Solutions of Practice Exercises for Math 301

1. p
Let u − v = (a1 , . . . , an) and v − wp = (b1 , . . ., bn), then
p u − w = (a1 + b1 , . . ., an + bn). We have to prove
(a1 + b1)2 + · · · + (an + bn )2 ≤ a21 + · · · + a2n + b21 + · · · + b2n . Squaring both sides and expanding
the left side, we get the equivalent inequality
q q
(a21 +· · ·+a2n )+2(a1 b1 +· · ·+an bn )+(b21 +· · ·+b2n ) ≤ (a21 +· · ·+a2n )+2 a21 + · · · + a2n b21 + · · · + b2n +(b21 +· · ·+b2n ).

Cancelling the common

p p terms and dividing by 2 on both sides, this is equivalent to a1b1 + · · · + an bn ≤
2 2
a1 + · · · + an b1 + · · · + b2n , which is the Cauchy-Schwarz inequality. So the triangle inequality is
2

true.
For those who forget the proof of the Cauchy-Schwarz inequality, here is the proof: Consider the
quadratic polynomial

P (x) = (a1 x + b1 )2 + · · · + (an x + bn )2 = (a21 + · · · + a2n ) x2 + 2(a1 b1 + · · · + anbn ) x + (b21 + · · · + b2n) .

| {z } | {z } | {z }
A B C

B2 B √ √
Since P (x) ≥ 0 for all x ∈ R, so B 2 − 4AC ≤ 0. Then ≤ AC, so ≤ A C, i.e. a1 b1 + · · ·+ an bn ≤
p p 4 2
a21 + · · · + a2n b21 + · · · + b2n .
p
2. If lim vi = L, then lim kvi − Lk = 0. Since 0 ≤ |xij − lj | ≤ (xi1 − l1 )2 + · · · + (xin − ln )2 = kvi − Lk
i→∞ i→∞
for all j = 1, . . ., n, by the sandwich theorem, we have lim |xij − lj | = 0, i.e. lim xij = lj for all j =
i→∞ i→∞
1, . . . , n.
Conversely, if lim xij = lj for all j = 1, . . . , n, then lim |xij − lj | = 0 for all j = 1, . . . , n. Adding these n
i→∞ i→∞ p
limits, we get lim (|xi1 − l1 | + · · · + |xin − ln |) = 0. Since kvi − Lk = (xi1 − l1 )2 + · · · + (xin − ln )2 ≤
i→∞
|xi1 − l1 | + · · · + |xin − ln | by the useful inequality, so lim kvi − Lk = 0 by the sandwich theorem.
i→∞

3. By the differentiation theorem, if a linear transformation satisfies the definition of the derivative of F
∂fi
at p, then the (i, j)-th entry of its matrix is (p) for i = 1, . . ., m; j = 1, . . . , n. There is exactly one
∂xj
linear transformation having that matrix with respect to the usual basis.

∂f1 ∂f1 ∂f1

4. The coordinate functions of F are f1 (x, y, z) = x2+ey , f2(x, y, z) = z cos x and = 2x, = ey , = 0,
∂x ∂y ∂z
∂f2 ∂f2 ∂f2
= −z sin x, = 0, = cos x. They are continuous everywhere on R3 . So F is C 1 at every point
∂x ∂y ∂z
of R3 . By the 1
 C theorem, F is differentiable
 at every (a, b, c) ∈ R3 . By the differentiation theorem,
2a eb 0
F 0(a, b, c) = . Then
−c sin a 0 cos a
   
x  b
 x  
2a e 0 2ax + eb y
DF (a, b, c)  y  = y = .
−c sin a 0 cos a −(c sin a))x + (cos a)z
z z

∂f1
5. (i) The coordinate functions of F are f1 (x, y, z) = xy + cos(y + z), f2 (x, y, z) = −zex and = y,
∂x
∂f1 ∂f1 ∂f2 ∂f2 ∂f2
= x − sin(y + z), = − sin(y + z), = −zex , = 0, = −ex . They are continuous ev-
∂y ∂z ∂x ∂y ∂z
erywhere on R3 . So F is C 1 at every point of R3 . By the C 1 theorem, F is differentiable at every
(a, b, c) ∈ R3.
  
b a − sin(b + c) − sin(b + c)
(ii) By the differentiation theorem, F 0(a, b, c) = . Then
−cea 0 −ea
   
x   x  
  b a − sin(b + c) − sin(b + c)  y  = bx + (a − sin(b + c))y − z sin(b + c)
DF (a, b, c) y = .
−cea 0 −ea −cea x − ea z
z z

∂F ∂F
6. For a 6= 0 and b 6= 0, = 2x, = 2y are continuous near (a, b). So F is C 1 at (a, b). By the C 1
∂x ∂y
theorem, F is differentiable there.
For a = 0 or b = 0, but (a, b) 6= (0, 0), F is not continuous at (a, b) because F (a, b) = 0, but
F (x, y) → a2 + b2 6= 0 as (x, y) → (a, b) from off the x or y-axes. So F is not differentiable at these
(a, b)’s.
∂F d  ∂F d 
For (0, 0), F (x, 0) = 0, F (0, y) = 0 imply (0, 0) = F (x, 0)x=0 = 0, (0, 0) = F (0, y)y=0
∂x dx ∂y dy
= 0. By the differentiation theorem, the only possible choice of the derivative of F at (0, 0) is T = 0. Now
kF (x, y) − F (0, 0) − T (x, y)k x2 + y 2 p
|F (x, y)| ≤ x2 + y2 for all (x, y) ∈ R2. So ≤ p = x2 + y 2 → 0
k(x, y)k x2 + y 2
as (x, y) → (0, 0). It follows F is differentiable at (0, 0).

∂f ∂f
7. f(x, 0) = 0, f(0, y) = 0 imply (0, 0) = 0, (0, 0) = 0. So we should check the definition with T = 0.
∂x ∂y
kf(x, y) − f(0, 0) − T (x, y)k |x|y2 |x|3 1
Now lim = lim 2 2 3/2
. Along y = x, lim 3/2 3 = 3/2 6= 0.
(x,y)→(0,0) k(x, y)k (x,y)→(0,0) (x + y ) x→0 2 |x| 2
So the limit is not 0 and f is not differentiable at (0, 0).

∂f ∂f ∂f ∂f
8. For x > 0, = x, = 2y are continuous. For x < 0, = 2x, = y are continuous. So f is C 1
∂x ∂y ∂x ∂y
(hence differentiable) at all (x, y), where x 6= 0.
1
For x = 0, y 6= 0, lim f(x, y) = y2 6= y2 = f(0, y), so f is not continuous (hence not differentiable)
x→0+ 2
at all (0, y), where y 6= 0. 
1 2 d 1 2  d 2  ∂f
For (0, 0), since f(x, 0) = 2 2x if x > 0 and ( x ) =0= (x ) , so (0, 0) = 0.
x if x ≤ 0 dx 2 x=0 dx x=0 ∂x
∂f d  d 1 2  

Also (0, 0) = f(0, y) = ( y ) = 0. So we should check the definition with T = 0. Now
∂y dy y=0 dy 2 y=0
kf(x, y) − f(0, 0) − T (x, y)k x2 + y 2 p
|f(x, y)| ≤ x2 + y2 for all (x, y) ∈ R2 . So ≤ p = x2 + y2 → 0 as
k(x, y)k x2 + y 2
(x, y) → (0, 0). It follows f is differentiable at (0, 0).
n 
x if x =
6 0 y if 0 6= y ∂f ∂f
9. We have f(x, 0) = = x and f(0, y) = = y. So (0, 0) = 1, (0, 0) = 1. We
0 if x = 0   0 if 0 = y ∂x ∂y
 
x x
should check the definition with T = (1 1) = x + y. As f(0, 0) = 0, we get
y y
( 2 2 p
kf(x, y) − f(0, 0) − T (x, y)k  x −y − (x + y) x2 + y 2 = 0 if x 6= y |x|
= x−y √ √ ≤ √ →0
k(x, y)k 2 2 2
|x + 2x − (x + x)|/ x + x = |x|/ 2 if x = y 2

as (x, y) → (0, 0). It follows f is differentiable at (0, 0).

∂F d 

10. (a) Note (0, b) = |xb| does not exist for b 6= 0. Since every open ball B((0, 0), r) contains
∂x dx x=0
points of the form (0, b) with b 6= 0, so F is not C 1 at (0, 0).
 ∂F d  ∂F d 
(b) We have (0, 0) = 0 = 0 and (0, 0) = 0 = 0. So we should check the definition
∂x dx x=0 ∂y dy y=0
x2 + y 2
with T = 0. Now |F (x, y)| = |xy| ≤ for all (x, y) ∈ R2 . So
2
p
kF (x, y) − F (0, 0) − T (x, y)k |xy| x2 + y 2 x2 + y 2
= p ≤ p = → 0 as (x, y) → (0, 0).
k(x, y)k x2 + y 2 2 x2 + y 2 2

It follows F is differentiable at (0, 0).

