 − 1 − λ = 0, (1)
0 1/4 −1/4
L2 = 3x y − 1 = 0, (2)
x = 1000. (3)

We obtain y = 34 x = 81x from (2) and y = 81000 from (3). Plugging

(x, y) = (1000, 81000) into (1), we obtain λ = 26. The solution is (x∗ , y ∗ , λ∗ ) =
(1000, 81000, 26).

51
 3. How much is the firm willing to pay to be allowed to use ε more units of input
x, for a small ε > 0?

Since the Lagrange multiplier is interpreted as the price the firm is willing to
pay for its constraint to be relaxed in “one” unit. So, if the firm is allowed to
use ε more units of input x, it is willing to pay 26ε.

Question 9.4 (20 √ points) Suppose that a firm produces a good with the production
function f (L, K) = LK. The firm wants to produce q, given that w and r are input
prices and p is output price. The firm thus minimizes the cost:
√
min C = wL + rK s.t. LK = q.
L,K

Answer the following questions.

1. Using the Lagrangian method, find the optimal choices, L∗ (w, r, q) and K ∗ (w, r, q).

We set up the Lagrangian:

L(L, K, λ) = wL + rK − λ(L1/2 K 1/2 − q).

The FOCs and the constraint of the Lagrangian are:

0 λ −1/2 1/2
L1 = w − L K = 0, (1)
2
0 λ
L2 = r − L1/2 K −1/2 = 0, (2)
2
L1/2 K 1/2 = q. (3)

From (1) and (2), we obtain

w L−1/2 K 1/2 w K
= 1/2 −1/2 ⇒ = .
r L K r L
w
Thus, K = r L. Plugging this into (3), we obtain
 w −1/2
L=q .
r
w
From K = r L, we further obtain
 w 1/2
K=q .
r
Hence, the optimal choices are
 w −1/2
L∗ (w, r, q) = q ,
r
 w 1/2
K ∗ (w, r, q) = q .
r

52
 Plugging the above optimal choices into (1), we obtain
 w −1/2
λ∗ = 2wq ,
r
which is positive.
0 0
We compute (g1 , g2 ) = (L−1/2 K 1/2 /2, L1/2 K −1/2 /2). Thus, we can confirm
0 0
(g1 (L∗ , K ∗ ), g2 (L∗ , K ∗ )) 6= (0, 0) so that the constraint qualification is satisfied.

2. Verify that the optimal choices derived in the previous question are the global
minimum point.

It suffices to show that the Lagrangian function given λ∗ , the Hessian

√ matrix
∗
associated with L(L, K, λ ) is positive semidefinite. Let g(L, K) = LK.

We compute all the first and second derivatives of g:

0 1 −1/2 1/2
g1 = L K ,
2
0 1 1/2 −1/2
g2 = L K ,
2
00 1
g11 = − L−3/2 K 1/2 ,
4
00 1 1/2 −3/2
g22 = − L K ,
4
00 00 1 −1/2 −1/2
g12 = g21 = L K .
4

Since L(L, K, λ∗ ) = wL + wK − λ∗ (g(L, K) − q), We form the Hessian matrix

of L:
 00 00 00 00   λ∗ −3/2 1/2 ∗
−λ∗ g11 −λ∗ g12 − λ4 L−1/2 K −1/2
  
L11 L12 4 L K
H= 00 00 = 00 00 = ∗ λ∗ 1/2 −3/2 ,
L21 L22 −λ∗ g21 −λ∗ g22 − λ4 L−1/2 K 1/2 4 L K

w −1/2
where λ∗ = 2wq


r > 0. We can check that

00 λ∗ −3/2 1/2
L11 = L K >0
4
00 λ∗ 1/2 −3/2
L22 = L K >0
4
00 00 00 00 (λ∗ )2 −1 −1 (λ∗ )2 −1 −1
L11 L12 − L12 L21 = L K − L K = 0.
16 16
Thus, H is positive semidefinite.

3. Define the minimal cost C ∗ = wL∗ (w, r, q) + rK ∗ (w, r, q). Find

∂C ∗ ∂C ∗ ∂C ∗
, and , when w = r = q = 1.
∂w ∂r ∂q

53
  w −1/2  w 1/2
C ∗ = wq + rq = 2qw1/2 r1/2 .
r r
We first compute the following:
∂C ∗
= qw−1/2 r1/2 = L∗ (w, r, q),
∂w
∂C ∗
= qw1/2 r−1/2 = K ∗ (w, r, q),
∂r
∂C ∗
= 2w1/2 r1/2 .
∂q
Hence,
∂C ∗ 
= 1,
∂w (w,r,q)=(1,1,1)

∂C ∗ 
= 1,
∂r (w,r,q)=(1,1,1)

∂C ∗ 
= 2.
∂q (w,r,q)=(1,1,1)


4. Assuming that the cost function C ∗ is twice continuously differentiable, show

that
∂K ∗ ∂L∗
= .
∂w ∂r
From the previous question, we derive the following relation:
∂C ∗
= L∗ (w, r, q),
∂w
∂C ∗
= K ∗ (w, r, q).
∂r
Since C ∗ is twice continuously differentiable, we further obtain the following
relationship:
∂2C ∗ ∂L∗ (w, r, q)
= ,
∂r∂w ∂r
∂2C ∗ ∂K ∗ (w, r, q)
= .
∂w∂r ∂w
Once again, since C ∗ is twice continuously differentiable, by Young’s Theorem,
we conclude
∂2C ∗ ∂2C ∗ ∂L∗ (w, q, r) ∂K ∗ (w, r, q)
= ⇔ = .
∂r∂w ∂w∂r ∂r ∂w
Question 9.5 (20 points) Consider the following utility maximization problem for
choosing two goods (x, y) subject to the budget constraint:
√
max u(x, y) = x + y subject to x + 4y = 100,
x≥0,y≥0

54
where the price of good x is 1 and that of good y is 4 and the consumer’s income is
100. Answer the following questions.

