Вы находитесь на странице: 1из 10

MA104 SOLUTIONS HW5 - HW9

1. HW5
Problem 17.8: (a) Show that min(f, g) =
1
2
(f + g)
1
2
|f g|.
Proof: The equality is true for all pairs of real numbers f and g, by examining
cases. If one wishes to think of them as living in the range of pair of functions,
then so be it.
To say an equation is true for two functions f and g is to say that it holds for
all numbers of the form f(x) and g(x), with x in the common domain of f and g.
(b) Show that min(f, g) = max(f, g).
Proof: The smaller of the two numbers f and g is bigger when you reverse the
sign. To recover the original number, one must again reverse the sign.
(c) Show that if f and g are continuous at x, then min(f, g) is continuous at x.
Proof: Take an arbitrary x
n
x; then
1
2
(f(x
n
) + g(x
n
))
1
2
|f(x
n
) g(x
n
)|
1
2
(f(x) + g(x))
1
2
|f(x) g(x)|.
This is a consequence of the limit law for multiplication and addition, together
with the fact that x
n
x |x
n
| |x|. Of course, we are using that f and g are
continuous when we assert f(x
n
) f(x) and g(x
n
) g(x). Translating via part
(a), we have veried that min(f, g) is continuous using the sequential denition of
continuity (def 17.1).
Problem 17.13: Show x
Q
(x) and x x
Q
(x) are discontinuous on R and
R

:= R {0}, respectively.
Proof: Recall the notation I am using from class... for any subset E R we can
dene the characteristic function of E. So
E
(x) = 1 if x E and 0 otherwise.
Also, we will use a simple D E when pointing from domain to range, and the
x f(x) arrow to specify the image of an element. Make sure you understand why
my statement of the problem and the books are identical.
Of course we are going to keep using the x
n
x notation for a sequence (x
n
)
converging to x. If you contest this based on the fact that the symbolism overlaps
with the domain range convention, then I remind you of your elementary
school days when f(x) was confused to mean f x.
1
2
Let x R. Then there exist both a sequence of rationals q
n
x and a sequence
of irrationals
n
x. (In fact, we could take
n
= q
n
+

2
n
.) Now

Q
(q
n
) = 1 and
Q
(
n
) = 0 n N.
Either
Q
(x) = 1 or 0, so we obtain failure of sequential continuity at x, using the
(q
n
) or (
n
) appropriately. For x x
Q
(x) proceed identically, for x = 0.
To see that it is indeed continuous at zero, let x
n
0. Then obviously
0 |x
n
||
Q
(x
n
)| |x
n
|
so that the squeeze theorem implies |x
n

Q
(x
n
)| 0. Since 0 0, we are done.
Perhaps a better way to think of this problem is involving density of Q. In
general, we have said that a set A is dense in B if every element of B can be
obtained as a limit of elements in A. (If we havent said this, then we should start
saying it.)
So once you have dened a function on a dense subset, if you would like to extend
this function to the whole set in a continuous manner, then there is at most one
way (possibly no way) to do this. For the extension to be sequentially continuous,
all decisions are already made by determining the value of the function on the
dense subset. The value of f(x) simply must be limf(x
n
) where x
n
x comes
from A.
Now consider an arbitrary
E
, E R. When restricted to E alone, it is in-
distinguishable from the constantly 1 function. In fact, the function R R given
by x 1 is a perfectly continuous extension of
E
|
E
to all of R. (As usual, the
notation f|
A
means the restriction of the domain of f to the set A.) However it is
not the same as extending it to be zero o of E, which gives the real
E
.
In fact, the uniqueness of such a continuous extension, and the fact that f(x) =
1 x R does extend the function
Q
|
Q
to all of R in a continuous manner, yields
a roundabout proof for this problem. In fact,
Q
disagrees with x 1 on a dense
subset, the irrationals.
Problem 17.9: (c) Show the function given by f(x) = xsin(
1
x
) for x = 0 and
f(0) = 0 is continuous at x = 0.
Proof : Note | sin( )| 1. Let > 0. Choose = . Then if 0 < |y 0| < , we
have
0 |y| | sin(1/y)| = |f(y) f(0)| |y| < = .
The proof is over, and you shouldnt need me to tell you that.
(d) Show f(x) = x
3
is continuous for all x.
Proof: Let > 0, and x x R. Choose
= min
_
1,