11. Define F : Rn → R2 by F (a) = (f(a), g(a)). Let P, Q, R, S be the functions in the examples following
the C 1 theorem. Since f, g are differentiable, F is also differentiable by the theorem just before the C 1
theorem. Also, P, Q, R, S are differentiable as shown in the examples. Now P ◦ F (a) = P (f(a), g(a)) =
f(a)+g(a) = (f +g)(a). Then f +g = P ◦F is differentiable by chain rule. Similarly, f −g = Q◦F, fg =
R◦F and f/g  = S◦F  are differentiable. Finally, by chain rule, (fg)0 (a) = (R◦F )0(a) = R0 (F (a))F 0(a) =
0
f (a)
( g(a) f(a) ) = g(a)f 0 (a) + f(a)g0 (a). So D(fg)(a) = g(a)Df(a) + f(a)Dg(a).
g0 (a)

12. Note g = F ◦γ : [0, 1] → R is continuous on [0, 1] and differentiable on (0, 1). By the mean value theorem,
g(1) − g(0) = g0 (s)(1 − 0) for some s ∈ (0, 1). Let c = γ(s), then F (b) − F (a) = F (γ(1)) − F (γ(0)) =
g(1) − g(0) = g0 (s) = (F ◦ γ)0 (s) = F 0(γ(s))γ 0 (s) = F 0(c)(b − a). (This exercise tells us that the mean
value theorem is still true for vector-valued functions F : Rm → Rk when k = 1.)

13. (a) F (2π) − F (0) = (1, 0) − (1, 0) = (0, 0), while (2π − 0)F 0(c) = 2π(− sin c, cos c) 6= (0, 0) for all c. (This
tells us that the mean-value theorem is not true in general for vector-valued functions F : Rm → Rk
when k > 1.)


(b) Let F (x) = (f1 (x), . . . , fn(x)). Consider φ : [0, 1] → R given by φ(t) = F (b) − F (a) · F (t) =
Xn


fi (b) − fi (a) fi (t). Since F is continuous on [a, b] and differentiable on (a, b), so f1 , . . . , fn are
i=1
continuous on [a, b] and differentiable on (a, b). Hence φ is continuous on [a, b] and differentiable on
0
(a, b). By the mean value
theorem, there exists
c ∈ (a, b) such that
φ(b) − φ(a) =
φ (c)(b − a). Now
φ(b)−φ(a) = F (b)−F (a) ·F (b)− F (b)−F (a) ·F (a) = F (b)−F (a) · F (b)−F (a) = kF (b)−F (a)k2.
By the Cauchy-Schwarz inequality,

X
n


kF (b) − F (a)k2 = φ(b) − φ(a) = φ0 (c)(b − a) = fi (b) − fi (a) fi0 (c)(b − a)
i=1
v v
u n u n
uX
2uX
≤ t fi (b) − fi (a) t fi0 (c)2 (b − a) = kF (b) − F (a)kkF 0(c)k(b − a).
i=1 i=1

Cancelling kF (b) − F (a)k on both ends, we get kF (b) − F (a)k ≤ kF 0(c)k(b − a).

∂u
14. Define F : R2 → R2 by F (x, y) = (u, v), where u = x2ey + cos xy + x and v = xexy + y2 . Now
∂x
y ∂u 2 y ∂v xy xy ∂v 2 xy
= 2xe − y sin xy + 1, = x e − x sin xy, = e + xye , = x e + 2y are continuous near (x, y) =
∂y ∂x  ∂y

1 0
(0, 1). So F is C 1 near (x, y) = (0, 1). Also, det F 0(0, 1) =   = 2 6= 0. By the inverse function the-
1 2
orem, near (x, y) = (0, 1), F −1(u, v) = (x, y) exists and is differentiable. Also
 ∂x   −1    
∂u (1, 1)
∂x
∂v (1, 1) −1 0 0

−1 1 0 1 2 0 1 0
= (F ) (F (0, 1)) = F (0, 1) = = = .
∂y
∂u (1, 1)
∂y
∂v (1, 1)
| {z } 1 2 2 −1 1 − 12 1
2
=(1,1)
 ∂u 1
Then (0, 1) = 0 6= −2 = ∂y
.
∂y ∂u (1, 1)

∂x ∂x ∂y ∂y
15. Define F : R2 → R2 by F (u, v) = (x, y), where x = u+v and y = u2+v. Now = 1, = 1, = 2u, =1
∂u ∂v  ∂u  ∂v
 1 1 
are continuous near (u, v) = (1, −1). So F is C 1 near (u, v) = (1, −1). Also, det F 0(1, −1) =  = −1 6= 0.
2 1
By the inverse function theorem, near (u, v) = (1, −1), F −1(x, y) = (u, v) exists and is differentiable.
Also  ∂u ∂u   −1  
∂x ∂y −1 0 0

−1 1 1 1 1 −1
∂v ∂v = (F ) (x, y) = F (u, v) = = .
∂x ∂y
2u 1 1 − 2u −2u 1

∂u 1 ∂2u ∂  1  ∂  1  ∂u 2 1
So = and = = = .
∂x 1 − 2u ∂x2 ∂x 1 − 2u ∂u 1 − 2u ∂x (1 − 2u)2 1 − 2u

∂x
16. Define F : R2 → R2 by F (v, w) = (x, y), where x = v + 2w + evw and y = e−vw + w + 2v. Now
∂v
∂x ∂y ∂y
= 1 + wevw , = 2 + vevw , = −we−vw + 2, = −ve−vw + 1 are continuous near (v, w) = (0, 0).
∂w ∂v ∂w 
1 2
1 0 
So F is C near (v, w) = (0, 0). Also, det F (0, 0) =   = −3 6= 0. By the inverse function theorem,
2 1
near (v, w) = (0, 0), F −1(x, y) = (v, w) exists and is differentiable. Also
 ∂v ∂v   −1    
∂x (1, 1) ∂y (1, 1) −1 0 0

−1 1 2 1 1 −2 − 13 2
= (F ) (F (0, 0)) = F (0, 0) = = = 3 .
2
∂w
∂x (1, 1)
∂w
∂y (1, 1) | {z } 2 1 −3 −2 1 3
− 13
=(1,1)

∂v ∂v 1 2 1 2  1  ∂w ∂w
Then (1, 1) + (1, 1) = − + = = + − = (1, 1) + (1, 1).
∂x ∂y 3 3 3 3 3 ∂x ∂y

17. Define F : R2 → R2 by F (v, w) = (x, y), where x = w cos v + v cos w + ew and y = w sin v + v sin w + ev .
∂x ∂x ∂y ∂y
Now = −w sin v + cos w, = cos v − v sin w + ew , = w cos v + sin w + ev , = sin v + v cos w
∂v ∂w ∂v ∂w 
1 2
1 0 
are continuous near (v, w) = (0, 0). So F is C near (v, w) = (0, 0). Also, det F (0, 0) =   = −2 6= 0.
1 0
By the inverse function theorem, near (v, w) = (0, 0), F −1(x, y) = (v, w) exists and is differentiable.
Also
 ∂v ∂v   −1    
∂x (1, 1) ∂y (1, 1) −1 0 0

−1 1 2 1 0 −2 0 1
= (F ) (F (0, 0)) = F (0, 0) = = = 1 .
∂w
∂x
(1, 1) ∂w
∂y
(1, 1) | {z } 1 0 −2 −1 1 2 − 12
=(1,1)

∂v ∂v ∂w ∂w 1  1
Then (1, 1) + (1, 1) + (1, 1) + (1, 1) = 0 + 1 + + − = 1.
∂x ∂y ∂x ∂y 2 2
∂x
18. Define F : R2 → R2 by F (u, v) = (x, y), where x = cos u + sin v and y = sin u + cos v. Now
∂u
∂x ∂y ∂y 1
= − sin u, = cos v, = cos u, = − sin v are continuous near (u, v) = (0, 0). So F is C near
∂v ∂u ∂v 
 0 1 
(u, v) = (0, 0). Also, det F 0(0, 0) =  = −1 6= 0. By the inverse function theorem, near (u, v) =
1 0
−1
(0, 0), F (x, y) = (u, v) exists and is differentiable. Also
 ∂u ∂u   −1  
∂x ∂y
−1 − sin u cos v 1 − sin v − cos v
∂v ∂v = (F −1)0 (x, y) = F 0(u, v) = = ,
∂x ∂y
cos u − sin v ∆ − cos u − sin u
  ∂u 2  ∂u 2 sin2 v cos2 v 1 cos2 u sin2 u
where ∆ = sin u sin v − cos u cos v. Then + = 2
+ 2
= 2 = + =
∂x ∂y ∆ ∆ ∆ ∆2 ∆2
 ∂v 2  ∂v 2
+ .
∂x ∂y
19. Consider F : R2 → R defined by F (x, y) = x3 + xy5 + y2 − 3. Then F (1, 1) = 0. Near (1, 1),
∂F ∂F ∂F
= 3x2 + y5 , = 5xy4 + 2y are continuous. So F is C 1 near (1, 1). Now (1, 1) = 7 6= 0. By
∂x ∂y ∂y
the implicit function theorem, near (1, 1), the equation can be expressed as y = f(x) for some C 1 (hence
dy dy
differentiable) function f. By implicit differentiation, we get 3x2 + y5 + 5xy4 + 2y = 0. Solving
dx dx
2 5
dy dy 3x + y dy 4
for , we get =− . So (1, 1) = − .
dx dx 5xy4 + 2y dx 7