1. Using the Lagrangian method, find the quantities demanded for the two goods.

We set up the Lagrangian:

L(x, y, λ) = x1/2 + y − λ(x + 4y − 100).

The FOCs of the Lagrangian and the constraint are:

0 1
L1 = x−1/2 − λ = 0, (1)
2
0
L2 = 1 − 4λ = 0, (2)
x + 4y = 100. (3)

We have λ = 1/4 from (2); x = 4 from λ = 1/4 and (1); and y = 24 from (3).
Therefore, the optimal choice is (x∗ , y ∗ ) = (4, 24).

2. Suppose income increases from 100 to 101. What is the exact increase in the
optimal value of u(x, y) based on the Lagrangian method? Compare with the
value found in the previous question for the Lagrangian multiplier.

First, we compute the value of u∗ = u(x∗ , y ∗ ):

√
u∗ = u(4, 24) = 4 + 24 = 2 + 24 = 26.

Next, we set up the Lagrangian for the income of 101:

L(x, y, λ) = x1/2 + y − λ(x + 4y − 101).

The FOCs of the Lagrangian and the constraint are:

0 1
L1 = x−1/2 − λ = 0, (1)
2
0
L2 = 1 − 4λ = 0, (2)
x + 4y = 101. (3)

We have λ = 1/4 from (2); x = 4 from λ = 1/4 and (1); and y = 97/4 =
24 + 1/4 from (3). Note that the constraint qualification is satisfied because
0 0
(g1 , g2 ) = (1, 4). So, the optimal choice becomes (x∗∗ , y ∗∗ ) = (4, 97/4). Third,
we compute the value of u∗∗ = u(x∗∗ , y ∗∗ ):

u∗∗ = 2 + 97/4 = 26 + 1/4.

Finally, we obtain u∗∗ − u∗ = 1/4, which is the same as the value of the
Lagrange multiplier we obtained.

55
3. Suppose we change the budget constraint to px + qy = m, but keep the same
utility function. We assume that p, q, and m are positive constants. Using
the Lagrangian method, derive the quantities demanded of the two goods if
m > q 2 /4p.

We set up the Lagrangian:

L(x, y, λ) = x1/2 + y − λ(px + qy − m).

The FOCs of the Lagrangian and the constraint are:

0 1
L1 = x−1/2 − λp = 0, (1)
2
0
L2 = 1 − λq = 0, (2)
px + qy = m. (3)

From (2), we have λ = 1/q. Plugging this into (1), we obtain x = q 2 /4p2 .
Furthermore, plugging this into (3), we obtain y = (m − q 2 /4p)/q, which is
positive because m > q 2 /4p.

4. Show that the solution candidate you found using the Lagrangian method in
the previous question is indeed the solution to the original constrained maxi-
mization problem.

Define g(x, y) = px + qy. Then,

∇g(x, y) = (p, q).

Note that the constraint qualification is satisfied because p, q > 0 so that

∇g(x, y) 6= (0, 0). Define

D1 = {(x, y) ∈ R2 | x ≥ 0}
D2 = {(x, y) ∈ R2 | y ≥ 0}
D3 = {(x, y) ∈ R2 |x + 4y = 100}

Define also D = D1 ∩ D2 ∩ D3 as the set of feasible points. Each D1 , D2 , and

D3 are clearly closed sets. Since any intersection of finitely many closed sets
is closed, D is also closed. Fix (x, y) ∈ D. Since y ≥ 0 in D, we have x ≤ 100.
Similarly, since x ≥ 0, we have y ≤ 25. This implies that for any (x, y) ∈ D,
we have 0 ≤ x ≤ 100 and 0 ≤ y ≤ 25.
Define B = {(x, y) ∈ R2 | x2 + y 2 ≤ 10625}. Fix (x, y) ∈ D. Then, we confirm

x2 + y 2 ≤ (100)2 + (25)2 = 10625.

This implies that (x, y) ∈ B as well. Thus, D ⊆ B. We confirm that D is

bounded. In sum, D is a nonempty compact set.

56
 √
Define f (x, y) = f 1 (x, y) + f 2 (x, y), where f 1 (x, y) = x and f 2 (x, y) = y.
Since both f 1 (x, y) and f 2 (x, y) are continuous functions, the sum of them
√
is also continuous, i.e., f (x, y) = x + y is a continuous function. By the
extreme-value theorem, the original constrained maximization problem has a
solution. Since the qualification constraint is also satisfied, the Lagrangian
method generates the solution if the solution exists. The preceding arguments
establish what we wanted.

