Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

Solutions to Fall 2006 Math 301 Midterm Problems

1. (a) F : R
n
R
m
is dierentiable at a = (a
1
, . . . , a
n
) R
n
if and only if there exists a linear transfor-
mation T : R
n
R
m
such that lim
xa
F(x) F(a) T(x a)
x a
= 0.
(b) Solution 1 for (0, 0). F(0, 0) = 0. F(x, 0) = 0 implies
F
x
(0, 0) = 0. Similarly, F(0, y) = 0 implies
F
y
(0, 0) = 0. So T(x, y) = 0. Now
F(x, y) F(0, 0) T(x 0, y 0)
_
x
2
+ y
2
=
|xy|
(x
2
+ y
2
)
3/2
=
| cos sin |
r
=
1
2r
+ as r 0, =

4
.
So F is not dierentiable at (0, 0).
Solution 2 for (0, 0). lim
(x,x)(0,0)
xx
x
2
+ x
2
=
1
2
= 0 = F(0, 0). So F is not continuous at (0, 0). Hence F
is not dierentiable at (0, 0).
Solution 1 for (1, 1). For (1, 1),
F
x
=
y(x
2
+ y
2
) xy(2x)
(x
2
+ y
2
)
2
=
y
3
x
2
y
(x
2
+ y
2
)
2
and
F
y
=
x(x
2
+ y
2
) xy(2y)
(x
2
+ y
2
)
2
=
x
3
xy
2
(x
2
+ y
2
)
2
are continuous in B((1, 1),

2). So F is C
1
at (1, 1). By C
1
-theorem, F is dierentiable
at (1, 1).
Solution 2 for (1, 1). F(1, 1) =
1
2
. F(x, 1) =
x
x
2
+ 1
implies
F
x
(1, 1) =
(x
2
+ 1) 2x
2
(x
2
+ 1)
2

x=1
= 0. Simi-
larly, F(1, y) =
y
1 + y
2
implies
F
y
(1, 1) =
(1 + y
2
) 2y
2
)
(1 + y
2
)
2

y=1
= 0. So T(x, y) = 0.
Using x 1 = r cos and y 1 = r sin , we have (x, y) (1, 1) is the same as r 0 and
F(x, y) F(1, 1) T(x 1, y 1)
_
(x 1)
2
+ (y 1)
2
=
|(xy)/(x
2
+ y
2
) 1/2|
_
(x 1)
2
+ (y 1)
2
=
(x y)
2
2(x
2
+ y
2
)
_
(x 1)
2
+ (y 1)
2
=
r(cos sin )
2
2
_
(1 + r cos )
2
+ (1 + r sin )
2
_ 0 as r 0.
So F is dierentiable at (1, 1).
2. Solution 1 (using inverse function theorem). Dene F : R
2
R
2
by F(x, y) = (v, w), where
v = e
x
2 cos y and w = sin x 2e
y
. Now
v
x
= e
x
,
v
y
= 2 siny,
w
x
= cos x,
w
y
= 2e
y
are continuous near (x, y) = (0, 0). So F is C
1
near (x, y) = (0, 0). Also, det F

(0, 0) =

1 0
1 2

= 2 = 0.
By the inverse function theorem, near (x, y) = (0, 0), F
1
(v, w) = (x, y) exists and is dierentiable.
Also,
_ x
v
(1, 2)
x
w
(1, 2)
y
v
(1, 2)
y
w
(1, 2)
_
=(F
1
)

( F(0, 0)
. .
=(1,2)
) =
_
F

(0, 0)
_
1
=
_
1 0
1 2
_
1
=
1
2
_
2 0
1 1
_
=
_
1 0

1
2
1
2
_
.
1
Solution 2 (using implicit function theorem). Dene F : R
4
R
2
by F(x, y, v, w) = (f
1
, f
2
),
where f
1
(x, y, v, w) = v e
x
+ 2 cos y and f
2
(x, y, v, w) = w sin x + 2e
y
. Let p = (0, 0, 1, 2), then
F(p) = (0, 0). Now
f
1
x
= e
x
,
f
1
y
= 2 sin y,
f
1
v
= 1,
f
1
w
= 0,
f
2
x
= cos x,
f
2
y
= 2e
y
,
f
2
v
= 0,
f
2
w
= 1
are continuous near p. So F is C
1
near p. Next
(f
1
, f
2
)
(x, y)
(p) =

