Sei sulla pagina 1di 4

Math 104 - Homework 10 Solutions

Lectures 2 and 4, Fall 2011


1. Suppose that f : R R has the property that |f(x) f(y)| |x y|
2
for all x, y R. Show
that f is constant.
Proof. Fix y R. For x = y we have

f(x) f(y)
x y

|x y|, so |x y|
f(x) f(y)
x y
|x y|.
Since lim
xy
|x y| = 0 and lim
xy
|x y| = 0, the squeeze theorem for limits implies that
lim
xy
f(x) f(y)
x y
= 0.
Thus f is dierentiable at y and f

(y) = 0. Since this holds for all y R, f must be constant as


claimed.
2. Ross, 29.18. Let f be dierentiable on R with a = sup{|f

(x)| | x R} < 1. Select s


0
R and
dene s
n
= f(s
n1
) for n 1. Prove that (s
n
) is a convergent sequence. You may use the version
of Problem 4 on Homework 4 where we replace
1
2
by any 0 < c < 1 without justication.
Proof. Fix n 1. By the Mean Value Theorem, there is some y such that
s
n+1
s
n
= f(s
n
) f(s
n1
) = f

(y)(s
n
s
n1
).
Taking absolute values gives
|s
n+1
s
n
| = |f

(y)||s
n
s
n1
| a|s
n
s
n1
|.
By the same reasoning, |s
n
s
n1
| a|s
n1
s
n2
|, so
|s
n+1
s
n
| a|s
n
s
n1
| a
2
|s
n1
s
n2
|.
Continuing in this manner shows that
|s
n+1
s
n
| a
n
|s
1
s
0
| for all n.
Now, if s
1
= s
0
, then the sequence
s
0
, f(s
0
), f(f(s
0
)), . . .
in question is constant and hence converges. If s
1
= s
0
, |s
1
s
0
| = 0 so

s
n+1
s
1
s
0

s
n
s
1
s
0

a
n
.
Since 0 < a < 1, Problem 4 of Homework 4 implies that the sequence
_
sn
s
1
s
0
_
converges, and hence
this sequence times the constant s
1
s
0
converges as well, which is what we wanted to show.
3. Ross, 31.5. Let g(x) = e
1/x
2
for x = 0 and g(0) = 0.
(a) Show that g
(n)
(0) = 0 for all n. You may use the fact that lim
x0
e
1/x
2
x
k
= 0 for all k.
(b) Show that the Taylor series for g about 0 agrees with g only at x = 0.
Solution. (a) First, g is clearly dierentiable for x = 0 since it agrees with the function e
1/x
2
near
such values. At x = 0, we have
lim
x0
g(x) g(0)
x 0
= lim
x0
e
1/x
2
x
= 0.
Thus g is dierentiable everywhere and
g

(x) =
_
_
_
2
x
3
e
1/x
2
if x = 0
0 if x = 0.
Now, g

is clearly dierentiable for x = 0 and


lim
x0
g

(x) g

(0)
x 0
= lim
x0
2e
1/x
2
x
4
= 0.
Hence g is twice dierentiable and
g

(x) =
_
_
_

6
x
4
e
1/x
2
+
4
x
6
e
1/x
2
if x = 0
0 if x = 0.
Since
lim
x0
g

(x) g

(0)
x 0
= lim
x0
_

6
x
5
e
1/x
2
+
4
x
7
e
1/x
2
_
= 0
and g

is clearly dierentiable for x = 0, g is three times dierentiable and


g

(x) =
_
_
_
24
x
5
e
1/x
2

12
x
7
e
1/x
2

24
x
7
e
1/x
2
+
8
x
9
e
1/x
2
if x = 0
0 if x = 0.
Continuing in this manner, we can now see that the limit dening the n-th derivative of g at 0 will
involve terms that have e
1/x
2
divided by some power of x. Such limits are all zero, and so we
conclude that g
(n)
(0) = 0 for all n.
(b) Based on part (a), the Taylor series of g around 0 is