10 Homework 10 (Due Date: Nov 21, 2019)

Question 10.1 (25 points) A consumer’s demands x, y, z for three goods are cho-
sen to maximize the utility function
√
U (x, y, z) = x + y − 1/z,

subject to the budget constraint px+qy+rz = m, where x ≥ 0, y > 0, z > 0; p, q, r > 0

√
and m ≥ pr + p2 /4q. Answer the following questions.

1. Write down the first-order conditions for a global maximum in the constrained
optimization problem.

We set up the Lagrangian:

L(x, y, z, λ) = x + y 1/2 − z −1 − λ(px + qy + rz − m).

The FOCs of the Lagrangian and the constraint are:

0
L1 = 1 − λp = 0, (1)
0 1
L2 = y −1/2 − λq = 0, (2)
2
0 1
L3 = 2 − λr = 0, (3)
z
px + qy + rz = m. (4)

2. Show that the constrained optimization problem has a solution.

Define
D = (x, y, z) ∈ R3 | px + qy + rz = m ,


as the set of feasible points. Since each px, qy, and rz is a continuous function,
the sum of these continuous functions is a continuous function of x, y, z. Thus,
D is a closed set. Choose k = max{m/p, m/q, m/r}. Since x > 0, y ≥ 0, z ≥ 0,
the budget constraint implies that x ≤ m/p, y ≤ m/q, and z ≤ m/r. Define
B = {(x, y, z) ∈ R3 | x2 + y 2 + z 2 ≤ 3k 2 }. We claim that D ⊆ B. Fix
(x, y, z) ∈ D arbitrarily. Then,
m2 m2 m2
x2 + y 2 + z 2 ≤ + 2 + 2 ≤ 3k 2 .
p2 q r

57
 Hence, D is a bounded set. The preceding argument concludes that D is a
compact set. The objective function U (x, y, z) = x + y 1/2 − z −1 is the sum
of x, y −1/2 , and z −1 , each of which is a continuous function. Thus, U (·) is
also a continuous function. By the extreme value theorem, the constrained
optimization problem has a solution.
3. Show that the first-order conditions you found above are sufficient for finding
a global maximum point.

Define g(x, y, z) = px + qy + rz = m. Compute

∇g(x, y, z) = (∂g/∂x, ∂g/∂y, ∂g/∂z) = (p, q, z) 6= (0, 0, 0).
Thus, the constraint qualification holds. By the previous question, we have
shown that the constrained optimization problem has a solution. Therefore,
the first-order condition for the Lagrangian is sufficient for finding a solution.
4. Find the utility-maximizing demands for all three goods as functions of the
four variables (p, q, r, m).

From (1), we obtain p λ = 1/p. Plugging this into (2) and (3), we obtain
2 2
y = p /4q and z = p/r, respectively. Finally, substituting these into (4),
we obtain
p2 √
 
1
x= m− − pr ,
p 4q
√
which is nonnegative because m ≥ pr + p2 /4q. This is consistent with x ≥ 0.
Note that the constraint qualification is satisfied unless (p, q, r) = (0, 0, 0),
√
which cannot be the case given the assumption we have. If m > pr + p2 /4q,
we have the interior solution. In this case, the unique candidate solution sat-
isfies the FOCs of the Lagrangian is indeed the solution to the constrained
optimization. The solution is summarized:
p2 √
 
∗ 1
x (p, q, r, m) = m− − pr ,
p 4q
y ∗ (p, q, r, m) = p2 /4q 2 ,
z ∗ (p, q, r, m) =
p
p/r.
√
If m = pr + p2 /4q, we have the boundary solution. But, even in this case,
the solution is expressed as the same as the one we provided above.
5. Show that the maximized utility is given by the following function:
r
m p r
U ∗ (p, q, r, m) = + −2 .
p 4q p

We simply compute U (x∗ , y ∗ , z ∗ ) where (x∗ , y ∗ , z ∗ ) is the solution we found in

the previous question:
m p p m p
U (x∗ , y ∗ , z ∗ ) = − − p−1/2 r1/2 + − p−1/2 r1/2 = + − 2p−1/2 r1/2 .
p 4q 2q p 4q

58
 6. Find ∂U ∗ /∂m.

We simply compute this:

∂U ∗
= 1/p.
∂m
Question 10.2 (25 points) Consider the following problem:
1
max 4 − x2 − 4y subject to 6x − 4y ≤ a,
x,y 2
where a ∈ R is a constant. Answer the following questions.

1. Write down the Kuhn-Tucker conditions for the constrained maximization

problem.

We set up the Lagrangian:

1
L(x, y, λ) = 4 − x2 − 4y − λ(6x − 4y − a).
2
The Kuhn-Tucker conditions are
0
L1 = −x − 6λ = 0, (1)
0
L2 = −4 + 4λ = 0, (2)
λ ≥ 0, (3)
6x − 4y ≤ a, (4)
λ(6x − 4y − a) = 0. (5)

2. Find a solution to the constrained maximization problem.

We obtain λ = 1 from (2). Plugging this into (1), we obtain x = −6. Since
λ = 1 > 0, we have 6x − 4y − a = 0 from (5). Plugging x = −6 into this
equality, we obtain y = −9 − a/4. So, we obtain (x, y) = (−6, −9 − a/4) as the
unique solution candidate by solving the Kuhn-Tucker conditions.
We compute the Hessian matrix associated with the Lagrangian function L
given λ = 1:  00 00   
L11 L12 −1 0
HL (x, y) = 00 00 = .
L21 L22 0 0
00 00 00 00 00 00
Since L11 = −1 < 0, L22 = 0, and L11 L22 − L12 L21 = 0, we conclude that HL
is negative semidefinite for all (x, y). This implies that (−6, −9 − a/4) is the
solution to the problem.
3. Derive V (a), which is the maximized objective function given a.