1 0
1 2

= 2 = 0. By the implicit
function theorem, near p, (x, y) can be expressed as a dierentiable function of (v, w). Using formula to
compute partial derivatives, we have
x
v
=
(f
1
, f
2
)
(v, y)
_
(f
1
, f
2
)
(x, y)
=

1 0
0 2

_
2 = 1,
x
w
=
(f
1
, f
2
)
(w, y)
_
(f
1
, f
2
)
(x, y)
=

0 0
1 2

_
2 = 0,
y
v
=
(f
1
, f
2
)
(x, v)
_
(f
1
, f
2
)
(x, y)
=

1 1
1 0

_
2 =
1
2
,
y
w
=
(f
1
, f
2
)
(x, w)
_
(f
1
, f
2
)
(x, y)
=

1 0
1 1

_
2 =
1
2
.
3. Solution 1. Since

k=1
(1)
k1
x
k
k
converges pointwise on [0, 1] to ln(1 +x), by the Abel limit theorem,
the convergence is uniform. Let h : [0, 1] [0, 1] be dened by h(x) = x
1/2
=

x. By the substitution
property, we have

k=1
(1)
k1
x
k/2
k
converges uniformly on [0, 1] to ln(1 +

x). By the integration
theorem for uniform convergence,
_
1
0
ln(1+

x) dx =
_
1
0

k=1
(1)
k1
x
k/2
k
dx =

k=1
_
1
0
(1)
k1
x
k/2
k
dx =

k=1
(1)
k1
k(
k
2
+ 1)
=

k=1
(1)
k1
2
k(k + 2)
.
Solution 2. Let

x = u. Then we have
_
1
0
ln(1 +

x) dx =
_
1
0
2uln(1 + u) du. From the Taylor series
of ln(1 + x), we see that

k=1
2(1)
k1
u
k+1
k
converges pointwise on [0, 1] to 2uln(1 + u). By the Abel
limit theorem, the convergence is uniform on [0, 1]. By the integration theorem for uniform convergence,
_
1
0
ln(1+

x) dx =
_
1
0
2uln(1+u) du =
_
1
0

k=1
2(1)
k1
u
k+1
k
du =

k=1
_
1
0
2(1)
k1
u
k+1
k
du =

k=1
(1)
k1
2
k(k + 2)
.
Solution 3. From the Taylor series of ln(1 +x), we have

k=1
(1)
k1
x
k/2
k
converges pointwise on [0, 1]
to S(x) = ln(1 +

x). Let S
n
(x) =
n

k=1
(1)
k1
x
k/2
k
. As convergent alternating series, S(x) is between
S
n
(x) and S
n+1
(x). For all x [0, 1], |S
n
(x) S(x)| |S
n+1
(x) S
n
(x)| =
|x|
(n+1)/2
n + 1

1
n + 1
= L
n
.
Since lim
n
L
n
= 0, by the L-test,

k=1
(1)
k1
x
k/2
k
converges uniformly on [0, 1] to ln(1 +

x). By the
integration theorem for uniform convergence,
_
1
0
ln(1+

x) dx =
_
1
0

k=1
(1)
k1
x
k/2
k
dx =

k=1
_
1
0
(1)
k1
x
k/2
k
dx =

k=1
(1)
k1
k(
k
2
+ 1)
=

k=1
(1)
k1
2
k(k + 2)
.
2
Remarks. Since

k=1
(1)
k1
k
= ln2 and
2
k(k + 2)
=
1
k

1
k + 2
, we have

k=1
(1)
k1
2
k(k + 2)
=

k=1
(1)
k1
_
1
k

1
k + 2
_
=

k=1
(1)
k1
k

k=1
(1)
k1
k + 2
= ln2
_
ln21+
1
2
_
=
1
2
.
4. (a) S
n
: E R converges uniformly on E to S : E R if and only if lim
n
S
n
S
E
= 0, where
f
E
= sup{|f(x)| : x E} (or for every > 0, there exists K N such that n K implies for every
x E, |S
n
(x) S(x)| < .)
(b) (i) By the continuity theorem for uniform convergence, S(x) is continuous on [a, b]. By the extreme
value theorem, there exists x
0
[a, b] such that m
0
= inf{S(x) : x [a, b]} = S(x
0
) > 0.
For = m
0
/2 > 0, there is K
1
N such that n K
1
implies for every x [a, b], |S
n
(x) S(x)| < m
0
/2.
Now S
n
(x) S(x) |S
n
(x) S(x)|. So for every n K
1
and x [a, b],
S
n
(x) S(x) |S
n
(x) S(x)| m
0
(m
0
/2) = m
0
/2 > 0.
(ii) For the K
1
in (i), we note that if n K
1
,

