Math 320, Real Analysis I

Midterm Exam Solutions

Part I: Show me one!

Give an example of each object described below, or explain why no such object exists. In the case where no object exists, you need only cite the relevant theorem. In the case where you do provide an example, you must explain why your example does what you claim it does. These questions are worth 5 points each, for a total of 30 points. 1. A bounded sequence that is not convergent. Example: (an ) = ((1)n ). This sequence is bounded by M = 1 > 0, for |an | = |(1)n | = 1 for all n N. However, the sequence diverges since its subsequences (1)2n +1 and (1)2n1 1 converge to dierent limits. 2. An unbounded sequence with a convergent subsequence. Example: Consider the sequence (an ) = (1, 1 , 2, 1 , 3, 1 , 4, 1 , 5, 1 , . . . ), i.e., 1 2 3 4 5 n + 1 if n is odd an = 2 2 if n is even n

Then (an ) is unbounded, since for any M > 0 there is a natural number N N such that N > M by the Archimedean Property, part (i), and so |a2N 1 | = |N | = N > M . However, 1 the subsequence (a2n ) = ( n ) 0 as we have shown many times. (For > 0, take N N 1 such that N < , which exists by the Archimedean Property, part (ii). Then, for n N , we 1 1 have |a2n 0| = | n 0| = n < , so lim a2n = 0 as claimed.) 3. A subset of R which has exactly four limit points.
1 1 Example: Consider A = { n : n N} {2 + n : n N} {4 + We claim that the set of limit points of A is LA = {0, 2, 4, 6}. 1 n 1 n

: n N} {6 +

: n N}.
1 n

1 1 1 First, 0 = lim n implies 0 LA . Similarly, 2 = lim 2 + n , 4 = lim 4 + n , and 6 = lim 6 + imply that 2, 4, 6 LA , so {0, 2, 4, 6} LA .

Now suppose c {0, 2, 4, 6}. We claim that this implies that c is not a limit point of A. Consider = min{|c 0|, |c 2|, |c 4|, |c 6|}, so > 0. Then, for 3 > 0, there exists N N 1 1 such that N < 3 , which implies that k + n V/3 (k) whenever n N , for k = 0, 2, 4, 6. Thus there are only nitely many elements of A not in V/3 (0) V/3 (1) V/3 (2) V/3 (3), and V/3 (c) V/3 (k) = for k = 0, 2, 4, 6. So, there are at most nitely many x A such that x V/3 (c), say x1 , x2 , . . . , xn , so let = min{/3, |c x1 |, |c x2 |, . . . , |c xn |} if c A or, if c A (say c = x1 , wlog) then let = min{/3, |c x2 |, . . . , |c xn |}. In either case, we nd that V (c) A = or {c}, depending as whether c A or c A. In either case, it is evident that c is not a limit point of A. Therefore, LA = {0, 2, 4, 6}, so A has exactly four limit points. 4. A compact subset of R that is disconnected. Example: The Cantor set, C, is compact but not connected. It is compact because it is closed (the arbitrary intersection of the closed sets Cn is closed) and bounded (by M = 1 > 0 since C [0, 1]). However, it is not connected since the sets A = C [0, 1/3] and B = C [2/3, 1] are separated and C = A B.

Math 320, Real Analysis I

Midterm Exam Solutions

5. A connected subset of R that is not compact. Example: The set E = (0, 1) is connected since it is an interval, but not compact since it is not closed. 6. A function f that is continuous at 1, 0, and 1, but is not continuous anywhere else. Example: Dene f (x) = Heres an attempt at the graph of f . 2 1 x3 x 0 xQ xQ

2 1 2

We claim that f is continuous at 1, 0 and 1. First observe that x3 x is a polynomial, and hence is continuous on R, which implies that
x1

lim x3 x = (1)3 (1) = 0,

x0

lim x3 x = (0)3 (0) = 0,

x1

lim x3 x = (1)3 (1) = 0.