We simply compute this as follows:

1  a
V (a) = f (−6, −9 − a/4) = 4 − × 36 − 4 −9 − = a + 22.
2 4

59
 0
4. Verify that V (a) = λ, where λ is the Lagrange multiplier in the previous
question.
0
We obtained V (a) = a + 22. So, we have V (a) = 1, which is the same as
the value of the Lagrange multiplier we obtained in the previous part of the
question.

Question 10.3 (25 points) Consider the following problem:

max x2 + 2y 2 − x subject to x2 + y 2 ≤ 1.
x,y

Answer the following questions.

1. Write down the Lagrangian for the constrained maximization problem.

L(x, y, λ) = x2 + 2y 2 − x − λ(x2 + y 2 − 1).

2. Find all pairs (x, y) that satisfy the Kuhn-Tucker conditions.

The Kuhn-Tucker conditions are

0
Lx = 2x − 1 − 2λx = 0, (1)
0
Ly = 4y − 2λy = 0, (2)
λ ≥ 0, (3)
2 2
x + y ≤ 1, (4)
λ(x2 + y 2 − 1) = 0. (5)

We have y(2 − λ) = 0 from (2). So, there are two cases: either λ = 2 or
y = 0. Assume that λ = 2. Plugging this into (1), we obttain x = −1/2.
Since λ = 2, we have x2 + 2
√y − 1 = 0 from (5). Plugging x = −1/2 into this
equality, we obtain y = ± 3/2. Therefore, there√are two solutions√that satisfy
the Kuhn-Tucker conditions: (x, y, λ) = (−1/2, 3/2, 2), (−1/2, − 3/2, 2).
Now, assume y = 0. If λ = 0, we have x = 1/2 from (1). So, (x, y, λ) =
(1/2, 0, 0) is a solution candidate satisfying the Kuhn-Tucker conditions. If
λ > 0, we have x = ±1 from (5). Plugging x = 1 into (1), we obtain λ = 1/2.
So, (x, y, λ) = (1, 0, 1/2) is a solution candidate satisfying the Kuhn-Tucker
conditions. Plugging x = −1, we obtain λ = 5/2. So, (x, y, λ) = (−1, 0, 5/2) is
a solution candidate satisfying the Kuhn-Tucker conditions. Note also that the
constraint qualification is satisfied because it can only be violated at (x, y) =
(0, 0).

3. Find the solution to the problem.

√ √
Let f (x, y) = x2 +2y 2 −x. We first observe that f (−1/2, − 3/2) = f (−1/2, 3/2) =
9 1
4 = 2 + 4 . Second, we observe that f (1/2, 0) = −1/2, f (1, 0) = 0, and

60
 √ √
f (−1, 0) = 2. Thus, both (x, y) = (−1/2, − 3/2), (−1/2, 3/2) are the only
solution candidates.
We can make the value of f as high as possible by choosing y 2 high enough.
This is simply impossible because we have the inequality constraint x2 +y 2 ≤ 1.
Therefore, the constraint must be binding. In this case, we have y 2 = 1 − x2 .
Since y 2 ≥ 0, we must have the condition that −1 ≤ x ≤ 1. Plugging y 2 = 1−x2
into f (x, y), we have the following:

1 2 9
p  
2 2 2 2
f (x, y) = f (x, ± 1 − x ) = x +2(1−x )−x = −x −x+2 = − x + + .
2 4
√ √
This implies that the solution to the problem is indeed (x, y) = (−1/2, − 3/2), (−1/2, 3/2).

Another way we can solve this is to use the Hessian matrix of the Lagrangian
function L(x, y, λ∗ = 2). We have L(x, y, 2) = x2 + 2y 2 − x − 2(x2 + y 2 − 1) =
−x2 − x + 2. So, the corresponding Hessian matrix is
 
−2 0
H= .
0 0
00 00
Since Lxx = −2 < 0, Lyy = 0, and |H| = 0, the Hessian matrix is negative
√ √
semidefinite for all (x, y). This implies that (−1/2, − 3/2), (−1/2, 3/2) is
the solution to the problem.

Question 10.4 (25 points) Consider the following utility maximization problem:

max u(x, y) = x + ln(1 + y) subject to 16x + y ≤ 495, x ≥ 0, y ≥ 0,

x,y

where u(x, y) denotes the utility of the consumer when he consumes a consumption
bundle (x, y). Answer the following questions.

1. Write down the necessary Kuhn-Tucker conditions (with non-negativeity con-

straints) for a point (x∗ , y ∗ ) to be a solution.

We set up the Lagrangian:

L(x, y, λ1 , λ2 , λ3 ) = x + ln(1 + y) − λ1 (16x + y − 495) − λ2 (−x) − λ3 (−y).