3(|x| + 1)
2
_
.
3
(Our choice here for is not the best one. However, it makes things easy, as you
may see below. The demand that also be less than 1 is a common trick. Where
does it get used??)
Then if y (x , x + ), we have
|f(y) f(x)| = |y
3
x
3
| = |y x||x
2
+ xy + y
2
| |y x| (|x|
2
+|x||y| +|y|
2
)
via applications of the triangle inequality and of multiplicative properties of absolute
value. Now we chose so that the inequality |y| < |x| + 1 will automatically hold.
Therefore, we can simplify the RHS extensively with the inequality
|y x| (|x|
2
+|x||y| +|y|
2
) < |y x| 3(|x| + 1)
2
which reveals the original motivation for choice of . Altogether, we have
|f(y) f(x)| < |y x| 3(|x| + 1)
2
< 3(|x| + 1)
2


3(|x| + 1)
2
3(|x| + 1)
2
= .
Problem 17.10: (a) Show f =
(0,)
is not continuous at zero.
Proof: Take x
n
= 1/n going to 0, then f(x
n
) = 1, but f(0) = 0.
(b) Show g(x) = sin(1/x) extended via 0 0 is not continuous at 0.
Proof: Take x
n
=
1
2n+/2
going to 0, and proceed as in (a).
(c) Show f(x) =
x
|x|
extended via 0 0 is not continuous at 0.
Proof: See (a).
(d) Any nonconstant integer valued function dened on all of R is discontinuous
somewhere. (You could take < 1/2, or work out a sequence.) Now think about
why dierentiability assumptions in discrete mathematics can lead to astrology-like
predictions if one is not careful.
The fact that a continuous integer valued function on a connected domain must
be constant is crucial to virtually every modern algebraic or dierential topology
result.
Problem 18.4: Show that if S R does not contain one of its limit points, then
there exists a continuous unbounded function with domain S. (Recall that a limit
point x for a set E has a sequence of points x
n
= x, x
n
E, such that x
n
x. In
general, the closure of a set of real numbers is the set together with its limit points,
that is
E = E limit points of E.
Equivalently, a set is closed if and only if it contains all of its limit points. So one
thing we get from the hypothesis is that S cannot be a closed set. Do not confuse
this with S being an open set; there are sets that are neither open nor closed.)
4
Proof: We will simply write the formula for such a function. For s S, dene
f(s) =
1
|sx|
, where x S S is a limit point of S not contained in S. Then f is
continuous on R {x}, being the composition of continuous functions
s s x, | |, and ( )
1
.
The only restriction is that we cannot invert zero; i.e. |s x| = 0; i.e. s = x.
Hence f is continuous on S R {x}. If x
n
x, x
n
S, x
n
= x, then
f(x
n
) =
1
|xnx|
+, since |x
n
x| 0
+
. In particular, f is unbounded.
Problem 18.9: Prove that a polynomial function of odd degree has at least one
real root.
Proof: First we argue that all polynomials are continuous. (It is okay if you simply
said this was obvious, but here is why it is obvious.) Note that a suitable deni-
tion of a polynomial is simply a nite composition of multiplication and addition
operations, either by variables or constants. In any event all of these maps preserve
limits of sequences; this is the content of the Limit Laws for multiplication and
addition. Hence all polynomials p satisfy x
n
x = p(x
n
) p(x). Further we
can conclude that if q is another polynomial, then x
p(x)
q(x)
is continuous on the
domain {x : q(x) = 0}.
Of interest to note is that we cannot prove at this stage that the zero set of
a polynomial must be nite, without resorting to an abstract algebra argument.
This is one of many problems completely resolved by the technical apparatus of the
calculus; i.e. derivatives, integrals, Rolles Theorem, etc. These topics will be the
last third of the course.
Write an arbitrary odd degree polynomial
p(x) = a
2k+1
x
2k+1
+ ... + a
0
,
where a
0
, ..., a
2k
R and a
2k+1
= 0 are the coecients. Noting that p(r) = 0 i
p(r)
a
2k+1
= 0 we may assume for simplicity that the polynomial is monic. That is, it
has leading coecient 1. Now write
p(x) = x
2k+1
_
1+
a
2k
x
+
a
2k1
x
2
+ ... +
a
0
x
n
_
.
As x , the term in parentheses goes to 1, being the sum of 1 and nitely
many terms that go to 0. In particular, we may choose X > 0 so that ( ) is
larger than 1/2. Hence
p(X) > X
2k+1

