Math 411 — Final Exam

Please sign the honor pledge on the front of the booklet. You must carefully justify all of
your statements. You may use anything from class or the book without proof.

1. (a) (5 pts.) State the Cauchy-Schwarz inequality.

(b) (15 pts.) Let f : Rn −
→ R be continuously differentiable function, f (0) = 0. Suppose
|f (u)| = 1 for some #u# = 1. Prove that there exists v such that #∇f (v)# ≥ 1.
Answer:
(a) For all vectors x, y ∈ Rn ,
|'x, y(| ≤ #x# · #y#.
It is not correct if you just write the inequality without explaining what x and y
are!!!
(b) Since f is C 1 on Rn , by the Mean Value Theorem (Thm 13.17) there exists 0 < θ < 1
such that
f (0 + u) − f (0) = '∇f (0 + θu), u(
Since f (0) = 0 and |f (u)| = 1, by Cauchy-Schwarz,
1 = |f (u)| = |f (0+u)−f (0)| = |'∇f (0+θu), u(| ≤ #∇f (0+θu)# ·#u# = #∇f (θu)#
Let v = θu.

2. (20 pts.) Prove that

sin xy + exz = 1
5y 2 − 3z 2 + x2 = 2
implicitly defines x and y as a C 1 function ψ(z) near the solution (0, 1, 1), and find the
best affine approximation of ψ at z = 1.
Answer: Let F(x, y, z) = (sin xy +exz , 5y 2 −3z 2 +x2 ). F is easily seen to be continuously
differentiable. Further
! "
y cos xy + zexz x cos xy xexz
DF(x,y,z) =
2x 10y −6z
so ! "
2 0 0
DF(0,1,1) =
0 10 −6
The left 2 × 2 minor is nonsingular, therefore by the Implicit Function Theorem there
exists δ > 0 and a continuously differentiable function ψ : Bδ (1) −→ R2 such that the
solutions to the system F(x, y, z) = (1, 2) are implicitly defined by (x, y) = ψ(z), for
(x, y) ∈ Bδ (0, 1). Further
! "−1 ! " ! "! " ! "
2 0 0 1/2 0 0 0
Dψ|z=1 = − =− =
0 10 −6 0 1/10 −6 3/5
 Thus the best affine approximation to ψ at z = 1 is

ψ(1 + h) ≈ ψ(1) + Dψ|z=1 h = (0, 1 + 3/5 · h)

# $n
3. (20 pts.) Let n > 0 be even. Maximize the function f (x, y, z) = x+y+z3 subject to the
constraint xn + y n + z n = 3. Conclude that for all real numbers a, b, c,
% &1/n
a+b+c an + bn + cn
≤ .
3 3

Answer: Let g(x, y, z) = xn + y n + z n be the constraint function. Notice that the

constraint set g(x, y, z) = 3 is compact: g is continuous and level sets of continuous
functions are closed, and the fact that n is even implies the level set is bounded. Thus f
attains a maximum on the constraint set, say at (x0 , y0 , z0 ).
Now
' % &n−1 % &n−1 % &n−1 (
n x+y+z n x+y+z n x+y+z
∇f (x, y, z) = , ,
3 3 3 3 3 3
# $
∇g(x, y, z) = nxn−1 , ny n−1 , nz n−1

Both functions are polynomials, so C 1 . The condition ∇g(x, y, z) = (0, 0, 0) implies

(x, y, z) = (0, 0, 0) which cannot happen given the constraint. Therefore we can apply the
Lagrange Multiplier Theorem and say there exists λ such that

∇f (x0 , y0 , z0 ) = λ∇g(x0 , y0 , z0 ).

If λ = 0 then (x + y + z)/3 = 0 which gives the minimum value of f (x, y, z) = 0.

Otherwise, since ∇f (x0 , y0 , z0 ) = ∗(1, 1, 1) for some ∗, this implies nxn−1
0 = ny0n−1 =
nz0 , therefore x0 = y0 = z0 since n − 1 is odd. Plugging back into the constraint gives
n−1

(x0 , y0 , z0 ) = ±(1, 1, 1), which gives the maximum value, 1, of the function f . That is

f (x, y, z) ≤ 1 for all xn + y n + z n = 3.

1/n
Now if a, b, c are arbitrary real numbers set S = ((an + bn + cn )/3) , and x = a/S,
y = b/S, z = c/S. Then (x, y, z) satisfies the constraint; we conclude that f (x, y, z) ≤ 1.
Writing this equation in terms of a, b, c and multiplying through by S n and taking the
nth root gives the result.
4. Let g : R3 −
→ R be a continuously differentiable function. Define the solution set

S = {x | g(x) ≤ c}.

Suppose that ∇g(x) -= 0 for all x ∈ S.