Let > 0 be arbitrary. Then there exist > 0, > 0, and > 0 such that |x (1)| < = |x3 x| = |(x3 x) 0| < , |x (0)| < = |x3 x| = |(x3 x) 0| < , |x (1)| < = |x3 x| = |(x3 x) 0| < . Now consider the function f near the points 1, 0, and 1. For |x (1)| < , we have |f (x) f (1)| = |f (x) (0)| = |f (x)| |x3 x| < , so f is continuous at 1. Similarly, for |x 0| < , |f (x) f (0)| = |f (x) (0)| = |f (x)| |x3 x| < , so f is also continuous at 0. When |x 1| < , we have |f (x) f (0)| = |f (x) (0)| = |f (x)| |x3 x| < , so f is continuous at 1 as well. However, we claim that f is not continuous anywhere else. Thus, let c R with c {1, 0, 1}. For each n N, by the Densities of Q and I in R, there is a rational number rn and an irrational 1 1 number zn such that c n < rn < c < zn < c + n . Clearly (rn ) c and (zn ) c, but
3 lim f (rn ) = lim(rn rn ) = c3 c = 0 = lim 0 = lim f (zn ),

so f is not continuous at c.

Math 320, Real Analysis I

Midterm Exam Solutions

Part II: -proofs

Give careful, detailed > 0, N N or > 0 proofs that the following limits have the stated values. These questions are worth 5 points each, for a total of 20 points. 1. lim 2n + 1 =2 n+4

7 Proof. Let > 0 be arbitrary. Then 4 is a real number, so there exists N N such that 7 N > 4 by the Archimedean Property, part (i). Then, for all n N , 7 4<n Therefore, lim = 7 < n+4 = 2n + 1 7 2 = < . n+4 n+4

2n + 1 = 2. n+4

2. lim

2n2 + 1 2 = 3n2 20 3 1 3 43 + 20 3 1 3 is a real number, so there is a natural

Proof. Let > 0 be arbitrary. Then number N N such that

N > max 2,

43 + 20 3

Then, for all n N , n > 2 so 3n2 20 > 0 and 1 3 43 + 20 3 <n = 43 < 3n2 20 3 = 2n2 + 1 2 43 = < . 2 20 2 20) 3n 3 3(3n

Therefore, lim

2n2 + 1 2 = . 2 20 3n 3

3. lim

x5

x5 1 = 2 25 x 10

Proof. Let > 0 be arbitrary. Let = min{1, 90}, so > 0. Then, for 0 < |x 5| < , we have 4 < x < 6 since 1, which implies that |x + 5| = x + 5 > 9. Therefore, when x5 1 0 < |x 5| < , x = 5 which implies that 2 = , so x 25 x+5 x5 1 1 1 [10] [x + 5] |x 5| 90 = = = < = . x2 25 10 x + 5 10 10(x + 5) 10|x + 5| 10[9] 90 Therefore, lim
x5

x5 1 = . x2 25 10

4. lim

x0

x5 1 = x2 25 5

Math 320, Real Analysis I

Midterm Exam Solutions

Proof. Let > 0 be arbitrary. Let = min{1, 20}, so > 0. Then, for 0 < |x 0| = |x| < 1, we have 1 < x < 1, so x = 5 and |x + 5| = x + 5 > 4. Therefore, when 0 < |x 0| = |x| < , x5 1 2 25 x 5 = = Therefore, lim [5(x 5)] [x2 25] 5x 25 x2 + 25 x(x 5) = = 2 25) 5(x 5(x 5)(x + 5) 5(x 5)(x + 5) |x| 20 < = . 5|x + 5| 5[4] 20

x0

x5 1 = . x2 25 5

Part III: Problems

The point value of each of the problems below is indicated. Choose problems that add up to 50 points (and no more!) to write up. 1. (10 points) Prove that the series 1+ converges. Proof. We observe that this is a rearrangment of the Alternating Harmonic Series, (1)n+1 1 1 1 1 1 = 1 + + + n 2 3 4 5 6 n=1