1
2
> 0.
Likewise, we may choose Y > 0 so that
p(Y ) < (Y )
2k+1

1
2
< 0.
Now apply the IVT to the domain interval [Y, X] and the range value 0
(p(Y ), p(X)).
Problem 18.12 (b) Show that the function g in 17.10(b) has the intermediate
value property on R.
5
Proof: Recall g(x) = sin(1/x) for all x = 0, and g(0) = 0. It suces to show g has
the IVP on all intervals of the form [a, b] for a, b 0, since if 0 / [a, b] then we
may simply apply the IVT to g on [a, b], as it is continuous everywhere but zero.
However, this is immediate because the range of g is [1, 1], and it takes on every
value in its range on any interval of the form [
1
2n
,
1
2n
] for n N, by applying
the usual IVT.
In other words, consider y between g(a) and g(b) and choose n large enough so
that
1
2n
< max(a, b). We may assume at least one of a or b is nonzero. Then y
[1, 1], so we may choose c [
1
2(n+1)
,
1
2n
] (or c [
1
2n
,
1
2(n+1)
] if necessary)
so that g(c) = y.
Problem 19.9: (a),(b),(c): Consider the function f as in 17.9 (c). Recall
f(x) = xsin(1/x) for x = 0
f(0) = 0.
Show that f is uniformly continuous on all of R.
Proof: Recall that, rewriting
xsin(1/x) =
sin(1/x)
1/x
,
we get that
lim
x
f(x) = 1,
from freshman calculus (L-H rule?!?!...). Now for any > 0, we may choose N >
1 > 0 so that
|x| > N = f(x)
_
1

2
, 1 +

2
_
.
Now f is uniformly continuous on [2N, 2N] since it is a closed bounded interval.
Choose < 1 according to uniform continuity with the desired precision . So, for
all x, y [2N, 2N] with |y x| < , we have |f(y) f(x)| < .
Now consider any pair x, y R with |y x| < . Then either both x and y are
outside the interval [N, N], or inside the interval [2N, 2N], so that this yields
uniform continuity on all of R.
2. HW7
Problem 19.1: I wont retype the functions, but (a), (b), (c), and (g) are all
uniformly continuous on the given domains. This is because if f : [a, b] R
is continuous, it is automatically uniformly continuous. So if D [a, b] is ever
considered, then f restricted to D will also be uniformly continuous. Reread the
denition of uniform continuity very carefully if that last sentence is confusing.
Now (e) and (f) are not uniformly continuous precisely because they cannot be
extended to the closures of their domains in a continuous manner. For f(x) = x
3
on all of R, we do not have uniform continuity. This example I will work in detail:
6
Consider the sequences x
n
= n and y
n
= n +
1
n
. Then d(x
n
, y
n
) =
1
n
0. (I am
using the metric space notation for |x
n
y
n
|!!) But
d(f(x
n
), f(y
n
)) =
_
n +
1
n
_
3
n
3
= 3n +
3
n
+
1
n
3
+.
So f cannot be uniformly continuous. (Remember the useful sequential criterion
that can conclude NOT uniform continuity... you may reference these things when
solving problems.)
Problem 19.6: Let f(x) =

x, f : [0, ) [0, ). Show f

(x) is unbounded
on (0, 1] but f is nonetheless uniformly continuous on (0, 1]. Also show that f is
uniformly continuous on [1, ).
Proof: From calculus, we know f

(x) =
1
2

x
, and this is obviously unbounded by
considering x 0. But the function [x

x] is continuous on all of [0, 1], so we
know it must be uniformly continuous on [0, 1], hence also on (0, 1] [0, 1]. This
shows we cannot have a converse to Theorem 19.6 in the book.
Technically the simple continuity of this function requires that we observe it is
the inverse to the strictly monotone [x x
2
], which is itself continuous by the limit
laws. Such an inverse must be continuous, c.f. Thm 18.4 in the text. However, I
think it is acceptable to claim this fact as obvious for exam or HW purposes.
To establish uniform continuity on [1, ), let > 0. Notice that, for x, y [1, ),
we have
|f(x) f(y)| = |

y|
|

x +

y|
|

x +

y|
=
|x y|
|

x +

y|
.
Since

x,

y 1, we also have that


1

x+

y

1
2
. Choose = 2 > 0. If |xy| < ,
combining the above yields
|f(x) f(y)|
|x y|
2
<

2
= .
Problem 19.7: Show that if a continuous f : [0, ) R is uniformly continuous
on [k, ) for some k > 0, then f is uniformly continuous on [0, ). Deduce that
x

x as in 19.6 is uniformly continuous on its domain.