61
 The Kuhn-Tucker conditions are
0
Lx = 1 − 16λ1 + λ2 = 0, (1)
0 1
Ly = − λ1 + λ3 = 0, (2)
1+y
λ1 , λ2 , λ3 ≥ 0, (3)
16x + y ≤ 495, (4)
x ≥ 0, (5)
y ≥ 0, (6)
λ1 (16x + y − 495) = 0, (7)
λ2 (−x) = 0, (8)
λ3 (−y) = 0. (9)

2. Find the solution to the problem.

We find all (x, y) that satisfy the Kuhn-Tucker conditions by considering the
following four cases.

Case 1: x > 0 and y > 0

We obtain λ2 = λ3 = 0 from (8) and (9). Plugging λ2 = 0 into (1), we obtain

λ1 = 1/16. Plugging λ3 = 0 and λ1 = 1/16 into (2), we obtain y = 15. Since
λ1 > 0, we have 16x + y − 495 = 0 from (4). Plugging y = 15 into this equality,
we obtain x = 30. Thus, (x, y) = (30, 15) is a solution.

Case 2: x > 0 and y = 0

Since x > 0, we have λ2 = 0 from (8). Plugging λ2 = 0 into (1), we obtain

λ1 = 1/16. Plugging y = 0 and λ1 = 1/16, we obtain λ3 = −15/16, which is
not consistent with (3). So, we ignore this case.

Case 3: x = 0 and y > 0

Since y > 0, we have λ3 = 0 from (9). Plugging λ3 = 0 into (2), we obtain

λ1 = 1/(1 + y) > 0 because y > 0. Since λ1 > 0, we have 16x + y − 495 = 0
from (7). Plugging x = 0 into this equality, we obtain y = 495. This implies
that λ1 = 1/496. Plugging this into (1), we conclude
1
1 − 16 × + λ2 = 0 ⇒ λ2 < 0,
496
which is inconsistent with (3). Thus, we ignore this case.

Case 4: x = 0 and y = 0

62
 We have 16x + y < 495 because x = y = 0. So, we have λ1 = 0 from (7).
Plugging λ1 = 0 into (1), we obtain λ2 = −1, which is inconsistent with (3).
Hence, we ignore this case.

We therefore conclude that there is the unique solution candidate (x, y, λ1 , λ2 , λ3 ) =

(30, 15, 1/16, 0, 0). Let g(x, y) = 16x + y. Since g is the unique constraint bind-
0 0
ing in this case (i.e., x > 0 and y > 0), we compute (g1 , g2 ) = (16, 4) 6= (0, 0).
Thus, the constraint qualification is satisfied.

We compute the Hessian matrix associated with the Lagrangian L given (λ1 , λ2 , λ3 ) =
(1/16, 0, 0):
 00 00  !
L11 L12 0 0
HL (x, y) = 00 00 = 1 .
L21 L22 0 − (1+y)2

00 00
Since L11 = 0, L22 = −1/(1+y)2 < 0, and |HL | = 0, HL is negative semidefinite
for all (x, y). Therefore, the unique solution candidate we found solving the
Kuhn-Tucker condition is indeed the solution to the problem.

3. Estimate by how much utility will increase if income is increased from 495 to
500.

Since the Lagrange multiplier corresponds to the marginal utility increase when
we relax the constraint in one unit. Thus, the utility increase as income is
increase from 495 to 500 is
λ1 × 5 = 5/16.

11 Homework 11 (Not for Submission)

Question 11.1 You may use the following convention: ∞ · 0 = 0 in this question.
Compute the following:

1. Z ∞
λe−λx dx
0

Fix b > 0. By integration by parts, we compute

Z b h ib
λe−λx dx = −e−λx = −(e−λb − 1).
0 0

Since e−λb → 0 as b → ∞, we have

Z ∞ Z b
λe−λx dx = lim λe−λx dx = 1.
0 b→∞ 0

63
 2. Z ∞
xλe−λx dx
0

Fix b > 0. By integration by parts, we compute the following.

Z b h ib Z b
xλe−λx dx = −xe−λx + e−λx dx
0 0 0
−λb 1 h −λx ib 1
= −(be − 0) − e = −be−λb − (e−λb − 1)
  λ 0 λ
1 1 −λb
= − +b e
λ λ
Since (1/λ + b)e−λb → 0 as b → ∞, due to the convention, we have
Z b
1
lim xλe−λx dx = < ∞.
b→∞ 0 λ
Therefore,
Z ∞ Z b
1
xλe−λx dx = lim xλe−λx dx = .
0 b→∞ 0 λ
3. Z ∞
(x − 1/λ)2 λe−λx dx
0

We compute
Z ∞ Z ∞
2 ∞ 1 ∞ −λx
Z Z
2 −λx
(x − 1/λ) λe dx = x2 λe−λx dx − xλe −λx
dx + e dx
0 λ 0 λ 0
Z0 ∞ i∞
2 1 h
= x2 λe−λx dx − − 2 e−λx
λ λ 0
Z0 ∞
2 1
= x2 λe−λx dx − + 2
0 λ λ
h i∞ 2 Z ∞ 2 1
= −x2 e−λx + xλe−λx dx − + 2
0 λ 0 λ λ
| {z }
=1
1
=
λ2
Question 11.2 Let f (x, y) = x3 − y 3 + 9xy. Answer the following questions.