1 1 1 1 1 1 1 1 + + + + + 3 2 5 7 4 9 11 6

but we cant give in to the temptation to use the Rearrangement Theorem and call it a day, for the Alternating Harmonic Series is only conditionally convergent. So we need another plan. Even though we cant say our given series is equal to the Alternating Harmonic Series, we will do well to recall that the Alternating Harmonic Series does converge by the Alternating Series Test. Call this sum S. Now, following the outline given by Abbott in Section 2.1, we also know that the series 1 2 n=1

(1)n+1 n

1 1 1 1 1 1 + + + 2 4 6 8 10 12

converges with sum 1 S by the Algebraic Limit Theorem for Series (Theorem 2.7.1), part (i). 2 Our next move will be to insert zeros in this series to space its terms out so that we can then do term-by-term addition of the Alternating Harmonic Series and its mutated relative, 0+ 1 1 1 1 1 1 0 +0+ 0 +0+ 0 + 2 4 6 8 10 12

which still converges to 1 S, since the zeros wont change the limit of the sequence of partial 2 sums. If we now add these series term-by-term, using the Algebraic Limit Theorem for Series (Theorem 2.7.1), part (ii), we have 1 1 1 1 1 1 1 1 1 1 1 (1+0)+( + )+( 0)+( )+( +0)+( + )+ = (1)+(0)+( )+( )+( )+(0)+ 2 2 3 4 4 5 6 6 3 2 5

Math 320, Real Analysis I

Midterm Exam Solutions

which must still converge with sum S + 1 S = 3 S. Then, just as we can insert zeros without 2 2 altering convergence or the value of the sum, we may also eliminate them, so we conclude that 1 1 1 1 1 1 + + + + 3 2 5 7 4 does converge with sum 3 S. 2 2. (10 points) Suppose an 0 for all n and a2 converges. Show that n an n also converges. Proof. We begin by recalling a device from Calculus III. If x = [x1 , x2 , . . . , xn ], y = [y1 , y2 , . . . , yn ] Rn , then their dot product is dened to be xy = ||x|| ||y|| cos , where is the angle between the vectors. While we dene the dot product in this way, we nd it much easier to compute using the formula x y = x1 y1 + x2 y2 + + xn yn . Therefore, |x1 y1 + x2 y2 + + xn yn | ||x|| ||y|| = x2 + x2 + + x2 n 1 2
2 2 2 y1 + y2 + + yn

This is known as the Cauchy-Schwarz Inequality. With this in hand, the problem isnt all that bad. an n converges. By denition, the series converges if and only if its sequence of partial sums, (sm ), converges, where m m an 1 = sm = an n n n=1 n=1 Suppose an 0 for all n and a2 converges. We want to prove that the series n for all m N. Heres where the Cauchy-Schwarz Inequality comes in, as well as our knowledge 1 that both of the series a2 and n n2 converge (the rst by hypothesis and the second as it is a p-Series with p = 2 > 1). Consider
m

|sm | =
n=1

an

1 n

a2 n
n=1 n=1

1 n2

a2 n
n=1 n=1

1 n2

and this last term is a nite constant that is independent of m. Hence, for all m N,
1 we have |sm | M , where M = a2 n n2 > 0 is a xed real number. Thus (sm ) is a bounded sequence. Furthermore, since we are told that an 0 for all n, so (sm ) is a ak+1 0 for all k implies sk+1 sk for all k. monotone increasing sequence since sk+1 sk = k+1 Therefore, by the Monotone Convergence Theorem, since (sm ) is monotone increasing an and is bounded, it converges. Hence the series also converges. n

Math 320, Real Analysis I

Midterm Exam Solutions

3. (20 points) [0, 1]. This have D0 D1 D2 D3 . . .