20. Define F : R3 → R2 by F (u, v, w) = (f1 , f2), where f1 (u, v, w) = euv + w and f2 (u, v, w) = u sin vw − v.
∂f1 ∂f1 ∂f1 ∂f2 ∂f2
Then F (1, 0, −1) = (0, 0). Now = veuv , = ueuv , = 1, = sin vw, = uw cos vw − 1,
∂u ∂v ∂w ∂u ∂v  
∂f2 ∂(f1 , f2)  1 1
= uv cos vw are continuous near (1, 0, −1). So F is C 1 near (1, 0, −1). Next (p) =   = 2 6= 0.
∂w ∂(v, w) −2 0 
By the implicit function theorem, near p, v and w can be expressed as differentiable functions of u. Now
d uv dv dw dv d dv dw
0= (e + w) = euv (v + u ) + and = (u sin vw) = sin vw + u cos vw( w + v ).
du du du du du du du
dv dw dv dv dv dw
At p, these become 0 = + and = − , which implies (p) = 0, (p) = 0.
du du du du du du
21. (i) Define F : R4 → R2 by F (x, y, u, v) = (f1 , f2), where f1 (x, y, u, v) = ex + ye − u + v − 1 and
f2 (x, y, u, v) = eu − ve − x + y − 1. Then F (0, 1, 0, 1) = (0, 0). Now
∂f1 ∂f1 ∂f1 ∂f1 ∂f2 ∂f2 ∂f2 ∂f2
= ex , = eye−1 , = −1, = −1, = −1, = 1, = eu , = −eve−1
∂x ∂y ∂u ∂v ∂x ∂y ∂u ∂v
 
∂(f1 , f2)  1 e
1
are continuous near (0, 1, 0, 1). So F is C near (0, 1, 0, 1). Next 
(0, 1, 0, 1) =   = 1 + e 6= 0.
∂(x, y) −1 1 
By the implicit function theorem, near (0, 1, 0, 1), (x, y) can be expressed as a differentiable function of
(u, v).
∂x ∂y ∂x ∂y
(ii) Taking partial derivatives with respect to u, we get ex + eye−1 − 1 = 0 and eu − + = 0.
∂u ∂u ∂u ∂u
∂x ∂y ∂x ∂y
When (u, v) = (0, 1), (x, y) = (0, 1) and we get (0, 1) + e (0, 1) − 1 = 0 and 1 − (0, 1) + (0, 1) = 0,
∂u ∂u ∂u ∂u
∂x ∂y
which implies (0, 1) = 1, (0, 1) = 0.
∂u ∂u
22. Since x, y are functions of w, z, taking partial differentiation of f1 (w, x, y, z) = 0 and f2 (w, x, y, z) = 0
with respect to w, we get
∂f1 ∂w ∂f1 ∂x ∂f1 ∂y ∂f1 ∂z ∂f2 ∂w ∂f2 ∂x ∂f2 ∂y ∂f2 ∂z
+ + + = 0 and + + + = 0.
∂w |{z}
∂w ∂x ∂w ∂y ∂w ∂z |{z}
∂w ∂w |{z}
∂w ∂x ∂w ∂y ∂w ∂z |{z}
∂w
=1 =0 =1 =0

∂f1 ∂x ∂f1 ∂y ∂f1 ∂f2 ∂x ∂f2 ∂y ∂f2 ∂x ∂y

Transfering terms, we get + =− and + =− . Treating ,
∂x ∂w ∂y ∂w ∂w ∂x ∂w ∂y ∂w ∂w ∂w ∂w
as unknowns and solving for them, we find
∂f2 ∂f1 ∂f1 ∂f2 ∂(f1 , f2) ∂f2 ∂f1 ∂f1 ∂f2 ∂(f1 , f2 )
− − − −
∂x ∂w ∂y ∂w ∂y ∂(w, y) ∂y ∂(x, w)
= = and = ∂x ∂w ∂x ∂w = .
∂w ∂f1 ∂f2 ∂f2 ∂f1 ∂(f1 , f2) ∂w ∂f1 ∂f2 ∂f2 ∂f1 ∂(f1 , f2 )
− −
∂x ∂y ∂x ∂y ∂(x, y) ∂x ∂y ∂x ∂y ∂(x, y)
23. Define F : R4 → R2 by F (x, y, z, w) = (f1 , f2), where f1 (x, y, z, w) = ex+y+z+w − cos(x + y) − sin(z + w)
and f2 (x, y, z, w) = e−x−y−z−w − sin(x + y) − cos(z + w). Let p = (1, −1, 1, −1), then F (p) = (0, 0). Now

∂f1 ∂f1 ∂f1 ∂f1

= ex+y+z+w + sin(x + y) = , = ex+y+z+w − cos(z + w) = ,
∂x ∂y ∂z ∂w

∂f2 ∂f2 ∂f2 ∂f2

= −e−x−y−z−w − cos(x + y) = , = −e−x−y−z−w + sin(z + w) =
∂x ∂y ∂z ∂w
 
∂(f1 , f2)  1 0 
are continuous near (1, −1, 1, −1). So F is C 1 near p. Next (p) =  = −1 6= 0. By the
∂(x, z) −2 −1 
implicit function theorem, near p, (x, z) can be expressed as a differentiable function of (y, w). Taking
partial derivatives of the equations with respect to w, we get
 ∂x
∂z  ∂x  ∂z 
ex+y+z+w +
+ 1 = − sin(x + y) + cos(z + w) +1
∂w ∂w ∂w ∂w
   ∂z 
−x−y−z−w ∂x ∂z ∂x
−e + + 1 = cos(x + y) − sin(z + w) +1 .
∂w ∂w ∂w ∂w
∂x ∂z ∂z ∂x ∂z ∂x ∂x ∂z
At p, we get + +1 = + 1, − − −1= . So = 0, = −1.
∂w ∂w ∂w ∂w ∂w ∂w ∂w ∂w
Alternatively, by exercise 22, at p, we get
 .
∂x ∂(f1 , f2) . ∂(f1 , f2)  0
 0 
=− = − (−1) = 0,
∂w ∂(w, z) ∂(x, z) −1 −1 
 .
∂z ∂(f1 , f2) . ∂(f1 , f2 )  1
 0 
=− = − (−1) = −1.
∂w ∂(x, w) ∂(x, z) −2 −1 

24. Define F : R4 → R2 by F (w, x, y, z) = (f1 , f2 ), where f1 (x, y, z, w) = x3 + y3 + z 3 + 3y − xw3 and

f2 (x, y, z, w) = x5 + y5 z 3 − xyz − x3w5 . Then F (p) = (0, 0). Now

∂f1 ∂f1 ∂f1 ∂f1

= −3xw2, = 3x2 − w3, = 3y2 + 3, = 3z 2 ,
∂w ∂x ∂y ∂z

∂f2 ∂f2 ∂f2 ∂f2

= −5x3 w4, = 5x4 − yz − 3x2w5, = 5y4 z 3 − xz, = 3y5 z 2 − xy
∂w ∂x ∂y ∂z
 
∂(f1 , f2 ) 2 3
1
are continuous near p. So F is C near p. Next 
(p) =   = −6 6= 0. By the implicit function
∂(x, y) 2 0
theorem, near p, (x, y) can be expressed as a differentiable function of (w, z). Taking partial derivatives
of the equations with respect to w, we get

∂x ∂y ∂y ∂x ∂x ∂y ∂x ∂y ∂x
3x2 + 3y2 +3 = w3 + 3xw2, 5x4 + 5y4 z 3 = yz + xz + 3x2w5 + 5x3w4.
∂w ∂w ∂w ∂w ∂w ∂w ∂w ∂w ∂w
∂x ∂y ∂x ∂x ∂x ∂x 5 ∂y 2
At p, we get 3 +3 = + 3 and 5 =3 + 5. So (p) = , (p) = − .
∂w ∂w ∂w ∂w ∂w ∂w 2 ∂w 3
Alternatively, by exercise 22, at p, we get
 
∂x ∂(f1 , f2) . ∂(f1 , f2)  −3 3  .
  (−6) = 5 ,
=− = −
∂w ∂(w, y) ∂(x, y) −5 0  2
 
∂y ∂(f1 , f2) . ∂(f1 , f2)  2 −3  .
 (−6) = − 2 .
=− = − 
∂w ∂(x, w) ∂(x, y) 2 −5  3
25. (a) Define F : R4 → R2 by F (w, x, y, z) = (f1 , f2 ), where f1 (x, y, z, w) = cos w + sin x + tan y − z − 1
and f2 (x, y, z, w) = ew + x − y − e−z . Then F (0, 0, 0, 0) = (0, 0). Now

∂f1 ∂f1 ∂f1 ∂f1 ∂f2 ∂f2 ∂f2 ∂f2

= − sin w, = cos x, = sec 2 y, = −1, = ew , = 1, = −1, = e−z
∂w ∂x ∂y ∂z ∂w ∂x ∂y ∂z
 
∂(f1 , f2) 1 1 
1
are continuous near (0, 0, 0, 0). So F is C near (0, 0, 0, 0). Next 
(0, 0, 0, 0) =   = −2 6= 0.
∂(x, y) 1 −1 
By the implicit function theorem, near p, (x, y) can be expressed as a differentiable function G of (w, z).
∂x ∂y
(b) Taking partial derivatives of the equations with respect to w, we get − sin w + cos x + sec2 y =0
∂w ∂w
∂x ∂y ∂x ∂y ∂x ∂y ∂x 1 ∂y 1
and ew + = . At (0, 0, 0, 0), we get + = 0 and 1 + = . Then (0, 0) = − , (0, 0) = .
∂w ∂w ∂w ∂w ∂w ∂w ∂w 2 ∂w 2

Alternatively, by exercise 22, at p, we get

 .
∂x ∂(f1 , f2) . ∂(f1 , f2) 0 1  1
=− = −  (−2) = − ,
∂w ∂(w, y) ∂(x, y) 1 −1  2
 