Proof: Let > 0, and let
1
> 0 be such that
x, y [k, ), |x y| <
1
|f(x) f(y)| < .
(Here we are using the assumption of uniform continuity.) We also know f is
continuous on the interval [0, k +
1
], hence it is uniformly continuous on that
interval. So we may choose
2
according to precision, and this time
x, y [0, k +
2
], |x y| <
2
|f(x) f(y)| < .
Now take = min(
1
,
2
), and we will show this > 0 suces to guarantee f is
uniformly continuous on all of [0, ).
7
Suppose x, y [0, ), and that |x y| < . Now x and y are at least within
1
of
each other, so that they must both lie in one of the sets [0, k +
1
], or [k, ). (Why?
Because x [0, k) and y (k +
1
, ) |x y| >
1
. Draw a picture if you are
confused.) In the case that x, y [0, k +
1
] we use the fact that |x y| <
2
,
and in the other case we use that |x y| <
1
. In both cases |f(x) f(y)| <
follows from the prior choice of
1,2
.
Hence, to establish that the square root function is uniformly continuous on [0, ),
we can combine the results of the previous problem to this one. In particular, we
take k to be 1.
Problems 20.1 - 20.9, 20.18: I will handwrite these and distribute in lecture
7/31. These are basic limit computations and should be considered review. Remem-
ber in the simplest denition of a general limit of a function, we just use sequences.
But one also must specify the set S (usually the domain of the function) from which
the sequences can be taken, as in the notation
lim
xa
S
f(x).
These techniques will need to be fresh on your mind especially when we start
derivatives and integrals, after midterm 2.
Problem 21.6: Show that the composition of continuous functions is continuous,
for arbitrary metric spaces.
Proof: Write f : X Y and g : Y Z. (We are suppressing the notation for the
particular metrics on X, Y, Z...because we do not need it! This argument holds for
general topological spaces, not just metric spaces.) We must show the function
g f : X Z
is continuous, assuming f and g are continuous. Let U be an arbitrary open set in
Z. Then we must show
(g f)
1
(U) = {x X : g(f(x)) U}
is an open set in X. Now
g(f(x)) U f(x) g
1
(U),
by denition. For simplicity, abbreviate the set g
1
(U) = V. We know V is open,
using the assumption that g is continuous. Again by denition, we have
f(x) V x f
1
(V ).
The set f
1
(V ) is open, since V is open, and f is continuous. The above equiva-
lences show that the set
{x : g(f(x)) U} = {x : f(x) g
1
(U)} = {x : x f
1
(g
1
(U))}
is exactly f
1
(V ), and so we are done.
Problem 21.11: Show that there do not exist surjective continuous functions
mapping [0, 1] (0, 1), and [0, 1] R.
8
Proof: These are both simple consequences of the fact(s) from class that the
continuous image of a compact set must be compact, together with
E R is compact E is closed and bounded.
The denition of compact we have been using is really sequential compactness. We
say E X is compact if, for all sequences (x
n
) taken from E, there is a convergent
subsequence (x
n
k
) whose limit is in E. Note that E compact automatically implies
E is closed and bounded; it is the converse that is not true in general (in particular,
we have seen it is not true for C[0, 1]). You should be able to prove this, that is,
compact closed and bounded.
One can also argue this by hand as follows:
Assume for contradiction that a surjective continuous f : [0, 1] (0, 1) exists.
Then
sup{f(x) : x [0, 1]} = sup (0, 1) = 1.
We know that c [0, 1] with f(c) = 1, since [0, 1] is closed and bounded, and f is
continuous. But then 1 f([0, 1]) = (0, 1), a contradiction.
The assumption that a surjective continuous f : [0, 1] R exists is even easier to
contradict, since the hypothesis implies its range is unbounded. This is not possible,
since we know continuous functions, on closed bounded intervals, are bounded.
3. HW8
Problem 22.12: See the Addendum to Problem 22.12, which I may or may not
have completed as you read this.
Problems 23.1-23.2: Again, these are basic interval of convergence calculations
from freshman calculus, meant as a review. I will handwrite solutions and distribute
them in lecture.
Problem 23.9: Let the sequence (f
n
) C[0, 1] be given by the formula
f
n
(x) = nx
n
.
One could also write this as
f
n
= [x nx
n
].
Show that f
n
(x) 0 x [0, 1). (Notice that this is an assertion about a sequence
of real numbers (f
n
(x)); there is one such sequence for every x. For x = 1 the
sequence is simply (n), which diverges.)
Observe also that
_
[0,1]
f
n
1. (This is again an assertion about a sequence of real
numbers, that (
_
f
n
) is the constant sequence of all 1s).
Proof:
9
We use the hint and observe that
|(n+1)x
(n+1)
|
nx
n
= |x|
n+1
n
|x| < 1, so that
limf
n
(x) = 0 for all such x. The point of this problem really is to provide you with
a good example of a function that converges pointwise to zero, but whose integrals
are always 1.
Note: For the following problems involving sequences of functions, I am using
the arrow notation from class, because it helps in summarizing the data that are
implicit in the phrase f is a function. Remember that the only way to tackle these
problems is through careful attention to denitions.
Problem 24.1: Show that the sequence of functions given by
f
n
=
_
x
1 + 2 cos
2
(nx)