Repeat the Cantor construction from Section 3.1 starting with the unit interval time, however, remove the middle fourth from each component. Thus we would 0 0 0 . . . . . . 9/64 15/64 3/8 3/8 5/8 5/8 49/64 55/64 . . . . . . 1 1 1

. . . Dening a

. . . . . . . . . Cantor-like set: D = Dn n=0

(a) Is the resulting set D = Dn compact? If yes, prove it. If no, explain why not. n=0 Yes, the resulting set D = Dn is compact. n=0 Proof. It is closed since, for each n the set Dn is the union of nitely many closed intervals and thus Dn is also closed, so D, which is the intersection of the closed sets Dn is also closed. Furthermore, D [0, 1] is bounded, so by the Heine-Borel Theorem, D is compact. (b) Is the resulting set D = Dn uncountable? If yes, prove it. If no, explain why not. n=0 Yes, the resulting set D = Dn is uncountable. n=0 Proof. We will show that D = is perfect, and hence is uncountable by Theorem 3.4.3. As we noted already in part (a), D is closed. Also, 0 Dn for all n implies that 0 D, so D = . Therefore, we only need to prove that D has no isolated points. Let x D. Then x Dn for all n N. For each n N, Dn is the union of closed intervals, so x must be in one, which we will call [an , bn ]. Dene a sequence (xn ) by xn = an bn if x = an if x = an

Then, xn D for all n, since the endpoints of intervals are never removed in the process of creating D, xn = x for all n by construction, and (xn ) x since the widths of the intervals [an , bn ] is decreasing to zero as n . Therefore, we conclude that x D is a limit point of D, so it is not an isolated point. Therefore, D does not contain any isolated points, since x D is arbitrary. Thus D is perfect and nonempty, so D is uncountable. (c) Compute the length of the Cantor-like set D = Dn . n=0 Solution: We compute the length of all of the removed intervals and then say that the length of D is 1 minus the sum of the lengths of the intervals removed. In the n-th step, n1 3n1 we remove a total of 2n1 intervals, each of length 1 3 = 23n1 , so the sum of the 4 8 lengths of the intervals removed is

2n1
n=1

3n1 3n1 3n1 1 = = = 3n1 2n n 2 2 4 4 n=1 n=1 n=1

3 4

n1

[1/4] = 1. 1 [3/4]

Therefore, the total length removed is 1, so the length of D must be 1 1 = 0.

Math 320, Real Analysis I

Midterm Exam Solutions

4. (20 points) Let f : A R and let c A be a limit point of A. Call the function f tinconuous at c if, for all > 0, there exists > 0 such that whenever x A and 0 < |x c| < , it follows that |f (x) f (c)| < . (a) If f is tinconuous at c, must f be continuous at c? If yes, prove it. If no, give a counterexample. No, if f is tinconuous at c, we cannot conclude that f is continuous at c. Heres a Counterexample: Let f : R R be the Dirichlet function, f (x) = 1 0 xQ xQ

and let c = 0. Then f is tinconuous at c = 0 because, for any > 0 we may be given, taking = 2 > 0 we have for any x R, |x c| = |x 0| < = |f (x) f (0)| = |f (x) 1| = |(1) 1| = 0 < 2 = x Q |(0) 1| = 1 < 2 = x Q

but f is not continuous at c = 0 since the Dirichlet function is nowhere-continuous on R. (b) If f is continuous at c, must f be tinconuous at c? If yes, prove it. If no, give a counterexample. No, if f is continuous at c, we cannot conclude that f is tinconuous at c. Heres a Counterexample: Let f : (0, ) R be given by f (x) = 1/x and let c = 1. Then, f is continuous at c = 1, as we have shown in class (or because it is a rational function whose denominator is not zero when c = 1). However, f is not tinconuous at c = 1, for when we let = 1 > 0, it is impossible to nd an > 0 such that 0 < |x 1| < = 1 1 implies |f (x) f (1)| = | x 1| < . To see this, suppose > 0 is given. Then there exists N N such that N > by the Archimedean Property, part (i). Then a = N 1 +1 satises a (0, ) and 0 < |a 1| < = 1, but |f (a) f (1)| = |(N + 1) (1)| = N > . Thus f (x) = 1/x is continuous at c = 1, but it is not tinconuous at c = 1. Call the function f toncinuous at c if there exists > 0 such that, for all > 0, 0 < |x c| < and x A implies |f (x) f (c)| < . (a) If f is toncinuous at c, must f be continuous at c? If yes, prove it. If no, give a counterexample. No, if f is continuous at c, we cannot conclude that f is toncinuous at c. Heres a Counterexample: Let f : R R be the Dirichlet function, f (x) = 1 0 xQ xQ