1. Find all stationary points of f (x, y).

We first check the FOCs:

∂f x2
(1) = 3x2 + 9y = 0 ⇒ y = −
∂x 3
∂f
(2) = −3y 2 + 9x = 0.
∂y

64
 Plugging y = −x2 /3 into (2), we obtain
x4
− + 9x = 0 ⇒ x(x3 − 27) = 0 ⇒ x(x − 3)(x2 + 3x + 9) = 0
3
Since x2 + 3x + 9 = (x + 3/2)2 + 27/4 > 0, we have two solutions: x = 0, 3.
Since y = −x2 /3, we have two stationary points
(x, y) = (0, 0), (3, −3).

2. Classify all stationary points you found in the previous question.

We compute the second-order derivatives of f (x, y):

∂2f
= 6x
∂x2
∂2f
= −6y
∂y 2
∂2f ∂2f
= = 9
∂x∂y ∂y∂x
We form the Hessian matrix associated with f (x, y):
 
6x 9
H(x, y) = .
9 −6y
At (0, 0), we have  
0 9
H(0, 0) =
9 0
We compute |H(0, 0)| = −81 < 0. This implies that H(0, 0) is neither positive
semidefinite nor negative semidefinite. Hence, (0, 0) is a saddle point.
At (3, −3), we have  
18 9
H(3, −3) =
9 18
We can check that |H1 (3, −3)| = 18 > 0 and |H2 (3, −3)| = 182 − 92 > 0.
Therefore, H(3, −3) is positive definite. Hence, (3, −3) is a local minimum
point.
3. Is there any local minimum point that corresponds to a global minimum point?
If so, verify it. If not, argue why.

In the previous question, we found (x, y) = (3, −3) as a local minimum point
of f (x, y). We compute the associated value of f (3, −3):
f (3, −3) = 27 + 27 − 81 = −27.
Compute
f (0, 4) = −43 = −64 < −27
This implies that (x, y) = (3, −3) is not a global minimum point of f (x, y).

65
Question 11.3 Consider the problem

max f (x, y) = x2 y subject to g(x, y) = 2x2 + y 2 = 3.

x,y

Solve this problem.

We set up the Lagrangian.

L(x, y, λ) = x2 y − λ(2x2 + y 2 − 3).

The FOCs are

0
(1) Lx = 2xy − 4λx = 0 ⇒ x(y − 2λ) = 0
0
(2) Ly = x2 − 2λy = 0
0
(3) Lλ = 2x2 + y 2 − 3 = 0.
√
We first√ consider the case that x = 0. By (3), we have y = ± 3. Since x = 0
and y = ± 3, we must have √ λ = 0 due√to (2). So, there are two solution candidates
in this case. (x, y, λ) = (0, 3, 0), (0, − 3, 0).
Next we consider the case that x 6= 0. Then, by (1), we have y = 2λ. Plug-
ging y = 2λ into (2), we obtain x2 = y 2 . Furthermore, plugging y 2 = x2 into
(3), we obtain x2 = 1, that is, x = ±1. Therefore, there are four solution candi-
dates (x, y) = (1, 1), (1, −1), (−1, 1), (−1, −1). By (1), we compute the associated
Lagrangian multiplier for each case. So,

(x, y, λ) = (1, 1, 1/2), (−1, −1, −1/2), (1, −1, −1/2), and (−1, 1, 1/2)

In sum, we have six candidates so far. We compute the maximum value for each
candidate:

f (1, 1) = f (−1, 1) = 1
f (1, −1) = f (−1, −1) = −1
√ √
f (0, 3) = f (0, − 3) = 0.

This implies that (1, 1) and (−1, 1) are the only two solution candidates.
We check the constraint qualification. Fix (x, y) as an arbitrary feasible point.
We compute the following:

∇g(x, y) = (∂g/∂x, ∂g/∂y) = (4x, 2y).

We claim that (4x, 2y) 6= (0, 0). Suppose, on the contrary, that (4x, 2y) = (0, 0).
This implies that x = y = 0. However, in this case, we obtain 2x2 + y 2 = 0 6= 3. So,
(x, y) = (0, 0) is not a feasible point. This is the desired contradiction. Thus, the
constraint qualification is satisfied.
We also confirm that the constrained optimization problem has a solution. Define
D = {(x, y) ∈ R2 |2x2 + y 2 = 3}. Since 2x2 and y 2 are each a continuous function

66
and the sum of two continuous functions is a continuous function, g(x, y) = 2x2 + y 2
is a continuous function. Hence, D is closed. Define B = {(x, y) ∈ R2 |x2 + y 2 ≤ 3}.
Fix (x, y) ∈ D. We compute the following.

x2 + y 2 ≤ 2x2 + y 2 = 3
|{z}
x2 ≥0

This implies that (x, y) ∈ B. So, D ⊆ B so that B is bounded. Hence, D is a

compact set. Since f (x, y) = x2 y is the product of two continuous functions, it is
also continuous. By the extreme value theorem, the solution candidates (1, 1) and
(−1, 1) are indeed the solutions.

Question 11.4 Consider the constrained maximization problem

1
max 2x + y − x3 − xy − y 2 subject to x ≥ 1
4 and x + y ≤ 3
x,y 3
Answer the following questions.

1. Write down the Kuhn-Tucker conditions.

We set up the Lagrangian:

 
1 3 2 1
L(x, y, λ1 , λ2 ) = 2x + y − x − xy − y − λ1 −x + − λ2 (x + y − 3).
3 4

The Kuhn-Tucker conditions are

0
(1) Lx = 2 − x2 − y + λ1 − λ2 = 0
0
(2) Ly = 1 − x − 2y − λ2 = 0
(3) λ1 ≥ 0, x ≥ 1/4, λ1 (−x + 1/4) = 0
(4) λ2 ≥ 0, x + y ≤ 3, λ2 (x + y − 3) = 0.