1
S
n
(x)

1
S(x)

=
|S
n
(x) S(x)|
S
n
(x)S(x)

|S
n
(x) S(x)|
m
2
0
/2
.
For every > 0, there is K
2
N such that n K
2
implies for every x [a, b], |S
n
(x) S(x)| < m
2
0
/2.
Let K = max(K
1
, K
2
). Then n K implies

1
S
n
(x)

1
S(x)

=
|S
n
(x) S(x)|
S
n
(x)S(x)
<
m
2
0
/2
m
2
0
/2
= .
5. Solution 1. By the M
k
theorem, there is a subsequence {x
rj
} such that lim
j
x
rj
= limsup
n
x
n
= a.
Similarly, by the m
k
theorem, there is a subsequence and {x
sk
} such that lim
k
x
sk
= liminf
n
x
n
= b.
Next let a < c < b. Assume there is no subsequence of {x
n
} converging to c. Then for some i N,
there is no term of {x
n
} in
_
c, c +
1
i
_
. Since lim
n
(x
n+1
x
n
) = 0, there is K N such that n > K
implies |x
n+1
x
n
| <
1
i
. Since lim
j
x
rj
= a < c, we can take a term x
rj
c with r
j
> K.
By induction, we shall show x
n
c for all n r
j
. The reason is as follow: the case n = r
j
is true.
Assuming x
n
c for n r
j
> K, we have x
n+1
< x
n
+
1
i
c +
1
i
and x
n+1

_
c, c +
1
i
_
, which imply
x
n+1
c.
Now x
n
c for all n r
j
implies there is no subsequence of {x
n
} converging to b > c, contradiction.
Solution 2. By the M
k
theorem, there is a subsequence {x
rj
} such that lim
j
x
rj
= limsup
n
x
n
= a.
Similarly, by the m
k
theorem, there is a subsequence and {x
sk
} such that lim
k
x
sk
= liminf
n
x
n
= b.
Next let a < c < b. For every i N, we claim there is x
ni
such that c x
ni
c +
1
i
and the indices
n
i
s are strictly increasing. (Then by the sandwich theorem, we will get lim
i
x
ni
= c.)
To get such a x
ni
, cut [a, b) into t subintervals [a, a + ), [a + , a + 2), . . . , [a + (t 1), a + t),
each with length =
b a
t

1
2i
. Since
_
c, c +
1
i
_
has length
1
i
2, it contains at least one of the
subintervals, say
_
c, c +
1
i
_
[a + d, a + (d + 1)). Note d = 0 since c > a.
Since lim
n
(x
n+1
x
n
) = 0, there is K N such that n > K implies |x
n+1
x
n
| < . Since
lim
j
x
rj
= a, there are innity many r
j
such that x
rj
(a, a+). Similarly, since lim
k
x
sk
= b, there
3
are innity many s
k
such that x
sk
(b, b+) = (a+(t1), b+). Take one such r
j
with r
j
> K. Then
take one such s
k
with s
k
> r
j
. Consider the terms x
rj
, x
rj+1
, x
rj+2
, . . . , x
sk
. Since x
rj
< a + a + d
and x
sk
> a+(t 1) a+d, there is a largest t {r
j
, r
j
+1, r
j
+2, . . . , s
k
1} such that x
t
< a+d.
Since t r
j
> K implies |x
t+1
x
t
| < , we must have x
t+1
[a + d, a + (d + 1))
_
c, c +
1
i
_
. Let
x
ni
= x
t+1
.
Finally, since t r
j
and there are innitely many r
j
we can choose, we may choose n
i
s strictly
increasing by choosing r
j
> n
i1
every time.
4

You might also like