∂y ∂(f1 , f2 ) . ∂(f1 , f2) 1 0.
 (−2) = 1 .
=− = − 
∂w ∂(x, w) ∂(x, y) 1 1 2

∂F
26. Since F is C 1 near (0, 0, 0), F (0, 0, 0) = 0,(0, 0, 0) 6= 0, by the implicit function theorem, near (0, 0, 0),
∂x

the equation F (x, y, z) = 0 can be expressed in the form x = f(y, z). So near (0, 0, 0), F f(y, z), y, z ≡ 0.
∂
∂F ∂x ∂F ∂y ∂F ∂z ∂x ∂F . ∂F ∂F
Hence 0 = F f(y, z) , y, z = + + . Then =− . ( Note 6= 0
∂y | {z } ∂x ∂y ∂y ∂y ∂z ∂y ∂y ∂y ∂x ∂x
=x |{z} |{z}
=1 =0
∂F ∂F
near (0, 0, 0) because (0, 0, 0) 6= 0 and is continuous near (0, 0, 0) as F is C 1 near (0, 0, 0).)
∂x ∂x
∂F ∂F
Similarly, (0, 0, 0) 6= 0 and (0, 0, 0) 6= 0, so it can also be expressed in the form y = g(x, z)
∂y ∂z
∂y ∂F . ∂F ∂z ∂F . ∂F
and z = h(x, y). As above, we get =− and =− . Multiplying these together,
∂z ∂z ∂y ∂x ∂x ∂z
∂x ∂y ∂z
we get = −1.
∂y ∂z ∂x

27. For the case x ∈ R, suppose lim xn 6= x. Then there exists ε > 0 such that for every K ∈ N, there is
n→∞
n ≥ K such that |xn − x| ≥ ε. Note |xn − x| ≥ ε if and only if xn ∈ I = [−∞, x − ε] ∪ [x + ε, +∞].
So for K = 1, there is n = n1 ≥ K = 1 with xn1 ∈ I. For K = n1 + 1, there is n = n2 ≥ K = n1 + 1
with xn2 ∈ I, etc. Thus we can get a subsequence xn1 , xn2 , xn3 , . . . ∈ I. Since subsequences of xnk are
subsequences of xn , so the L set for xnk is a subset of the L set of xn. Also, since xnk ∈ I, it is a subset
of I. These imply the L set for xnk is {x} and x ∈ I, a contradiction. The case x = −∞ or x = +∞ is
similar.

n2 1 − cos e−n n2 1 − cos θ 2 sin θ

28. (a) lim n2(1 − cos e−n) = lim n −n
= lim n lim = lim n lim = 0. There-
n→∞ n→∞ e e n→∞ e θ→0 θ n→∞ e θ→0 1
fore, liminf xn = 0 = limsup xn .
n→∞ n→∞

1
(b) By calculus, we see that f(x) = x ln x has an absolute minimum value − attained at x = e−1 and
e
 1  n cos(nπ/2)−1 1
lim f(x) = +∞. So L ⊆ − , +∞ . For n = 1, 3, 5, . . ., we have xn = f(e ) = f(e−1 ) = − .
x→+∞ e e
 1
So x1, x3, x5, . . . converges to − . For n = 4, 8, 12, . . ., en cos(nπ/2)−1 = en−1 → +∞, which implies
e
1
x4, x8, x12, . . . converges to +∞. So liminf xn = − and limsup xn = +∞.
n→∞ e n→∞

1 1 − cos θ
29. (i) Note |xn| ≤ n(1 − cos n1 ). Setting θ = 1
n
, we have lim n(1 − cos ) = lim = lim sin θ = 0.
n→∞ n θ→0 θ θ→0
By sandwich theorem, lim xn = 0. So liminf xn = 0 = limsup xn .
n→∞ n→∞ n→∞

(ii) Note −1 ≤ xn ≤ 1. So L ⊆ [−1, 1]. Now x5 = x15 = x25 = · · · = −1 implies x5, x15, x25, . . . converges
to −1. Also x10 = x20 = x30 = · · · = 1 implies x10, x20, x30, . . . converges to 1. Therefore, liminf xn = −1
n→∞
and limsup xn = 1.
n→∞

n/2 1 1 1  1 1
30. (i) Note |xn| ≤ = . So − ≤ xn ≤ . Then L ⊆ − , . Now observe that if n = 2k + 1,
2 2 2 2 2 2
−k 1 k 1
then xn = and so x1, x3, x5, . . . converges to − . Also, if n = 2k, then xn = = and so
2k + 1 2 2k 2
1 1 1
x2, x4, x6, . . . converges to . Therefore, liminf xn = − and limsup xn = .
2 n→∞ 2 n→∞ 2
1 1
1
1
1 1 1 1
(ii) Expand xn = 1(1 + ) 1 + 2 1 + 4 · · · 1 + 2n−2 to get 1 + + 2 + 3 + · · · + 2n−1 −1 .
π π π π π π π π
1 1 1 1 1
So lim xn = 1 + + 2 + 3 + · · · = . Therefore, liminf xn = = limsup xn.
n→∞ π π π 1 − (1/π) n→∞ 1 − (1/π) n→∞

n sin(e−n ) n sin θ 1
31. (i) Note lim n sin(e−n ) = lim = lim n lim = lim n = 0. So liminf xn = 0 limsup xn.
n→∞ n→∞ en e−n n→∞ e θ→0 θ n→∞ e n→∞ n→∞

(ii) Note 0 < xn < +∞. So L ⊆ [0, +∞]. Now observe that if n = 7(2k + 1), then xn = 2−7(2k+1) and
so x7 , x21, x35, . . . converges to 0. Also, if n = 14k, then xn = 214k and so x14, x28, x42, . . . converges to
+∞. Therefore, liminf xn = 0 and limsup xn = +∞.
n→∞ n→∞
    
 an+1  (n + 1)2 /(2n) if n is odd  an+1 
32. Note 0 <    
 = (n + 1)/(2n2) if n is even < +∞. For odd n → ∞,  an  → +∞. For even n →
  an     √
 an+1   an+1   an+1  p n
n/2 if n is odd

∞,   
→ 0. Therefore, liminf   
= 0 and limsup   n
= +∞. Next, |an| = √ .
an  n→∞ an  n→∞ an  ( n
n) 2
/2 if n is even
√ p 1 p 1 p
Since lim n n = 1, so lim n |an| = . Therefore, liminf n |an| = = limsup n |an|.
n→∞ n→∞ 2 n→∞ 2 n→∞

33. (a) Let Mk = sup{ak , ak+1, ak+2, . . .} and Nk = sup{bk , bk+1, bk+2, . . .}, then ak +bk , ak+1 +bk+1 , ak+2 +
bk+2, . . . are at most Mk + Nk . So sup{ak + bk , ak+1 + bk+1, ak+2 + bk+2, . . .} ≤ Mk + Nk . Taking limit
as k → ∞ on both sides, we get limsup(an + bn) ≤ limsup an + limsup bn.
n→∞ n→∞ n→∞
n
(b) Let an = (−1) , then −1 ≤ an ≤ 1. So L ⊆ [−1, 1]. Since x2 = x4 = x6 = · · · = 1, so x2 , x4, x6, . . .
converges to 1. Therefore, limsup an = 1. Similarly, let bn = (−1)n+1 = −an , then limsup bn = 1. We
n→∞ n→∞
get limsup(an + bn) = 0 6= 2 = limsup an + limsup bn .
n→∞ n→∞ n→∞

34. (a) The fourth inequality is proved in the previous exercise and the first inequality follows similarly. For
the third inequality, let mk = inf{ak , ak+1, ak+2, . . .} and let limsup bn = lim bnj . Then
n→∞ j→∞

mnj + bnj ≤ anj + bnj ≤ sup{anj + bnj , anj +1 + bnj +1 , anj +2 + bnj +2 , . . .}.

So liminf an + limsup bn = lim mnj + lim bnj ≤ lim (sup{anj + bnj , anj +1 + bnj +1 , . . .}) = limsup(an + bn).
n→∞ n→∞ j→∞ j→∞ j→∞ n→∞
The second inequality follows similarly.
 1 1
(b) Since 1 = lim (an + bn) = limsup(an + bn) ≥ liminf an + limsup bn ≥ + limsup bn, so ≥ limsup bn .
n→∞ n→∞ n→∞ n→∞ 2 n→∞ 2 n→∞
1 1 1
Also, limsup bn ≥ liminf bn ≥ . Therefore, both limsup bn and liminf bn equal , which implies lim bn = .
n→∞ n→∞ 2 n→∞ n→∞ 2 n→∞ 2
1
Similarly, lim an = .
n→∞ 2

35. Let Mk = sup{ak , ak+1, ak+2, . . .} and Nk = sup{bk , bk+1, bk+2, . . .}. Since an, bn are nonnegative
real numbers, ak bk , ak+1bk+1, ak+2bk+2, . . . are at most Mk Nk . So sup{ak bk , ak+1bk+1, ak+2bk+2 , . . .} ≤
Mk Nk . Taking limit as k → ∞ on both sides, we get limsup(an bn ) ≤ (limsup an )(limsup bn ).
n→∞ n→∞ n→∞

36. Let an be the sequence 0, 1, 0, 1, 0, 1, . . . and bn be the sequence 1, 0, 1, 0, 1, 0, . . .. Then an bn = 0 for all
n. So limsup(an bn) = 0 6= 1 = (limsup an )(limsup bn).
n→∞ n→∞ n→∞