2. Show that the solution to the Kuhn-Tucker conditions has both Lagrange mul-
tipliers equal to zero.

We investigate the solution candidates by considering the following cases.

Case 1: λ1 > 0 and λ2 > 0

In this case, by (3) and (4), both constraints are binding. Hence, (x, y) =
(1/4, 11/4). Plugging this into (2), we obtain
1 11 19
1− − − λ2 = 0 ⇒ λ2 = − ,
4 2 4
which implies λ2 < 0. This contradicts (4). So, this case is invalid.

67
Case 2: λ1 > 0 and λ2 = 0

By (3), we have x = 1/4. Plugging x = 1/4 and λ2 = 0 into (2), we obtain

1 3
1− − 2y = 0 ⇒ y =
4 8
Plugging (x, y) = (1/4, 3/8) and λ2 = 0 into (1), we obtain
1 3 25
2− − + λ1 = 0 ⇒ λ1 = − .
16 8 16
However, this contradicts (3). So, this case is invalid.
Case 3: λ1 = 0 and λ2 > 0

Plugging λ1 = 0 into (1) and (2), we obtain

0
(1 ) λ2 = 2 − x2 − y
0
(2 ) λ2 = 1 − x − 2y
0 0
Solving (1 ) and (2 ), we obtain

y = x2 − x − 1.

Since λ2 > 0, by (4), we have x + y = 3. Plugging y = x2 − x − 1 into x + y = 3,

we obtain
x + x2 − x − 1 = 3 ⇒ x2 = 4 ⇒ x = ±2.
By (3), x = −2 is invalid. If x = 2, we have y = 1. Plugging (x, y) = (2, 1)
0
into (2 ), we have
λ2 = 1 − 2 − 2 = −3,
which contradicts λ2 ≥ 0, which is part of the requirement of (4). So, this case
is invalid.
Case 4: λ1 = 0 and λ2 = 0

Plugging λ1 = λ2 = 0 into (1) and (2), we obtain

00
(1 ) 2 − x2 − y = 0 ⇒ y = 2 − x2
00
(2 ) 1 − x − 2y = 0
00
Plugging y = 2 − x2 into (2 ), we obtain

1 − x − 2(2 − x2 ) = 0 ⇒ 2x2 − x − 3 = 0 ⇒ (x + 1)(2x − 3) = 0.

There are two solutions. x = −1, 3/2. We must have x ≥ 1/4 by (3), x = −1 is
invalid. On the other hand, x = 3/2 ≥ 1/4. Plugging x = 3/2 into y = 2 − x2 ,
we obtain y = −1/4. In this case, x + y = 3/2 − 1/4 = 5/4 < 3. Therefore, we
have the unique solution candidate: (x, y) = (3/2, −1/4).

68
 3. Solve the constrained maximization problem.

From the previous question, we have (x, y, λ1 , λ2 ) = (3/2, −1/4, 0, 0) as the

unique solution candidate. We claim that this solution candidate is indeed a
solution. This will proved by showing that the Hessian matrix associated with
the Lagrangian function L(x, y, λ1 = 0, λ2 = 0) is negative semidefinite for all
(x, y). We compute the second-order derivatives of the Lagrangian:
00
L11 = −2x
00
L22 = −2
00 00
L12 = L21 = −1.
Thus, the associated Hessian matrix is given as follows:
 00 00   
L11 L12 −2x −1
HL (x, y) = 00 00 =
L21 L22 −1 −2
00 00
Since x ≥ 1/4, we have L11 = −2x < 0, L22 = −2 < 0, and
|HL (x, y)| = 4x − 1 ≥ 0,
because x ≥ 1/4. Thus, the associated Hessian matrix is negative semidefinite
for all (x, y). Therefore, the solution candidate we found using the Kuhn-
Tucker conditions is indeed a solution to the constrained maximization prob-
lem.
Question 11.5 Consider the following problem:
max xey−x − 2ey subject to y ≤ 1 + x/2, x ≥ 0, y ≥ 0.
x,y

Assume that the constrained maximization problem has a solution. Then, solve this
constrained optimization problem using the Kuhn-Tucker conditions.
We set up the Lagrangian:
L(x, y, λ1 , λ2 , λ3 ) = xey−x − 2ey − λ1 (−x/2 + y − 1) − λ2 (−x) − λ3 (−y).
The Kuhn-Tucker conditions are:
0
Lx = ey−x − xey−x + λ1 /2 + λ2 = 0, (1)
0
Ly = xey−x − 2e − λ1 + λ3 = 0, (2)
λ1 , λ2 , λ3 ≥ 0, (3)
−x/2 + y − 1 ≤ 0, (4)
x ≥ 0, (5)
y ≥ 0, (6)
λ1 (−x/2 + y − 1) = 0, (7)
λ2 (−x) = 0, (8)
λ3 (−y) = 0. (9)