n
_
converges uniformly to the identically zero function (denoted 0). So what you are
required to show can be rewritten:
d

__
x
1 + 2 cos
2
(nx)

n
_
, [x 0]
_
= d

(f
n
, 0) 0.
We would say the sequence f
n
converges to 0 = [x 0] in the metric space
C
b
(R), d

.
What does the b mean? If all of R is considered,
d

(f, g) := sup{|f(x) g(x)| : x R}


may be innite, because we are not guaranteed that arbitrary continuous functions
on R are bounded (e.g. f(x) = x). This condition is automatic for continuous func-
tions on a closed bounded interval. Most of the time this technicality is addressed
by considering C
b
(S), the set of continuous functions on S that are also bounded.
Notice that when S = R, this has the unfortunate consequence of outlawing many
nice functions, like the polynomials. So one way to succinctly collect these facts
into a mathematical statement would be C
b
[0, 1] = C[0, 1], but C
b
(R) = C(R).
(However, C
b
(S) C(S) for all sets S, trivially.)
However, these considerations are moot in our particular example, so it doesnt
really matter.
Proof:
Let > 0, and choose N N such that n > N
3

n
< . Then, observing
|1 + 2 cos()|

n

|1| + 2| cos()|

n

3

n
we have that d

(f
n
, 0) 0 since the bound on the RHS is independent of x.
Problem 24.2: Let f
n
= [x
x
n
]. For each x R, compute limf
n
(x). Let the
function f = [x limf
n
(x)]. Show that f
n
converges uniformly to f on [0, 1], but
not on all of [0, ).
10
Proof:
For any xed x R, the sequence
x
n
0. (This is the Archimedean property!).
For x [0, 1] we have the bound
|x|
n

1
n
and so we conclude that d

(f
n
, 0) goes to zero if the supremum is taken over
x [0, 1]. Notice that where x was on the real line played a crucial role... the
bound |x| 1 is chosen because the domain is [0, 1]
This gives us a clue that f
n
does not converge uniformly on all of R. Indeed, if
it did, it would have to converge to its pointwise limit function. However,
sup
_
|x|
n
: x R
_
= +,
so this is not the case.
Problem 24.6: For
f
n
=
_
x (x 1/n)
2

, (f
n
) C[0, 1],
determine the pointwise limit function
f = [x limf
n
(x)].
Do the f
n
converge uniformly to f?
Response: The pointwise limit function is clearly [x x
2
]. We compute, for
x [0, 1]:
|(x 1/n)
2
x
2
| = |2 x 1/n 1/n
2
| 2/n + 1/n
2
.
Again, the bound on the RHS is independent of x and tends to 0, so we conclude
that f
n
does converge uniformly to f(x) = x
2
on [0, 1].
4. HW9
Problem 27.2: Show that, for any continuous f : R R, there exists a sequence
(p
n
) of polynomials such that p
n
converges uniformly to f on any given bounded
subset of R.
Proof: For each n N, choose a polynomial p
n
such that |p
n
(x) f(x)| <
1
n
for
all x [n, n]. This is possible by the Stone Weierstrass Theorem.
Now let E R be any bounded subset, and select N
1
>> 0 so that E
[N
1
, N
1
]. Let > 0 and choose N
2
so that
1
N2
< . Then for all n > max{N
1
, N
2
}
we have that |p
n
(x) f(x)| <
1
n
< for all x E [N
1
, N
1
] [n, n], so we are
done.

Вам также может понравиться