37. The middle inequality is trivial. We will prove the rightmost inequality. (The leftmost inequal-
ity is similar.) Now limsup an = lim Mk = M, where Mk = sup{ak , ak+1, . . .}. If M = +∞, then
n→∞ k→∞
limsup σn ≤ +∞ = M = limsup an . If M < +∞, then for every real number r > M, there is Mk
n→∞ n→∞
such that Mk < r. So ak , ak+1, ak+2, . . . ≤ Mk < r. Let c = a1 + · · · + ak−1. Then for every n > k,
(a1 + · · · + ak−1) + (ak + · · · + an ) c (n − k + 1)r c (n − k + 1)r 
σn = ≤ + . So limsup σn ≤ limsup + .
n n n n→∞ n→∞ n n
c (n − k + 1)r 
Since lim + = r, we have limsup σn ≤ r. Letting r → M +, we get
n→∞ n n n→∞

limsup σn ≤ M = limsup an .
n→∞ n→∞

 (x − 1)k+1 (2k)!  |x − 1|
 
38. (a) lim   = lim = 0 < 1 for all x. So the domain is (−∞, +∞).
k→∞ (2(k + 1))! (x − 1)k k∞ (2k + 1)(2k + 2)
 xk+1 ln k  |x| ln k |x|/k
 
(b) lim   = lim = lim = |x| < 1 ⇔ x ∈ (−1, 1). For x = 1, we have
k→∞ ln(k + 1) xk k∞ ln(k + 1) k→∞ 1/(k + 1)
X∞ X∞ X∞
1 1 (−1)k
≥ , which diverges by p-test. For x = −1, converges by the alternating series
ln k k ln k
k=2 k=2 k=2
1
test as k % ∞, ln k % ∞, & 0. So the domain is [−1, 1).
ln k
 (2x − 3)k+1 4k  |2x − 3| 1 7
1 X (2x − 3)k
∞ X∞
 
(c) lim   = < 1 ⇔ x ∈ − , . For x = − , = (−1)k ,
k→∞ 4(k+1) (2x − 3)k 4 2 2 2 4k
k=1 k=1
7 X (2x − 3)k X
∞ ∞
which diverges by term test. For x = , = 1, which diverges. So the domain is
2 4k
k=1 k=1
1 7

− , .
2 2

 e(k+1)x k + 1   < 1 if x ∈ (−∞, 0)
  x
(d) (This is not a power series!) lim   = e = 1 if x = 0 . For x = 0, we have
k→∞ k + 2 ekx 
> 1 if x ∈ (0, +∞)
X∞
ekx X
∞
1
= , which diverges by integral test. So the domain is (−∞, 0).
k+1 k+1
k=1 k=1

39. Let f(x) = tan x. Then

f(0) = 0, f 0 (0) = sec2 0 = 1, f 00 (0) = 2 sec2 0 tan 0 = 0, f 000(0) = 4 sec2 0 tan2 0 + 2 sec4 0 = 2,
 f (iv) (0) = 8 sec2 0 tan3 0 + 16 sec 4 0 tan 0 = 0, f (v) (0) = 16 sec2 0 tan4 0 + 88 sec 4 0 tan2 0 + 16 sec6 0 = 16.
1 2
So the first three nonzero term of the Taylor series of tan x with center at c = 0 is x + x3 + x5.
3 15
Next
x3 x5
sin x x− + − ··· 1 2
= 6 120 = x + x3 + x5 + · · · .
cos x 2 4 3 15
x x
1− + −···
2 24

X
∞ X
∞ X
∞ X
∞
40. The geometric series xk and −xk have centers at c = 0 and radii R = 1. Their sum (1 − 1)xk = 0xk
k=0 k=0 k=0 k=0
= 0 has radius R = ∞.

1 + cos 2x  cos2 x if n = 0
41. For f(x) = cos2 x = , we have f (n) (x) = (−1)k 22k−2 sin 2x if n = 2k − 1, k ∈ N . For x =
2 
 (−1)k 22k−1 cos 2x if n = 2k, k ∈ N
1 if n = 0 X∞
(−1)k 22k−1 2k
0, f (n) (0) = 0 if n = 2k − 1, k ∈ N . Then the Taylor series of f at c = 0 is 1 + x .
 (2k)!
(−1)k 22k−1 if n = 2k, k ∈ N k=1
Since |f (n) (x)| ≤ 1 · 2n for all x ∈ (−∞, +∞), by the Taylor series theorem,the Taylor series converges
pointwise on (−∞, +∞) to cos2 x. (We may also get
X
∞
(−1)k (2x)2k
1+ ∞
1 + cos 2x k=0
(2k)! X (−1)k 22k−1
cos2 x = = =1+ x2k .
2 2 (2k)!
k=1

(−1)k 22k−1
However, it needs to be shown that are the Taylor coefficients of cos2 x at c = 0.)
(2k)!
2 2
42. Step 1: we will show lim tn e−t = 0 for n = 0, 1, 2, . . . by induction. For n = 0, we have lim e−t = 0.
t→∞ t→∞
t  1 
For n = 1, we have lim t2 = lim 2 = 0 by l’Hopital’s rule. For n ≥ 2, by l’Hopital’s rule and
t→∞ e t→∞ 2tet
case n − 2, we have
tn  ntn−1 n tn−2 
lim t2 = lim 2 = lim 2 = 0 .
t∈∞ e t→∞ 2tet 2 t→∞ et
1
−1/x2
Step 2: we will show that for x 6= 0, n = 1, 2, 3, . . ., f (n) (x) = Pn e , where Pn (x) is a polynomial
x
2 2

1 −1/x2
by induction. For n = 1, f 0 (x) = 3 e−1/x = P1 e , where P1(x) = 2x3. For n ≥ 2, assuming
x x
case n − 1,
d 1
−1/x2   1 0 1
1
2  −1/x2 1
−1/x2
f (n) (x) = Pn−1 e = − 2 Pn−1 + Pn−1 3
e = Pn e ,
dx x x x x x x
where Pn(x) = −x2Pn−1
0
(x) + 2x3Pn−1(x) is a polynomial as Pn−1(x) is a polynomial by case n − 1.
Step 3: we will show f (n) (0) = 0 for n = 1, 2, 3, . . . by induction. For n = 1,
f(x) − f(0) 1 2 2
f 0 (0) = lim = lim e−1/x = lim te−t = 0
x→0 x−0 x→0 x t→∞

by step 1. For n ≥ 2, assuming f (n−1) (0) = 0, by step 2,

f (n−1) (x) − f (n−1) (0) 1 1
−1/x2 2

f (n) (0) = lim = lim Pn−1 e = lim tPn−1(t)e−t = 0
x→0 x−0 x→0 x x t→∞
 because tPn−1(t) is a polynomial and the limit can be calculated termwise using step 1.

xn
43. (a) The pointwise limit function on [0, 0.99] is f(x) = lim = 0. Now we have lim kfn − fk[0,0.99]
n→∞ 1 + xn n→∞
0.99n
= lim = 0. So we have uniform convergence.
n→∞ 1 + 0.99n
2
(b) The pointwise limit function on [0.1, ∞) is g(x) = lim nxe−nx = 0 by l’Hopital’s rule. By calculus,
n→∞
−nx2 1 1 1
we see that gn (x) = nxe is decreasing when x > √ . Since lim √ = 0, when √ < 0.1, the
2n n→∞ 2n 2n
maximum of |gn(x)| on [0.1, ∞) is 0.1ne−0.01n. So lim kgn − gk[0.1,∞) = lim 0.1ne−0.01n = 0 and we
n→∞ n→∞
have uniform convergence.
xn rn
(c) The pointwise limit function on [0, r] is h(x) = lim cos xn = 1. Now | cos xn − 1| = 2 sin2 ≤ 2 sin2
n→∞
n
2 2
2 r
= Ln for all x ∈ [0, r]. Since lim Ln = lim 2 sin = 0, by L-test, we have uniform convergence. Al-
n→∞ n→∞ 2
ternatively, we can also use the mean-value theorem, | cos xn − 1| = |(sin c)(xn − 0)| ≤ |xn| ≤ rn → 0 as
n → ∞.

Remarks. Parts (a) and (c) can also be done by using Dini’s theorem since the settings are on closed
and bounded intervals. For part (a), as above fn (x) and f(x) are continuous and for every x ∈ [0, 0.99],
d  w  1 xn
as n % ∞, xn & 0 and = > 0 implies fn (x) = & 0. So Dini’s theorem
dw 1 + w (1 + w)2 1 + xn
implies uniform convergence. For part (c), as above hn (x) and h(x) are continuous and for every x ∈
[0, r], as n % ∞, xn & 0 and hn (x) = cos xn % 1. Again Dini’s theorem implies uniform convergence.