and let c = 0. Then f is toncinuous at c = 0 because, if we take = 2 > 0, then for any > 0 we have |x c| = |x 0| < = |f (x) f (0)| = |f (x) 1| = |(1) 1| = 0 < 2 = x Q |(0) 1| = 1 < 2 = x Q

Math 320, Real Analysis I

Midterm Exam Solutions

but f is not continuous at c = 0 since the Dirichlet function is nowhere-continuous on R. (b) If f is continuous at c, must f be toncinuous at c? If yes, prove it. If no, give a counterexample. No, if f is continuous at c, we cannot conclude that f is toncinuous at c. Heres a Counterexample: Let f : R R be given by f (x) = x and let c = 0. Then f is continuous at c = 0 since f is a polynomial. However, f is not toncinuous at c = 0. Suppose it were, and let > 0 be chosen as in the denition so that, for all > 0 we have x R and 0 < |x 0| < implies |f (x) f (0)| < . In particular, for = + 1 > 0, it must be the case that |f (x) f (0)| < whenever 0 < |x 0| < = + 1. Consider a = + 0.5, which satises a R and 0 < |a 0| < + 1. However, |f (a) f (0)| = |a 0| = |a| = + 1 > , which contradicts our assumption. Therefore, 2 f is not toncinuous at c = 0. 5. (20 points) Suppose A R, f is a function from A to R, and c is a limit point of A. (a) Give a formal denition of the one-sided limits
xc+

lim f (x)

and

xc

lim f (x).

Denitions: We say that limxc f (x) = L if, for every > 0, there exists > 0 such that for any x A, c < x < c (i.e., < x c < 0) = |f (x) L| < .

We say that limxc+ f (x) = L if, for every > 0, there exists > 0 such that for any x A, c < x < c + (i.e., 0 < x c < ) = |f (x) L| < . (b) Prove that lim f (x) exists if and only if both one-sided limits exist are they are equal.
xc

Proof. (=) Suppose that limxc f (x) exists and is equal to L. Let > 0 be arbitrary. Then there exists a > 0 such that for any x A, 0 < |x c| < = |f (x) L| < .

Therefore, if x A and < x c < 0, then 0 < |x c| < , so |f (x) L| < . Since our > 0 is arbitrary, we conclude that limxc f (x) = L exists. Similarly, for our given > 0 and the > 0 above, whenever x A and 0 < x c < , so 0 < |x c| < , we have |f (x) L| < . Hence limxc+ f (x) = L as well. Therefore, if limxc f (x) exists, the both limxc f (x) and limxc+ f (x) exist and are both equal to L, and thus to one another. (=) Suppose both one-sided limits exist and are equal to L. Let > 0 be arbitrary. Then, as limxc f (x) = L, there exists 1 > 0 such that 1 < x c < 0 and x A implies that |f (x) L| < . Similarly, since limxc+ f (x) = L, there exists 2 > 0 such that for any x A with 0 < x c < 2 , we have |f (x) L| < . Now take = min{1 , 2 }. Suppose x A and 0 < |x c| < . Then either 1 < x c < 0 or 0 < x c < 2 . In either case, we conclude |f (x) L| < , so limxc f (x) = L, since > 0 was arbitrary.