(a) (5 pts.) Prove that ∂S = {x | g(x) = c} (Hint: Prove ⊂ then ⊃).
 (b) (5 pts.) Prove that for every x ∈ ∂S, there exists an open set U 0 x such that
∂S ∩ U has Jordan content 0. It follows that if ∂S is bounded, then it has Jordan
content 0 (don’t prove this).
(c) (5 pts.) Define what it means for a set to have volume. Explain why B1 (0, 0, 0) has
volume and set up (but do not evaluate) an iterated integral that computes it.
(d) (5 pts.) Find a bounded set A ⊂ R3 , and a function f : A −
→ R such that f is
continuous on A, but f is not integrable.
Answer:
(a) (⊂) Assume x ∈ ∂S. By definition, for every k > 0 there exists uk , vk ∈ Bε (x)
such that uk ∈ S, vk -∈ S. In other words, g(uk ) ≤ c and g(vk ) > c. g is C 1 , so
continuous, so by comparison,
g(x) = lim g(uk ) ≤ c, and, g(x) = lim g(vk ) ≥ c.
k−
→∞ k−
→∞
We conclude that g(x) = c.
(⊃) Assume g(x) = c. The condition ∇g(x) -= 0 implies that x is not a relative
extremum. In other words, for every ε > 0 there exists u, v ∈ Bε (x) such that
g(u) ≥ g(x) = c and g(v) ≤ g(x) = c. Thus u ∈ S and v -∈ S; since ε was arbitrary
we conclude that x ∈ ∂S.
(b) Let x = (x, y, z). Since ∇g(x, y, z) -= 0 we may assume after relabeling that
D3 g(x, y, z) -= 0. Thus we can apply the Implicit Function Theorem to say there is
δ > 0 and a C 1 function ψ : Bδ (x, y) −
→ R such that for z # ∈ Bδ (z)
g(x# , y # , z # ) = c ⇐⇒ z # = ψ(x# , y # ).
Let U = Bδ (x, y) × Bδ (z) ⊂ R3 . The previous statement exactly says that ∂S ∩ U
is exactly the graph of the function ψ. ψ is C 1 , so continuous, so integrable on the
Jordan domain Bδ (x, y). Thus ∂S ∩ U = (graph of ψ) has Jordan content 0 (Cor
18.26).
(c) A set S is said to have volume if the function f : S − → R which is constantly 1 is
integrable. By parts (a) and (b), B1 (0, 0, 0) is a Jordan domain because it is of the
form given with g(x, y, z) = x2 + y 2 + z 2 and c = 1. Following Example 19.11 in the
) 2π ) π ) 1
book, the volume of the ball is given by the integral θ=0 ϕ=0 ρ=0 ρ2 sin θ dρ dϕ dθ.
(d) The exam should have said “bounded set”. Certain if A is not bounded then no
function on A is integrable, just because we never defined what that meant. An
example of a bounded set A is the set of points in I = [0, 1] × [0, 1] × [0, 1] with a
rational component, and f is the function that is constantly 1 on A. If f* is the zero
extension of f to I, then M (f*, J) = 1 and m(f*, J) = 0 for every subrectangle J ⊂ I,
by the density of the rationals and irrationals in R. It follows that U (f*, P) = 1 and
L(f*, P) = 0 for every partition of I, so f is not integrable because f* is not.

5. (a) (10 pts.) Let I be a generalized rectangle in Rn and suppose that the function
f) : I −
→ R is continuous. Assume that f (x) ≥ 0 for all points x in I. Prove that if
I
f = 0, then f (x) = 0 for all points x in I.
(b) (10 pts.) Define f : [0, 1] × [0, 1] −
→R
+
1 x=y
f (x, y) =
0 otherwise
)
Prove that [0,1]×[0,1]
f = 0 using the Archimedes-Riemann theorem.
Answer:

(a) Prove the contrapositive. Assume f (x0 ) > 0 for some x0 ∈ I (say f (x0 ) = ε).
To simplify things, by continuity we may assume that x0 is in the interior of I. By
continuity there exists δ such that Bδ (x0 ) ⊂ I and if x ∈ Bδ (x0 ) then |f (x)−f (x0 )| <
ε/2 (i.e., f (x) > ε/2). Define g : I −
→ R by
+
ε/2 x ∈ Bδ (x0 )
g(x) =
0 otherwise

By construction, f (x) ≥ g(x) for all x ∈ I. We need to mention that g is integrable

because it is the a multiple
) of the
) characteristic function of the Jordan domain
Bδ (x0 ). By monotonicity, I f ≥ I g = ε/2 · vol Bδ (x0 ) > 0.
(b) Let Pk be the standard partition of I = [0, 1] × [0, 1] into k 2 equally sized squares.
These squares fall into two types: those J that intersect the diagonal and those J#
that don’t. There are exactly k +2(k −1) = 3k −1 squares that intersect the diagonal
(I don’t mind if you say there are k). Since f is constantly 0 on the squares J# ,

M (f, J) = 1
m(f, J) = 0
M (f, J# ) = m(f, J# ) = 0

so
, , 1 k− →∞
U (f, Pk ) = 1 · vol J + 0 · vol J = (3k − 1) · 2
−−−−→ 0
k
J J
, , k−
→∞
L(f, Pk ) = 0 · vol J + 0 · vol J = 0 −−−−→ 0
J J

Since these limits agree, Pk is an Archimedean )sequence of partitions; by the

Archimedes-Riemann theorem, f is integrable and I f = limk−
→∞ U (f, Pk ) = 0.