44. The pointwise limit of Sn (x) on [0, 1) is S(x) = lim xn = 0. Since

n→∞

kSn − Sk[0,1) = sup{|Sn (x) − S(x)| : x ∈ [0, 1)} = sup{xn : x ∈ [0, 1)} = 1,

lim kSn − Sk[0,1) = 1 6= 0. So Sn (x) does not converge uniformly on [0, 1).
n→∞

 sin kx  1 X∞
1/e
 
45. (a)  k  ≤ Mk = k for all x ∈ (−∞, +∞) and Mk = , so by Weierstrass’ M -test, the
e e 1 − (1/e)
k=1
series converges uniformly on (−∞, +∞).
 xk  Z ∞ X∞
  1 dx 1
(b)   ≤ Mk = for all x ∈ [−1, 1] and = implies Mk converges by
k(ln k)2 k(ln k)2 2 x(ln x)2 ln 2
k=1
the integral test. So by Weierstrass’ M -test, the series converges uniformly on [−1, 1].
p ∞
X
(c) |ke−kx| ≤ Mk = ke−k for all x ∈ [1, ∞) and lim k
Mk = e−1 < 1 implies Mk converges by the
k→∞
k=1
root test. So by Weierstrass’ M -test, the series converges uniformly on [1, ∞).
1 1
(d) By calculus, we have max |x ln x| = max (−x ln x) = . So |(x ln x)k | ≤ Mk = k for all x ∈ (0, 1].
x∈(0,1] x∈(0,1] e e
X
∞
1/e
Now Mk = , so by Weierstrass’ M -test, the series converges uniformly on (0, 1].
1 − (1/e)
k=1
  p X
∞
 
(e) k2 xk  ≤ Mk = k2 rk for all x ∈ [−r, r] and lim k Mk = r < 1 implies Mk converges by root
k→∞
k=1
test. So by Weierstrass’ M -test, the series converges uniformly on [−r, r].
  xk  rk p X∞
 
(f)   ≤ Mk = for all x ∈ [0, r) and lim k Mk = r < 1 implies Mk converges by root test. So
k k k→∞
k=1
by Weierstrass’ M -test, the series converges uniformly on [0, r).
kx 1 1 1

(g) By calculus, kx has a maximum value on [0, +∞) at x = and it is decreasing on , +∞ . As
e e k k
1 1 1
lim = 0, there is a least positive integer N such that < r. For k = 1, 2, . . ., N − 1, we have r ≤ ,
k→∞ k N k
 kx  1 1  kx  kr
   
so  kx  ≤ Mk = on [r, +∞). For k = N, N + 1, . . . , we have < r, so  kx  ≤ Mk = kr on [r, +∞).
e e k r e e
X∞
1 X kr
∞
k kr 1
Then Mk = (N − 1) + converges by root test as lim = r < 1. So by Weierstrass’
e ekr k→∞ ekr e
k=1 k=N
M -test, the series converges uniformly on [r, +∞).

n 0 if x 6= π/2
46. (a) On [0, π], Sn (x) converges pointwise to S(x) = lim sin x = . Suppose Sn (x) con-
n→∞ 1 if x = π/2
verge uniformly on [0, π]. Since each Sn (x) is continuous on [0, π], by the continuity theorem, S(x)
should be continuous on [0, π], but S(x) is not continuous at π/2. We have a contradiction.
X∞ n
(b) On (0, 1], xk (1 − x) converges pointwise to S(x) = (1 − x)/(1 − x) = 1 if x 6= 1 . Suppose the
k=0
0 if x = 1
X∞
series xk (1 − x) converge uniformly on (0, 1]. Since each term is continuous on (0, 1], by the continuity
k=0
theorem, S(x) should be continuous on (0, 1], but S(x) is not continuous at 1. We have a contradiction.
2
(c) On [0, 1], Sn (x) converges pointwise to S(x) = lim nxe−nx = 0 by l’Hopital’s rule. Suppose Sn (x)
n→∞
converge uniformly on [0, 1]. Since each Sn (x) is integrable on [0, π], by the integration theorem,
Z 1 Z 1 Z 1
2 1 1 1
0= S(x) dx = lim Sn (x) dx = lim nxe−nx dx = lim (− e−n + ) = ,
0 n→∞ 0 n→∞ 0 n→∞ 2 2 2
which is a contradiction.

47. Let Sn (x) = (−1)n for all x ∈ [0, 1]. Then Sn (x) does not converge pointwise on [0, 1] and so it cannot
converge uniformly on [0, 1]. However, Sn2 (x) = 1 for all x ∈ [0, 1] and it converges uniformly on [0, 1] to
S(x) = 1.

1
48. Let fn (x) = x and gn (x) = on E = R. Then fn (x) converges uniformly on E to x as kfn −xkE = kx−
n
1 x
xkE = 0 and gn(x) converges uniformly on E to 0 as lim kgn − 0kE = lim = 0. Now fn (x)gn (x) =
n→∞ n→∞ n n
x
converges pointwise on E to lim = 0, but the convergence is not uniform as kfn gn − 0kE = ∞ for
n→∞ n
every n, which does not have limit 0.

49. For every x ∈ E,

|fn(x)gn (x) − f(x)g(x)| = |fn (x)gn(x) − f(x)gn (x) + f(x)gn (x) − f(x)g(x)|
≤ |fn (x) − f(x)||gn (x)| + |f(x)||gn(x) − g(x)|
≤ kfn − fkE kgnkE + kfkE kgn − gkE .

Then kfn gn − fgkE ≤ kfn − fkE kgnkE + kfkE kgn − gkE . Since f, g are bounded on E, kfkE , kgkE < ∞.
Since gn converges uniformly on E to g, so lim kgn − gkE = 0, which means there exists K such that
n→∞
n ≥ K implies kgn − gkE < 1. Then n ≥ K implies kgnkE ≤ kgn − gkE + kgkE < 1 + kgkE . So
kfn gn − fgkE ≤ kfn − fkE (1 + kgkE ) + kfkE kgn − gkE → 0 as n → ∞.
 By sandwich theorem, lim kfn gn − fgkE = 0. Therefore, fn gn converges uniformly on E to fg.
n→∞

50. Since fn converges uniformly on E to f, so lim kfn − fkE = 0, which means there exists K such that
n→∞
n ≥ K implies kfn − fkE < 1. Then for every x ∈ E and n ≥ K, we have

|fn (x)| ≤ kfnkE = kfn − f + f − fk + fk kE ≤ kfn − fkE + kf − fK kE + kfK kE ≤ 2 + MK .

Let M = max(M1 , . . . , MK−1, 2 + MK ), then for n < K and x ∈ E, we have |fn(x)| ≤ Mn ≤ M and for
n ≥ K and x ∈ E, we have |fn(x)| ≤ 2 + MK ≤ M. Therefore, fn is uniformly bounded on E.

51. For the first statement, we have k(fn +gn )−(f +g)kE ≤ kfn −fkE +kgn −gkE → 0 as n → ∞. Therefore,
lim k(fn + gn ) − (f + g)kE = 0. For the second statement, by the previous exercise, there exist M, N ∈
n→∞
R such that |fn (x)| ≤ M and |gn(x)| ≤ N for every n and x ∈ E. Then |f(x)| = lim |fn (x)| ≤ M for
n→∞
every x ∈ E. So kfkE ≤ M. Now

kfn gn − fgkE ≤ kfn gn − fgn kE + kfgn − fgkE

≤ kfn − fkE kgnkE + kfkE kgn − gkE
≤ kfn − fkE N + M kgn − gkE → 0 as n → ∞

by the uniform convergence of fn on E to f and gn on E to g. By sandwich theorem, lim kfn gn − fgkE = 0.

n→∞
Therefore, fn gn converges uniformly on E to fg.

X
∞
52. (a) If gk (x) converges uniformly on [0, 1], then by term test lim gk (x) = 0 for every x ∈ [0, 1]. So
k→∞
k=1
∞
X
lim cos gk (x) = 1 for every x ∈ [0, 1]. Then cos gk (x) cannot converge at any x ∈ [0, 1] by term test.
k→∞
k=1

(b) By the mean value theorem, | cos a − cos b| = |(− sin θ)(a − b)| ≤ |a − b| for some θ between a and
b. So k cos Sn − cos Sk[0,1] ≤ kSn − Sk[0,1] → 0 as n → ∞. Therefore, cos Sn (x) converges uniformly on
[0, 1] to cos S(x) by definition.

Kxn
53. Since F0(x) is bounded on [0, 1], let |F0(x)| ≤ K for all x ∈ [0, 1]. We will prove |Fn(x)| ≤
n!
for all x ∈ [0, 1] byZ mathematical Zinduction. The case n = 0 is true. Suppose the case n is true.
 x  x
Ktn Kxn+1
 
Then |Fn+1(x)| =  Fn (t) dt ≤ dt = for all x ∈ [0, 1], completing the induction.
0 0 n! (n + 1)!
K K
Since |Fn(x)| ≤ for all x ∈ [0, 1] and lim = 0, so the pointwise limit function on [0, 1] is
n! n→∞ n!
K
F (x) = lim Fn (x) = 0 by the sandwich theorem. Now |Fn(x) − F (x)| ≤ = Ln and lim Ln = 0.
n→∞ n! n→∞
K X∞
By L-test, Fn (x) converges uniformly on [0, 1] to F (x) = 0. Next |Fk (x)| ≤ = Mk and Mk = Ke.
k!
k=0
X∞
So Fk (x) converges uniformly on [0, 1] by Weierstrass’ M -test.
k=0

∞
X ∞
X ∞
X
54. Since |cnI(x−xn )| ≤ |cn| = Mn and Mn = |cn| < ∞, by Weierstrass’s M -test, f(x) = cnI(x − xn )
n=1 n=1 n=1
converges uniformly on (a, b). For every x 6= xn, all the terms I(x − xn ) are continuous at x. By the
∞
X
continuity theorem, f(x) = cnI(x − xn ) is continuous at x.
n=1
 Remarks. For the case (a, b) = (−∞, ∞), cn = 0.5n and xn is a countably infinite set S (for example
S = Q) arranged in a sequence, then f(x) above is a function that is continuous on R \ S. Note that this
k−1
X X∞
function is not continuous at every xk because cnI(x − xn ) + cnI(x − xn) is continuous at xk
n=1 n=k+1
by a similar proof as above and the term ck I(x − xk ) is discontinuous at xk . Hence f(x) is continuous
precisely on R \ S only. Can you find a function that is continuous on Q only? No. Why?