Math 320, Real Analysis I

Midterm Exam Solutions

6. (20 points) A real-valued function dened on an interval I is said to be convex if, given any two points x1 , x2 I and any t [0, 1], we have f ( (1 t)x1 + tx2 ) (1 t)f (x1 ) + tf (x2 ). (a) Interpret this inequality geometrically in terms of the graph of f . Solution: It says that the line segment joining the points (x1 , f (x1 )) and (x2 , f (x2 )) always lies on or above the graph of y = f (x). (b) Let a I be a xed point; for any x I, x = a, consider ma (x) = f (x) f (a) , xa

so that ma (x) is the slope of the secant line connecting (a, f (a)) to (x, f (x)). Show that if f is convex, then ma (x) is an increasing1 function. Proof. Suppose f is convex and let a I be a xed point. Consider the function ma (x) = f (x)f (a) . Suppose x, y A and x y. We need to show that ma (x) ma (y), xa i.e., that ma (y) ma (x) 0. Suppose a < x < y. Then there exists t (0, 1) such that x = (1 t)a + ty, so convexity implies f (x) = f ((1 t)a + ty) (1 t)f (a) + tf (y) = f (a) + t[f (y) f (a)]. Thus f (x) f (a) t[f (y) f (a)]. Also x = (1 t)a+ ty = a+ t(y a) implies x a = t(y a), so ma (y) ma (x) = f (y) f (a) f (x) f (a) f (y) f (a) f (x) f (a) = ya xa ya t(y a) f (y) f (a) t[f (y) f (a)] = 0. ya t(y a)

Next assume x < y < a. Then there exists s (0, 1) such that y = (1 s)x + sa, or ay = (1s)[xa]. Then f (y) (1s)f (x)+sf (a), so f (y)f (a) (1s)[f (x)f (a)] or (1 s)[f (a) f (x)] f (a) f (y). Thus ma (y) ma (x) = f (a) f (y) f (a) f (x) f (y) f (a) f (x) f (a) = ya xa ay ax (1 s)[f (a) f (x)] f (a) f (x) = 0. (1 s)[a x] ax

Finally, we consider the case where x < a < y, so there exists r (0, 1) such that a = (1 r)x + ry. Then f (a) (1 r)f (x) + rf (y), so ma (y) ma (x) = = f (y) f (a) f (x) f (a) ya xa f (y) [(1 r)f (x) + rf (y)] f (x) [(1 r)f (x) + rf (y)] y [(1 r)x + ry] x [(1 r)x + ry] (1 r)[f (y) f (x)] r[f (x) f (y)] f (y) f (x) f (x) f (y) = = 0. (1 r)[y x] r[x y] yx xy

A function g : A R is said to be increasing if, for all x, y A, xy = f (x) f (y).

Math 320, Real Analysis I

Midterm Exam Solutions

(c) Conversely, show that if ma (x) is increasing for every a I, then f is convex. Proof. Suppose that ma (x) is an increasing function for every a I. Let x, y I and let t (0, 1). Without loss of generality, suppose x < y and set z = (1 t)x + ty. Consider mz , which is increasing. Thus mz (x) mz (y) = = = = = f (x) f (z) f (y) f (z) xz yz f (z) f (x) f (y) f (z) zx yz f (z) f (x) f (y) f (z) [(1 t)x + ty] x y [(1 t)x + ty] f (z) f (x) f (y) f (z) t[y x] (1 t)[y x] (1 t)[f (z) f (x)] t[f (y) f (z)]

= f (z) (1 t)f (x) + tf (y) which is what we wanted to show. Therefore, since x, y I and t (0, 1) are arbitrary, we conclude that f is a convex function. (d) Show that if f is convex, then it is continuous. Proof. Let a I. Suppose a is an interior point of I and select a small enough > 0 such that a [a , a + ] I. Since ma (x) is an increasing function, by part (b), for x [a , a + ] the maximum value of |ma (x)| is either |ma (a )| or |ma (a + )|. Call this maximum slope M . Then, for x (a , a + ), we have |f (x) f (a)| = |ma (x) [x a]| M |x a|, which implies that f is continous at a. If a is an endpoint of I, dont worry about it.