n=0
g
(n)
(0)
n!
x
n
= 0 + 0 + 0 + = 0.
This series converges to 0 for all values of x (since it is just the constant zero series), and hence
only agrees with g at x = 0 since g is only zero at x = 0.
4. Ross, 32.3. Let g(x) = x
2
for rational x and g(x) = 0 for irrational x.
(a) Calculate the upper and lower Darboux integrals for g on the interval [0, b].
(b) Is g integrable on [0, b]?
Proof. (a) Let P be any partition of [0, b]. By the denseness of R\Qin R, any subinterval determined
by P will contain an irrational number, and hence the inmum of g on any such subinterval will
be 0. Thus L(g, P) = 0 for any partition P and so L(g) = 0.
Now, we claim that U(g, P) = U(h, P) where h is the function dened by h(x) = x
2
for all x.
Indeed, on any subinterval determined by P, the supremum of g will be the same as the supremum
of h by the denseness of Q in R. To be precise, suppose that [x
i1
, x
i
] is such a subinterval. If the
right endpoint x
i
is rational, then the supremum of g and h on [x
i1
, x
i
] is x
2
i
. If x
i
is irrational,
certainly x
2
i
is an upper bound for the values of g on [x
i1
, x
i
]. For any 0 < < x
2
i
, there is a
rational r such that
_
x
2
i
< r < x
i
.
Then for this rational we have x
2
i
< g(r) = r
2
< x
2
i
, showing that x
2
i
satises the -
characterization of the supremum of g on [x
i1
, x
i
]. Thus sup g = sup h on [x
i1
, x
i
], so U(g, P) =
U(h, P) for any partition P. Hence the set of numbers of which U(g) is the inmum of is the same
as that for which U(h) is the inmum of, so U(g) = U(h) =
b
3
3
.
(b) Since L(g) = 0 =
b
3
3
= U(g), g is not integrable on [0, b].
5. Ross, 32.6. Let f be a bounded function on [a, b]. Suppose there exist sequences (U
n
) and
(L
n
) of upper and lower Darboux sums for f such that lim(U
n
L
n
) = 0. Show f is integrable and
_
b
a
f = limU
n
= limL
n
.
Proof. To be precise, suppose that U
n
is the upper sum corresponding to the partition P
n
and that
L
n
is the lower sum corresponding to the partition Q
n
. Let > 0 and pick N large enough so that
U
N
L
N
= U(f, P
N
) L(f, Q
N
) < .
Then for the partition P = P
N
Q
N
, we have U(f, P) U(f, P
N
) and L(f, Q
N
) L(f, P), so
U(f, P) L(f, P) U(f, P
N
) L(f, Q
N
) < ,
showing that f satises the -characterization of Darboux integrability. Hence f is integrable, so
U(f) = L(f).
Now, x > 0 and pick N such that U
n
L
n
< for n N. Since U(f) = L(f) is the
supremum of all lower sums, we have
U
n
U(f) U
n
L
n
< for n N.
This shows that (U
n
) converges to U(f). Similarly, since L(f) = U(f) is the inmum of all upper
sums, we have
L(f) L
n
U
n
L
n
< for n N,
so L
n
L(f). We conclude that
_
b
a
f = U(f) = L(f) = limU
n
= limL
n
as claimed.
6. Let f : [0, 1] R be my favorite function, dened for x [0, 1] by
f(x) =
_

_
0 if x R\Q
1
q
if x =
p
q
where p, q Z have no common factors, q > 0, and p = 0
1 if x = 0.
Show that f is integrable and compute
_
1
0
f(x) dx.
Proof. Fix > 0. Then there are only nitely many positive integers smaller than 2/, and hence
only nitely many rational numbers in [0, 1] whose denominators are smaller than 2/. (The reason
is the same as that given in the proof of problem 2 on homework 7.) Denote the rationals whose
denominators are smaller than 2/ by
p
1
q
1
, ,
p
N
q
N
,
so that there are N of them. Note that we then have f(p
i
/q
i
) = 1/q
i
> /2 for each i = 1, . . . , N.
Surround each of these N points by an interval of length less than /2N and not containing any
of these other N points. (This is possible because there is a distance of at least 1/ max{q
1
, . . . , q
N
}
between any of these N points, as in the proof of problem 2 on homework 7.) Let P = {x
0
, . . . , x
m
}
be the partition given by the endpoints of these intervals. Note that since any interval of positive
length contains an irrational number, we have that
inf
t[x
i
,x
i+1
]
f(t) = 0
for any points x
i
and x
i+1
in our partition. Thus the lower Darboux sum of f with respect to P is
0.
Now, split up the upper Darboux sum into a sum over those intervals which do contain one of
the p
i
/q
i
and a sum into those which do not:
U(f, P) =
m1

k=0
_
sup
t[x
k
,x
k+1
]
f(t)
_
(x
k+1
x
k
)
=

intervals containing a p
i
/q
i
_
sup
t[x
k
,x
k+1
]
f(t)
_
(x
k+1
x
k
) +

rest
_
sup
t[x
k
,x
k+1
]
f(t)
_
(x
k+1
x
k
).
All of the supremums in this rst piece are <= 1 and there are N such intervals, so the rst piece
is less than

intervals containing a p
i
/q
i
(x
k+1
x
k
) <
N

k=1

2N
=

2
.
Note that here we have used the fact that the intervals containing some p
i
/q
i
have length smaller
than /2N.
All the supremums in the second piece are <= /2 since the intervals in the rst sum contain
all the rationals where this fails. Since the sum of the lengths in the intervals in this second piece is
less than the length of the entire interval [0, 1], we have that the second piece is less than or equal
to

rest

2
(x
k+1
x
k
) =

2

rest
(x
k+1
x
k
) <=

2
.
Thus the full upper Darboux sum is less than /2 + /2 = , and hence for this partition P,
U(f, P) L(f, P) = U(f, P) 0 = U(f, P) < ,
showing that f is integrable. Looking at the lower Darboux sums, we see that L(f) = 0, so
_
1
0
f(x) dx = 0.

Potrebbero piacerti anche