1 X n X ∞
55. First we consider the case when both sides are finite. Let gi ( ) = aij and gi(0) = aij . Then gi
n
j=1 j=1
is continuous on E = {0, 1, 12 , 13 , . . .}. Since aij ≥ 0, we have |gi(x)| ≤ gi (0) = Mi for every x ∈ E.
X∞ ∞ X
X ∞  X∞
Now Mi = aij < ∞. So gi (x) converges uniformly on E. By the continuity theorem for
i=1 i=1 j=1 i=1
uniform convergence,
∞ X
X ∞  X∞ X
∞ X
∞ X
∞
1
aij = gi(0) = lim gi(x) = lim gi (x) = lim gi( )
x→0 x→0 n→∞ n
i=1 j=1 i=1 i=1 i=1 i=1
∞ X
X n  n X
X ∞  ∞ X
X ∞ 
= lim aij = lim aij = aij .
n→∞ n→∞
i=1 j=1 j=1 i=1 j=1 i=1

In the case one of the side is infinite, say the left side. Then for every r > 0, there exists K such that
XK X∞  X∞ X ∞  X ∞ X K  X K X∞ 
aij > r. So aij ≥ aij = aij > r. Letting r → ∞, we get the right
i=1 j=1 j=1 i=1 j=1 i=1 i=1 j=1
side is infinite.

56. From the graph of f we see 0 ≤ f(x) < 1 and f is discontinuous only at the integers. Then on
1 1 2 k−1 1
[0, 1], k f(kx) is bounded and discontinuous at , , . . . , , 1. So k f(kx) is integrable on [0, 1].
2 k k k 2
1  1 X∞ X∞
1
 
Now  k f(kx) ≤ k = Mk and Mk = 1. So g(x) = f(kx) converges uniformly on [0, 1] by
2 2 2k
k=1 k=1
Weierstrass’ M -test. By the integration theorem, g(x) is integrable on [0, 1].

57. See the solution of Homework 2.

1 1 1
58. By calculus, fn (x) has a maximum value √ when x = √ and a minimum value − √ when
2 n n 2 n
1 1
x = − √ . For x ∈ R, f(x) = lim fn (x) = 0. Since |fn(x) − f(x)| ≤ Ln = √ → 0, by L-test, fn (x)
n n→∞ 2 n
0 1 − nx2
converges uniformly on R to f(x) = 0. Next fn (x) = . For x 6= 0,
(1 + nx2)2

1 − nx2
f 0 (x) = 0 = lim = lim fn0 (x).
n→∞ (1 + nx2 )2 n→∞

However, lim fn0 (0) = 1 6= 0 = f 0 (0).

n→∞

  X∞
1 − cos(x2 ) 1 − cos θ sin θ 1 (−1)k w2k
59. By l’Hopital’s rule, lim = lim = lim = . Since cos w =
x→0 x4 θ→0 θ2 θ→0 2θ 2 (2k)!
k=0
2
1 − cos(x ) 1  X
∞ k 2 2k 
(−1) (x ) X∞
(−1) k+1 4k−4
x
for all w ∈ R, so = 4 1− = for every x ∈ R \ {0}.
x4 x (2k)! (2k)!
k=0 k=1
 1 X∞
(−1)k+1 x4k−4
The power series also converges to when x = 0. So, by Abel’s limit theorem,
2 (2k)!
k=1
converges uniformly on [0, 1]. By the integration theorem for uniform convergence, we have
Z 1 Z ∞
1X X ∞ Z 1 X (−1)k+1 ∞
1 − cos(x2) (−1)k+1 x4k−4 (−1)k+1 x4k−4
dx = dx = dx = .
0 x4 0 k=1 (2k)! (2k)! (2k)!(4k − 3)
k=1 0 k=1

! ∞
2
ex − 1 2xex
2
X wk
60. By l’Hopital’s rule, lim = lim = 0 . Since ew = for all w ∈ R, so substituting
x→0 x x→0 1 k!
k=0
1 X (x2)k  X
2 ∞ ∞
2 ex − 1 x2k−1
w = x , we have = −1 = for every x ∈ R \ {0}. The power series also
x x k! k!
k=0 k=1
X ∞
x2k−1
converges to 0 when x = 0. So, by Abel’s limit theorem, converges uniformly on [0, 1]. By the
k!
k=1
integration theorem for uniform convergence, we have
Z 1 2 Z ∞
1X X ∞ Z 1 X 1 ∞
ex − 1 x2k−1 x2k−1
dx = dx = dx = .
0 x 0 k=1 k! 0 k! k!(2k)
k=1 k=1

2
x2k−1 ex − 1
Remarks: Since the terms are nonnegative continuous and the sum (with 0 at x = 0) is
k! x
continuous on [0, 1], uniform convergence on [0, 1] also follows from Dini’s theorem.

1 1
61. (a) On (0, 12 ], f(x) = |x ln(2x)| = −x ln(2x). So f 0 (x) = − ln(2x) − 1 = 0 implies x = . There |x ln(2x)| = .
2e 2e
ln(2x)  1/x  1
Next lim f(x) = lim = lim = 0 and f( 12 ) = 0. So the maximum value is .
x→0+ x→0+ −1/x x→0+ 1/x2 2e
∞
x ln(2x) X
(b) From (a), we will interpret x ln(2x) to be 0 at x = 0. On [0, 12 ], = (−1)k xk+1 ln(2x) by the
1+x
k=0

 1 X∞
1
geometric series formula. Since |(−1)k xk+1 ln(2x)| = xk x ln(2x)  ≤ k+1 = Mk and Mk = , the
2 e e
k=0
Z 1 k+2  1 Z 1 k+2
2 x 2 2 x 1 1
series converges uniformly on [0, 12 ]. Since xk+1 ln(2x) dx = ln(2x) − dx = − k+2 ,
0 k + 2 0 0 k + 2 x 2 (k + 2)2
by the integration theorem,
Z 1 Z 1 ∞
X ∞
X Z 1 ∞
X
2 x ln(2x) 2 2 (−1)k+1
dx = (−1)k xk+1 ln(2x) dx = (−1)k xk+1 ln(2x) dx = .
0 1+x 0 0 2k+2 (k
+ 2)2
k=0 k=0 k=0

Remarks: In (b), the uniform convergence part also follows from (a) that x ln(2x) is bounded on [0, 12 ]
and the geometric series converges uniformly on [0, 12 ] by Abel’s limit theorem.

∞
X 1
62. (a) Since xj converges pointwise on (−1, 1) to , by differentiation of power series,
j=0
1−x

d  1  d X j X d j X j−1 X
∞ ∞ ∞ ∞
1
2
= = x = x = jx = (k + 1)xk .
(1 − x) dx 1 − x dx dx
j=0 j=0 j=1 k=0
 √
∞
X h 1i √ x 2
2+k
(b) From part (a), we see that (k + 1)x converges pointwise on 0, to the function .
2 (1 − x)2
k=0 √
√ √
1
2+k h 1i X∞
(1/2) 2
2+k
Since |(k + 1)x | ≤ (k + 1) = Mk on 0, and Mk = , by Weierstrass’ M -test,
2 2
k=0
(1 − 12 )2
the convergence is uniform. So by the integration theorem,
Z 1 √ Z 1 ∞
X ∞ Z
X 1 ∞
X
2 x 2 2 √
2+k
2 √
2+k k+1
dx = (k + 1)x dx = (k + 1)x dx = √ √ .
0 (1 − x)2 0 0 ( 2 + k + 1)2 2+k+1
k=0 k=0 k=0
√
Remarks: In (b), the uniform convergence part also follows from the facts that x 2 is bounded on
X
∞
[0, 12 ] and the power series (k + 1)xk converges pointwise on [0, 12 ], hence uniformly by Abel’s limit
k=0
theorem. √
√
2+k x 2
Even easier, since the terms (k + 1)x are nonnegative continuous and the sum is
(1 − x)2
h 1i
continuous on 0, , uniform convergence also follows from Dini’s theorem.
2
sin x
63. (a) Since lim = 1 and lim x ln x = 0, so lim sin x ln x = 0.
x→0+ x x→0+ x→0+

d k e−1
(b) (x ln x) = xk−1(k ln x + 1) = 0 implies x = e−1/k , there the value is − . Also, lim xk ln x = 0
dx k x→0+
 (−1)n x2n+1 ln x  e−1
Mn+1
 
and 1k ln 1 = 0. So  ≤ = Mn for all x ∈ [0, 1]. Now lim =
(2n + 1)! (2n + 1)!(2n + 1) n→∞ Mn