69
Case 1: x > 0 and y > 0

We obtain λ2 = λ3 = 0 from (8) and (9). Assume that λ1 > 0. In this case, we
have y = 1 + x/2 from (7). Plugging λ2 = λ3 = 0 into (1) and (2), respectively, we
obtain
2xey−x − 2ey−x = xey−x − 2e ⇔ (x − 2)ey−x = −2e (I)
Plugging y = 1 + x/2 into (I), we obtain
(2 − x)e1−x/2 = 2e.
Therefore, we have x = 0 as the solution to the above equality. However, this
contradicts our assumption that x > 0. So, we ignore the case of λ1 > 0. Now, we
assume λ1 = 0. Plugging λ1 = λ2 = 0 into (1), we obtain (1 − x)ey−x = 0. We
thus have x = 1. Furthermore, plugging λ1 = λ3 = 0 and x = 1 into (2), we obtain
ey−1 = 2e. That is, we obtain y = ln 2 + 2. However, we plug x = 1 and y = ln 2 + 2
into the left hand side of (4) so that we obtain
−x/2 + y − 1 = −1/2 + ln 2 + 2 − 1 = ln 2 + 1/2 > 0,
which contradicts (4). We ignore Case 1.

Case 2: x > 0 and y = 0

Assume that λ1 > 0. In this case, we have −x/2 + y − 1 = 0 from (7). Plugging
y = 0 into this equality, we obtain x = −2, which contradicts (5). Hence, we ignore
the case of λ1 > 0. Now, assume that λ1 = 0. Since x > 0, we have λ2 = 0 from (8).
Plugging λ1 = λ2 = 0 and y = 0 into (1), we obtain
e−x − xe−x = 0 ⇔ (1 − x)e−x = 0.
Therefore, we obtain x = 1. We also compute
−x/2 + y − 1 = −1/2 − 1 = −3/2 < 0,
which satisfies (4). Plugging x = 1, y = 0, and λ1 = 0 into (2), we obtain
e−1 − 2e + λ3 = 0 ⇔ λ3 = 2e − e−1 > 0,
which satisfies (3). Therefore, (x, y) = (1, 0) is a solution candidate associated with
λ1 = 0, λ2 = 0, and λ3 = 2e − e−1 .

Case 3: x = 0 and y > 0

Assume that λ1 > 0. Then, we have −x/2 + y − 1 = 0 from (7). Plugging x = 0

into this equality, we obtain y = 1. Since y > 0, we have λ3 = 0 from (9). Plugging
λ3 = 0, x = 0, and y = 1 into (2), we obtain
−2e − λ1 = 0 ⇔ λ1 = −2e < 0,

70
which contradicts (3). So, we ignore the case of λ1 > 0. Now, assume λ1 = 0.
Plugging λ1 = λ3 = 0 and x = 0 into (2), we obtain

−2e = 0,

which is a contradiction. Thus, we ignore Case 3.

Case 4: x = 0 and y = 0

In this case, we have −x/2 + y − 1 < 0 so that we have λ1 = 0 from (7). Plugging
x = 0, y = 0, and λ1 = 0 into (1), we obtain

1 + λ2 = 0 ⇔ λ2 = −1,

which contradicts (3). Thus, we ignore Case 4.

Considering all the four cases, we conclude that there is the only one solution
candidate (x, y) = (1, 0) where only y ≥ 0 constraint is binding.
Define g 1 (x, y) = −x/2 + y − 1 ≤ 0, g 2 (x, y) = −x ≤ 0, and g 3 (x, y) = −y ≤ 0.
We compute the following:

∇g 1 (x, y) = (−1/2, 1);

∇g 2 (x, y) = (−1, 0);
∇g 3 (x, y) = (0, −1).

We check the following cases to verify that the constraint qualification holds: Fix
(x, y) as an arbitrary feasible point satisfying all the inequality constraints.
Case 1: g 1 , g 2 , and g 3 are not binding at (x, y)
In this case, the constraint qualification holds automatically.
Case 2: g 1 is binding but g 2 and g 3 are not binding at (x, y)
In this case, we have ∇g 1 (x, y) = (−1/2, 1) 6= (0, 0). So, the constraint qualifica-
tion holds.
Case 3: g 2 is binding but g 1 and g 3 are not binding at (x, y)
In this case, we have ∇g 2 (x, y) = (−1, 0) 6= (0, 0). So, the constraint qualification
holds.
Case 4: g 3 is binding but g 1 and g 2 are not binding at (x, y)
In this case, we have ∇g 3 (x, y) = (0, −1) 6= (0, 0)
Case 5: g 1 and g 2 are binding but g 3 is not binding at (x, y)
In this case, we have ∇g 1 (x, y) = (−1/2, 1) and g 2 (x, y) = (−1, 0), which are
clearly not parallel to each other. Hence, the constraint qualification holds.

71
Case 6: g 2 and g 3 are binding but g 1 is not binding at (x, y)
In this case, we have ∇g 2 (x, y) = (−1, 0) and g 3 (0, −1), which are clearly not
parallel to each other. Hence, the constraint qualification holds.
Case 7: g 1 and g 3 are binding but g 2 is not binding at (x, y)
In this case, we have ∇g 1 (x, y) = (−1/2, 1) and g 3 (x, y) = (0, −1), which are
clearly not parallel to each other. Hence, the constraint qualification holds.
Exhausting all the cases, we conclude that the constraint qualification is satisfied.
Since we assume that it has a solution, we indeed found the solution to the problem,
which is (x, y) = (1, 0).

72