2n + 1 X∞ X∞
(−1)n x2n+1 ln x
lim 2
= 0 < 1 implies M n converges. By Weierstrass’ M -test,
n→∞ (2n + 2)(2n + 3) (2n + 1)!
n=0 n=0
converges uniformly on [0, 1].
Z 1
xk+1 1 1 xk+1 1 1
(c) Since xk ln x dx =  −  =− , by (b) and the integration theorem,
0 k + 1 0 k + 1 k + 1 0 (k + 1)2
Z 1 Z ∞
1X X∞ Z 1 X∞
(−1)n x2n+1 ln x (−1)nx2n+1 ln x (−1)n+1
sin x ln x dx = dx = dx = .
0 0 n=0
(2n + 1)! n=0 0
(2n + 1)! n=0
(2n + 1)!(2n + 2)2

 2
1 − cos 2x  sin x if n = 0
64. (a) For f(x) = sin2 x = , we have f (n) (x) = (−1)k+1 22k−2 sin 2x if n = 2k − 1, k ∈ N . At
2 
 (−1)k+1 22k−1 cos 2x if n = 2k, k ∈ N
0 if n = 0 or 2k − 1, k ∈ N
x = 0, we get f (n) (0) = . Then the Taylor series of sin2 x at
(−1)k+1 22k−1 if n = 2k, k ∈ N
X∞
(−1)k+1 22k−1 2k
c = 0 is x . Since |f (n) (x)| ≤ 1 · 2n for all x ∈ (−∞, +∞), by the Taylor series
(2k)!
k=1
theorem,the Taylor series converges pointwise on (−∞, +∞) to sin2 x.
sin2 x X (−1)k+122k−1 2k−2
∞
(b) By (a), = x for every x ∈ (0, 1]. At x = 0, the series converges to 1.
x2 (2k)!
k=1
So the power series converges pointwise on [0, 1], hence uniformly by Abel’s limit theorem. By the
integration theorem,
Z 1 Z 1X
∞ X ∞ Z 1 X (−1)k+1 22k−1
∞
sin2 x (−1)k+1 22k−1 2k−2 (−1)k+1 22k−1 2k−2
dx = x dx = x dx = .
0 x2 0 k=1 (2k)! (2k)! (2k)!(2k − 1)
k=1 0 k=1
 ln(1 − x2 ) X x2k−2 ∞
65. (a) For x ∈ [ 13 , 23 ], = − . By Abel’s limit theorem (or Dini’s theorem), the power
x2 k
k=1
series converges uniformly on [ 13 , 23 ]. By the integration theorem,

Z 2 Z 2 ∞
X ∞ Z
X 2 ∞
X 22k−1 − 1
3 ln(1 − x2) 3 x2k−2 3 x2k−2
dx = − dx = − dx = − .
1 x2 1 k 1 k k(2k − 1)32k−1
3 3 k=1 k=1 3 k=1

 xk+3 k(k + 2)  X∞
xk+2

(b) Since lim  k+2  = |x| < 1 for x ∈ (−1, 1), so f(x) = converges
k→∞ (k + 1)(k + 3) x k(k + 2)
k=1
1 X∞
1
pointwise on (−1, 1). For x = , we get , hence it converges. By the differentiation of
2 k(k + 2)2k+2
k=1
X∞
0 xk+1
power series, f (x) = = −x ln(1 − x) on (−1, 1). So
k
k=1

X
∞ 1 Z 1 Z 1  12 Z 12 t2 −1
1 2
0
2 t2 
k+2
=f = f (t) dt = −t ln(1 − t) dt = − ln(1 − t) − − dt
k(k + 2)2 2 0 0 2 0 0 2 1−t
k=1
Z

 12
1
ln 2 1 2 1
ln 2 1 t2 5 − 6 ln 2
= + t+1− dt = + + t + ln(1 − t)  = .
8 2 0 1−t 8 2 2 0 16

Z 1 1 Z 1

66. (a) Applying integration by parts twice, we will see that u2 sin u du = u2(− cos u) + 2u cos u du
0
  Z 1 
0 0
1
= − cos 1 + 2 u sin u − sin u du = − cos 1 + 2(sin 1 + cos 1 − 1) = 2 sin 1 + cos 1 − 2.
0 0

X
∞
(−1)k u2k+3
(b) Using the Taylor series of sin x, we see u2 sin u = for all u ∈ R. By Abel’s limit
(2k + 1)!
k=0
theorem, the series converges uniformly on [0, 1]. By the integration theorem,
Z 1 Z 1X
∞ X ∞ Z 1 X ∞
(−1)k u2k+3 (−1)k u2k+3 (−1)k
u2 sin u du = du = du = .
0 0 k=0 (2k + 1)! (2k + 1)! (2k + 1)!(2k + 4)
k=0 0 k=0

X
∞ Z 1
(−1)k
Therefore, using (a), we get =2 u2 sin u du = 4 sin 1 + 2 cos 1 − 4.
(2k + 1)!(k + 2) 0
k=0

x3 x2
sin x − x cos x (x − 6 + · · ·) − x(1 − 2 + · · ·) x3 ( 13 + · · ·)
67. For x ∈ [−1, 1], tan x − x = = 2 . Note thex2
=
cos x 1 − x2 + · · · 1− 2 + · · ·
1 (sin x − x cos x)/x3 if x 6= 0
factor + · · · in the numerator converges pointwise on [−1, 1] to . By
3 1/3 if x = 0
tan x − x 1
. x2
1
Abel’s limit theorem, it is continuous at x = 0. So lim 3
= lim +··· lim 1 − +··· = .
x→0 x x→0 3 x→0 2 3
Similarly,
2
1 − cos x 1 − (1 − x2 + · · ·) 1
1
lim 2
= lim 2
= lim + ··· = ,
x→0 x x→0 x x→0 2 2
3
x − sin x x − (x − x6 + · · ·) 1
1
lim 3
= lim 3
= lim + ··· = .
x→0 x x→0 x x→0 6 6
 tan x − x tan x − x
. 1 − cos x
1.1 2 x − sin x x − sin x
. tan x − x

As x → 0, = → = and =
x(1 − cos x) x3 x2 3 2 3 tan x − x x3 x3
1.1 1
→ = .
6 3 2
∞
X
68. For k = 1, by the differentiation of power series, f 0 (x) = kak (x − c)k−1 for every x ∈ (a, b). So
k=1
f 0 (c) = a1 . Next, suppose the case k = n is true for all power series. For the case k = n + 1, since
X
∞ X
∞
f 0 (x) = kak (x − c)k−1 = (j + 1)aj+1(x − c)j , by the k = n case,
k=1 j=0

(f 0 )(n) (c) f (n+1) (c) f (n+1) (c)

(n + 1)an+1 = = ⇒ an+1 = .
n! n! (n + 1)!

 kek  1 X∞
1  xk+1 k 
   
69. Since lim   = < 1, by the ratio test, converges. For f(x), since lim   = |x| < 1
k→∞ (k + 1)ek+1 e kek k→∞ k + 1 xk
k=1
X ∞
xk
for x ∈ (−1, 1), so f(x) = converges pointwise on (−1, 1). By differentiation of power series,
k
k=1
X
∞
1
f 0 (x) = xk−1 = on (−1, 1). So
1−x
k=1
∞
X Z 1/e Z 1/e
1 1
1
0 1 1

k
= f = f − f(0) = f (t) dt = dt = − ln 1 − .
ke e e 0 0 1−t e
k=1

|x|2n+3 (2n − 1)(2n + 1) 2n − 1 2

70. (a) Applying the ratio test, lim 2n+1
= lim |x| = |x|2 < 1 for
n→∞ (2n + 1)(2n + 3) |x| n→∞ 2n + 3
∞ 
X (±1)2n+1  X 1
n+1 ∞
 (−1)
x ∈ (−1, 1). At x = ±1,  ≤ , which converges by p-test. The domain
n=1
(2n − 1)(2n + 1) n=1
n2
is [−1, 1].
X∞
(−1)n+1 x2n
(b) By the differentiation of power series, for x ∈ (−1, 1), f 0 (x) = = x arctan x.
n=1
2n − 1
(c) Since f(0) = 0, for every w ∈ (−1, 1), we have
Z w  w 1 Z w x2
x2  w2 1
f(w) = x arctan x dx = arctan x − 2
dx = arctan w − (w − arctan w).
0 2 0 2 0 1 + x 2 2

1 π 1 π π 1
Applying Abel’s limit theorem on [0, 1], we have f(1) = lim f(w) = − 1− = − .
w→1− 24 2 4 4 2
 
 x4(k+1) (4k)!  |x|4

4k  = lim
71. By ratio test, since lim  = 0 < 1 for every
k→∞  4(k + 1) ! x  k→∞ (4k + 4)(4k + 3)(4k + 2)(4k + 1)
X∞
x4k X∞
1
x ∈ (−∞, ∞), the series f(x) = converges pointwise on (−∞, ∞), in particular f(1) =
(4k)! (4k)!
k=0 k=0
converges. Applying differentation of power series four times, we find f(0) = 1, f 0 (0) = 0, f 00(0) =
0, f 000(0) = 0 and

d4  X x4k  X x4k−4 X x4j

∞ ∞ ∞
f (4) (x) = 4
1+ = = = f(x).
dx (4k)! (4k − 4)! (4j)!
k=1 k=1 j=0
By the given fact, f(x) = aex + be−x + c cos x + d sin x for some a, b, c, d ∈ R. Then

1 = f(0) = a + b + c, 0 = f 0 (0) = a − b + d, 0 = f 00 (0) = a + b − c, 0 = f 000 (0) = a − b − d.

∞
X
1 1 1 1 1 1 1
Solving these, we find a = , b = , c = , d = 0. Therefore, = f(1) = e + e−1 + cos 1.
4 4 2 (4k)! 4 4 2
k=0